Research on Women's Issues in Transportation - Volume 2: Technical Papers (2005)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

INJURY PREVENTION AND ERGONOMICS 98709mvpTxt 89_154 9/20/05 5:13 PM Page 125

98709mvpTxt 89_154 9/20/05 5:13 PM Page 126

1 2 7 Safety of Pregnant Drivers in the United Kingdom B. Serpil Acar and Alix M. Weekes, Loughborough University, United Kingdom It is widely accepted that women have a differentdriving style and travel patterns from those of men,whereas pregnant women have yet again a different set of travel patterns and preferences. Pregnancy can cause a wide range of symptoms and physical changes that are not limited to the abdominal region. Research to date has not considered the real-life experiences and problems of car travel during pregnancy. The project Automotive Design: Incorporating the Needs of Preg- nant Women at Loughborough University addresses issues such as seat belt safety, behaviors, and needs in a holistic manner for the first time and provides explicit information about pregnant women. A pregnancy and driving questionnaire is used to investigate how U.K. women’s experiences of driving and using passive safety systems (seat belts, airbags, and head restraints) are affected during car travel. The main safety concerns found in 450 completed questionnaires were low levels of correct seat belt and head restraint positioning and proximity to the steering wheel and airbags. The cor- rect position for the shoulder section of the seat belt is between the breasts and for the lap section around the abdomen and across the hips underneath the abdomen. Some U.K. pregnant women used the correct position for the shoulder belt, and others positioned the lap belt correctly across the hips. However, the seat belts are designed to protect the car occupant when used cor- rectly, not correctly in part. Therefore this study is focused on correct usage of the entire seat belt. Certain factors seem to influence correct seat belt positioning positively, and this information could be used to target schemes to provide seat belt information. Targeting information to women in their first pregnancy will improve seat belt positioning for that first pregnancy and will help in subsequent pregnancies. Pregnant women commonly reported concern that the seat belt was incorrectly positioned, and they felt unsafe while using the seat belt. In some cases women took action to alleviate this fear, for example, by ceasing to use the seat belt or by holding it. This is evidence that women modify their seat belt behavior for protection during pregnancy but may actually put themselves at greater risk of injury. The majority of women in their third trimester of pregnancy were seated with their abdomen less than 25 cm from the steering wheel because of abdominal protrusion. This problem is counteracted by moving the seat rearward, but that results in difficulty reaching the pedals. More suitable designs would help women to increase their steering wheel clearance while maintaining their ability to reach the pedals. All the information about pregnant women’s experiences of using passive safety systems is presented as an informa- tion catalog for automotive designers as part of this project. This catalog includes guidelines to aid future vehicle design concepts with the aim of improving car travel for pregnant women. Detailed findings of this project have been submitted for publication or can be obtained from the authors. 98709mvpTxt 89_154 9/20/05 5:13 PM Page 127

1 2 8 Pregnant Women and Safety Belts What Do We Know? Laurie F. Beck, Ruth A. Shults, and Brenda Colley Gilbert, Centers for Disease Control and Prevention Injuries are a leading cause of death among pregnant women, and motor vehicle crashes are a leading cause of hospitalized injuries during pregnancy. The protective effect of safety belts for pregnant women and fetuses has been well documented. Self-reported data from two pop- ulation-based surveys were used to examine safety belt use among reproductive-aged women and prenatal coun- seling about safety belts during pregnancy. The preva- lence of safety belt use among reproductive-aged women ranged from 70% to 91% across 19 states. The preva- lence of counseling about safety belts during pregnancy ranged from 37% to 57%. Younger, non-Hispanic black, and less educated reproductive-aged women were less likely to use seat belts. Pregnant women with these char- acteristics were more likely than older, non-Hispanic white, and more educated women to receive counseling about safety belt use. Population-based data on safety belt use among pregnant women are needed. Because belt use may change as the pregnancy advances, it should be measured during various stages of pregnancy. Adher- ence to counseling guidelines is low and should be increased. Provider counseling should be used in combi- nation with effective tools such as legislation and high- visibility law enforcement, and the impact of counseling should be rigorously evaluated. There are more than 4 million live births in theUnited States each year (Ventura et al. 2004).Every year, more than 1 million women of reproductive age (15 to 44 years) are treated in emer- gency departments for motor vehicle crash injuries, and more than 6,000 women are killed [Centers for Disease Control and Prevention (CDC) 2002]. A woman’s greatest risk for motor vehicle injury occurs during this life stage (CDC 2002). Eighty-five percent of pregnancies occur during the ages 15 to 34, and motor vehicle crashes are the leading cause of death for women in this age group (CDC 2002; Ventura et al. 2004). Historically, it has been difficult to quantify the bur- den of motor vehicle crash injuries involving pregnant women. Maternal deaths by definition include only those that are “related to or aggravated by pregnancy or its management” and exclude deaths from accidental causes such as motor vehicle crashes [Committee on Fetus and Newborn, American Academy of Pediatrics (AAP), and Committee on Obstetric Practice, American College of Obstetricians and Gynecologists (ACOG) 2002]. As a result, crash-related deaths do not appear in maternal mortality statistics. However, some reports suggest that injuries are a leading cause of death for pregnant and postpartum women in the United States (Dannenberg et al. 1995; Fildes et al. 1992), and motor vehicle crashes have emerged as a leading cause of injury-related hospitalizations during pregnancy (Weiss et al. 2002). There are also risks for the fetus involved in a motor vehicle crash. In recent years, the burden of crash- related fetal loss has been described at a population level. Weiss et al. (2001) reviewed fetal death certificates in 16 states and found that motor vehicle crashes caused more than 80% of the injury-related fetal deaths. They 98709mvpTxt 89_154 9/20/05 5:13 PM Page 128

also found that the crash-related fetal death rate (2.3 per 100,000 live births) was approximately one-half that of the crash-related infant death rate (4.9 per 100,000 live births). Safety belts were introduced for occupant protection in the United States in the 1960s (Graham 1989). Although evidence of the protective effect of safety belt use during pregnancy began to appear during the early 1970s, case reports were also appearing in the literature that described instances of injury to or even death of the fetus as a result of the safety belt, the lap belt in particu- lar (Handel 1978; Matthews 1975; McCormick 1968; Pepperell et al. 1977; Raney 1970; Rubovits 1964; White- house 1972). These case reports may have fueled early concerns about wearing safety belts during pregnancy. In issuing guidelines regarding safety belts and pregnancy with AAP in 1992, ACOG reported data suggesting that belt use among pregnant women in the early 1970s was approximately half that of all women (ACOG 1992). Reasons given for not wearing a safety belt during preg- nancy included concerns about harming the infant as well as discomfort or not being in the habit of wearing one (Johnson and Pring 2000; McGwin et al. 2004; Pearlman and Phillips 1996). Evidence has continued to accumulate over the years supporting the use of a properly positioned safety belt during pregnancy (Crosby and Costiloe 1971; Hyde et al. 2003; Wolf et al. 1993). At the same time, safety belt use in the United States has increased dramatically from less than 20% in the early 1980s to 81% in 2002 (Beck et al. 2004; Williams and Lund 1986). However, few data exist about safety belt use during pregnancy. Clinic-based surveys conducted in 1993, 1997, and 2001 reported that 45% to 86% of pregnant women in the United States always wear safety belts (McGwin et al. 2004; Pearlman and Phillips 1996; Tyroch et al. 1999). In addition, approximately 25% to 50% of preg- nant women are not aware of the proper positioning of safety belts during pregnancy (Johnson and Pring 2000; McGwin et al. 2004; Tyroch et al. 1999). AAP and ACOG have recognized the importance of occupant safety during pregnancy by issuing guidelines for health care providers to counsel all pregnant women on the proper use of safety belts during pregnancy (Committee on Fetus and Newborn, AAP, and Commit- tee on Obstetric Practice, ACOG 2002; ACOG 1992). These guidelines are consistent with general occupant safety recommendations from the U.S. Preventive Ser- vices Task Force, which call for providers to counsel all patients to use occupant restraints (DiGuiseppi et al. 1996). Although approximately 25% of pregnant women receive less-than-adequate prenatal care (calcu- lated as a function of month of initiation and number of visits, adjusted for gestational age at delivery), 99% of pregnant women in the United States receive at least some prenatal care (Martin et al. 2003). Prenatal care visits are thus an ideal mechanism for educating women about safety belt use. Given the impact of motor vehicle crashes on preg- nant women and fetuses, it is important to monitor health-promoting behaviors that can improve crash out- comes. Although limitations existed in the available data sets, it was possible to examine population-based data on safety belt use among women of reproductive age and prenatal counseling about wearing safety belts during pregnancy. A previous paper (Beck et al. 2005) uses these data to evaluate physician adherence to AAP- ACOG counseling guidelines, the results of which are summarized briefly here. DATA SOURCES AND ANALYTIC METHODS Data were from two ongoing, population-based surveil- lance systems administered by the CDC. The Behavioral Risk Factor Surveillance System (BRFSS) collects self- reported data on a variety of health-related topics. All 50 states, the District of Columbia (DC), and three ter- ritories participate. A disproportionate stratified sample of adults (aged at least 18 years) was selected for the 50 states and DC. Data were collected with telephone inter- views. Details of the BRFSS methodology are described elsewhere (Mokdad et al. 2003). In 2002 the median response rate, as defined by the Council of American Survey Research Organizations (White 1984), was 58% (range across states: 42% to 83%). In 2002 BRFSS respondents were asked how often they used safety belts when they drove or rode in a car. For this analysis, safety belt use was dichotomized as “always wears” versus “does not always wear” (i.e., nearly always, sometimes, rarely, or never) because a per- son may be involved in a crash during any given vehicle trip and must therefore wear a safety belt on each and every trip. Respondents who reported that they never rode in cars were excluded from the analysis (0.2%). The analysis was limited to women of reproductive age (18 to 44 years) in the 50 states and DC. The exclu- sion of women less than 18 years old was dictated by the BRFSS sampling design. In 2002, fewer than 1% of girls aged 10 to 17 years gave birth and 0.03% of women aged 45 to 54 years gave birth in the United States (Martin et al. 2003; Census Bureau 2002). Pregnancy status was assessed at the time of the interview (currently pregnant or not). Safety belt use among women of reproductive age was examined by sociodemographic variables (age, race or ethnicity, education) and by type of safety belt legisla- tion (primary versus secondary enforcement) in the state in 2002. Washington upgraded from secondary to pri- mary enforcement during the study period and was excluded from this portion of the analysis (Insurance 1 2 9PREGNANT WOMEN AND SAFETY BELTS 98709mvpTxt 89_154 9/20/05 5:13 PM Page 129

Institute for Highway Safety n.d.). All missing observa- tions were excluded from the analysis (missing data for the variables examined ranged from 0.0% to 1.3%). The Pregnancy Risk Assessment Monitoring System (PRAMS) collects self-reported data on maternal behav- iors and experiences that occur before, during, and after pregnancy. Women who deliver live-born infants are sam- pled from birth certificates at 2 to 6 months postpartum. Data are collected with mailed, self-administered surveys or with telephone interviews. Details of the PRAMS methodology are described elsewhere (Colley Gilbert et al. 1999). In 2000, 19 states (Alabama, Alaska, Arkansas, Col- orado, Florida, Hawaii, Illinois, Louisiana, Maine, Nebraska, New Mexico, New York, North Carolina, Ohio, Oklahoma, South Carolina, Utah, Washington, West Virginia) participated in PRAMS. New York data did not include New York City. The median response rate, weighted to reflect the sampling design, was 78% (range across states: 72% to 86%). In 2000 survey respondents were asked whether, dur- ing any prenatal care visit, a doctor, nurse, or other health care worker talked with them about using a safety belt during pregnancy. Safety belt counseling was examined by selected indicators. Maternal age, race or ethnicity, educa- tion, and parity were obtained from birth certificates. All other variables were self-reported on the PRAMS ques- tionnaire: payment source for prenatal care, type of pre- natal care provider, and timing of entry into prenatal care. Women who did not receive any prenatal care were excluded from all analyses; the prevalence of not receiving any prenatal care ranged from 0.3% to 1.5% across states. All missing observations were also excluded (missing data for the variables examined ranged from 0.02% to 6.8%). To accommodate the complex survey designs, SUDAAN software was used for analysis. For the BRFSS data, prevalence estimates and 95% confidence intervals (CIs) were calculated to provide national estimates of safety belt use among all reproductive-aged women and for the subset of pregnant women. State-based estimates of safety belt use were calculated only for women of reproductive age. Numbers were not sufficient to report state-based esti- mates of safety belt use among pregnant women. For the PRAMS data, prevalence estimates and 95% CIs for being counseled to use a safety belt during pregnancy were calcu- lated by state. With aggregated data for 19 states, risk ratios (and 95% CIs) were calculated to examine the asso- ciation between sociodemographic characteristics and the outcome variables (counseling and safety belt use). FINDINGS The BRFSS analysis was restricted to women of repro- ductive age (18 to 44 years), and the mean age of this group was 31.5 years. The majority (61%) of respon- dents had more than a high school education. Two- thirds of the respondents were non-Hispanic white, 16% were Hispanic, 13% were non-Hispanic black, and 6% were other race. Five percent of the respondents were pregnant at the time of the interview. In 2002 self-reported safety belt use in the United States (50 states and DC) was 83.8% (95% CI: 83.2, 84.4) for reproductive-aged women and 84.1% (95% CI: 81.9, 86.3) for pregnant women. Safety belt use among reproductive-aged women ranged from 59.8% to 94.1% across all states. Among the 19 states that also participated in PRAMS, safety belt use among reproductive-aged women ranged from 70% to 91% (Table 1). Subsequent analyses were restricted to women in these 19 states. The prevalence of always wearing safety belts was higher in states with primary enforcement laws than in states with secondary enforcement laws (85.2% versus 79.3%, respectively). Non-Hispanic black women were slightly less likely to wear safety belts than were non- Hispanic white women, women aged 29 years or younger were slightly less likely to wear safety belts than older women, and women with a high school or less than a high school education were slightly less likely to wear safety belts than women with more than a high school education (Table 2). PRAMS respondents were younger (mean age: 27 years) and less educated (48% with more than high school education) than the BRFSS respondents. Sixty- four percent of the PRAMS respondents were non- Hispanic white, 14% were Hispanic, 17% were non-Hispanic black, and 5% were other race. 1 3 0 RESEARCH ON WOMEN’S ISSUES IN TRANSPORTATION TABLE 1 Prevalence of Safety Belt Use Among Women of Reproductive Age: 19 States (BRFSS, 2002) Prevalence State Sample N % (95% CI) Alabamaa 848 85.6 (82.7–88.5) Alaska 800 73.9 (69.6–78.2) Arkansas 916 69.5 (66.0–73.0) Colorado 1085 81.1 (78.2–84.0) Florida 1463 84.6 (82.4–86.8) Hawaiia 1398 91.4 (89.2–93.6) Illinois 692 75.0 (69.5–80.5) Louisianaa 1427 80.9 (78.5–83.3) Maine 614 76.6 (72.7–80.5) Nebraska 1136 76.3 (73.4–79.2) New Mexicoa 1086 89.7 (87.5–91.9) New Yorka 1217 83.1 (80.6–85.6) North Carolinaa 1738 91.3 (89.3–93.3) Ohio 1070 78.8 (75.9–81.7) Oklahomaa 1601 82.1 (79.7–84.5) South Carolina 1160 76.8 (73.5–80.1) Utah 1122 78.7 (75.4–82.0) Washington 1294 89.9 (87.9–91.9) West Virginia 800 76.1 (72.6–79.6) a State had primary enforcement belt law in effect January 1, 2002. 98709mvpTxt 89_154 9/20/05 5:13 PM Page 130

The prevalence of pregnant women who reported prenatal counseling to use a safety belt was 48.2% overall; the prevalence ranged from 36.7% to 56.5% across the 19 states (Table 3). Women who were least likely to report having been counseled to wear safety belts were at least 30 years of age, were non-Hispanic white, had more than a high school education, were not receiving Medicaid, or were receiving prenatal care from a private provider (Table 4). The prevalence of being counseled decreased as level of education increased (<high school, 56.6%; high school, 50.5%; and >high school, 43.4%). DISCUSSION OF RESULTS It was found that 84% of women of reproductive age always wore safety belts in 2002. These findings are consistent with recent reports of belt use among women. The National Highway Traffic Safety Administration (NHTSA) found 79% belt use in a 2002 observational survey (Glassbrenner 2003), and the 2000 Motor Vehi- cle Occupant Safety Survey found a self-reported preva- lence of 88% (Block 2001). The patterns of safety belt use observed by age, race or ethnicity, and education were similar to those reported in other studies (Block 2001; Glassbrenner 2003; Lerner et al. 2001; Nelson et al. 1998). Of interest is the fact that the type of enforce- ment law in the state had an impact on safety belt use among reproductive-aged women. Numerous evalua- tions comparing primary and secondary enforcement laws have shown that adult use of safety belts is higher in states with primary laws. On average, belt use tends to be about 8% to 14% higher in states with primary laws than in those with secondary laws (Beck et al. 2004; Dinh-Zarr et al. 2001; NHTSA 2003). The analy- sis in this study, although limited to women of repro- ductive age, still showed a difference of six percentage points. Given the evidence about the benefits of safety belts during pregnancy, it is important to monitor the preva- 1 3 1PREGNANT WOMEN AND SAFETY BELTS TABLE 2 Associations Between Safety Belt Use Among Women of Reproductive Age and Sociodemographic Characteristics: 19 States (BRFSS, 2002) Risk Ratio Characteristic (Unadjusted) 95% CI Maternal race/ethnicity Hispanic 0.99 0.95–1.04 Non-Hispanic black 0.95 0.92–0.99 Non-Hispanic other 1.00 0.96–1.05 Non-Hispanic white referent Maternal age (years) 18–24 0.90 0.88–0.93 25–29 0.95 0.93–0.98 30–44 referent Maternal education <HS 0.92 0.87–0.97 HS 0.95 0.92–0.97 >HS referent Type of safety belt law Primary 1.07 1.05–1.10 Secondary referent Note: HS = high school. TABLE 3 Prevalence of Prenatal Counseling to Use Safety Belts During Pregnancy: 19 States (PRAMS, 2000) Prevalence State Sample N % (95% CI) Alabama 1536 49.9 (46.8–53.0) Alaska 1430 49.9 (47.0–52.8) Arkansas 1604 36.7 (33.2–40.2) Colorado 2118 48.3 (45.6–51.0) Florida 1957 45.9 (42.6–49.2) Hawaii 2443 48.8 (46.6–51.0) Illinois 1936 50.3 (47.9–52.7) Louisiana 2220 52.2 (49.7–54.7) Maine 1123 55.4 (52.1–58.7) Nebraska 2066 50.8 (48.1–53.5) New Mexico 1571 55.7 (53.2–58.2) New Yorka 1220 39.0 (35.5–42.5) North Carolina 1764 55.9 (52.8–59.0) Ohio 1611 46.7 (43.4–50.0) Oklahoma 1932 42.9 (39.4–46.4) South Carolina 1563 51.9 (48.0–55.8) Utah 1610 42.8 (39.7–45.9) Washington 1540 56.5 (53.0–60.0) West Virginia 1273 46.9 (43.6–50.2) a Data do not include New York City. TABLE 4 Associations Between Prenatal Counseling to Use Safety Belts During Pregnancy and Sociodemographic Characteristics : 19 States (PRAMS, 2000) Risk Ratio Characteristic (Unadjusted) 95% CI Maternal race/ethnicity Hispanic 1.25 1.18–1.32 Non-Hispanic black 1.37 1.32–1.43 Non-Hispanic other 1.23 1.15–1.32 Non-Hispanic white referent Maternal age (years) <17 1.19 1.09–1.29 18–24 1.17 1.12–1.23 25–29 1.10 1.04–1.16 ≥30 referent Maternal education <HS 1.31 1.24–1.37 HS 1.16 1.11–1.22 >HS referent Parity 1st birth 1.02 0.98–1.06 2nd or higher birth referent PNC payer Medicaid 1.21 1.16–1.25 Non-Medicaid referent PNC provider Public 1.25 1.20–1.30 Private referent Entry into PNC 2nd/3rd trimester 1.01 0.96–1.06 1st trimester referent Note: HS = high school; PNC = prenatal care. 98709mvpTxt 89_154 9/20/05 5:13 PM Page 131

