Rob van Hest Capture-recapture Methods in Surveillance - RePub ...
Rob van Hest Capture-recapture Methods in Surveillance - RePub ...
Rob van Hest Capture-recapture Methods in Surveillance - RePub ...
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
Chapter 11<br />
overestimation of the number of malaria, Legionnaires’ disease and tuberculosis patients<br />
<strong>in</strong> the Netherlands.<br />
Perfect record-l<strong>in</strong>kage assumption<br />
For all three <strong>in</strong>fectious diseases studied, unique identification numbers, such as a social<br />
security number, used <strong>in</strong> all registers, allow<strong>in</strong>g optimal record-l<strong>in</strong>kage, did not exist. In all<br />
studies record-l<strong>in</strong>kage was manual, through almost similar procedures described <strong>in</strong><br />
chapters 4 to 6, us<strong>in</strong>g patients’ identifiers such as date of birth and postcode. The malaria<br />
and tuberculosis studies were performed before 1999 when <strong>in</strong> the Netherlands a new<br />
Infectious Diseases Act came <strong>in</strong>to force, 3 record<strong>in</strong>g only year of birth <strong>in</strong>stead of date of<br />
birth, effectively rul<strong>in</strong>g out reliable record-l<strong>in</strong>kage between the notification and other<br />
registers. Therefore still sufficient demographical, geographical and microbiological<br />
identifiers could be collected at source for adequate record-l<strong>in</strong>kage. In addition, for<br />
tuberculosis record-l<strong>in</strong>kage was relaxed and near-l<strong>in</strong>ks were double-checked. Despite<br />
these efforts still <strong>in</strong>dications for imperfect record-l<strong>in</strong>kage exist. First, the notification<br />
register and the Netherlands Tuberculosis Register (NTR) should overlap completely but<br />
we found only 91.1% overlap after record-l<strong>in</strong>kage. Second, some culture-positive<br />
tuberculosis patients could not be found <strong>in</strong> the notification register but were l<strong>in</strong>ked to the<br />
NTR. Although before cross-validation misclassification of tuberculosis patients was<br />
considered to be m<strong>in</strong>imal, this study demonstrated that imperfect record-l<strong>in</strong>kage can<br />
exceed expectations, with considerable impact on the capture-<strong>recapture</strong> estimates. Initially<br />
unidentified <strong>in</strong>dications for misclassification could not be <strong>in</strong>vestigated for malaria and<br />
Legionnaires’ disease.<br />
The capture-<strong>recapture</strong> study on the <strong>in</strong>cidence of Legionnaires’ disease was<br />
performed after the new Infectious Diseases Act came <strong>in</strong>to force. Therefore <strong>in</strong>formation<br />
on the date of birth of the patients could no longer be collected from the national<br />
notification register at source but had to be obta<strong>in</strong>ed from the local Public Health<br />
Services process<strong>in</strong>g the notifications, creat<strong>in</strong>g more opportunities for clerical errors and<br />
rema<strong>in</strong><strong>in</strong>g <strong>in</strong>complete <strong>in</strong>formation on personal identifiers <strong>in</strong> some records, jeopardis<strong>in</strong>g<br />
reliable record-l<strong>in</strong>kage. This may have caused more misclassification of patients over the<br />
registers compared to the malaria and tuberculosis capture-<strong>recapture</strong> studies.<br />
The impact of changes <strong>in</strong> the <strong>in</strong>fectious disease legislation <strong>in</strong> the Netherlands on<br />
the validity of capture-<strong>recapture</strong> estimates was demonstrated <strong>in</strong> a follow-up capture<strong>recapture</strong><br />
study on malaria <strong>in</strong>cidence <strong>in</strong> the Netherlands between 1995 and 2003. 4 After<br />
1999 the malaria <strong>in</strong>cidence estimates more than doubled, almost certa<strong>in</strong>ly reflect<strong>in</strong>g<br />
overestimation as a result of imperfect record-l<strong>in</strong>kage. Inaccurate record-l<strong>in</strong>kage, i.e.<br />
<strong>in</strong>correctly establish<strong>in</strong>g the <strong>recapture</strong>, can substantially alter the observed and unobserved<br />
fractions. 5<br />
Depend<strong>in</strong>g on the number of missed l<strong>in</strong>ks and misl<strong>in</strong>ks, violation of the perfect<br />
record-l<strong>in</strong>kage assumption can result <strong>in</strong> overestimation or underestimation of the number<br />
of malaria, Legionnaires’ disease and tuberculosis patients <strong>in</strong> the Netherlands.<br />
160