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.
Summary<br />
1019-1715). To reduce possible bias due to heterogeneity among the patients, a covariate<br />
capture-<strong>recapture</strong> model was specified, i.e. a capture-<strong>recapture</strong> model <strong>in</strong>clud<strong>in</strong>g<br />
categorical covariates associated with the probability of capture <strong>in</strong> a register, because we<br />
expected and observed regional differences <strong>in</strong> the <strong>in</strong>cidence rate of Legionnaires’ disease.<br />
The covariate model <strong>in</strong>clud<strong>in</strong>g “region” as covariate estimated 886 Legionnaires’ disease<br />
patients (95%CI 827-1022), result<strong>in</strong>g <strong>in</strong> an estimated completeness of notification of<br />
42.1%. The notified, ascerta<strong>in</strong>ed and estimated average annual <strong>in</strong>cidence rates of<br />
Legionnaires’ disease were 1.2, 2.4 and 2.8 per 100 000 <strong>in</strong>habitants respectively but higher<br />
<strong>in</strong> the southern region of the Netherlands. We conclude that covariate capture-<strong>recapture</strong><br />
analysis, acknowledg<strong>in</strong>g regional differences of Legionnaires’ disease <strong>in</strong>cidence, appears to<br />
reduce bias <strong>in</strong> the estimated national <strong>in</strong>cidence rate <strong>in</strong> the Netherlands. In Chapter 6 we<br />
describe a systematic process of record-l<strong>in</strong>kage, cross-validation, case-ascerta<strong>in</strong>ment and<br />
capture-<strong>recapture</strong> analysis to assess the quality of tuberculosis registers and to estimate<br />
the completeness of notification of <strong>in</strong>cident tuberculosis cases <strong>in</strong> 1998. A saturated logl<strong>in</strong>ear<br />
capture-<strong>recapture</strong> model <strong>in</strong>itially estimated an unexpectedly high number of 2053<br />
(95%CI 1871-2443) tuberculosis cases, result<strong>in</strong>g <strong>in</strong> an estimated completeness of<br />
notification of 63.2%. After adjustment for possible imperfect record-l<strong>in</strong>kage and<br />
rema<strong>in</strong><strong>in</strong>g false-positive hospital cases a more parsimonious and better fitt<strong>in</strong>g capture<strong>recapture</strong><br />
model estimated 1547 (95%CI 1513-1600) tuberculosis cases, result<strong>in</strong>g <strong>in</strong> a<br />
completeness of notification of 86.4%. Truncated population estimators gave similar<br />
results. In this chapter we demonstrate the possible impact of violation of the perfect<br />
record-l<strong>in</strong>kage and perfect positive predictive value assumptions on capture-<strong>recapture</strong><br />
estimates.<br />
In the context of the second research question of this thesis, after chapter 6<br />
describ<strong>in</strong>g capture-<strong>recapture</strong> analysis to estimate completeness of notification of<br />
tuberculosis dur<strong>in</strong>g one year at the national level <strong>in</strong> the Netherlands, <strong>in</strong> Chapter 7 we<br />
estimate tuberculosis <strong>in</strong>cidence and completeness of tuberculosis registration systems<br />
dur<strong>in</strong>g one year at the regional level <strong>in</strong> the Piedmont Region of Italy. A parsimonious<br />
capture-<strong>recapture</strong> model estimated 704 (95%CI 688-728) tuberculosis patients, result<strong>in</strong>g<br />
<strong>in</strong> an estimated completeness of notification of 79.1%. The overall estimated tuberculosis<br />
<strong>in</strong>cidence rate <strong>in</strong> the Piedmont Region is 16.7 cases per 100 000 population but varies<br />
between different subsets of the population. We conclude that when multiple record<strong>in</strong>g<br />
systems are available, record-l<strong>in</strong>kage can improve case-detection and capture-<strong>recapture</strong><br />
analysis can be used to assess tuberculosis <strong>in</strong>cidence and the completeness of notification,<br />
contribut<strong>in</strong>g to a more accurate surveillance of local tuberculosis epidemiology. In<br />
Chapter 8 we estimated the completeness of tuberculosis notification <strong>in</strong> England at the<br />
national level for four years to assess the performance of the Enhanced Tuberculosis<br />
<strong>Surveillance</strong> (ETS) system, <strong>in</strong>troduced <strong>in</strong> 1999. Due to the scale of this study (28 678<br />
observed patients), for record-l<strong>in</strong>kage of the hospitalised tuberculosis cases sophisticated<br />
record-l<strong>in</strong>kage computer software was required and the proportion of false-positive cases<br />
among the unl<strong>in</strong>ked hospital-derived tuberculosis records was estimated through a logistic<br />
regression population mixture model. Accord<strong>in</strong>g to a saturated capture-<strong>recapture</strong> model<br />
the estimated completeness of notification is 56.2%, highly <strong>in</strong>consistent with prior<br />
estimates. A truncated population estimator, Zelterman’s truncated Poisson-mixture<br />
model, estimated the completeness of notification at 79.5%. We conclude that record-<br />
176