lence of safety belt use among pregnant women. Knowl- edge of who is not wearing safety belts can help direct resources to promote belt use during pregnancy. Exist- ing data regarding safety belt use during pregnancy are limited for several reasons. Several published studies that have examined pregnancy-related behavior are clinic-based and may not be generalizable to the broader population of pregnant women (McGwin et al. 2004; Pearlman and Phillips 1996; Tyroch et al. 1999). Although the BRFSS data are population-based, the major limitation of using these data to measure behav- iors of pregnant women is that the surveillance system targets the general population of U.S. adults and not pregnant women. Pregnancy status is assessed at the time of the interview, but the survey is not designed to measure safety belt practices during pregnancy specifi- cally. The survey question (“How often do you use seat belts when you drive or ride in a car?”) measures behav- ior over a general time frame. The wording of the ques- tion assumes a constant pattern of behavior over time, which may not be valid when women become pregnant. Thus, it is difficult to interpret the meaning of the preva- lence of safety belt use among currently pregnant women. In addition to this measurement issue, the num- ber of pregnant women in the sample is not sufficient to conduct an in-depth analysis or to examine belt use among pregnant women at the state level. Although PRAMS does not provide national esti- mates, it is a potential source of population- and state- based data on safety belt use during pregnancy. Questions on occupant safety for women and infants are available for participating states to add to their sur- veys; one of these questions measures safety belt use during the last 3 months of pregnancy. Measuring behavior during the latter stages of pregnancy is impor- tant because this is a time when previously reported con- cerns about belt use (i.e., discomfort, fear of harming the infant) may become more pronounced (because of the woman’s growing abdomen). Thirty-one states cur- rently participate in PRAMS; however, only one state (Utah) currently uses this question. Two states (Mary- land and Vermont) used this question for 2001 to 2003. Available 2001 data from Maryland indicate that 85% of women always wore safety belts during the last 3 months of pregnancy (Maryland Department of Health and Mental Hygiene 2004). Some limitations of self-reported surveys should be discussed. One issue is the inability to assess whether the belt is properly positioned. The lap belt should be placed under the abdomen and across the upper thighs. The shoulder belt should be positioned between the breasts, which may require adjusting the seat position. However, direct observation of belt use by pregnant women may not be feasible. In lieu of direct observation, women could be surveyed about their knowledge of proper belt placement as well as their frequency of belt use, similar to the methods used by McGwin et al. (2004). Respon- dents are asked to identify the proper belt position from text descriptions and photographs of various belt posi- tions. Social desirability is another concern with self- reported surveys, that is, whether respondents will answer survey questions according to what they believe to be the socially acceptable response. However, there is evidence that social desirability has a minimal impact on measures of safety belt use in the United States (Nelson 1996). Recall bias may also be a concern, particularly for the PRAMS survey, which women complete several months after delivery. Since 1992 ACOG has recommended that prenatal care providers counsel all patients about safety belt use during pregnancy. Verbal discussion of this issue is specifically recommended, as opposed to the distribu- tion of written materials such as brochures. Therefore, the survey question asked women to report only discus- sions with their providers. Although certain groups of women were more likely to receive counseling than oth- ers, the overall prevalence of counseling was low (less than 50%). The significance of these findings is dis- cussed in more detail elsewhere (Beck et al. 2005). The reasons for the low levels of provider counseling on this topic are not clear. The prevalence of counseling for many other health-related behaviors (e.g., smoking, drinking, breastfeeding, nutrition) is much higher (>80%) (Petersen et al. 2001). Because of the data issues for reporting of maternal mortality statistics, prenatal providers may be less aware of the potential risks for their patients. However, pregnant women may have as many as 12 to 14 prenatal visits during a routine preg- nancy, and virtually all (99%) receive at least one pre- natal visit (CDC 2000). Therefore, prenatal providers have a unique opportunity to educate women about occupant safety. Providers can correct misconceptions about the protective effects for the woman and fetus and instruct women about the proper position of the safety belt. It may be advisable to devote resources toward edu- cating prenatal providers about the risks faced by their clients. Increased awareness of the issue may lead to increased prevalence of prenatal counseling about belt use during pregnancy. In addition, obstetricians and gynecologists as well as pediatricians could become advocates for strengthening laws in those states with secondary enforcement safety belt laws. Precedent exists for the role of physicians in advocating for the safety of their patients. For example, pediatricians were an important force in the passage of child restraint laws in the 1980s (Graham 1989). To fully inform the discussion about occupant safety for pregnant women, population-based data on proper safety belt use among pregnant women are needed. 1 3 2 RESEARCH ON WOMEN’S ISSUES IN TRANSPORTATION 98709mvpTxt 89_154 9/20/05 5:13 PM Page 132

Because belt use may change as the pregnancy advances, particularly during the last trimester, belt use should be measured during various stages of pregnancy. Provider counseling should be used as a means to educate women about the proper positioning of the safety belt to protect the woman and fetus in the event of a crash. In addition, the impact of counseling on safety belt use and knowl- edge of proper belt positioning should be rigorously evaluated. Finally, findings here about the impact of pri- mary enforcement laws suggest that effective strategies to increase safety belt use among the general population can be effective for reproductive-aged women as well. To promote the safety of pregnant motor vehicle occu- pants, prenatal counseling should be used in combina- tion with strategies such as legislation and high-visibility law enforcement. REFERENCES ACOG. 1992. Automobile Passenger Restraints for Children and Pregnant Women (ACOG Technical Bulletin 151). International Journal of Gynaecology and Obstetrics, Vol. 37, pp. 305–308. Beck, L. F., K. A. Mack, and R. A. Shults. 2004. Impact of Primary Laws on Adult Use of Safety Belts—United States, 2002. Morbidity and Mortality Weekly Report, Vol. 53, pp. 257–260. Beck, L. F., R. A. Shults, and B. Colley Gilbert. 2005. Preva- lence of Seat Belt Use Among Reproductive-Aged Women and Prenatal Counseling To Wear Seat Belts. American Journal of Obstetrics and Gynecology, Vol. 192, pp. 580–585. Block, A.W. 2001. 2000 Motor Vehicle Occupant Safety Sur- vey, Volume 2: Seat Belt Report. Report DOT HS-809- 389, U.S. Department of Transportation. CDC. 2000. Entry into Prenatal Care—United States, 1989–1997. Morbidity and Mortality Weekly Report, Vol. 49, pp. 393–398. CDC. 2002. Web-Based Injury Statistics Query and Report- ing System (WISQARS). U.S. Department of Health and Human Services. www.cdc.gov/ncipc/wisqars. Accessed Aug. 8, 2004. Census Bureau. 2002. Population by Age, Sex, Race, and His- panic or Latino Origin for the United States: 2000 (PHC- T-9). Census 2000 Gateway website: www.census.gov/ main/www/cen2000.html. Accessed March 10, 2005. Colley Gilbert, B., H. Shulman, C. Johnson, M. Rogers, and L. Fischer. 1999. The Pregnancy Risk Assessment Mon- itoring System (PRAMS): Methods and 1996 Response Rates from 11 States. Maternal and Child Health Jour- nal, Vol. 3, pp. 199–209. Committee on Fetus and Newborn (AAP) and Committee on Obstetric Practice (ACOG). 2002. Guidelines for Peri- natal Care, 5th ed. Washington, D.C. Crosby, W. M., and J. P. Costiloe. 1971. Safety of Lap-Belt Restraint for Pregnant Victims of Automobile Colli- sions. New England Journal of Medicine, Vol. 284, pp. 632–636. Dannenberg, A.L., D. M. Carter, H. W. Lawson, D. M. Ash- ton, S. F. Dorfman, and E. H. Graham. 1995. Homicide and Other Injuries as Causes of Maternal Death in New York City, 1987 Through 1991. American Journal of Obstetrics and Gynecology, Vol. 172, pp. 1557–1564. DiGuiseppi, C., D. Atkins, S. H. Woolf, and D. B. Kamerow (eds.). 1996. Guide to Clinical Preventive Services (CPS), 2nd ed. U.S. Preventive Services Task Force, Alexandria, Va. Dinh-Zarr, T. B., D. A. Sleet, R. A. Shults, S. Zaza, R. W. Elder, J. L. Nichols, et al. 2001. Reviews of Evidence Regarding Interventions to Increase the Use of Safety Belts. American Journal of Preventive Medicine, Vol. 21 (4S), pp. 48–65. Fildes, J., L. Reed, N. Jones, M. Martin, and J. Barrett. 1992. Trauma: The Leading Cause of Maternal Death. Jour- nal of Trauma, Vol. 32, pp. 643–645. Glassbrenner, D. 2003. Safety Belt Use in 2002: Demographic Characteristics. NHTSA Research Note. Report DOT HS-809-557, U.S. Department of Transportation. Graham, J. 1989. Auto Safety: Assessing America’s Perfor- mance. Auburn House Publishing Company, Boston, Mass. Handel, C. K. 1978. Case Report of Uterine Rupture After an Automobile Accident. Journal of Reproductive Medi- cine, Vol. 20, No. 2, pp. 90–92. Hyde, L. K., L. J. Cook, L. M. Olson, H. B. Weiss, and J. M. Dean. 2003. Effect of Motor Vehicle Crashes on Adverse Fetal Outcomes. Obstetrics and Gynecology, Vol. 102, pp. 279–286. Insurance Institute for Highway Safety. n.d. Safety Belt Use Laws. Highway Loss Data Institute. www.hwysafety.org/ safety_facts/state_laws/restrain3.htm. Accessed Nov. 21, 2003. Johnson, H. C., and D. W. Pring. 2000. Car Seatbelts in Preg- nancy: The Practice and Knowledge of Pregnant Women Remain Causes for Concern. British Journal of Obstet- rics and Gynaecology, Vol. 107, pp. 644–647. Lerner, E. B., D. V. K. Jehle, A. J. Billittier, R. M. Moscati, C. M. Connery, and G. Stiller. 2001. The Influence of Demographic Factors on Seatbelt Use by Adults Injured in Motor Vehicle Crashes. Accident Analysis and Pre- vention, Vol. 33, pp. 659–662. McCormick, R. D. 1968. Seat Belt Injury: Case of Complete Transection of Pregnant Uterus. Journal of the Ameri- can Osteopathic Association, Vol. 67, pp. 1139–1141. McGwin, G., S. R. Russell, R. L. Rux, C. A. Leath, F. Valent, and L. Rue. 2004. Knowledge, Beliefs, and Practices Concerning Seat Belt Use During Pregnancy. Journal of Trauma, Vol. 56, pp. 670–675. Martin, J. A., B. E. Hamilton, P. D. Sutton, S. J. Ventura, F. Menacker, and M. L. Munson. 2003. Births: Final Data 1 3 3PREGNANT WOMEN AND SAFETY BELTS 98709mvpTxt 89_154 9/20/05 5:13 PM Page 133

for 2002. National Vital Statistics Reports, Vol. 52, No. 10, pp. 1–113. Maryland Department of Health and Mental Hygiene. 2004. Maryland PRAMS Report: 2001 Births. Family Health Administration. Matthews, C. D. 1975. Incorrectly Used Seat Belt Associated with Uterine Rupture Following Vehicular Collision. American Journal of Obstetrics and Gynecology, Vol. 121, pp. 1115–1116. Mokdad, A. H., D. F. Stroup, and W. H. Giles. 2003. Public Health Surveillance for Behavioral Risk Factors in a Changing Environment: Recommendations from the Behavioral Risk Factor Surveillance Team. Morbidity and Mortality Weekly Report, Vol. 52 (RR09), pp. 1–12. Nelson, D. E. 1996. Validity of Self Reported Data on Injury Prevention Behavior: Lessons from Observational and Self-Reported Surveys of Safety Belt Use in the U.S. Injury Prevention, Vol. 2, pp. 67–69. Nelson, D. K., J. Bolen, and M. Kresnow. 1998. Trends in Safety Belt Use by Demographics and by Type of State Safety Belt Law, 1987 Through 1993. American Journal of Public Health, Vol. 88, pp. 245–249. NHTSA. 2003. Safety Belt Use in 2003. Report DOT-HS-809- 646. U.S. Department of Transportation. Pearlman, M. D., and M. E. Phillips. 1996. Safety Belt Use During Pregnancy. Obstetrics and Gynecology, Vol. 88, pp. 1026–1029. Pepperell, R. J., E. Rubinstein, and I. A. MacIsaac. 1977. Motor-Car Accidents During Pregnancy. Medical Jour- nal of Australia, Vol. 1, pp. 203–205. Petersen, R., A. Connelly, S. L. Martin, and L. L. Kupper. 2001. Preventive Counseling During Prenatal Care: Pregnancy Risk Assessment Monitoring System (PRAMS). American Journal of Preventive Medicine, Vol. 20, pp. 245–250. Raney, E. H. 1970. Fetal Death Secondary to Nonpenetrating Trauma to the Gravid Uterus. American Journal of Obstetrics and Gynecology, Vol. 106, pp. 313–314. Rubovits, F. E. 1964. Traumatic Rupture of the Pregnant Uterus from “Seat Belt” Injury. American Journal of Obstetrics and Gynecology, Vol. 90, pp. 828–829. Tyroch, A. H., K. L. Kaups, J. Rohan, S. Song, and K. Beingesser. 1999. Pregnant Women and Car Restraints: Beliefs and Practices. Journal of Trauma, Vol. 46, pp. 241–245. Ventura, S. J., J. C. Abma, W. D. Mosher, and S. Henshaw. 2004. Estimated Pregnancy Rates for the United States, 1990–2000: An Update. National Vital Statistics Reports, Vol. 52, No. 23, pp. 1–10. Weiss, H. B., T. J. Songer, and A. Fabio. 2001. Fetal Deaths Related to Maternal Injury. Journal of the American Medical Association, Vol. 286, pp. 1863–1868. Weiss, H. B., B. Lawrence, and T. Miller. 2002. Prevalence and Risk of Hospitalized Pregnant Occupants in Car Crashes. Annual Proceedings of the Association for the Advance- ment of Automotive Medicine, Vol. 46, pp. 355–366. White, A. A. 1984. Response Rate Calculation in RDD Tele- phone Health Surveys: Current Practice. In 1983 Pro- ceedings of the Section on Survey Research Methods, American Statistical Association, Washington, D.C., pp. 277–282. Whitehouse, D. B. 1972. Hazard to Fetus from Safety Har- ness. British Medical Journal, Vol. 1, p. 510. Williams, A. F., and A. K. Lund. 1986. Seat Belt Use Laws and Occupant Crash Protection in the United States. Ameri- can Journal of Public Health, Vol. 76, pp. 1438–1442. Wolf, M. E., B. H. Alexander, F. P. Rivara, D. E. Hickok, R. V. Maier, and P. M. Starzyk. 1993. A Retrospective Cohort Study of Seatbelt Use and Pregnancy Outcome After a Motor Vehicle Crash. Journal of Trauma, Vol. 34, pp. 116–119. 1 3 4 RESEARCH ON WOMEN’S ISSUES IN TRANSPORTATION 98709mvpTxt 89_154 9/20/05 5:13 PM Page 134

1 3 5 Protecting the Pregnant Occupant and Fetus in Motor Vehicle Crashes Biomechanical Perspective Kathleen DeSantis Klinich, Jonathan D. Rupp, and Lawrence W. Schneider, University of Michigan Transportation Research Institute Mark D. Pearlman, University of Michigan Health Systems Providing effective protection for fetuses of pregnant occupants in motor vehicle crashes (MVCs) poses a chal- lenge to automotive safety engineers because of limited data on the causes of fetal loss and injury. Recent studies have improved the understanding of biomechanical fac- tors leading to adverse fetal outcomes in MVCs and have resulted in tools to evaluate restraint system performance for pregnant occupants. An anthropometry study of seated pregnant occupants throughout gestation has pro- vided data on the size and shape of the pregnant abdomen relative to steering wheels and belt restraints. In-depth investigations of 42 crashes involving pregnant occupants resulted in logistic regression models that estimate the risk of adverse fetal outcome on the basis of crash sever- ity and maternal restraint use. Data from these studies were used to develop a pregnant abdomen and injury ref- erence values for the small female Hybrid III crash dummy. Highest-priority areas for future research are to monitor fetal outcomes after MVCs systematically, improve instrumentation for the pregnant crash dummy, expand the database of MVCs involving pregnant women, and measure material property characteristics of uterine and placental tissue. Each year in the United States, approximately128,000 of the 4 million pregnant women (10%of women aged 15 to 45) are involved in tow- away crashes (1). The actual number of fetal losses from motor vehicle crashes (MVCs) each year in the United States is unknown because mortality databases do not explicitly include maternal involvement in MVCs as a cause of fetal death. However, analysis of information available on fetal death certificates leads to a conservative estimate that 370 fetal deaths occur each year because of maternal involvement in an MVC (2), which is twice the number of infants under age 1 killed each year in MVCs. Moreover, although the number of children with physical and mental dis- abilities resulting from trauma sustained in utero dur- ing MVCs is also unknown, it is expected to be substantially higher than the estimated number of fetal losses. Despite these statistics, few studies have addressed the unique transportation safety needs of pregnant occupants and their fetuses. EARLY BIOMECHANICS RESEARCH Early biomechanics research on fetal loss resulting from MVCs involved sled tests performed on pregnant baboons in the 1960s (3–5). Results were limited but suggested that use of a three-point belt was better than a lap belt alone. Culver and Viano (6) attempted to esti- mate the anthropometry of a pregnant occupant for use in motor vehicle design by using scaling techniques to generate small, average, and large pregnant occupants and assuming that abdomen size varied with maternal stature. In the early 1990s, Pearlman and Viano (7) developed the first pregnant crash test dummy by mod- ifying a standard Hybrid III small female dummy to accommodate fetal and uterine components that were instrumented to measure fetal acceleration and load 98709mvpTxt 89_154 9/20/05 5:13 PM Page 135

