Record Linkage Techniques -- 1997 Proceedings of an International Workshop and Exposition (1999) / Chapter Skim
Currently Skimming:
Chapter 10 Invited Session on More Record Linkage Applications in Epidemiology
Chapter 10 Invited Session on More Record Linkage Applications in Epidemiology
Pages 293-332
The Chapter Skim interface presents what we've algorithmically identified as the most significant single chunk of text within every page in the chapter.
Select key terms on the right to highlight them within pages of the chapter.
From page 293... ...
Kunitz, Clara Lee, and Rene C KozZof/; Kunitz and Associates, and Harvey Schwartz, Agency for Health Care Policy and Research Christian Home, Jean-Marie Berthelof, Pierre Davidt, and[MichaeZ WoZfson, Statistics Can adla Cam Mustard and Leslie Roos, University of Manitoba Steve Kendrick, National HeaZfh Service, Scotland 293 ~
|
From page 295... ...
Kozioff; Kunitz and Associates, Inc. Harvey Schwartz, Agency for Health Care Policy and Research '~''ons.'o.2,'-r,-,.e,-d,''::.,''a.~-'p e~.~ · - .
|
From page 296... ...
The primary objectives of this record linkage methodology project were to: link two patient- level related datasets that contain racial and ethnic descriptor; and assess the value of the linked data to address medical effectiveness research questions that focus on the quality, effectiveness, and outcomes from care for minority populations. Data Sets AHCPR's contractor, KAI, a health research firm, identified data sets to use for assessing the value of linking administrative health related data bases to support medical effectiveness research in minority populations.
|
From page 297... ...
The ambulatory surgery files were not segregated by year and contain slightly more than two million records. The CSRS data files are also summarized by year and contain a considerably smaller number of records because of the more narrow focus of the records on cardiac surgery.
|
From page 298... ...
The cohort was to be extracted from the linked SPARCS and CSRS datasets. The linked datasets were to contain records for 3 years, 1991-1993.
|
From page 299... ...
-- Race Code Conversions Description ~ SPARCS Race | CSRS Race Asian orPacificIslander | 1 | 8 Black 2 2 Hispanic 3 Native American 4 Other 5 White l 6 | 1 Linkage Objective The linkage objective was to build a longitudinal, comprehensive patient history that captured clinical encounters over time and across care settings. Thus, records for the same patient were linked in two ways: matches were performed within each of the three data sets; and matches were performed between the DDA/UBF files and CSRS and between the DDA/UBF and Ambulatory Surgery files.
|
From page 300... ...
The responses are fairly consistent across DDA and UBF subfiles and between linked and unlinked records with slight differences in reimburser and diagnoses, which could be a function of the research question reflected in the linked files. The DDA variables were selected for matching and were compared for linked and unlinked patient records, because of their tendency to be more reliable in the clinical area.
|
From page 301... ...
The research question, while resulting in a complex subject identification procedure, was typical of many medical effectiveness questions. The amount of time, then, needed for progressing from a linked data set to analyses for outcomes research, is several months and should be built into the research planning process.
|
From page 302... ...
When the initial match between the DDA/UBF file and the Ambulatory Surgery file took place there were no matches. The resolution involved the recreation of the Ambulatory Surgery file using only the first ten characters of the MRN in the encryption process.
|
From page 303... ...
Linking is only the first step when the data are to be used to address research questions. The linkage process identifies a set of unique indexes for each of the patient records in each of the linked files.
|
From page 304... ...
· Recognize Time Needed for Research. -- Research efforts using linked data sets must allocate sufficient time and manpower resources to identify and extract the suitable subpopulation for a specific research question.
|
From page 305... ...
, Oregano Head Survey of 1990, Enquete Sand Quebec of 1987 and 1992-93, Health and Activity Limitation Survey of 1986 and 1991 (Stalistics Canada, 1988) , Canadian Health and Disability Survey of 1983-84 (Statistics Canada, 1986a)
|
From page 306... ...
in addidon, all activities with He linked data set are covered by a memorandum of understanding including Statistics Canada, He Universe of Manitoba end He Manitoba Ministry of Heals. Data The detailed questionnaire (questionnaire 2B)
|
From page 307... ...
developed at Statistics Canada, was used for He pairing stage. CANLINK is a probabilistic matching software that pairs records from two sets of data by using He discriminatory power of He common variables available.
|
From page 308... ...
