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 ...
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
Chapter 8<br />
bacteriological confirmation, records three months before and after the study period were<br />
also exam<strong>in</strong>ed.<br />
Record-l<strong>in</strong>kage<br />
Duplicate entries with<strong>in</strong> each of the three data sources were excluded. Hospital records<br />
were l<strong>in</strong>ked to the previously l<strong>in</strong>ked Notification and Laboratory records. Record-l<strong>in</strong>kage<br />
software developed by the Centre for Infections establishes a likelihood of association<br />
between two records based on a core set of identifiers (date of birth, age, full postcode<br />
and sex of the patient and proximity of date of notification, <strong>in</strong>itial mycobacterial isolate or<br />
hospital admission). It allows for visual <strong>in</strong>spection of available additional <strong>in</strong>formation on<br />
geographical location, site of disease, ethnicity and smear, culture or histopathology<br />
results (when performed). All cases with <strong>in</strong>complete or miss<strong>in</strong>g <strong>in</strong>formation on both the<br />
date of birth and age were labelled as “<strong>in</strong>sufficient identifiers” and excluded.<br />
The software allocates an a priori determ<strong>in</strong>ed maximum number of po<strong>in</strong>ts to<br />
each core identifier for complete agreement, reflect<strong>in</strong>g the perceived relative importance<br />
of that identifier. Record-pairs with full agreement of all core identifiers are automatically<br />
assigned as true l<strong>in</strong>ks. Po<strong>in</strong>ts are deducted proportionally to the presumed loss of<br />
<strong>in</strong>formation for <strong>in</strong>creas<strong>in</strong>g deviation from perfect l<strong>in</strong>kage of each identifier to generate an<br />
aggregate score, reflect<strong>in</strong>g the likelihood of association between two patient records. All<br />
categories of candidate-l<strong>in</strong>ks other than automatically assigned l<strong>in</strong>ks were visually<br />
<strong>in</strong>spected and either accepted or rejected. L<strong>in</strong>ked cases were allocated to the year of first<br />
known date of notification, culture-confirmation or hospital admission.<br />
False-positive records and correction<br />
All laboratory-confirmed cases reported through MycobNet were assumed true<br />
tuberculosis cases, as previously found <strong>in</strong> a local capture-<strong>recapture</strong> study <strong>in</strong> England. 17<br />
Notification and Hospital records not l<strong>in</strong>ked with Laboratory could potentially <strong>in</strong>clude<br />
three groups of false-positive records:<br />
1. Cases ultimately diagnosed with an <strong>in</strong>fection with Mycobacteria Other Than<br />
Tuberculosis (MOTT)<br />
2. Cases with a f<strong>in</strong>al diagnosis other than tuberculosis or MOTT <strong>in</strong>fection.<br />
3. Cases misclassified or miscoded.<br />
The proportion of unl<strong>in</strong>ked Hospital cases attributable to MOTT <strong>in</strong>fection was<br />
estimated by l<strong>in</strong>k<strong>in</strong>g Hospital data from 2003 with a MOTT database which began <strong>in</strong> that<br />
same year and used to correct the number of unl<strong>in</strong>ked Hospital cases <strong>in</strong> all years under<br />
study us<strong>in</strong>g a formula expla<strong>in</strong>ed below, assum<strong>in</strong>g the annual proportion is similar.<br />
In order to estimate the proportion of cases with a f<strong>in</strong>al diagnosis other than<br />
tuberculosis or MOTT <strong>in</strong>fection Notification cases not known to Laboratory were l<strong>in</strong>ked<br />
with Treatment Outcome Monitor<strong>in</strong>g data, conta<strong>in</strong><strong>in</strong>g data on Notification cases with a<br />
f<strong>in</strong>al diagnosis other than tuberculosis. At the time of this study Treatment Outcome<br />
Monitor<strong>in</strong>g data were only available for 2001. The proportion of false-positive<br />
112