transmitted through the abdomen. However, the utility of this dummy is limited by an unrealistic abdominal shape, a stiff force-deflection response, and a lack of correlation between measurements and the risk of adverse fetal outcome. RECENT BIOMECHANICS RESEARCH Research at University of Michigan Transportation Research Institute In the late 1990s, General Motors funded several proj- ects to improve automotive safety for pregnant occu- pants (8). The objectives of these studies were to obtain a better understanding of the anthropometry of the pregnant motor vehicle occupant, to determine biome- chanical factors surrounding fetal trauma to pregnant occupants involved in real-world crashes, and to develop an improved pregnant crash test dummy for evaluating the effectiveness of vehicle restraint and crashworthiness technologies in reducing fetal loss and disabling trauma in MVCs. Anthropometry of Pregnant Occupants In this study, the automotive-seated anthropometry and vehicle seat positioning of 22 women was investigated four times during their pregnancies (9). Subjects in five different stature groups were tested in the adjustable seating buck shown in Figure 1, which was configured to different vehicle interior package geometries with varying belt anchorage locations. Data collected included preferred seating positions of pregnant drivers, proximity of the pregnant occupant to the steering wheel and airbag module, contours of the subjects’ tor- sos and abdomens relative to the seat-belt centerline, and subject perceptions of their seated posture and proximity to vehicle components. The anthropometry data showed that abdominal size does not vary signifi- cantly with stature. This finding revealed a problem with the abdominal size of the original pregnant dummy, which was developed with the assumption that abdominal size does vary with stature. Figure 2 com- pares the mean midline abdominal contour of all sub- jects at 7 months’ gestation with the abdominal contour of the original pregnant dummy. Figure 3 shows the position of a lap belt relative to the midline abdominal contour and the anterior-superior iliac spines (ASIS), which suggests that even with the lap belt positioned as is recommended (as low as possible over the pelvis), there is potential for lap-belt loading of the uterus. Fig- ure 4 shows a postural representation of a subject and illustrates how the clearance between the abdomen and steering wheel rim decreases as gestation progresses because pregnant occupants do not adjust their seat position rearward during pregnancy. Investigations of Crashes Involving Pregnant Occupants In-depth investigations of 42 real-world crashes involv- ing pregnant occupants were performed over a 2-year period (10). Investigators collected information about the crash circumstances and conditions, measured and 1 3 6 RESEARCH ON WOMEN’S ISSUES IN TRANSPORTATION FIGURE 1 Adjustable laboratory seating buck used in study of pregnant anthropometry. 0 100 200 300 400 –100 0 100 200 300 x-axis (mm) z-axis (mm) Average 30-week pregnant woman Original pregnant dummy Top of sternum Bottom of sternum Pubic symphysis FIGURE 2 Mean midline abdominal contour of pregnant subjects at 30 weeks’ gestation compared with contour of original pregnant crash test dummy. 98709mvpTxt 89_154 9/20/05 5:13 PM Page 136

photographed the crash scene and vehicle damage, and obtained detailed information regarding the occupant and fetal injuries from medical records and subject inter- views. Results of the study are consistent with the previ- ously reported observation that the leading cause of fetal death from MVCs is placental abruption, the pre- mature separation of the placenta from the uterus, which prevents transfer of oxygen and nutrients to the fetus (11, 12). Data in the literature and in the current study also indicate that placental abruption occurs more frequently than other common adverse fetal outcomes such as uterine rupture and direct fetal injury. Analysis of these crash-injury data demonstrated clear associations between adverse fetal outcome (fetal loss, placental abruption, uterine laceration, delivery at less than 32 weeks, or direct fetal injury) and higher crash severities and the lack of maternal belt restraint use. Fig- ure 5 shows the estimated risk of adverse fetal outcome by crash severity and restraint condition, with points used to develop the curves by means of logistic regres- sion analysis included. Overall, improperly restrained occupants (no restraint, improper belt use, airbag only) were 5.7 times more likely to have adverse fetal out- comes than properly restrained occupants (properly positioned three-point belts, with or without airbags). However, belt use does not affect the likelihood of adverse fetal outcome in high-severity crashes. Development of Pregnant Crash Test Dummy The results of the anthropometric and crash investiga- tion studies were implemented in the design, develop- ment, and validation of the new pregnant crash dummy shown in Figure 6, the Maternal Anthropomorphic 1 3 7PROTECTING THE PREGNANT OCCUPANT AND FETUS IN CRASHES z-axis (mm) ASIS Lap belt path Midline abdominal contour 1100 1000 900 800 700 600 500 400 500 <front x-axis rear> (mm) 600 700 800 FIGURE 3 Midline contour of pregnant abdomen compared with path of lap belt and ASIS. 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500 –100 0 100 200 300 400 500 600 700 800 900 1000 x (mm) z (mm) ~13 weeks ~22 weeks ~30 weeks ~37 weeks FIGURE 4 Postural representation of 162-cm woman throughout gestation. 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% 0 10 20 30 40 50 60 Crash Severity (mph) Probability of Adverse Fetal Outcome Restrained Fetal Outcome: Restrained Not Properly Restrained Fetal Outcome: Not Properly Restrained FIGURE 5 Estimated risk of adverse fetal outcome as a function of crash severity and maternal restraint. 98709mvpTxt 89_154 9/20/05 5:13 PM Page 137

Measurement Apparatus, Version 2B (MAMA-2B). The MAMA-2B can be used to assess restraint system per- formance for pregnant occupants and their fetuses (13). As part of the pregnant dummy development program, two primary mechanisms of placental abruption in MVCs were hypothesized based on modeling and lim- ited biomechanical testing of uterine and placental tis- sues. The first is that the uterine-placental interface (UPI) fails in shear because of local deformation of the uterus in an area where the placenta is attached. The second is that the UPI fails in tension because of the inertia of the amniotic fluid and the dynamic viscous coupling of the uterus to the lumbar spine. The MAMA-2B is implemented in the Hybrid III small female crash test dummy and uses a fluid-filled silicone- rubber bladder to represent the uterus at 30 weeks’ gesta- tion. The MAMA-2B incorporates improved anthro- pometry with a midline contour that is based on the results of the study of automotive-seated pregnant anthropometry. The MAMA-2B abdomen was designed with a humanlike mechanical response to dynamic rigid- bar, belt, and close-proximity airbag loading and was instrumented with intrauterine pressure transducers located at the anterior and posterior surfaces of the blad- der. To develop injury criteria for the MAMA-2B, peak uterine pressure was measured in a series of sled test sim- ulations of frontal crashes. The crash severity and restraint conditions for the sled tests are shown in Table 1 and were based on the injury risk curves developed from the crash investigation study (shown in Figure 5). As shown in Figure 7, peak anterior uterine pressure mea- sured in the tests is correlated with the likelihood of adverse fetal outcome and can differentiate between load- ing conditions as shown in Figure 8. With proper belt restraint of a pregnant occupant, lower peak pressures at a given crash severity result than with improper restraint, and tests in the passenger mode (i.e., without the steering wheel) have lower peak pressures than tests run in the driver position. The intrauterine pressure data from the sled tests also suggest that airbags are beneficial to preg- nant occupants. Other Recent Research A limited number of tests have been performed to characterize the mechanical properties of placental 1 3 8 RESEARCH ON WOMEN’S ISSUES IN TRANSPORTATION FIGURE 6 MAMA-2B set up for sled test impact simulation of frontal crash. TABLE 1 Test Matrix for Developing MAMA-2B Criteria Risk of Adverse Fetal Test DV Outcome No. (km/h) Position Restraint (%) 1 13 Driver None 36 2 13 Driver Three-point belt 9 3 25 Driver Three-point belt 26 4 25 Driver Three-point belt + airbag 26 5 25 Driver Three-point belt + airbag 26 6 20 Driver None 54 7 13 Passenger Three-point belt 9 8 35 Passenger Three-point belt 51 9 35 Driver Three-point belt 51 10 35 Driver Three-point belt 51 11 35 Driver None 86 12 45 Driver Three-point belt 76 13 45 Driver Three-point belt + airbag 76 14 55 Driver Three-point belt 90 15 55 Passenger Three-point belt 90 y = 15.552x0.445 R*R = 0.6605 0 20 40 60 80 100 120 140 160 180 0 20 40 60 80 100 Risk of Adverse Fetal Outcome (%) Peak Anterior Pressure (kPa) FIGURE 7 Correlation between anterior pressure and risk of adverse fetal outcome. 98709mvpTxt 89_154 9/20/05 5:13 PM Page 138

and uterine tissue for use in computer modeling stud- ies of injury mechanisms (8). Preliminary results indicate that the two tissues have similar character- istics, but failure data were not measured. Acar and Weekes (14) performed a study of the anthropome- try of pregnant occupants to identify issues in vehi- cle design relevant to pregnant occupants. Their results are being used to generate guidelines for accommodating the needs of pregnant occupants in vehicles. However, their data are limited because they do not track changes throughout gestation on individual subjects. Researchers at Virginia Polytechnic Institute and State University (15) developed a finite-element model of a 7-month pregnant uterus and incorporated it into a MADYMO model of a small female driver. Simula- tions of unrestrained, three-point-belt, and three- point-belt-plus-airbag conditions were performed at crash severities between 13 and 55 km/h. Peak uterine strain was correlated with the likelihood of adverse fetal outcome predicted by the University of Michigan Transportation Research Institute crash investigation data (R2 = 0.846). The simulations also indicate that the lowest levels of uterine strain at a given crash severity occur for the three-point-belt-plus-airbag condition, suggesting that this is the safest restraint for the 7-month pregnant occupant involved in a frontal crash. Additional research (16) using this model suggests that uterine compression at the site of placental attachment is strongly correlated with peak strain at the UPI. Other researchers have developed models of pregnant occupants (17), but little is known about the details of their work since results have not been published in a technical journal. FUTURE RESEARCH NEEDS Although these recent studies have made substantial contributions to improving automotive safety for preg- nant occupants, further research and development efforts are needed to address the problem of fetal trauma from MVCs. In particular, research is needed to • Implement requirements for hospital trauma records and fetal death certificates to include maternal motor vehicle trauma as the cause for fetal injury or death so that more precise estimates of the magnitude of adverse fetal outcomes from MVCs can be made; • Perform additional in-depth investigations of MVCs involving pregnant occupants to establish an enhanced crash-injury database that can be used to develop better models for predicting the risk of adverse fetal outcomes under different conditions, for exam- ple, drivers versus passengers and front versus side impacts; • Perform biomechanical testing on uterine and pla- cental tissues to more accurately define the dynamic response and failure tolerances of these materials, which would help to develop more realistic physical and com- puter models that could be used to improve automotive safety for pregnant occupants and their fetuses; and • Implement additional design and instrumentation enhancements to the MAMA-2B so it can become a more reliable tool to assess the effects of changes in occupant restraint systems relative to improving protec- tion for the fetus. ACKNOWLEDGMENTS Some of the research discussed in this paper was funded by General Motors (GM) pursuant to an agreement between GM and the U.S. Department of Transporta- tion. The authors acknowledge Bethany Eby and Jamie Moore of the University of Michigan Transportation Research Institute and Steve Moss of First Technology Safety Systems for their contributions to this work. REFERENCES 1. Klinich, K. D., L. W. Schneider, J. L. Moore, and M. D. Pearlman. Investigations of Crashes Involving Pregnant Occupants. Final Report UMTRI-99-29. University of Michigan Transportation Research Institute, Ann Arbor, 1999. 2. Weiss, H. B., T. J. Songer, and A. Fabio. Fetal Deaths Related to Maternal Injury. Journal of the American Medical Association, Vol. 286, No. 15, 2001, pp. 1862–1868. 1 3 9PROTECTING THE PREGNANT OCCUPANT AND FETUS IN CRASHES 0 50 100 150 200 250 0 20 40 60 Impact Severity (km/h) Anterior Pressure (kPa) 3-pt pass 3-pt driver Unrestr. driver FIGURE 8 Variation in anterior pressure with restraint and impact severity. 98709mvpTxt 89_154 9/20/05 5:13 PM Page 139

3. Snyder, R. G., C. C. Snow, W. M. Crosby, P. Hanson, J. Fineg, and R. Chandler. Impact Injury to Pregnant Female and Fetus in Lap Belt Restraint. Proc., 10th Inter- national Stapp Car Crash Conference, Society of Auto- motive Engineers, Warrendale, Pa., 1966, pp. 249–259. 4. Crosby, W. M., R. G. Snyder, C. C. Snow, and P. G. Han- son. Impact Injuries in Pregnancy, I: Experimental Stud- ies. American Journal of Obstetrics and Gynecology, Vol. 101, No. 1, 1968, pp. 100–110. 5. King, A. I., W. M. Crosby, L. C. Stout, and R. H. Eppinger. Effects of Lap Belt and Three-Point Restraints on Pregnant Baboons Subjected to Deceleration. Proc., 15th Stapp Car Crash Conference, Society of Automo- tive Engineers, Warrendale, Pa., 1972, pp. 68–83. 6. Culver, C. C. , and D. C. Viano. Anthropometry of Seated Women During Pregnancy: Defining a Fetal Region for Crash Protection Research. Human Factors, Vol. 32, No. 6, 1990, pp. 625–636. 7. Pearlman, M. D., and D. Viano. Automobile Crash Sim- ulation with the First Pregnant Crash Test Dummy. American Journal of Obstetrics and Gynecology, Vol. 175, 1996, pp. 977–981. 8. Pearlman, M. D., K. D. Klinich, L. W. Schneider, J. Rupp, and S. Moss. A Comprehensive Program to Improve Safety for Pregnant Women and Fetuses in Motor Vehicle Crashes: A Preliminary Report. Ameri- can Journal of Obstetrics and Gynecology, Vol. 182, No. 6, 2000, pp. 1554–1564. 9. Klinich, K. D., L. W. Schneider, J. Rupp, B. Eby, and M. Pearlman. Challenges in Frontal Crash Protection of Pregnant Drivers Based on Anthropometric Considera- tions. In Occupant Protection, SAE 1999-01-0711, SAE-SP-1432, SAE International, Warrendale, Pa., 1999, pp. 105–127. 10. Klinich, K. D., L. W. Schneider, J. L. Moore, and M. D. Pearlman. Investigation of Crashes Involving Pregnant Occupants. Proc., 44th Annual Conference of the Asso- ciation for the Advancement of Automotive Medicine, Association for the Advancement of Automotive Medi- cine, Chicago, 2000, pp. 37–55. 11. Pearlman, M. D., J. E. Tintinalli, and R. P. Lorenz. Blunt Trauma During Pregnancy. New England Journal of Medicine, Vol. 323, No. 23, 1990, pp. 1609–1613. 12. Klinich, K. D., L. W. Schneider, J. L. Moore, and M. D. Pearlman. Injuries to Pregnant Occupants in Automo- tive Crashes. Proc., 42nd Annual Conference of the Association for the Advancement of Automotive Medi- cine, Association for the Advancement of Automotive Medicine, Des Plaines, Ill., 1998, pp. 57–91. 13. Rupp, J. D., K. D. Klinich, S. Moss, J. Zhou, M. D. Pearlman, and L. W. Schneider. Development and Test- ing of a Prototype Pregnant Abdomen for the Small- Female Hybrid III ATD (SAE 2001-22-0003). Stapp Car Crash Journal, Vol. 45, 2001, pp. 61–77. 14. Acar, B. S., and A. M. Weekes. Pregnant Driver Behav- iour and Safety. In Driver Behaviour and Training (L. Dorn, ed.), Ashgate, Stratford-upon-Avon, United King- dom, 2003, pp. 123–134. 15. Moorcroft, D. M., J. D. Stitzel, G. G. Duma, and S. M. Duma. Computational Model of the Pregnant Occu- pant: Predicting the Risk of Injury in Automobile Crashes. American Journal of Obstetrics and Gynecol- ogy, Vol. 189, No. 2, 2003, pp. 540–544. 16. Duma, S. M., D. M. Moorcroft, J. D. Stitzel, and G. G. Duma. Evaluating Pregnant Occupant Restraints: The Effect of Local Uterine Compression on the Risk of Fetal Injury. Proc., 48th Annual Conference of the Associa- tion for the Advancement of Automotive Medicine, Association for the Advancement of Automotive Medi- cine, Chicago, 2004, pp. 104–114. 17. Jordan, M. Improving Car Safety for Pregnant Women. Wall Street Journal, June 14, 2004. 1 4 0 RESEARCH ON WOMEN’S ISSUES IN TRANSPORTATION 98709mvpTxt 89_154 9/20/05 5:13 PM Page 140

1 4 1 Computational Model of Pregnant Motor Vehicle Occupant Stefan M. Duma, Dave Moorcroft, Joel Stitzel, and Greg Duma, Virginia Polytechnic Institute and State University A validated model of a 30-week-pregnant motor vehicle occupant is presented and the risk of fetal injury in frontal crashes is examined. A model of the pregnant uterus was imported into MADYMO 6.0 and included in the fifth-percentile female human body model by using membrane elements to serve as ligaments and facet surfaces for the overlying skin. A simulation matrix of 15 tests was developed to predict fetal out- come and included frontal crash impulses from minor (<24 km/h) to moderate (24 to 48 km/h) and severe (>48 km/h) crashes for the driver and passenger occupant positions. The test matrix included various restraint combinations: no restraint, lap belt, three-point belt, three-point with airbag, and airbag only. Overall, the risk of adverse fetal outcome was found to increase with increasing crash severity and to be higher for properly restrained drivers than for passengers. The peak uterine strain was reduced by 26% to 54% for the passenger position versus the driver position. This difference was due primarily to driver interaction with the steering wheel. For both occupant positions, the maternal injury indices were greatest for the unrestrained occupant. The current modeling effort has verified previous experi- mental findings regarding the importance of proper restraint use for the pregnant occupant. Automobile crashes are the largest single causeof death for pregnant women (Attico et al.1986) and the leading cause of traumatic fetal injury mortality in the United States (Weiss and Strot- meyer 2002). Each year in the United States 160 preg- nant women are killed in motor vehicle crashes and an additional 800 to 3,200 fetuses are killed although the mother survives (Klinich et al. 1999a, 1999b). Unfor- tunately, fetal injury in motor vehicle crashes is diffi- cult to predict because real-world crash data are limited and cadaver studies are not feasible. In an effort to reduce the risk of injury to pregnant occupants in car crashes, a pregnant anthropometric test dummy (ATD) was developed at the University of Michigan Transportation Research Institute (Rupp et al. 2001a). The Maternal Anthropomorphic Measure- ment Apparatus Version 2B (MAMA-2B) is a second- generation prototype ATD that is a retrofitted Hybrid III small female dummy. One of the primary limitations of the pregnant dummy is the lack of injury criteria for the fetus. The MAMA-2B was designed to measure anterior and posterior pressure in the fluid-filled abdomen insert as well as the strain on the perimeter of the insert. However, only the anterior pressure measure- ments were repeatable (Rupp et al. 2001a). Therefore, it would be beneficial to have an injury criterion for the pregnant dummy that utilizes currently established ATD measurement methods. One leading example would be to measure overall abdominal compression in a manner similar to that used to measure chest compression. For example, this measurement could be done by using a string potentiometer as is done in the chest. The most common cause of fetal death from motor vehicle accidents is placental abruption, which is the premature separation of the placenta from the uterus (Klinich et al. 1999b). Both the pregnant dummy and 98709mvpTxt 89_154 9/20/05 5:13 PM Page 141