It should be understood Hat As ideating Cl-~ via vet -~ __ ~ ^~ ~ ^~ ~ information was not used to determine the vagary or specific matches, but only to estimate actual matching rates at aggressed levels. Names and addresses were compared manually with those on the mircrofilmed 2B questionnaires kept at Statistics Canada.
|
From page 309... ...
Since it was Me 1986 Census file that entirely detenn~nes ache composition of this population, all the stratification variables were either taken directly Tom Mat file or derived from it. The final number of strata for private households was 61 I
|
From page 310... ...
10 20 30 40 50 60 70 80 90100 Age Table 1. -- Match Rate According to Mobility: Private Households Only Mobility Match Rate l % l 5_~ 1~_~o ~81.7 Same CD 65.8 Over CD 62.5 CD: Census Division, a geographic unit used by the Census.
|
From page 311... ...
-- Match Rate According to Family She: Private Households Only Family Size 1 1 1 2 1 3 1 4 1 5 1 6 ~ 7 1 8 1 9 1 10+ Match rate % 1 66.4 1 78.5 1 74.7 1 79.8 1 77.1 1 70.0 55.9 1 51.4 1 39.2 1 467 Resulis for Evaluation of Concordance of Pairs Formed Table 4 shows that overall, more than 95% of Me definde matches retained represent Me same individual. As Me sardine of 20,000 units was drawn Dom deflate matches only, this meant Mat Me matching was of exceptional quality.
|
From page 312... ...
The MH file contained no infonnation on residents of several c~egones of collective dwellings for which medical services were provided by He federal government, such as military camps and some Indian reserves, whereas He Census considered~ese persons to be residents of Manitoba. For purposes of comparison, we excluded persons living in nursing homes (an ~ns~ional collective dwelling)
|
From page 313... ...
-- Accuracy of the Sample by Age Group Versus MH: Private And Non-Ir~stitu~ional Collective Households Males Difference Females Difference Total Difference Age MH Tom Sample MH Tom Sample MH from Sample % % onto O to 4 years 32 743 ~2.57 3 ~ 105 1.98 63 848 2.28 5 to 14 years 78 076 1.47 73 912 2.87 151 988 2.15 15 to 24 Years 86 722 -~.24 82 971 1.61 169 693 0.15 25 to 44 years 165 783 -2.84 159 458 1.92 325 241 -0.51 45 to 64 years 96 989 -~.71 98 997 0.92 195 986 -0.38 65 years and + 57 904 -1.34 74 129 -1.10 132 033 -~.20 Total 518 217 -1.20 520 572 1.39 1 038 789 0.10 Tables 6, 7 and ~ compare He mortally rate, medical care mili~ion and hospital care Minion by whether they were eshm~ Dom our sa-le or Tom the MH file. It should be noted Hat He death rates reported ~ He literature (Statistics Canada, 1994b3 were slightly higher Pan those preserved In Table 6, ~ the difference increasing with age.
|
From page 314... ...
Table 7. ~ Number and Costs of Medical Services, 19X6-87 Fiscal Year: Private and Non-Institutional Collective Households | Number ofS rvices L Costs ofServ~ces ($)
|
From page 315... ...
For example, cane of the projects proposed byte MCHPE consists of analys~ng morbidity with respect to an indiv~dual's occupation and by examining He extent to which He head care utilization for a particular class of illnesses is related to He basic occupational group. The census data can be used to classify individuals according to the reported occupation, or according to Nether or not they are employed and whether or not they are in He labour force.
|
From page 316... ...
A study to examine the impact of parental socioeconomic stays on the use of hospital and ambulatory medical care services during Me first year of life has just been co - leted. It showed Mat after c~troHing for low bird weight, maternal age and the joint effects of education and income; for hospital care, education was significantly negatively associated and exhibited a threshold effect between Me lowest quartile and all Ever quartiles; for ambulatory treatment care, income was significantly associated and exhibited a linear effect; for preventive care, both income and education were associated and exhibited a threshold effect between The lowest quartiles and all Aver quartiles.
|
From page 317... ...
. The Impact of SocimEconomic Inequity on the Health Care Utilization Practices of Ants During He First Year of Life, Symposium on ~ergenerahonal Equity in Canada, Statistics Canada, Ottawa.
|
From page 318... ...
No. 84-208 Statistics Canada (1988J, Me Health and Acidity Limitation Survey, Selected Data for Canada, Provinces and Terntones,Cat.No.
|
From page 319... ...
-ease bitt fJ ~S=~st~cs~of~1 In i ~r ~thei:~i~ ..d,omh.~'.,'.b..'.'2"'~.,.'.~.',n''~' ''I '' ' ' '' 'h' 'hi - ,,, Introduction n ecord linkage using probability matching, like many fields of human endeavour, has progressed as a highly fruitful interplay between theory and experiment, axioms and pragmatism. One viewpoint would see record linkage as primarily a highly practical enterprise based on common-sense and close attention to the empirical characteristics of the data sets involved in any linkage.
|
From page 320... ...
The decision was taken that from 1968 all hospital discharge records, cancer registrations and death records would be held centrally in machine readable form and would contain patient identifying information (names, dates of birth, area of residence etc.~. The decision to hold patient identifying information was taken with probability matching in mind and reflected familiarity web the early work of Howard Newcombe in Canada and close contact between Scotland and the early stages of the Oxford record linkage initiative.
|
From page 321... ...
(Kendrick and Clarke, 1993~. The largest currently contains aD hospital discharge data, cancer registrations and Registrar General's death records from 1981 to 1995 (over 14 million records relating to just over 4 million individuals)
|
From page 322... ...
Making the linkage decision. -- How do we convert the probability weights representing relative odds into absolute odds which will support He linkage decision?
|
From page 323... ...
No matter how few newcomer records are involved, it is still necessary to sort all the central catalog records for the years of interest. If only a few years are involved, and especially if linkage is restricted to a subset of the central records e.g., cancer registrations, the exercise is feasible but immensely inefficient.
|
From page 324... ...
If both match Men we proceed to Bill probability matching between the catalog and newcomer records. If neither or only one match then no further action is taken.
|
From page 325... ...
The calculation of probability weights aims to provide a mathematical grounding for this decision. However, it is a fundamental characteristic of the odds represented by probability weights that they are relative odds rather than absolute odds.
|
From page 326... ...
~ this case we have relatively little leverage to imp rove the terms of conversion between relative odds and absolute odds. If however we are linking a file of hospital discharge records to a file of death records we can obtain some "structural leverage." Death only occurs once and assuming that this is reflected in there being only one dead record per person In He file of dead records, He linkage becomes many-to-one.
|
From page 327... ...
- Reflecting as Hey did the entire Scottish population as well as deaths and transfers these were large files: approximately 6.3 million NHSCR records against 7.S million CHI records.
|
From page 328... ...
There was a match between the Soundex/NYSIIS code of surname, first initial, date of birth, sex and NHS number. For the remaining 75O,OOO CHI records, probability matching was earned out Resources for clerical checking were limited and such checking was limited to a sample of best link pairs to determine a probability weight which would represent absolute odds for the correctness of a linkage which were sufficiently high for administrative purposes.
|
From page 329... ...
Thus, it is not possible for He linkage to bring together records In He target file by "bridging" between Hem -- this would involve restructuring or resorting. As we saw earlier, as patient record sets In He main linked database grew larger, He false positive rate crept upwards, oRen because of illegitimate bridging by new records.
|
From page 330... ...
linkage is not about He mechanical application of complex and abstract rules. As circumstances change and data sets vary there is unlikely ever to be one definitive best memos of ca~y~ng out record linkage using probability matching.
|
From page 331... ...
. Medical Record Linkage in Scotland, Health Bulletin (Edinburgh)
|
From page 332... ...
. The Best-Link Principle in He Probability Matching of Population Data Sets: The Scottish Eminence In Linking He Commune Health Index to He National Heal Service Central Register, Methods of Information in Medicine fin press)
|
Key Terms
This material may be derived from roughly machine-read images, and so is provided only to facilitate research.
More
information on Chapter Skim is available.