the pregnant model presented in this study utilize this injury mechanism to predict fetal outcome (Moorcroft et al. 2003a). However, because of the difficulties in measuring this mechanism in the pregnant dummy, such as tissue strain and pressure, a computational model is desired that can accurately predict fetal injury risk. Therefore, a validated model of the pregnant occupant is presented here to examine the risk of fetal injury in frontal crashes for a range of restraint configurations in both driver and passenger occupant positions. METHODOLOGY Motor vehicle crashes were simulated with the MADYMO software package developed by the Nether- lands Institute of Applied Geoscience (TNO). In order to create the pregnant occupant, a finite element model of a pregnant uterus was inserted into the abdomen of a multibody human model (Figure 1) (Moorcroft et al. 2003a, 2003b, 2003c). The finite element uterine model is designed to represent an occupant in her 30th week of gestation. The abdomen consists of the uterus, placenta, and amniotic fluid. A fetus was not included because the injury mechanisms that predominantly contribute to fetal loss, as described by Rupp et al. (2001a), are inde- pendent of the fetus. In other words, placental abrup- tion is not effected by direct fetal loading of the placenta. The uterus is approximately ellipsoidal with a major axis of 27 cm and 18 cm, and it is 1 cm thick. The placenta is located at the fundus of the uterus and is 2 cm thick. The remainder of the interior of the uterus is filled with the amniotic fluid. The human model is a fifth-percentile female (5 ft tall, 110 lb) and the weight of the pregnant occupant model is 135 lb. The multi- body human model provides biofidelic response of an occupant in a motor vehicle crash while reducing the computational time compared with a full finite element human model. The anthropometry of a pregnant woman was quantified by Klinich et al. (1999a); the abdominal contour of the pregnant model matches their data. The uterine model is supported to the human model by the uterosacral and round ligaments, as well as the cervix. The bottom four nodes of each ligament are con- strained to move with the pelvis for both translation and rotation. The uterine model is also surrounded by fat to represent the boundary conditions involving the spine, abdominal organs, and the pelvis. All uterine bod- ies were modeled as linear elastic solids. Although the uterus and placenta are considered viscoelastic and anisotropic (Conrad et al. 1966; Pearsall and Roberts 1978; Mizrahi and Karni 1975), sufficient data were not available to accurately apply these material types. The amniotic fluid was modeled as a solid because MADYMO did not utilize fluid elements at the time of model development. Tension tests on human uterus tissue have been reported byRupp et al. (2001b), Pearsall and Roberts (1978), and Wood (1964). The Young’s modulus ranged from 20.3 kPa to 1,379 kPa, with an average of 566 kPa. The Poisson’s ratio is set to 0.40 since the uterus is a muscular organ, and the density is 1,052 kg/m3 (Moorcroft et al. 2003a). Rupp et al. (2001b) reported the results of five tension tests on placental specimens. The average modulus was 33 kPa, with a high of 63 kPa. Testing was not taken to failure. The highest mod- ulus is used in the pregnant model because it is expected that the placenta is stiffer than the fat. The Poisson’s ratio is assumed to be 0.45 because the tissue is muscu- lar (n = 0.40) engorged with blood (n = 0.50). The den- sity of the placenta is 995 kg/m3 (Moorcroft et al. 2003a). The amniotic fluid, which is 99% water and therefore incompressible, was assumed to have a negli- gible Young’s modulus and a Poisson’s ratio of 0.49. The Young’s modulus of 20 kPa is used for the fluid because moduli of lower values produced unstable results. The computational model uses peak von Mises strain in the uterus near the placenta as the measure for predicting risk of injury. High risk is associated with the presumed 60% strain tissue limit, allowing the predic- tion of fetal injury based on the strain. This strain limit is based on tissue tests of the uterine-placental interface (UPI) (Rupp et al. 2001b). Material properties of the ligaments connecting the uterus to the pelvis were not available in the literature. A brief search of general ligament properties showed that the elastic modulus of ligaments is typically two orders of magnitude greater than that of the uterus (Iwamoto et al. 1999; Zhang et al. 2001; Yamada 1970). Therefore, the elastic modulus of the uterosacral and round ligaments is set to 100 times the modulus of the uterus. The density and Poisson’s ratio were also taken from general ligament data (Iwamoto et al. 1999; Zhang et al. 2001). An isotropic representation of fatty 1 4 2 RESEARCH ON WOMEN’S ISSUES IN TRANSPORTATION Seat Back Floor Steering Wheel Knee Bolster Placenta Amniotic Fluid Round Ligament Head Restraint FIGURE 1 Pregnant occupant in driver-side interior. 98709mvpTxt 89_154 9/20/05 5:13 PM Page 142

tissue was used by Todd and Thacker (1994) in model- ing the human buttocks. The Young’s modulus for a seated female is 47 kPa with a Poisson’s ratio of 0.49. This Poisson’s ratio represents a nearly incompressible material. Contacts were created such that the fluid inte- rior of the uterus was free to move within the uterus, with contact friction. However, the fluid could not pen- etrate the uterus or placenta. Default master-slave con- tact treatments within MADYMO were used for all contacts. Four techniques were used to validate the pregnant model. First, a global biofidelity response was evaluated by using a seat belt to compress the pregnant abdomen dynamically (Moorcroft et al. 2003b). The results for force versus compression were within the published cor- ridors from scaled cadaver tests (Hardy et al. 2001). Second, a similar validation procedure was performed with a rigid bar (Moorcroft et al. 2003b) and these results were also consistent with previous data (Hardy et al. 2001). The third technique involved validating the model against real-world crashes in order to investigate the model’s ability to predict injury. When fatal crashes from pregnant occupants were used (Klinich et al. 1999b), the model showed strong correlation (R2 = 0.85) between peak strain at the UPI as measured in the model compared with risk of fetal demise as reported in the real-world crashes over a range of impact velocities and restraint conditions (Moorcroft et al. 2003a). The fourth method compared the physiological failure strain from placental tissue tests with the failure strain mea- sured in the model. Tissue tests by Rupp et al. (2001b) suggested approximately a 60% failure strain for UPI tissues, which is in agreement with the model’s predic- tion of 75% risk of fetal loss at a 60% strain in the UPI. In summary, the global-, injury-, and tissue-level valida- tion techniques all indicate that the model is good at predicting injurious events for the pregnant occupant. The simulations presented here were chosen to deter- mine the effect of restraint use and occupant position on the response of the pregnant occupant. The test matrix consisted of 15 simulations with occupant position and occupant restraint variations (Table 1). The applied sled pulse is a half-sine wave imposed for a duration of 100 ms. Acceleration is defined with respect to time; therefore the area under the curve corresponds to the change of velocity of the crash. Two interiors were used in the sim- ulations, a standard driver-side interior and a passenger- side interior. The driver and passenger interiors are typical MADYMO interiors made up of rigid planes to represent the seat, vehicle floor, and knee bolster. Positioning of the pregnant occupant was based on the seated anthropometry of a pregnant woman in her 30th week of pregnancy as defined by Klinich et al. (1999a). Four parameters were chosen to define the position of the occupant on the driver side by using the parameter values that correspond to the small female group in the study by Klinich et al. (average height, 5 ft; average weight, 134 lb). The abdominal clearance, defined as the distance between the abdomen and the bottom of the steering wheel, is 38 mm. The mean overlap of the uterus to the steering wheel is 12%, where the overlap is defined as the ratio of the vertical height of the uterus above the bottom of the steering wheel to the total vertical height of the uterus. The seatback angle, relative to vertical, is 13 degrees, and the steering wheel tilt is 29 degrees from ver- tical. Standard MADYMO finite element belts were used for the three-point restraint condition, with no force- limiting properties and no pretensioner. For the driver airbag tests, a MADYMO 600-mm driver airbag (vol- ume, 35 L) is used, with inflation triggered 15 ms into the 1 4 3COMPUTATIONAL MODEL OF PREGNANT MOTOR VEHICLE OCCUPANT TABLE 1 Pregnant Model Test Parameters and Results Risk of Maximum Crash Fetal Strain in Chest Speed Injurya the Uterine Deflection Occupant Restraint (km/h) (%) Wall (%) HICb V*Cb (m/s) (mm) Driver None 13 44 23.3 1 0.12 38.6 Driver None 20 65 36.6 13 0.31 39.1 Driver None 25 77 44.6 41 0.47 39.4 Driver None 35 100 60.8 156 0.72 39.7 Driver 3-pt belt 13 32 15.5 4 0.03 43.4 Driver 3-pt belt 25 51 27.9 62 0.09 47.1 Driver 3-pt belt 35 89 52.6 185 0.12 52.4 Driver 3-pt belt 45 99 58.7 211 0.13 54.3 Driver 3-pt belt 55 100 61.2 310 0.17 58.2 Driver 3-pt belt + airbag 25 52 28.1 49 0.22 45.1 Driver 3-pt belt + airbag 35 59 33.0 114 0.24 48.2 Driver 3-pt belt + airbag 45 80 46.6 173 0.20 49.0 Passenger None 35 52 28.2 2820 0.33 32.7 Passenger 3-pt belt 35 60 33.7 181 0.30 51.5 Passenger 3-pt belt + airbag 35 46 24.4 140 0.27 47.8 a Based on the peak strain at the uterine placental interface as measured in the model. b HIC represents the head injury criterion and V*C represents the viscous criterion. 98709mvpTxt 89_154 9/20/05 5:13 PM Page 143

simulation. The knees are approximately 11 cm from the knee bolster. For the passenger position, the seat was moved backward to match the distance seen in the typical passenger compartment and representing a fully aft seat position (abdominal clearance to dashboard, 40 cm). The same finite element belts were used for the passenger tests as for the driver tests, with the shoulder belt repositioned to extend from the right shoulder to the left hip. A stan- dard MADYMO passenger airbag (volume, 135 L) was used and triggered at 15 ms. There is no knee bolster interaction on the passenger side given the large distance between the knees and bolster (28 cm). RESULTS For the pregnant driver occupant, the unrestrained occu- pant resulted in substantially higher risk of abdominal and head trauma compared with the fully restrained driver in a similar crash (Figure 2). For all simulations both strain in the uterus and maternal responses were considered with respect to fetal outcome (Table 1). In particular, the risk of fetal injury was determined on the basis of the correlation between peak strain at the UPI and risk of fetal death in real-world crash investigation (Klinich et al. 1999b; Moorcroft et al. 2003c). Simula- tions in which the occupant was positioned in the pas- senger-side interior resulted in lower peak uterine strains measured at the UPI compared with those for the driver- side interior for all restraints tested. Substantial reduc- tions were seen for the unrestrained and three-point belt cases for similar crash speeds. For belted simulations, the peak strain is 26% to 36% less in passenger-side simula- tions compared with driver-side simulations even though the forward motion of the occupant is roughly equal between simulations with the same restraint. The key difference in the tests is the presence or absence of the steering wheel. In the driver-side configuration, the occupant con- tacts the steering wheel to some degree in all the config- urations tested. A lower peak strain is recorded in the unrestrained cases because the abdomen does not con- tact the steering wheel because of the seatbelts in the belted cases and there is no contact between the head and the windshield in the unrestrained cases. For the two types of belted cases, the occupant does not approach the dashboard, and therefore strain is primar- ily due to inertial effects as the belt restrains the lower abdomen and the upper abdomen continues to trans- late. The main effect of varying the occupant position therefore appears to be to alter the abdominal loading pattern from one of contact in the driver-side cases to one of inertia in the passenger-side cases. The importance of examining the maternal response is highlighted in the unrestrained passenger-side case. Although this simulation produced a low peak strain, based on the head injury criterion (HIC) value of 2820, it is reasonable to predict maternal death. This elevated value is the result of severe contact between the occu- pant’s head and the windshield. HIC scores for the remaining simulations were generally low and consistent between seating positions. Thorax response for the unre- strained occupant shows the same trend as that for the strain, in which the limited contact between the thorax and any vehicle surface reduces the passenger injury risk as compared with the driver response. For the restrained occupant, a slight increase is seen in thorax injury risk with the removal of the steering wheel. This finding is a result of the contact between the steering wheel and the pregnant abdomen, which reduces the load applied to the shoulder belt in driver-side simulations as compared with that in the passenger-side simulations. 1 4 4 RESEARCH ON WOMEN’S ISSUES IN TRANSPORTATION (a) (b) FIGURE 2 (a) Unrestrained pregnant driver in simulated 35-km/h crash; (b) fully restrained pregnant driver in simulated 35-km/h crash. 98709mvpTxt 89_154 9/20/05 5:13 PM Page 144

DISCUSSION OF RESULTS Overall, there is a high probability that placental abrup- tion would occur in the driver-side, unrestrained, frontal impact simulation. In the passenger-side simulation, there is a near 100% risk of life-threatening maternal brain injury in the similarly unrestrained condition and therefore a high risk for fetal death. The use of a three- point belt as well as an airbag reduces the risk to the pregnant women and the fetus. The difference in abdom- inal clearance between the driver-side and passenger-side simulations played an important role in peak strain in the uterine wall. The strain was 26% to 54% less in passenger-side simulations, primarily because of the presence or absence of the steering wheel. On the basis of the results of this study, it is recommended that when it is practical, the pregnant woman ride in the passenger seat with a three-point seat belt and airbag and the seat positioned as far rearward as possible. Placental abruption is believed to occur when the strain in the uterine wall exceeds 60%. The risk for pla- cental abruption is largest for high strains that occur near the placenta, which can be dramatically influenced by the lap belt position. Simulations have demonstrated that the vertical position of the lap belt can increase fetal risk by a factor of 3 (Figure 3) (Moorcroft et al. 2004). As the lap belt approaches the height of the placenta, which is located at the top of the uterus, the observed strain increases for a given crash pulse. Simulations with the lap belt directly loading the uterus at the placental location produced the highest recorded strain. Once the lap belt height is above the placenta, the strain decreases, with the strain for the top belt position matching that seen for the recommended belt location. However, it is important to note that there is increased risk to the mother with incorrect lap belt placement, including elevated head and chest injury response. This fact is important because the best way to protect the fetus is to protect the mother. Predicting fetal injury from abdominal deflection is loosely analogous to using chest deflection to predict tho- racic injury. As a simple comparison, chest deflection for the small female is limited to 52 mm by federal safety standards (Eppinger et al. 1999). A chest deflection of 52 mm is approximately 35% compression, which corre- sponds to approximately 40% risk of an Abbreviated Injury Scale 3 injury or greater (Mertz et al. 1991, 105–119). Given the obvious anatomical differences between the thorax and pregnant uterus, it is interesting that 35% compression of the uterus is also the higher limit of injury (Duma et al. 2004). The abdominal deflec- tion could be measured in the same manner as the chest deflection by using a string potentiometer or chest band or through processing of digital video. It is important to note that the measurements need to be taken from a preg- nant dummy with the same anthropometry and abdomi- nal force-deflection response as those of a pregnant woman. Overall, it is important to note that previous simula- tions indicate that for all frontal impacts it is safest for the pregnant occupant to ride in the passenger seat while wearing a three-point seat belt and utilizing the frontal airbag when appropriate (Moorcroft et al. 2004). The results of the current study support these earlier findings. As with all computational models, this model is lim- ited by the accuracy of input and simplifications made. The tissue data, from which the failure strain is derived, are sparse, and simplifications are made to use that data as a material model. In addition, the boundary condi- tions and geometry can and should be improved in future generations of the model. Furthermore, the model only looks at injury at the UPI. In cases with very large deflections, direct injury to the fetus may occur at injury rates different from those for placental abruption. In order to investigate this injury risk, a fetus would need to be added to the model. It is recommended that the methods in this study be applied to future generations of the pregnant occupant model to provide a continually improved understanding of pregnant occupant injury risk prediction. CONCLUSIONS A finite element model of the pregnant abdomen was created to predict fetal outcome following a motor vehi- cle crash. The model was incorporated into a human body model in a dynamic solver and validated with data from previous studies. The model is sensitive to changes in restraint conditions such as inertial, steering wheel, seat belt, airbag, and combined loading. Peak uterine strain was reduced by 30% to 50% in passenger-side 1 4 5COMPUTATIONAL MODEL OF PREGNANT MOTOR VEHICLE OCCUPANT (a) (b) FIGURE 3 Simulations at 35 km/h showing uterine compression for (a) correctly positioned belt and (b) incorrectly positioned belt. 98709mvpTxt 89_154 9/20/05 5:13 PM Page 145

simulations versus driver-side simulations, primarily because of increased distance between the abdomen and the nearest vehicle surface, namely, the steering wheel for driver-side tests and the dashboard for passenger-side tests. Overall, the model has verified previous experi- mental findings regarding the importance of proper restraint use for the pregnant occupant. The model can be used to run numerous tests and design advanced restraint systems specifically for pregnant occupants. ACKNOWLEDGMENTS The authors thank Altair Engineering, Inc., for the use of HyperWorks as the pre- and postprocessor and TNO for the use of MADYMO. The authors also thank Kathy Klinich, at the University of Michigan Transportation Research Institute, for her assistance in the development of the pregnant model. REFERENCES Attico, N.B., R. J. Smith III, M. B. Fitzpatrick, and M. Keneally. 1986. Automobile Safety Restraints for Preg- nant Women and Children. Journal of Reproductive Medicine, Vol. 31, No. 3, pp. 187–92. Conrad, J. T., W. K. Kuhn, and W. L. Johnson. 1966. Stress Relaxation in Human Uterine Muscle. American Jour- nal of Obstetrics and Gynecology, Vol. 95, No. 2, pp. 254–265. Duma, S. M., D. M. Moorcroft, J. D. Stitzel, and G. G. Duma. 2004. Evaluating Pregnant Occupant Restraints: The Effect of Local Uterine Compression on the Risk of Fetal Injury. Proc., 48th Annual Conference of the Associa- tion for the Advancement of Automotive Medicine, Association for the Advancement of Automotive Medi- cine, Chicago, Oct., pp. 104–114. Eppinger, R., E. Sun, F. Bandak, M. Haffner, N. Khaewpong, M. Maltese, S. Kuppa, T. Nguyen, E. Takhounts, R. Tannous, A. Zhang, and R. Saul. 1999. Development of Improved Injury Criteria for the Assessment of Advanced Automotive Restraint Systems, II. NHTSA, U.S. Department of Transportation. Hardy, W. N., L. W. Schneider, and S. W. Rouhana. 2001. Abdominal Impact Response to Rigid-Bar, Seatbelt, and Airbag Loading. Stapp Car Crash Journal, Vol. 45, pp. 1–32. Iwamoto, M., K. Miki, K. H. Yang, and A. I. King. 1999. Development of a Finite Element Model of the Human Shoulder. Presented at AMERI-PAM ’99 User Confer- ence, Detroit, Mich. Klinich, K. D., L. W. Schneider, B. Eby, J. Rupp, and M. D. Pearlman. 1999a. Seated Anthropometry During Preg- nancy. UMTRI-99-16. University of Michigan Trans- portation Research Institute, Ann Arbor. Klinich, K. D., L. W. Schneider, J. L. Moore, and M. D. Pearl- man. 1999b. Investigations of Crashes Involving Preg- nant Occupants. UMTRI-99-29. University of Michigan Transportation Research Institute, Ann Arbor. Mertz, H. J., J. D. Horsch, G. Horn, and R. W. Lowne. 1991. Hybrid III Sternal Deflection Associated with Thoracic Injury Severities of Occupants Restrained with Force- Limiting Shoulder Belts. SAE Technical Paper 910812. Society for Automotive Engineering, Warrendale, Pa. Mizrahi, J., and Z. Karni. 1975. A Mechanical Model for Uterine Muscle Activity During Labor and Delivery. Israel Journal of Technology, Vol. 13, pp. 185–191. Moorcroft, D. M., S. M. Duma, J. D. Stitzel, and G. G. Duma. 2003a. Computational Model of the Pregnant Occu- pant: Predicting the Risk of Injury in Automobile Crashes. American Journal of Obstetrics and Gynecol- ogy, Vol. 189, No. 2, pp. 540–544. Moorcroft, D. M., S. M. Duma, J. D. Stitzel, and G. G. Duma. 2003b. A Finite Element and Multi-Body Model of the Pregnant Occupant for the Analysis of Restraint Effec- tiveness. SAE Technical Paper 2003-01-0157. Society for Automotive Engineering, Warrendale, Pa. Moorcroft, D. M., J. D. Stitzel, S. M. Duma, and G. G. Duma. 2003c. The Effects of Uterine Ligaments on the Fetal Injury Risk in Frontal Automobile Crashes. Journal of Automobile Engineering, Vol. 217, Part D, pp. 1049–1055. Moorcroft, D. M., S. M. Duma, J. D. Stitzel, and G. G. Duma. 2004. The Effect of Pregnant Occupant Position and Belt Placement on the Risk of Fetal Injury. SAE Techni- cal Paper 2004-01-0324. Society for Automotive Engi- neering, Warrendale, Pa. Pearsall, G. W., and V. L. Roberts. 1978. Passive Mechanical Properties of Uterine Muscle (Myometrium) Tested in Vitro. Journal of Biomechanics, Vol. 11, pp. 167–176. Rupp, J. D., K. D. Klinich, S. Moss, J. Zhou, M. D. Pearlman, and L. W. Schneider. 2001a. Development and Testing of a Prototype Pregnant Abdomen for the Small-Female Hybrid III ATD. Stapp Car Crash Journal, Vol. 45, pp. 61–78. Rupp, J. D., L. W. Schneider, K. D. Klinich, S. Moss, J. Zhou, and M. D. Pearlman. 2001b. Design, Development, and Testing of a New Pregnant Abdomen for the Hybrid III Small Female Crash Test Dummy. Final Report UMTRI 2001-07-11. University of Michigan Transportation Research Institute, Ann Arbor. Todd, B. A., and J. G. Thacker. 1994. Three-Dimensional Computer Model of the Human Buttocks, in Vivo. Jour- nal of Rehabilitation Research and Development, Vol. 31, No. 2, pp. 111–119. Weiss, H. B., and S. Strotmeyer. 2002. Characteristics of Preg- nant Women in Motor Vehicle Crashes. Injury Preven- tion, Vol. 8, No. 3, pp. 207–214. 1 4 6 RESEARCH ON WOMEN’S ISSUES IN TRANSPORTATION 98709mvpTxt 89_154 9/20/05 5:13 PM Page 146

Wood, C. 1964. Physiology of Uterine Contractions. British Journal of Obstetrics and Gynecology, Vol. 71, pp. 360–373. Yamada, H. 1970. Strength of Biological Materials. Williams & Wilkins Company, Baltimore, Md. Zhang, L., K. H. Yang, R. Dwarampudi, K. Omori, T. Li, K. Chang, W. N. Hardy, T. B. Khalil, and A. I. King. 2001. Recent Advances in Brain Injury Research: A New Human Head Model Development and Validation. Stapp Car Crash Journal, Vol. 45, pp. 369–394. 1 4 7COMPUTATIONAL MODEL OF PREGNANT MOTOR VEHICLE OCCUPANT 98709mvpTxt 89_154 9/20/05 5:13 PM Page 147

1 4 8 What Are Young Female Drivers Made Of? Differences in Driving Behavior and Attitudes of Young Women and Men in Finland Sirkku Laapotti, University of Turku, Finland Young female drivers in Finland are described by com- paring the driving behavior and attitudes of young women and men. The study also questioned whether the traffic behavior and attitudes of female drivers have changed to resemble those of male drivers more closely during the past 20 years. The study used questionnaires to collect data from about 40,000 drivers on their atti- tudes and behavior (quantity and quality of driving, number and type of accidents, number of violations). Data on traffic offenses were also gathered by question- naires from 30,275 drivers on an official register. Acci- dent databases covering three levels of severity were used in the study: self-reported accidents, accidents in which claims were made to insurance companies, and fatal accidents investigated by the Road Accident Inves- tigation Teams in Finland. The results showed that on the whole, female drivers hold more positive attitudes toward traffic regulations and safety. They committed fewer traffic offenses and were involved in accidents less often than men (exposure controlled for). Typical female drivers’ accidents were those involving backing up and minor single-vehicle accidents. It is concluded that traffic attitudes and accident patterns of female drivers have not changed to resemble those of men more closely during the past 20 years in Finland. Compared with men, women have traditionallyhad a subordinate role in motor vehicle traffic:women drive less, are less likely to own a car, and are not as interested in motor vehicles and occupa- tions connected with traffic and cars. They are also less frequently involved in accidents and commit fewer traf- fic offenses than men. However, equality of the sexes is emerging in the field of motor vehicle traffic, at least as far as driver licensing and the amount of driving are concerned. In terms of research, the driving behavior of women has not attracted much attention until now, when the number of female drivers’ accidents is on the increase. Hierarchical models of driving behavior try to take into account several focuses of human behavior in explaining driver behavior (Michon 1984; Mikkonen and Keskinen 1980; van der Molen and Bötticher 1988). In Finland, Mikkonen and Keskinen had devel- oped the theory of internal models in the control of traffic behavior by 1980 but unfortunately it was only in Finnish. The model consists of three levels of internal models according to their extent: the largest model is called the route model. It contains knowledge about roads and events between a start and a goal. A route model is divided into several visual scenes, each of which is called a sight model. This level includes infor- mation needed in current traffic situations. A sight model is divided into several maneuvering models. Internal models include only the information needed for interplay between the road user and the environ- ment. Although the theory of Mikkonen and Keskinen (1980) is a cognitive theory, the authors suggested a close connection between internal models and the moti- vational and emotional system. The role of motives and emotions is to control the use of the internal models (Keskinen et al. 2004). 98709mvpTxt 89_154 9/20/05 5:13 PM Page 148

Evidence with regard to young (male) drivers led to calls to attribute a more important role to emotions, motives, and personality factors in explaining driving behavior (Evans 1991; Jessor 1987; Jonah 1986; Näätä- nen and Summala, 1976). Young men are skillful drivers as far as vehicle maneuvering is concerned: they require less training than any other driver group to pass the dri- ving test. However, young men still have the highest acci- dent and offense rates. Different persons with different motives have different internal models in traffic. A young man who seeks praise and admiration from his friends by showing off his driving skills has a different kind of inter- nal model than a middle-aged man who only uses the car to get from one place to another. The whole driving style differs depending on what the goal of driving is. A driver’s circumstances and overall behavior in life, as well as factors related to his or her personality, are also reflected in traffic behavior (Tillman and Hobbs 1949; Evans 1991; Jessor 1987). Keskinen (1996) and Hatakka et al. (2002) enlarged the theory of internal models in the control of traffic behavior by adding a fourth level (the largest, highest level) to the model. In this model the role of motives and emotions is more heavily emphasized, and driving behavior is seen in a larger context. The highest level is called goals for life and skills for living, which refers to general motives and attitudes in life, the importance of cars and driving for a driver’s personal development, and skills for self-control. STUDY GOALS This study describes young female drivers by comparing the driving behavior and attitudes of young women and men. The quantity and quality of driving, the number and type of accidents, and the number of violations are studied as driving behavior. The second aim of the study is to find possible changes in the difference between male and female drivers’ behavior and attitudes during the past 20 years. The focus of this study is on 18- to 25- year-old drivers. Differences in female and male drivers’ behavior in traffic are discussed in the framework of Keskinen’s hierarchical model of driving behavior. MATERIALS, SUBJECTS, AND METHODS The materials of this study were gathered in conjunction with several research projects conducted by the Traffic Research Group at the University of Turku, in Finland, during the past 15 years (Table 1). The main focus of many of the research projects was on driver training, and therefore the subjects of the studies were mostly novice drivers. The current study concludes the findings concerning female drivers; a more detailed description may be found elsewhere (Laapotti 2003). Self-Reported Information Amount of Driving and Type and Number of Accidents During First 18 Months as Driver The sample was drawn from the official register of driv- ing licenses. The sampling procedure made it possible to compare drivers of different ages who had received their license at the same time and who had the same amount of driving experience. The sample covered about 25% of 1 4 9WHAT ARE YOUNG FEMALE DRIVERS MADE OF? TABLE 1 Summary of Study Materials Study Year and Response Source of Original Paper Sample Description Rate Information Main Contents of the Study 1978 and 2001 Novice drivers aged 80% in 1978 Questionnaire Driving behavior and attitudes Elio et al., 1978 18 to 20 years, Comparison between years Laapotti et al., 2001 driving time 6–18 months 65% in 2001 Change Laapotti et al., 2003 1989 and 1990 Novice drivers aged 75% Questionnaire Self-reported driving kilometrage and Keskinen et al., 1992; 18 to 50 years, Driver’s license register accidents Laapotti et al., 2001 driving time 6–18 months Traffic offenses n = 30,275 1978 to 2001 Drivers aged 18 to 21 years — Culpable parties in Fatal motor car loss-of-control Laapotti and n = 413 fatal accidents fatal accidents, accidents: background factors Keskinen, 1998 Road Accident Investigation Teams 1984 to 2000 18-, 20-, 25-, 35-, — Culpable parties in Type of accidents Laapotti and 45- and 55-year-old drivers accidents Change Keskinen, 2004 n = 140,800 accidents Finnish Motor Insurers´ Centre 2002 Novice drivers, aged 48% Questionnaire Self-reported quantity and quality of Hatakka et al., 18 to 59 years, driving and accidents 2003 driving time 1–48 months n = 6,800 98709mvpTxt 89_154 9/20/05 5:13 PM Page 149

all novice drivers in Finland during 1989 and 1990. The information on driving experience and accidents was gathered by using a mailed questionnaire and covered the whole independent driving career of the drivers (range 6 to 18 months, mean 12.2 months, and mode 13 months). The drivers were asked to report all the acci- dents they had been involved in, whether they were at fault or not. They were asked to include all accidents, even minor ones, that had resulted in at least some mate- rial damage. The questionnaire allowed respondents to provide details on up to four accidents. The drivers were also asked to report on how many kilometers they had driven with a car (total mileage). Driving Behavior and Attitudes During First 18 Months as Driver The study replicated the traffic attitude survey con- ducted in 1978 in Finland, called Survey78 (Elio et al. 1978) and compared the results of Survey78 with the replicated survey in 2001 (called Survey01). The ques- tionnaire included 83 attitude-related items. Eighteen of these items concerned driver training and education. The remaining 65 items focused on issues such as traffic regulations, personal driving style, occupational driv- ing, road maintenance, and vehicles. In the explorative factor analysis, four factors were found. Four summary variables were named by using the criterion of factor loadings of at least 0.40: obeying traffic rules and driv- ing safely (16 items), pleasure and confidence in one’s driving (four items), attitudes toward occupational driv- ing (five items), and attitudes toward road maintenance and other road users (eight items). This study reports the main results of the first summary variable (obeying traffic rules and driving safely). The comparison of all attitude-related items between 1978 and 2001 is reported elsewhere (Laapotti et al. 2003). Quantity and Quality of Driving and Accidents During First 4 Years as Driver The subjects were asked how much they had driven (in kilometers) by car during the intermediate phase of driv- er training. (In Finland the intermediate phase of driver training is a period of independent training between licensing and the second phase of driver training.) The reasons for driving and the driving environment during the intermediate phase were determined by requesting the subjects to evaluate what percentage of their driving was to or from work or school, for errands, for an occu- pation, at work, on leisure time trips, or just for fun. Further, the subjects were asked to estimate what per- centage of their driving time was in the dark, in built-up areas, alone, in the evening, and at night. The number of accidents and violations during the intermediate phase of driver training was also determined. Traffic Offense Rates from Drivers’ License Register It was possible to calculate offense rates (offenses per kilometers driven) for those novice drivers who returned the questionnaire concerning accidents and kilometers driven (n = 30,275). The register of drivers’ licenses is maintained by the Vehicle Administration in Finland and includes all traffic offenses apart from parking tick- ets for which a penalty has been imposed. The follow- up period for traffic offenses was about 2 years. The current study reports the results concerning minor offenses. Minor offenses are not based on a court deci- sion and do not result in a withdrawal of the license. Most of the minor offenses are speeding. Number and Type of Accidents from Accident Databases All Accidents The Finnish Motor Insurers’ Centre maintains accident statistics covering all accidents for which damages have been paid. The data cover drivers found guilty of an accident, including information about when and where the accident happened, the type of accident, and the age and sex of the driver. The drawback with the all acci- dents data is that claims to insurance companies are usu- ally not made for minor—especially single-vehicle— accidents with low damages because the loss of the no- claims bonus results in a higher insurance premium. Drunk driving accidents are also underrepresented in this data set because insurance companies do not pay damages in such cases. Fatal Accidents In Finland, an accident is defined as fatal if someone involved in the accident dies within 30 days as a conse- quence of the accident. All fatal motor vehicle accidents in Finland are investigated by the Road Accident Investi- gation Teams. These teams prepare in-depth, on-the-spot crash reports on the basis of their findings. Each investi- gation team consists of a police officer, a traffic safety engineer, an automobile inspection engineer, a medical expert, and nowadays also an expert from the field of behavioral science [more details may be found elsewhere (Hakamies-Blomqvist 1994)]. The reports produced by 1 5 0 RESEARCH ON WOMEN’S ISSUES IN TRANSPORTATION 98709mvpTxt 89_154 9/20/05 5:13 PM Page 150

the investigation teams include descriptions of the type of the accident and how it took place, the time and loca- tion, weather and road conditions, speed used by all par- ties concerned, and speed limits. These reports also contain testimony by the driver (if he or she survived), the other driver in the case of a collision, passenger or passengers, and eyewitnesses. The members of the teams work together to produce a final report on the course of the accident and its probable causes and to suggest the means to prevent such accidents in the future. Both coded data material and the original team reports are available for research. Researchers have widely utilized data collected by these teams because they allow the use of disaggregated data covering a large amount of variables; reports in English include those by Hakamies-Blomqvist (1994), Hernetkoski and Keskinen (1998), Laapotti and Keskinen (1998, 2004), and Rajalin (1994). Statistical Methods Statistical analysis was conducted using SAS, Version 8.2. Analysis of variance (genmod) was used to find the effects of age, sex, and mileage on the number of acci- dents and offenses. Chi-square testing was used in group comparisons. Logistic regression analysis (logistic) was used in the modeling of categorical variables: the preva- lence of certain accident types was explained by the sex and age of the drivers and by calendar year. RESULTS Driving Exposure Women are less likely to have a driving license than men in Finland. In 2001, 62% of all 18- to 19-year-old women compared with 76% of all 18- to 19-year-old men were licensed. However, the difference between the sexes in driving license ownership has decreased sharply among young people and will decrease among older people as today’s middle-aged female drivers continue to age. Not only are women less likely to have a driving license, but they also drive less than men. Women drive about half as much as men. For example, at the begin- ning of the 1990s, young men drove about 26,000 km annually and women of the same age about 13,000 km annually. As far as the purpose for driving was concerned, female drivers reported more driving on errands than men. Young men reported the most “just for fun” driv- ing (22% of total driving). Young drivers reported more leisure time driving and driving for fun than middle-aged drivers. Driving conditions also varied between driver groups. Young female drivers reported less driving on slippery roads than young male drivers. Young drivers, both men and women, reported more driving with pas- sengers, in urban areas, on weekends, and at night than middle-aged drivers. Attitudes and Confidence in One’s Personal Driving Skills On the basis of the summary variable, it was found that female drivers had more positive attitudes toward traffic rules and safe driving than men did (df = 1, F = 410.08, p < .001). It was found that young drivers in 2001 claimed to have been less respectful of traffic rules and driving safely than young drivers in 1978 (df = 1, F = 66.97, p < .001). However, there was an interaction effect between the drivers’ sex and the survey year in that the negative change in attitudes toward traffic rules and safe driving between Survey78 and Survey01 was smaller among women than among men (df = 1, F = 4.49, p < .05). The comparison between 1978 and 2001 revealed that confidence in one’s personal driving skills had increased among both young women (c2 = 167.14, p < .001) and young men (c2 = 774.08, p < .001). However, the differ- ence between women and men remained the same: women rated their driving skills lower than those of men (Figure 1). Traffic Offenses Women commit fewer traffic offenses than men, middle- aged drivers fewer than young drivers, and those with low driving mileage fewer than those with high driving mileage (Figure 2). However, the age of the driver and the amount of driving had a weaker effect on the number of offenses among women than among men. For exam- ple, women with high mileage had even fewer traffic offenses than men with low mileage. Number of Accidents Female drivers had fewer traffic accidents than men, and the result held true even when driving mileage was controlled for (Figure 3). The difference in the number of accidents between women and men becomes bigger when the more serious accidents are concerned. For example, women are responsible for about 25% of all accidents for which damages have been paid in 2003 in Finland but only for about 16% of all fatal accidents in 2003 in Finland (Finnish Motor Insurers’ Centre 2004). 1 5 1WHAT ARE YOUNG FEMALE DRIVERS MADE OF? 98709mvpTxt 89_154 9/20/05 5:13 PM Page 151

Types of Accidents Accidents while backing up and minor single-vehicle accidents were typical types of crashes for women. Of all young female drivers’ crashes, 27% were accidents while backing up, whereas for young men the percent- age was 21%. It was more typical for men than women to have rear-end collisions. Of all male drivers’ crashes, 33% were rear-end crashes, whereas the corresponding figure for women was 27%. Of all female drivers’ fatal accidents, 57% were head- on collisions (compared with 36% of male drivers’ fatal accidents). A typical male drivers’ fatal accident was driving off the road. Of all male drivers’ fatal accidents, 47% were off-the-road accidents, whereas the propor- tion for women was 26%. Men lost the control of their vehicle typically at high speeds, in good weather and road conditions, and when drunk (40% of all their loss- of-control accidents). For female drivers those kinds of accidents were rare (7% of all their loss-of-control acci- 1 5 2 RESEARCH ON WOMEN’S ISSUES IN TRANSPORTATION 0 10 20 30 40 50 60 70 Ba d Be low av era ge Av era ge Ab ov e a ve rag e Ex ce llen t Ba d Be low av era ge Av era ge Ab ov e a ve ra ge Ex ce llen t Males Females % Survey 1978 Survey 2001 FIGURE 1 Self-evaluation of driving skills in Survey78 and Survey01. 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 18–20 21–30 31–50 Male high mileage Female high mileage Male low mileage Female low mileage M ea n Age FIGURE 2 Mean number of offenses in different age, sex, and mileage groups (high mileage = 20,000 to 29,999 km/year, low mileage = 100 to 9,999 km/year); analysis of variance: significant main effects on number of offenses: age, sex, and mileage (p < .001); coeffect: sex * mileage (p < .001). 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 M ea n 18–20 21–30 31–50 Age Male high mileage Female high mileage Male low mileage Female low mileage FIGURE 3 Mean number of self-reported accidents in differ- ent age, sex, and mileage groups (high mileage = 20,000 to 29,999 km/year, low mileage = 100 to 9,999 km/year); analysis of variance: significant main effects on number of offenses: age, sex, and mileage (p < .001); coeffects: age * sex (p < .001), sex * mileage (p < .05). 98709mvpTxt 89_154 9/20/05 5:13 PM Page 152

dents). Female drivers lost control of their vehicle typi- cally in slippery road conditions while sober and using moderate speed (48% of all their loss-of-control acci- dents). No change in the difference in accident patterns between men and women was found from 1984 to 2000 (Laapotti and Keskinen 2004). DISCUSSION OF RESULTS According to the results, female drivers drove less per year than male drivers, which suggests that female driv- ers are less experienced than men. The amount of driv- ing as such is connected to the skills at the lower levels of driving behavior, that is, vehicle maneuvering and mastering traffic situations (Keskinen 1996; Hatakka et al. 2002). The current study found that the proportion of minor single-vehicle and backing-up accidents was higher for women than for men. Further, female drivers lost control of their vehicle proportionally more often in bad road and weather conditions than men did. These types of accidents may be regarded as signs of problems at the lower levels in the hierarchy of driving behavior. Vehicle handling was found to be more problematic for women than for men in earlier studies as well (Karpf and Williams 1983; Rolls et al. 1991; Storie 1977). The study found that the goals and context of driving differed between men and women. Driving on errands was typical for female drivers of all ages. Young men were an exceptional group in the sense that nearly one- fourth of their driving was just for fun, whereas for other driver groups the proportion of such driving was 16% or lower. This result supports the finding that young men are more interested in cars and driving than other driver groups are (Rolls et al. 1991). The study found that young women had more posi- tive attitudes toward traffic rules and safe driving than men did. Women committed fewer traffic violations than men. Further, speeding and drunk driving were sel- dom the background factors explaining the fatal acci- dents of women, whereas for young men speeding and drunk driving were typical. Young women have fewer problems at the higher levels of driving behavior than young men. It may be concluded that women manage well in traffic as far as safety is concerned. Driver licensing most often takes place at the age of 18 to 19 in Finland. At this age, young persons in indus- trialized countries are still in the process of developing their adult identity. They are questioning the values and interests of adult society and are choosing their own for themselves. The stage of adolescence, both physiologi- cal and psychological, tends to start and end a little bit earlier among girls. Young men may still be rebellious at 18 and 19, whereas women already behave in a more adult fashion by then (Roberts et al. 2001). This expla- nation may be why women and men differ from each other in terms of driving behavior more at a young age than later on in life. Female drivers rated their own driving skills lower than those of male drivers, both in 1978 and in 2001. This find- ing may be in connection with the traditional view that a good driver is one who is skillful in vehicle handling and in mastering traffic situations. The current study aims to stress that a good driver is also safety oriented. Evidence thus far supports the conclusion that young female drivers are more safety oriented than their male counterparts. This study indicates that there are no major changes to this status quo. Although the number of driving licenses and the amount of driving by women have increased rapidly during the last 20 to 30 years, the difference in traffic attitudes and driving behavior between the sexes still remains. REFERENCES Elio, K., V. Koivisto, and K. Lasonen. 1978. Vähäisen ajokokemuksen omaavien kuljettajien liikennearvot ja - asenteet, niiden yhteyksiä kuljettajia kuvaaviin tausta- muuttujiin sekä liikennekäyttäytymiseen. [Traffic Values and Attitudes of Drivers with Low Driving Experience and Connections Between Traffic Values and Attitudes and Drivers’ Background Factors and Traffic Behaviour.] Research Report No. 75. Depart- ment of Education, University of Jyväskylä, Finland. Evans, L. 1991. Traffic Safety and the Driver. Van Nostrand Reinhold, New York. Finnish Motor Insurers’ Centre. 2004. Statistics of Traffic Accidents in Finland in 2003. Helsinki, Finland. Hakamies-Blomqvist, L. E. 1994. Older Drivers in Finland: Traffic Safety and Behavior. Report 40. Central Organ- isation for Traffic Safety, Helsinki, Finland. Hatakka, M., E. Keskinen, C. Baughan, C. Goldenbeld, N. P. Gregersen, H. Groot, S. Siegrist, G. Willmes-Lenz, and M. Winkelbauer (eds.). 2003. Basic Driver Training: New Models. EU Project, Final Report. Department of Psychology, University of Turku, Finland. Hatakka, M., E. Keskinen, N. P. Gregersen, A. Glad, and K. Her- netkoski. 2002. From Control of the Vehicle to Personal Self-Control: Broadening the Perspective to Driver Educa- tion. Transportation Research, Vol. 5F, pp. 201–215. Hernetkoski, K., and E. Keskinen. 1998. Self-Destruction in Finnish Motor Traffic Accidents in 1974–1992. Acci- dent Analysis and Prevention, Vol. 30, pp. 697–704. Jessor, R. 1987. Risky Driving and Adolescent Problem Behav- ior: An Extension of Problem Behavior Theory. Alco- hol, Drugs and Driving, Vol. 3, pp. 1–11. Jonah, B. A. 1986. Accident Risk and Risk-Taking Behaviour Among Young Drivers. Accident Analysis and Preven- tion, Vol. 18, pp. 225–271. 1 5 3WHAT ARE YOUNG FEMALE DRIVERS MADE OF? 98709mvpTxt 89_154 9/20/05 5:13 PM Page 153

Karpf, R. S., and A. F. Williams. 1983. Teenage Drivers and Motor Vehicle Deaths. Accident Analysis and Preven- tion, Vol. 15, pp. 55–63. Keskinen, E. 1996. Why Do Young Drivers Have More Acci- dents? In Junge Fahrer und Fahrerinnen [Young Dri- vers] (in German and in English), Berichte der Bundesanstalt für Strassenwesen, Mensch und Sicher- heit, Heft M 52, Bergisch Gladbach, Germany. Keskinen, E., M. Hatakka, A. Katila, and S. Laapotti. 1992. Onnistuiko kuljettajaopetuksen uudistus? Seurantapro- jektin loppuraportti. [Was the Renewal of the Driver Training Successful? Final Report of the Follow-Up Group.] Psychological Report No. 94. University of Turku, Finland (English abstract). Keskinen, E., M. Hatakka, S. Laapotti, A. Katila, and M. Peräaho. 2004. Driver Behaviour as a Hierarchical Sys- tem. In Traffic and Transport Psychology: Theory and Application (T. Rothengatter and R. D. Huguenin, eds.), Elsevier, Amsterdam, Netherlands. Laapotti, S. 2003. What Are Young Female Drivers Made of? Differences in Atttitudes, Exposure, Offenses and Acci- dents Between Young Female and Male Drivers. Doc- toral thesis. Annales Universitatis Turkuensis, Sarja B, Osa 264, Painosalama Oy. Laapotti, S., and E. Keskinen. 1998. Differences in Fatal Loss- of-Control Accidents Between Young Male and Female Drivers. Accident Analysis and Prevention, Vol. 30, pp. 435–442. Laapotti, S., and E. Keskinen. 2004. Has the Difference in Accident Patterns Between Male and Female Drivers Changed Between 1984 and 2000? Accident Analysis and Prevention, Vol. 36, No. 4, pp. 577–584. Laapotti, S., E. Keskinen, M. Hatakka, and A. Katila. 2001. Novice Drivers’ Accidents and Violations—A Failure on Higher or Lower Hierarchical Levels of Driving Behaviour. Accident Analysis and Prevention, Vol. 33, pp. 759–769. Laapotti, S., E. Keskinen, and S. Rajalin. 2003. Comparison of Young Male and Female Drivers’ Attitude and Self- Reported Behaviour in Finland in 1978 and 2001. Jour- nal of Safety Research, Vol. 34, pp. 579–587. Michon, J. A. 1984. A Critical View of Driver Behavior Mod- els: What Do We Know, What Should We Do? In Human Behavior and Traffic Safety (L. Evans and R. Schwing, eds.), Plenum Press, New York. Mikkonen, V., and E. Keskinen. 1980. Sisäisten mallien teoria liikennekäyttäytymisestä. [Internal Models in the Con- trol of Traffic Behavior.] General Psychology Mono- graphs No. B1. University of Helsinki, Finland. Näätänen, R., and H. Summala. 1976. Road-User Behavior and Traffic Accidents. North-Holland Publishing Com- pany, Amsterdam, Netherlands. Rajalin, S. 1994. The Connection Between Risky Driving and Involvement in Fatal Accidents. Accident Analysis and Prevention, Vol. 26, pp. 555–562. Roberts, B. W., A. Caspi, and T. E. Moffitt. 2001. The Kids Are Alright: Growth and Stability in Personality Devel- opment from Adolescence to Adulthood. Journal of Per- sonality and Social Psychology, Vol. 81, pp. 670–683. Rolls, G. W. P., R. D. Hall, R. Ingham, and M. McDonald. 1991. Accident Risk and Behavioural Patterns of Younger Drivers. AA Foundation for Road Safety Research, Hampshire, England. Storie, V. J. 1977. Male and Female Car Drivers: Differences Observed in Accidents. Report 761. U.K. Transport and Road Research Laboratory, Berkshire. Tillmann, W. A., and G. E. Hobbs. 1949. The Accident-Prone Automobile Driver. A Study of the Psychiatric and Social Background. American Journal of Psychiatry, Vol. 106, pp. 321–331. van der Molen, H. H., and A. M. T. Bötticher. 1988. A Hier- archical Risk Model for Traffic Participants. Ergonom- ics, Vol. 31, pp. 537–555. 1 5 4 RESEARCH ON WOMEN’S ISSUES IN TRANSPORTATION 98709mvpTxt 89_154 9/20/05 5:13 PM Page 154

1 5 5 Pedestrian–vehicle crashes are examined for patterns by gender. The analysis focuses on how the pedestrian crashes of men and women vary by personal character- istics (age, condition, injury) and physical characteris- tics of the crash area (location type, density, land use, pedestrian activity). The data for this study are pedes- trian–vehicle crashes in Baltimore City, Maryland, from the State of Maryland Motor Vehicle Accident Report. The results from the analysis presented here suggest that, in general, there are few significant gender effects in the majority of pedestrian crashes. Women tend to be involved in fewer pedestrian crashes overall, and when they are involved, they appear to exhibit fewer risk-tak- ing behaviors, such as violating traffic laws and con- suming alcohol or drugs. Women were slightly less likely to be injured in a crash and less likely to die as a result. The effects of land use on pedestrian crash rates were not significant by gender. However, a higher percentage of women’s crashes occur in areas with high pedestrian activity, which may be reflective of the distribution of areas in which women walk. The National Highway Traffic Safety Administra-tion (NHTSA) (2001) reports that althoughpedestrians are involved in a small proportion of vehicle crashes (2% to 3%), they represent a much higher proportion of crash fatalities (11% to 13%). In a majority of these deaths (55%), improper pedestrian behavior was a contributing factor. A variety of other studies have shown that male pedestrians are more likely to be involved in a crash and die as a result than their female counterparts [Hayakawa et al. 2000; Hebert Martinez and Porter 2004; Hijar et al. 2003; Khan et al. 1999; World Health Organization (WHO) 2003]. From these results, it would appear that women are less at risk for involvement in a pedestrian–vehicle crash. Contributing factors to the lower crash rates for women may include fewer risk-taking behaviors such as alcohol consumption, illegal midblock crossing or jaywalking, and crossing under unsafe conditions (low visibility or high vehicular volume). However, less research has focused on those factors related to women’s involvement in pedestrian crashes and the per- sonal and environmental factors associated with them. Gender differences in pedestrian–vehicle crashes are examined for the city of Baltimore, Maryland. The objec- tive of the analysis was to determine the relationships, if any, between characteristics of pedestrians and the physi- cal characteristics of crash location with particular emphasis on women’s involvement in crashes. BACKGROUND LITERATURE There is a long list of research on the travel and trans- portation differences between men and women; how- ever, most of this work has focused on motorized travel. Studies focused exclusively on the gender dif- ferences in pedestrian trips are sparse and the few that do exist show mixed results. For example, evidence shows that women are more dependent on transit, tend to make more household maintenance trips, and Women’s Involvement in Pedestrian–Vehicle Crashes Influence of Personal and Environmental Factors Kelly J. Clifton, Carolina Burnier, and Kandice Kreamer Fults, University of Maryland, College Park 98709mvpTxt 155_210 9/20/05 5:42 PM Page 155

engage in more trip chaining (Rosenbloom 1997). Transit use is linked to pedestrian trips, since walking is frequently an access mode for transit. Because women have a greater reliance on transit, one would expect them to make more pedestrian trips connected with transit use. There is some evidence, however, that household responsibilities and resulting travel pat- terns may not accommodate much additional pedes- trian activity. For example, Handy (1996) found that women walked to the store fewer times per month than men. In contrast, some studies on women’s walk- ing patterns show that women walk farther than men do (Carlsson-Kanyama et al. 1999) and make more walking trips (Root et al. 2000). Another thread of literature important to the topic of pedestrian activity is the interaction between urban form and travel. Like the research on women’s travel, the land use and transportation connection has a rich and varied literature, a review of which may be found elsewhere (Crane 2000), but not without some short- comings, particularly when the relationship among land use, transportation choices, and gender differences is examined. The literature has less to say about how or if these land use variables may affect women’s travel dif- ferently from that of men; a more in-depth treatment has been made by Clifton and Dill (2005). Likewise, studies examining women’s involvement in pedestrian crashes are limited. In 2002, 4,808 pedestrians were killed in traffic crashes in the United States. However, data from NHTSA (2004) reveal that although traffic fatalities have increased over the past 10 years, pedestrian fatali- ties have steadily decreased. Female pedestrian fatalities represent a minority of crashes in the United States and account for 32% of all pedestrian fatalities (NHTSA 2004; Demetriades et al. 2004). Regarding fatal pedes- trian crash rates, men are involved in greater numbers of fatal crashes per population than women in every age category (NHTSA 2004). Much research has been dedicated to the differences between men and women in risk-taking behaviors. This research may help explain the differences in crash involvement by gender. Men may be placing themselves at greater risk by crossing an intersection improperly, disobeying a pedestrian signal, or crossing at a nonin- tersection (midblock) location. In a study of Baltimore and Washington, D.C., pedestrian–vehicle crashes, Preusser et al. (2002) found that pedestrians were judged culpable in 50% of the crashes. Alcohol and sub- stance abuse by the pedestrian are involved in 15% of pedestrian crashes (Stutts et al. 1996) and men are more likely than women to be drinking heavily and using illicit drugs (Thom 2003). Because 75% of all pedestrian crashes occur in urban areas (Hebert Martinez and Porter 2004), it is important to understand the interaction between pedes- trian crashes and the built environment. Intuitively, it appears that the built environment could have signifi- cant influence on pedestrian crash rates. Pedestrian generators, such as high-density development and com- mercial land uses, would expect to experience more crashes because they have more pedestrian demand. Women may be at an increased risk in these areas because they may be more likely to be shopping or run- ning errands. However, as stated before, women are consistently found to have lower crash involvement rates than men. As shown by previous research, men and women have different travel patterns. These differences in travel behavior, including mode choice, trip purpose, and time of day, may lead to distinct patterns in walk- ing behavior. In turn, these travel patterns, along with other factors, may result in different patterns of crash involvement. The literature documents, although to a lesser extent than the coverage of differences in travel behavior, men’s disproportionate involvement in pedestrian collisions. This finding, in turn, raises questions about disparities in pedestrian crashes between the sexes and the relationship to underlying travel behavior and the environment. An examination of pedestrian–vehicle crashes cannot be used to make the link between these differences in the travel behav- ior of men and women because the data are insuffi- cient. However, study of these negative outcomes for pedestrians can suggest ways in which the sexes are affected differently by their own behavior and their environment. The purpose of this study was to examine how per- sonal and environmental factors differ between men and women involved in police-reported pedestrian crashes. Specifically, this study aimed to examine gender differ- ences in pedestrian crashes with respect to involvement, severity of injury, personal factors and behaviors, and environmental conditions. To accomplish this aim, pedestrian crash data for the city of Baltimore, Mary- land, were examined by using descriptive statistics and multivariate analysis. DATA AND METHODOLOGY This study presents a detailed analysis of gender, pedestrian crashes, and land use variables by using a variety of statistical tests. The data for this study are pedestrian–vehicle crashes in Baltimore City, Mary- land, from the State of Maryland Motor Vehicle Acci- dent Report. The data on pedestrian–vehicle collisions include more than 3,000 individual crash records over a 3-year period from 2000 to 2002. The data con- tained information on the pedestrian such as age, gen- 1 5 6 RESEARCH ON WOMEN’S ISSUES IN TRANSPORTATION 98709mvpTxt 155_210 9/20/05 5:42 PM Page 156

der, clothing, obedience to pedestrian signals, and presence of alcohol. In addition, the data included crash time and location information such as presence of signal, turning movement, road condition, weather, and time of day. The location of the crash was recorded as the nearest intersection. These data were supplemented with characteristics of the crash loca- tion, including measures of land use and pedestrian activity. The socioeconomic characteristics of the area surrounding the crash were obtained from the 2000 U.S. census, including information such as median income, total population and population density, num- ber of households, total employment, and percentage of population that walks to work. All measures were aggregated to the block-group level. In addition five measures of the built environment were calculated: single-family residential dwelling unit density, percent- age of single-family dwellings within 1⁄4 mi of a bus stop, percentage of single-family dwellings within 1⁄4 mi of commercial locations, percentage of land dedicated to parks, and population density. The city of Baltimore is the largest city in Maryland and has a population of over 650,000 residents with a population density of over 8,000 persons per square mile. The median household income for Baltimore City, $30,078, is lower than the national median of $41,994 and significantly lower than the Maryland state median of $52, 868. Twenty-three percent of Bal- timore City residents are below the poverty level, sig- nificantly higher than the national average of 12%. Notably, unlike the national and state averages, Balti- more City has a higher percentage of women than men in the workforce: 52.6% women versus the national average of 46.5% and Maryland state average of 48.8%. More than 6% of Baltimore City’s employees walk to work, more than twice the national average. In addition, Baltimore City provides numerous public transit options and facilities: bus routes, light rail, a subway system, and a commuter rail connecting to Washington, D.C. Baltimore’s high number of pedestrians, population with a greater reliance on alternative modes, variation in urban structure across the metropolitan area, and high number of pedestrian–vehicle crashes make it a compelling case for closer investigation. Figure 1 shows the spatial distribution of the pedestrian–vehicle crashes differentiated by gender of the pedestrian involved in the crash. The clustering of crashes around major arte- rials can be seen on this map. Of the 3,009 crashes, women accounted for 40%. With the data just described, the research design uti- lized multivariate statistical analysis to examine the dif- ferences in crash rates by gender, including the effect of personal characteristics and environmental conditions. The data were analyzed using descriptive statistics to examine the differences in crash involvement for the sexes based on severity of injury, age of pedestrian, pres- ence of alcohol or other substance, pedestrian obedience to traffic signals, age of pedestrian, time of day, and land use. A model of crash densities was employed to esti- mate the effects of personal and environmental fac- tors on the spatial differences in crash involvement of men and women. In order to better understand the effects of the pedestrian, location, and environmental characteristics, multivariate analysis of the crashes was performed. Ordinary least squares linear regres- sion analysis was used to determine the effects of each variable on the number of crashes per square mile per population. The spatial unit of analysis was the block- group level. The model was segmented by gender: one was estimated for the male crash data and one for the female crash data. The dependent variable was the natural log of the number of crashes per square mile per population, and the independent variables were the characteristics of the population and crash and the land use variables. The location characteristics include median income and percentage of nonwhite population, children in the area, and vehicle owner- ship. The variables of interest that could have an impact on future land use policies are pedestrian activity, commercial accessibility, transit accessibility, and roadway density. Table 1 shows the model speci- fication and estimation for both men and women. There was no significant multicollinearity among the variables in the model. In addition, these models were corrected for heteroscedasticity. 1 5 7WOMEN’S INVOLVEMENT IN PEDESTRIAN–VEHICLE CRASHES Gender Male Female Baltimore City Street Network FIGURE 1 Spatial distribution of pedestrian crashes by gender. 98709mvpTxt 155_210 9/20/05 5:42 PM Page 157

RESULTS Pedestrian and Crash Characteristics As shown in Table 2, the majority of crashes involved adults aged 16 to 64. A very few (less than 5%) crashes involved the elderly (65+). Thirty-six percent of pedes- trian crashes occurred with children (0 to 15 years of age). Men and women involved in a pedestrian–vehicle collision have similar age profiles. Seventy-eight percent of the pedestrian crashes were reported to have occurred at locations with malfunc- tioning or no signalization, and almost 80% of the fatal crashes occurred in areas with no traffic signal, either because there was no signal at the intersection or pedes- trians were crossing midblock. To explore this issue fur- ther, additional statistics were computed for severity of injury by compliance with traffic laws. Results show that pedestrians who did not obey a traffic signal were slightly more likely to sustain injuries than those who crossed with a signal (56% for the former and 51% for the latter). Of those involved in a pedestrian–vehicle collision, women were more likely to have been crossing in accordance with traffic laws (13% for women com- 1 5 8 RESEARCH ON WOMEN’S ISSUES IN TRANSPORTATION TABLE 1 Model Estimation Results: Natural Logarithm of Crash Density per Population Female Male Total Pedestrian activity 0.013 ** 0.002 0.008 Income (in $1000) 0.003 –0.005 0.000 % park –0.015*** –0.011 –0.011** Commercial accessibility 0.010*** 0.006** 0.009*** Transit accessibility –0.018*** –0.011*** –0.015*** Race 0.702*** 0.995*** 0.900*** % pop. < 16 years old 0.158 0.215 0.016 Education –0.930* –0.040 –0.458 Vehicle ownership 0.171 –0.132 –0.431 Density of roads 0.033*** 0.036*** 0.037*** N 339 422 473 R2 0.412 0.431 0.430 Corrected for heteroscedasticity. *Statistically significant at the 10% level. **Statistically significant at the 5% level. ***Statistically significant at the 1% level. TABLE 2 Pedestrian Crashes by Pedestrian Characteristics Characteristics of Crash or Statistical Tract Where Crash Is Located Total N Total % Male Female Significance Number of crashes 3009 100.0% 1778 1207 Age 0.168 Child (0 to 15 years of age) 1001 35.7% 37.1% 33.7% Adult (16 to 64 years of age) 1685 60.1% 58.8% 62.0% Elderly (over 65 years of age) 118 4.2% 4.1% 4.3% Total 2804 100.0% 100.0% 100.0% Traffic law obedience 0.000 No or malfunctioning signal 1971 78.1% 79.6% 75.8% Obeyed signal 229 9.1% 6.2% 13.3% Disobeyed signal 324 12.8% 14.2% 10.9% Total 2524 100.0% 100.0% 100.0% Severity of injury 0.064 No injury 1463 49.0% 47.3% 51.5% Nonfatal injury 1478 49.5% 51.1% 47.2% Fatality 44 1.5% 1.6% 1.2% Total 2985 100.0% 100.0% 100.0% Substance abuse 0.000 None 1200 90.7% 87.4% 95.7% Substance present 93 7.0% 9.3% 3.6% Substance contributed 30 2.3% 3.3% 0.8% Total 1323 100.0% 100.0% 100.0% Time of day 0.029 Daylight 2051 69.1% 67.4% 71.6% Dawn or dusk 145 4.9% 5.5% 4.0% Dark 772 26.0% 27.1% 24.4% Total 2968 100.0% 100.0% 100.0% 98709mvpTxt 155_210 9/20/05 5:42 PM Page 158

pared with 6% for men). These findings are consistent with research on the increased propensity of men to engage in risk-taking behavior. Regarding severity of injury, over half of the crashes resulted in injuries and 1.5% of reported crashes resulted in pedestrian fatalities. Women were slightly less likely to be injured in a crash with 51% of the crashes involving no injuries. Consistent with previous research, men were more likely to be injured in a crash and slightly more likely to die as a result. However, these gender differences in severity of injury were slight and only significant at the 90% confidence interval. For the pedestrians involved in the crash, the pres- ence of alcohol, medication, and illegal substances was reported in three categories: no substance detected, sub- stance present, and substance contributed to crash. In over 90% of the crashes involving pedestrians, no illegal substance was detected. Seven percent of pedestrian crashes involved pedestrians who had medication, alco- hol, or illegal substances present in their system, and in 2% the pedestrians had substances in their system that contributed to the crash. Of these, men proved more likely to test positive for some substance. In over 3% of all crashes involving men, alcohol, medication, or illegal substances contributed to the crash compared with less than 1% for women. In addition, men were more likely to have alcohol or other drugs or medications present at the scene of the accident and were more likely to have alcohol listed as a contributing factor to the crash. To explore this issue further, the results indicate that the severity of injury increased with presence of the ille- gal substance. Table 3 shows pedestrian crashes by sever- ity of injury and traffic obedience and substance abuse for both men and women. Seventy percent of pedestrian crashes in which a substance contributed to the crash resulted in injuries or fatalities. Alcohol was present in 8% of pedestrian crashes but accounted for more than 12% of fatalities. Also, an illegal drug was present in only 0.9% of pedestrian crashes but accounted for 3% of all fatalities. The data show differences in the rela- tionship between severity of injury and presence of sub- stance by gender. Specifically, for pedestrians involved in fatal crashes, 10.5% of men showed that a substance contributed to the crash, whereas no female fatalities were reported with a contributing substance. Almost 70% of all pedestrian crashes occurred dur- ing daylight hours and accounted for about 61% of fatalities. In contrast, more than 30% of pedestrian crashes occurred after dark but accounted for nearly 40% of the fatalities. These results indicate that night- time pedestrian crashes were more likely to result in more severe injuries. Men and women had similar dis- tributions of crashes by time of day; women (72%) are involved in slightly more crashes during daylight than men (67%). Although these differences were small, the analysis shows that the relationship between time of day and gender of the pedestrian is statistically significant. Land Use and Crash Characteristics Further gender analysis of the crashes was performed by using land use characteristics for the location of the pedes- trian crash; the results are shown in Table 4. Journey-to- work data at the tract level from the U.S. census were used as one indicator of pedestrian activity. Areas were catego- rized into low pedestrian activity (less than 10% of work- ers walk), medium (10% to 27% walk), and high (27% 1 5 9WOMEN’S INVOLVEMENT IN PEDESTRIAN–VEHICLE CRASHES TABLE 3 Pedestrian Crashes by Severity of Injury and Pedestrian Characteristics Severity of Injury Statistical Total % Total N No Injury Nonfatal Fatality Significance Traffic law obedience—male 0.393 No or malfunctioning signal 79.6% 1196 80.6% 78.5% 88.6% Obeyed signal 6.2% 93 6.6% 6.1% 0.0% Disobeyed signal 14.2% 213 12.8% 15.4% 11.5% Total 100.0% 1502 100.0% 100.0% 100.0% Traffic law obedience—female 0.750 No or malfunctioning signal 75.8% 775 76.8% 75.2% 61.5% Obeyed signal 13.3% 136 12.9% 13.5% 23.1% Disobeyed signal 10.9% 111 10.3% 11.3% 15.4% Total 100.0% 1022 100.0% 100.0% 100.0% Substance abuse—male 0.004 None 87.4% 692 91.6% 84.7% 68.4% Substance present 9.3% 74 6.2% 11.5% 21.1% Substance contributed 3.3% 26 2.2% 3.8% 10.5% Total 100.0% 792 100.0% 100.0% 100.0% Substance abuse—female 0.095 None 95.7% 508 97.5% 94.7% 84.6% Substance present 3.6% 19 2.1% 4.3% 15.4% Substance contributed 0.8% 4 0.4% 1.1% 0.0% Total 100.0% 531 100.0% 100.0% 100.0% 98709mvpTxt 155_210 9/20/05 5:42 PM Page 159

and more). Surprisingly, the results show that crashes have a negative correlation with levels of pedestrian activity. Tracts with a low percentage of the people who walk to work had the highest percentage (72%) of pedestrian–vehicle collisions compared with tracts with medium pedestrian activity (17% of crashes) and high activity (11% of crashes). Men were involved in slightly higher percentages of pedestrian crashes than women in tracts with low pedestrian activity; however, women were involved in a higher percentage of crashes in areas with medium and high pedestrian activity than men were. These differences are small and they are only statistically significant at the 90% confidence level. These results may indicate that neighborhoods with lower pedestrian activ- ity are more prone to having pedestrian crashes because motorists are not expecting pedestrians to be present. Walking to work is just one indicator of pedestrian activ- ity, and additional data on pedestrian demand may reveal different patterns. Figure 2 shows the spatial distribution of the pedestrian crashes by percentage of workers who walk to work. Similarly, tracts were categorized by the amount of parkland as low (less than 10% of tract area), medium (10% to 40% of tract area), and high (greater than 40%). The results were counterintuitive; tracts with the least amount of parkland have the highest number of pedestrian–vehicle crashes. Eighty-five percent of all crashes occurred in tracts with a low percentage of parks, and only 4% of all crashes occurred in tracts with high levels of park area. Again, motorists driving near park areas may be more attentive to pedestrian traffic. Gender of the pedestrian involved in crashes did not appear to have a significant relationship with parkland. The combination of age and presence of parkland had a notable but not significant effect on crashes. A low per- centage (3%) of adults and elderly were involved in crashes in areas with a high provision of parks. In con- trast, nearly twice as many pedestrian crashes (almost 6%) involving children occurred at locations with a high density of parkland. Population density and income levels of neighbor- hoods had an effect on the frequency of pedestrian crashes. Areas with medium levels of population density (10,000 to 20,000 persons per square mile) had the high- est percentage (43%) of pedestrian crashes compared with 38% in low-density areas (fewer than 10,000 per- sons per square mile) and 19% in high-density areas (more than 20,000 persons per square mile). There were no significant gender effects by the income level of the tract. The lowest percentage of crashes (10%) occurred in high-income areas, pointing to the generally lower levels of walking for transporta- tion purposes in these areas. However, nearly 16% of fatal crashes occur in high-income areas. The reasons 1 6 0 RESEARCH ON WOMEN’S ISSUES IN TRANSPORTATION TABLE 4 Pedestrian Crashes by Land Use Characteristics of Crash or Statistical Tract Where Crash Is Located Total N Total % Male Female Significance Pedestrian activity 0.081 Low (less than 10%) 2143 71.8% 73.1% 69.8% Medium (10% to 27%) 506 17% 16.6% 17.5% High (>27%) 336 11.3% 10.3% 12.7% Total 2985 100.0% 100.0% 100.0% % parkland 0.241 Low (0% to 10%) 2534 84.9% 84.4% 85.7% Medium (10% to 40%) 322 10.8% 10.8% 10.8% High (>40%) 129 4.3% 4.8% 3.6% Total 2985 100.0% 100.0% 100.1% Population density 0.413 Low (0 to 10,000 persons/mile2) 1160 38.9% 39.7% 37.6% Medium (10,001 to 20,000 persons/mile2) 1271 42.6% 42.4% 42.9% High (>20,001 persons/mile2) 554 18.6% 17.9% 19.5% Total 2985 100.0% 100.0% 100.0% Median household income 0.912 Low (<$25,000) 1201 40.2% 40.6% 39.8% Medium ($25,000 to $40,000) 1474 49.4% 49.1% 49.8% High (>$40,000) 310 10.4% 10.3% 10.4% Total 2985 100.0% 100.0% 100.0% % within 1⁄4 mile of commercial parcel 0.714 Low (0% to 55%) 164 5.5% 5.7% 5.1% Medium (55% to 85%) 299 10.0% 10.2% 9.8% High (>85%) 2522 84.5% 84.1% 85.1% Total 2985 100.0% 100.0% 100.0% % within 1⁄4 mile of bus stop 0.890 Low (40% to 80%) 93 4.3% 4.4% 4.2% Medium (80% to 95%) 374 17.4% 17.7% 17.1% High (95% to 100%) 1681 78.3% 77.9% 78.8% Total 2148 100.0% 100.0% 100.1% 98709mvpTxt 155_210 9/20/05 5:42 PM Page 160

for this finding may be related to the more suburban nature of these areas, with wider roads, higher vehicle speeds, and less pedestrian infrastructure. The relationship between pedestrian–vehicle crashes and pedestrian access to commercial land uses was examined. The percentage of single-family residential units in the tract within 1⁄4 mi of commercial uses was employed as a measure of pedestrian access to retail and services and categorized as high (over 85% of house- holds live within 1⁄4 mi of commercial area), medium (55% to 85%), and low (less than 55%). Eighty-four percent of all pedestrian crashes occurred in areas with high levels of pedestrian access to commercial sites. The relationship between transit access and pedes- trian crashes was another land use variable analyzed in this study. The transit access measure was calculated as the percentage of single-family residential units in a tract that have a transit stop within 1⁄4 mi. As with much of the analysis, gender did not seem to have a signifi- cant effect on this relationship. But interestingly, areas with high transit access had the highest percentage of crashes (over 78%) and a higher percentage of crashes involving children occurred in areas of high transit access (81%). These results may indicate that areas with good accessibility to bus stops have more pedes- trians and therefore more opportunity for pedestrian crashes. Regression Analysis An examination of the pedestrian crash densities for each gender as a function of a variety of personal and environ- mental factors was made by using multivariate regression models. The models’ estimations, shown in Table 1, show a negative relationship between crashes and percentage of parkland available at the block-group level. This relation- ship was statistically significant for the female crash model but not for the male model. Since parks are an attractor for pedestrian activity, this negative relationship may be indicative that drivers are more careful around areas in which there would be a higher chance of pedestrians, specifically children, crossing the streets. Although the differences were not statistically signifi- cant, the models seem to indicate a gender difference based on income. The female model indicates that more of their crashes were associated with locations with higher income levels. This finding may be explained by the fact that these higher-income areas may have a greater percentage of the population who can afford cars. How- ever, the male model shows a negative relationship between income and pedestrian crashes. The race variable is defined as the percentage of non- white population in the area. This variable is positively associated with male and female crashes. This statisti- cally significant result showed that locations with a greater percentage of minority groups were more likely to have pedestrian crashes for both models. The interest of this study is the relationship between the land use variables and pedestrian crashes. The land use variables of interest in the model are those that show the different levels of population and infrastructure density. The variables for density of roads and commercial accessi- bility were statistically significant for both models; how- ever, the pedestrian activity variable is only significant for the female model. These models show that the propensity of crashes increases with the increase in density of roads (over 3%), pedestrian activity (approximately 1% for women and 0.2% for men), and commercial accessibility (approximately 1% for women and 0.6% for men). The reasons for these findings may be related to the greater number of pedestrians in that area and therefore greater opportunity for a pedestrian–vehicle crash. In the more suburban areas, with lower housing densities, there is usu- ally less pedestrian infrastructure, more vehicle ownership, and therefore less pedestrian activity. It is interesting that transit accessibility has a statisti- cally significant and negative impact on crashes. The log models show that a one-unit increase in transit accessi- bility results in a decrease of crashes by over 1%. This result may indicate that in areas with high transit acces- sibility, more people tend to walk and drivers may be more attuned to the pedestrian nature of the area. The downtown area has the greatest transit access and is an 1 6 1WOMEN’S INVOLVEMENT IN PEDESTRIAN–VEHICLE CRASHES FIGURE 2 Spatial distribution of pedestrian crashes by per- centage of workers who walk to work. 98709mvpTxt 155_210 9/20/05 5:42 PM Page 161

area with typically lower travel speeds for vehicles, which may contribute to the lower crash rates. The variable for education, defined as percentage of the population who attended college, was only statistically significant for the female model. The education variable was negatively related to crash rates, which may indicate that people may depend more on walking as a mode of transportation in locations with a less-educated popula- tion. However, this explanation does not address why this variable was significant only for the female model. CONCLUSIONS The evidence from the analysis presented here suggests that, in general, there are few significant gender effects in the majority of pedestrian crashes. Women tend to be involved in fewer pedestrian crashes overall and, when they are involved, appear to exhibit fewer risk-taking behaviors, such as violating traffic laws and consuming alcohol or drugs. Women were slightly less likely to be injured in a crash and less likely to die as a result. The effects of land use on pedestrian crash rates did not show significant effects by gender. However, a higher percentage of women’s crashes occur in areas with high pedestrian activity, which may be reflective of the distribution of areas in which women walk. These findings are not new but do confirm the evidence presented in previous research. Because there are no corresponding data on pedestrian demand and behaviors, one can only speculate about potential causes for these gender differences in crashes. Fortunately, it does not appear overall that women are at a particular disadvantage in terms of pedestrian safety. Improvements to walking environments are likely to have a similar effect for the safety of men and women. How- ever, policy and treatment interventions in term of risk- taking behavior may prove more effective if men are targeted. Before one dismisses the issue of gender differ- ences in pedestrian–vehicle collisions as insignificant, a comprehensive and complementary analysis of pedestrian demand and behavior would provide a more thorough understanding of women’s pedestrian safety issues. REFERENCES Carlsson-Kanyama, A., A.-L. Linden, and A. Thelander. 1999. Gender Differences in Environmental Impacts from Pat- terns of Transportation: A Case Study from Sweden. Society and Natural Resource, Vol. 12, pp. 355–369. Clifton, K. J., and J. Dill. 2005. Women’s Travel Behavior and Land Use: Will New Styles of Neighborhoods Lead to More Women Walking? In Conference Proceedings 35: Research on Women’s Issues in Transportation, Vol. 2, Transportation Research Board of the National Acade- mies, Washington, D.C., pp. 89–99. Crane, R. 2000. The Influence of Urban Form on Travel: An Interpretive Review. Journal of Planning Literature, Vol. 15, No. 1, pp. 3–23. Demetriades, D., J. Murray, M. Martin, G. Velmahos, A. Salim, K. Alo, and P. Rhee. 2004. Pedestrian Injured by Automobiles: Relationship of Age to Injury Type and Severity. Journal of the American College of Surgeons, Vol. 199, No. 3, pp. 382–387. Handy, S. 1996. Understanding the Link Between Urban Form and Nonwork Travel Behavior. Journal of Planning Education and Research, Vol. 15, No. 3, pp. 183–198. Hebert Martinez, K. L., and B. E. Porter. 2004. The Likelihood of Becoming a Pedestrian Fatality and Drivers’ Knowl- edge of Pedestrian Rights and Responsibilities in the Commonwealth of Virginia. Transportation Research, Vol. 7F, pp. 43–58. Hayakawa, H., P. S. Fischbeck, and B. Fischhoff. 2000. Traf- fic Accident Statistics and Risk Perceptions in Japan and the United States. Accident Analysis and Prevention, Vol. 32, pp. 827–835. Hijar, M., J. Trostle, and M. Bronfman. 2003. Pedestrian Injuries in Mexico: A Multi-Method Approach. Social Science and Medicine, Vol. 57, pp. 2149–2159. Khan, F. M., M. Jawaid, H. Chotani, and S. Luby. 1999. Pedes- trian Environment and Behavior in Karachi, Pakistan. Accident Analysis and Prevention, Vol. 31, pp. 335–339. NHTSA. 2001. Traffic Safety Facts 2000. Report DOT-HS- 809-337. U.S. Department of Transportation. NHTSA. 2004. Traffic Safety Facts 2002: Pedestrians. U.S. Department of Transportation. www-nrd.nhtsa. dot.gov/pdf/nrd-30/NCSA/TSF2002/2002pedfacts.pdf. Accessed July 2004. Preusser, D. F., J. K. Wells, A. F. Williams, and H. B. Weinstein. 2002. Pedestrian Crashes in Washington, DC and Baltimore. Accident Analysis and Prevention, Vol. 34, pp. 703–710. Root, A., L. Schintler, and K. Button. 2000. Women, Travel and the Idea of “Sustainable Transport.” Transport Reviews, Vol. 20, No. 3, pp. 369–383. Rosenbloom, S. 1997. Trends in Women’s Travel Patterns. In Women’s Travel Issues: Proceedings from the Second National Conference, Oct. 1996, Report FHWA-PL-97- 024, FHWA, U.S. Department of Transportation. Stutts, J. C., W. W. Hunter, and W. E. Pein. 1996. Pedestrian–Vehicle Crash Types: An Update. In Transporta- tion Research Record 1538, TRB, National Research Council, Washington, D.C., pp. 68–74. Thom, B. 2003. Risk-Taking Behavior in Men: Substance Use and Gender. Health Development Agency, London. WHO. 2003. The Injury Chartbook: A Graphical Overview of the Burden of Injuries. www.who.int/violence_ injury_prevention/injury/chartbook/chartb/en/. Accessed March 2, 2004. 1 6 2 RESEARCH ON WOMEN’S ISSUES IN TRANSPORTATION 98709mvpTxt 155_210 9/20/05 5:42 PM Page 162

1 6 3 Nonfatal Transportation-Related Injury Among Women Differences in Injury Patterns and Severity by Age Ann M. Dellinger, Centers for Disease Control and Prevention Transportation ranks among the leading causes of death and injury for women in the United States. National esti- mates of nonfatal injury were based on weighted data from 31,144 women aged 15 years and older treated in hospital emergency departments during 2002. These injuries were compared with data from 31,733 men col- lected in the same system. Injuries were classified by age, disposition (e.g., treated and released, hospitalized), per- son type (e.g., occupant, pedestrian), body area injured, and type of injury (diagnosis). The 31,144 injuries repre- sented an estimated 1.8 million transportation injuries to women in the United States. The majority (93.6%) of injured women were treated and released. Data were divided into five categories: motor vehicle occupant, pedestrian, pedal cyclist, motorcyclist, and all other transportation-related injuries. An estimated 1,495,884 female occupants were injured during 2002. Women (1,280) had a higher occupant injury rate per 100,000 population than men (1,127). Men had higher rates as pedestrians, pedal cyclists, and motorcyclists. This analy- sis demonstrates the heavy burden of transportation injury among women, along with notable differences in injury severity and injury patterns. The need for personal vehicles to meet dailyrequirements for independent living has risen asthe U.S. population has spread beyond urban areas. The convenience of personal vehicles has allowed the distance from where one lives and works, shops, and enjoys leisure time to grow. Additional time on the road increases the risk of crashes and their accompanying injuries. In 2003 there were 97.9 mil- lion female licensed drivers (1), driving an average of 44 min a day (2). Transportation ranks among the leading causes of death and nonfatal injury for women in the United States (3). Evans (4) has shown that in comparable crashes, women are at greater risk of death than are men of the same age; this relationship appears to hold true for persons in the mid-teen years through the late fifties. In addition to the risk of death, after major trauma women are at greater risk for lower quality-of-life out- comes and increased psychologic sequelae such as depression (5, 6). These findings point to an important health threat to women in the United States. Although the number of transportation-related deaths has been well studied, the extent of nonfatal injury has garnered less attention. The purpose of this study was to assess the size of the burden of nonfatal transportation-related injury among women in the United States. DATA SOURCES AND METHODS Data were obtained from the National Electronic Injury Surveillance System-All Injury Program (NEISS-AIP) operated by the U.S. Consumer Product Safety Commis- sion. This system consists of a nationally representative sample of hospital emergency department visits `from 66 hospitals. Data were collected from medical records, and the most severe injuries were recorded. The most severe injury was based on the principal diagnosis as recorded 98709mvpTxt 155_210 9/20/05 5:42 PM Page 163

on the emergency department record. Data were weighted by the inverse of the probability of selection to provide national estimates; only the initial visit for any nonfatal injury was recorded. Visits were excluded if there was no diagnosis of injury—for example, pain only—or if the victim was pronounced dead on arrival. This study includes injuries categorized as transportation-related sustained by motor vehicle occu- pants (drivers and passengers), pedestrians, pedal cyclists, and motorcyclists. National estimates were based on weighted data from 31,144 nonfatal transportation- related injuries among women aged 15 years and older treated in emergency departments during the period from January to December 2002. Data on these injuries were compared with data from 31,733 men collected in the same system. Injuries were classified by age, disposition (e.g., treated and released, hospitalized), person type (e.g., motor vehicle occupant, pedestrian), primary body area injured, and type of injury (primary diagnosis). Popula- tion data were from postcensal bridged-race population estimates. Confidence intervals were designed to take into account the complex survey design and sample weights. FINDINGS Demographic Characteristics of Sample The 31,144 injuries represented an estimated 1.8 mil- lion nonfatal transportation-related injuries to women in the United States during 2002. More than half of the injuries (53.4%) were among those aged 25 to 54, 6.3% were among teenagers aged 15 to 19, and 8.0% were among women aged 65 and older. The age distribution for men was similar. The majority (91.8%) of those injured were treated and released, although a slightly higher proportion of women were treated and released (93.6%) than men (90.1%). Consequently, women showed a lower proportion (4.7%) of hospitalizations than men (7.6%). These hospitalizations represented more than 86,000 women with severe transportation- related injuries (Table 1). The proportion hospitalized was consistently low at 3% to 4% until ages 55 to 64, when the proportion hospitalized began to rise with age, reaching a high of 18.1% among women aged 85 and older. Data from the emergency department visits were divided into five categories: motor vehicle occupant, pedestrian, pedal cyclist, motorcyclist, and all other transportation-related injuries. Most (74.0%) injuries were to motor vehicle occupants, although the propor- tion differed for women (81.8%) compared with men (66.3%). An estimated 1,495,884 female motor vehicle occupants were injured during 2002 (Table 1). The proportion of pedestrian injuries was similar for women (3.0%) and men (4.0%); however, there were differences in the proportion of pedal cyclists (women, 2.7%; men, 8.8%) and motorcyclists (women, 1.3%; men, 7.8%). Although the proportion of each of these categories of injuries was less than 5% among women, 1 6 4 RESEARCH ON WOMEN’S ISSUES IN TRANSPORTATION TABLE 1 Weighted Estimates of Nonfatal Injuries Treated in Emergency Departments by Age and Sex, United States, 2002 Females Males Sample Weighted 95% Confidence Sample Weighted 95% Confidence Characteristic Size Estimate Interval Size Estimate Interval 31,144 1,828,442 1,540,964–2,115,921 31,733 1,878,056 1,600,406–2,155,705 Age 15–19 5,078 298,204 243,925–352,483 5,376 320,401 265,470–375,333 20–24 4,743 279,047 225,801–332,293 5,100 306,230 252,432–360,028 25–34 6,869 392,331 328,946–455,717 7,595 433,804 369,469–498,139 35–44 6,015 349,756 289,152–410,360 6,213 363,109 312,461–413,757 45–54 3,970 233,772 195,260–272,285 3,742 221,002 187,215–254,789 55–64 2,148 129,032 105,988–152,075 1,873 114,338 93,917–134,759 65–74 1,195 74,964 61,982–87,946 1,014 64,055 53,066–75,043 75–84 855 54,114 41,673–66,555 645 42,716 33,490–51,942 85+ 271 17,221 12,967–21,475 175 12,400 8,923–15,876 Disposition Treated/released 28,875 1,711,471 1,439,144–1,983,798 28,079 1,692,615 1,450,763–1,934,468 Hospitalized 1,870 86,175 48,961–123,389 3,115 143,144 76,827–209,461 Transferred 226 18,884 14,015–23,753 311 27,370 20,614–34,126 Other* 173 11,913 7,026–16,799 228 14,926 9,809–20,044 Type Occupant 25,928 1,495,884 1,240,092–1,751,675 21,959 1,245,314 1,048,166–1,442,462 Pedestrian 1,059 55,210 42,575–67,846 1,393 75,475 57,039–93,911 Pedal cyclist 743 49,989 31,988–67,989 2,625 165,868 122,707–209,030 Motorcyclist 342 23,408 18,112–28,704 2,147 146,559 113,314–179,804 Other transport** 3,072 203,952 175,553–232,350 3,609 244,839 206,350–283,329 *Includes observation only and unknown disposition. **Includes all other transportation-related injury such as off-road vehicles, buses, nonmoving vehicles (hand in car door), boats, etc. 98709mvpTxt 155_210 9/20/05 5:42 PM Page 164

the estimated number of injuries was substantial: 55,210 as pedestrians, 49,989 as pedal cyclists, and 23,408 as motorcyclists (Table 1). Types of Injury Body Part Overall, nearly half (44%) of those injured sustained head or neck injuries, 47% among women and 42% among men. Upper and lower trunk injuries were more prevalent than leg or foot or arm or hand injuries. The head and neck area was most commonly injured, but this amount differed by age group, with the proportion declining with age. For example, 49% of women 15 to 19 years old sustained head or neck injury as their most severe injury compared with 42% among those 55 to 64 and 28% among those 85 and older. The upper trunk was the second most common area injured but this injury was more likely with increasing age: 13% among women 15 to 19, 17% among those 45 to 54, and 23% among those 85 and older. This general pattern also was seen among men (Table 2). Diagnosis The five most common types of nonfatal transportation injury were strains and sprains followed by contusions and abrasions, fractures, lacerations, and internal injury (Figures 1–5). This ranking was similar for women and men, although the proportions differed. Women were more likely to sustain strains and sprains (48%) than men (37%) (Figure 1), whereas the proportions of con- tusions and abrasions were similar (women 29%, men 27%) (Figure 2). Men had more fractures (12%) than women (7%) (Figure 3). The distribution of injury types differed by age. Strains and sprains were more common in younger age groups and represented more than half of the injuries among women ages 20 to 54, a third of the injuries among women ages 65 to 74, and less than a quarter of the injuries among women ages 75 and older. The pro- portion of fractures increased with age, reaching a high of 24% among women 85 and older. Internal injuries 1 6 5NONFATAL TRANSPORTATION-RELATED INJURY AMONG WOMEN TABLE 2 Proportion of Nonfatal Injuries Treated in Emergency Departments by Body Part, Age, and Sex, United States, 2002 Body Part Injured Head/Neck Upper Trunk Lower Trunk Leg/Foot Arm/Hand Other/Unknown Females 15–19 49 13 12 13 11 2 20–24 51 13 14 11 11 1 25–34 48 14 15 11 10 2 35–44 47 15 15 12 10 2 45–54 46 17 13 13 9 2 55–64 42 22 10 15 10 2 65–74 36 21 12 17 13 2 75–84 34 24 13 15 12 2 85+ 28 23 12 14 23 1 Males 15–19 44 13 8 16 17 2 20–24 44 14 12 15 13 2 25–34 41 16 14 15 12 2 35–44 41 18 15 14 11 2 45–54 39 20 14 14 11 2 55–64 42 20 12 12 12 2 65–74 43 24 11 11 10 2 75–84 38 23 12 12 12 2 85+ 39 26 10 12 11 2 Both sexes 44 16 13 14 12 2 0 15–19 20–24 24–34 35–44 45–54 55–64 65–74 75–84 85+ 10 20 30 40 50 60 Age Group Percentage Women Men FIGURE 1 Most common types of nonfatal injury for persons treated in emergency departments, United States, 2002: strains and sprains. 98709mvpTxt 155_210 9/20/05 5:42 PM Page 165

were most common among women ages 75 to 84 (7%) and lowest among women 85 and older (4%) (Figure 5). Nonfatal Injury Rates Nonfatal injury rates per 100,000 population varied by category (e.g., occupant, pedestrian) and by sex. The overall transportation injury rate was higher among men (1,700) than among women (1,565). Motor vehicle occu- pants had higher rates of injury than pedestrians, pedal cyclists, motorcyclists, and all other transportation- related categories. Women (1,280) had a higher occupant injury rate than men (1,127), but men had higher injury rates for all other categories (pedestrian, pedal cyclist, motorcyclist, and other transport). The pedestrian injury rate was 1.5 times higher in men than women (68 versus 47), the pedal cyclist injury rate was 3.5 times higher in men (150 versus 43), and the motorcyclist injury rate was 6.5 times higher in men (133 versus 20) (Table 3). DISCUSSION OF RESULTS This analysis demonstrates the heavy burden of transportation-related injury among women in the United States, along with notable differences in injury severity and injury patterns. The 1.8 million injuries represent an average 5,000 women a day treated at hos- pital emergency departments for transportation-related injuries; the majority, nearly 1.5 million, were motor vehicle occupants. Although most women were treated 1 6 6 RESEARCH ON WOMEN’S ISSUES IN TRANSPORTATION 0 5 10 15 20 25 30 35 40 Percentage 15–19 20–24 24–34 35–44 45–54 55–64 65–74 75–84 85+ Age Group Women Men FIGURE 2 Most common types of nonfatal injury for persons treated in emergency departments, United States, 2002: contusions and abrasions. 0 2 4 6 8 10 12 14 16 18 Percentage 15–19 20–24 24–34 35–44 45–54 55–64 65–74 75–84 85+ Age Group Women Men FIGURE 4 Most common types of nonfatal injury for persons treated in emergency departments, United States, 2002: lacerations. 0 5 10 15 20 25 Percentage 15–19 20–24 24–34 35–44 45–54 55–64 65–74 75–84 85+ Age Group Women Men FIGURE 3 Most common types of nonfatal injury for persons treated in emergency departments, United States, 2002: fractures. 0 1 2 3 4 5 6 7 8 9 10 15–19 20–24 24–34 35–44 45–54 55–64 65–74 75–84 85+ Age Group Percentage Women Men FIGURE 5 Most common types of nonfatal injury for persons treated in emergency departments, United States, 2002: internal injuries. 98709mvpTxt 155_210 9/20/05 5:42 PM Page 166

and released, an estimated 86,000 were hospitalized and another 18,000 were transferred for more special- ized care. These 100,000+ women with more serious injuries were not evenly distributed across age groups; the proportions of both hospitalized and transferred increased with age. The head and neck were by far the most commonly injured areas for women, and this proportion decreased with age. Strains and sprains were the most common diagnoses, and this proportion also decreased with age. Taken together, these results imply that problems such as whiplash are frequent and might be expected less fre- quently among older women. This finding is consistent with previous research (7, 8) and a separate analysis of the NEISS-AIP database that found that women and younger persons were at higher risk of whiplash injury (9). An alternative explanation may be the nature of the data collection system, which records only the most severe injury. If older women have other injuries that are more severe than neck strain, the distribution of neck strain would be skewed toward younger age groups. The finding that women aged 75 to 84 had the highest proportion of internal injuries whereas those 85 and older had the lowest was surprising. It may be that because the incidence of hip fracture rises dramatically with age, the most severe injury would be recorded (the hip fracture) and the proportion of recorded internal injuries would decline. Alternatively, women in the oldest age group may be more likely to die of internal injuries and not be recorded in this database of nonfatal injury. There were differences in injury rates by sex. Men had higher rates of injury in all categories except motor vehicle occupant. Rates were based on age- and sex- specific population estimates, so there was no adjust- ment for exposure to the traffic environment. It is likely that men have higher rates because of their increased exposure as pedal cyclists, motorcyclists, and off-road vehicle users (all-terrain vehicle users were included in the “other transport” category). There is also evidence that men have higher rates of risk-taking behavior (10, 11) and that they are less likely to seek care than women. If men are less likely to seek care for injuries that do not necessarily require care, strains and sprains, for example, their proportion of injuries that typically require care, say fractures, would be higher. This theory is consistent with the finding here that men had a 60% higher proportion of fractures than women. There are both disadvantages and advantages to the use of the NEISS-AIP for the study of transportation- related injury. The NEISS-AIP database provides national estimates but cannot provide state or local esti- mates; therefore, analyses of potential geographic varia- tions in injury distributions are not possible. In addition, certainly nonfatal injuries treated in outpatient clinics and physicians’ offices will not be captured in this sys- tem. Last, variables other than age also affect the sever- ity and outcome of injury. Emergency department medical records did not consistently include informa- tion about restraint use, seating position, alcohol use, or medication use. Enhancements to the data collected in emergency departments would improve the ability to develop effective prevention programs. For example, more complete information on seating position would 1 6 7NONFATAL TRANSPORTATION-RELATED INJURY AMONG WOMEN TABLE 3 Nonfatal Injury Rates for Persons Treated in Emergency Departments by Type of Person and Sex, United States, 2002 Weighted Injury Rate per Estimate 100,000 Population 95% Confidence Interval Females Occupant 1,495,884 1280 1061–1499 Pedestrian 55,210 47 36–58 Pedal cyclist 49,989 43 27–58 Motorcyclist 23,408 20 16–25 Other transport* 203,952 175 150–199 Overall 1,828,442 1565 1319–1811 Males Occupant 1,245,314 1127 949–1305 Pedestrian 75,475 68 52–85 Pedal cyclist 165,868 150 111–189 Motorcyclist 146,559 133 103–163 Other transport* 244,839 222 187–256 Overall 1,878,056 1700 1448–1951 Both sexes Occupant 2,741,883 1206 1009–1403 Pedestrian 130,796 58 44–71 Pedal cyclist 215,901 95 69–122 Motorcyclist 170,011 75 58–91 Other transport* 448,910 198 170–225 Total 3,707,502 1631 1387–1875 *Includes all other transportation-related injury such as off-road vehicles, buses, nonmoving vehicles (e.g., hand in car door), boats, etc. 98709mvpTxt 155_210 9/20/05 5:42 PM Page 167

enable estimates that are specific to drivers and to pas- sengers instead of a combined occupant category. If it was found that drivers had a different pattern of injury, programs to address driver issues could target drivers instead of using a broader occupant approach. Targeted programs tend to increase efficiency. The main advan- tage of the NEISS-AIP is that the large sample size and representative nature of these data allow for excellent estimates of the public health burden of nonfatal trans- portation injury among women. Moreover, these med- ical data can reveal the most important injuries by age group and serve to guide prevention efforts. REFERENCES 1. Highway Statistics 2003. FHWA, U.S. Department of Transportation, 2004. 2. NHTS 2001 Highlights Report. BTS03-05. Bureau of Transportation Statistics, U.S. Department of Trans- portation, 2003. 3. Web-based Injury Statistics Query and Reporting Sys- tem (WISQARS). National Center for Injury Prevention and Control, Centers for Disease Control and Preven- tion, 2003. www.cdc.gov/ncipc/wisqars. Accessed Aug. 29, 2004. 4. Evans, L. Traffic Safety. Science Serving Society, Bloom- field Hills, Mich., 2004. 5. Holbrook, T. L., and D. B. Hoyt. The Impact of Major Trauma: Quality-of-Life Outcomes Are Worse in Women Than in Men, Independent of Mechanism and Injury Severity. Journal of Trauma, Vol. 56, 2004, pp. 284–290. 6. Holbrook, T. L., D. B. Hoyt, M. B. Stein, and W. J. Sieber. Gender Differences in Long-Term Posttraumatic Stress Disorder Outcomes After Major Trauma: Women Are at Higher Risk of Adverse Outcomes Than Men. Journal of Trauma, Vol. 53, 2002, pp. 882–888. 7. Farmer, C. M., J. K. Wells, and J. V. Werner. Relation- ship of Head Restraint Positioning to Driver Neck Injury in Rear-End Crashes. Accident Analysis and Pre- vention, Vol. 31, 1999, pp. 719–728. 8. Chapline, J. F., S. A. Ferguson, R. P. Lillis, A. K. Lund, and A. F. Williams. Neck Pain and Head Restraint Posi- tion Relative to the Driver’s Head in Rear-End Colli- sions. Accident Analysis and Prevention, Vol. 32, 2000, pp. 287–297. 9. Quinlan, K. P., J. L. Annest, B. Myers, G. Ryan, and H. Hill. Neck Strains and Sprains Among Motor Vehicle Occupants—United States, 2000. Accident Analysis and Prevention, Vol. 36, 2004, pp. 21–27. 10. Turner, C., and R. McClure. Age and Gender Differ- ences in Risk-Taking Behaviour as an Explanation for High Incidence of Motor Vehicle Crashes as a Driver in Young Males. Injury Control and Safety Promotion, Vol. 10, 2003, pp. 123–130. 11. Howland, J., R. Hingson, T. W. Mangione, N. Bell, and S. Bak. Why Are Most Drowning Victims Men? Sex Dif- ferences in Aquatic Skills and Behaviors. American Journal of Public Health, Vol. 86, No. 1, pp. 93–96. 1 6 8 RESEARCH ON WOMEN’S ISSUES IN TRANSPORTATION 98709mvpTxt 155_210 9/20/05 5:42 PM Page 168