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.
General discussion<br />
Perfect positive predictive value assumption<br />
The positive predictive value of the notification and laboratory registers for malaria is<br />
considered to be high. Malaria is a specific disease, often requir<strong>in</strong>g a history of recent<br />
travell<strong>in</strong>g <strong>in</strong> tropical areas, and unlikely to be diagnosed without confirmation or strong<br />
suspicion. The laboratory plays a crucial role <strong>in</strong> the diagnosis through thick and th<strong>in</strong><br />
smear microscopy and serological antigen tests, all with a high specificity. Malaria has a<br />
specific ICD-9 code and although a number of patients could have been admitted to<br />
hospital for observation after develop<strong>in</strong>g fever follow<strong>in</strong>g a tropical journey without a f<strong>in</strong>al<br />
diagnosis of malaria, compared to Legionnaires’ disease and tuberculosis, the positive<br />
predictive value of the malaria hospital episode register is also expected to be high.<br />
For the notification of Legionnaires’ disease the criteria require a cl<strong>in</strong>ical picture<br />
compatible with pneumonia and a confirmed or probable microbiology laboratory<br />
diagnosis, accord<strong>in</strong>g to the European Work<strong>in</strong>g Group for Legionella Infections (EWGLI)<br />
def<strong>in</strong>ition. We demonstrate after record-l<strong>in</strong>kage that these criteria apparently are not<br />
applied uniformly over all registers. More than malaria, the registers could use different,<br />
less specific, case-def<strong>in</strong>itions, result<strong>in</strong>g <strong>in</strong> a proportion of false-positive cases <strong>in</strong> these<br />
registers, e.g. as the result of the absence of pneumonia, the low positive predictive value<br />
of the s<strong>in</strong>gle Legionella antibody titre test or the absence of a specific Legionnaires’<br />
disease code <strong>in</strong> ICD-9. 6,7<br />
For tuberculosis the number of false-positive cases is assumed to be zero <strong>in</strong> the<br />
reference laboratory register and limited <strong>in</strong> the notification register. The latter is due to a<br />
good organisation of tuberculosis surveillance <strong>in</strong> the Netherlands and identification of<br />
false-positive cases with an <strong>in</strong>fection caused by non-tuberculous mycobacteria or another<br />
diagnosis than non-tuberculous mycobacteriosis or tuberculosis through our crossvalidation.<br />
However, foreign reports <strong>in</strong>dicate considerable contam<strong>in</strong>ation of tuberculosis<br />
hospital registers with false-positive cases. 8,9 The results of our study support these<br />
observations as 62.4% of the unl<strong>in</strong>ked hospital cases could not be verified through crossvalidation,<br />
compared to 7.6% of the unl<strong>in</strong>ked notified cases. These possibly rema<strong>in</strong><strong>in</strong>g<br />
false-positive cases likely contribute to considerable bias <strong>in</strong> the capture-<strong>recapture</strong> estimate.<br />
Cross-validation and identification of assumed false-positive cases could not be<br />
performed for malaria and Legionnaires’ disease. Violation of the perfect positive<br />
predictive value assumption results <strong>in</strong> overestimation of the number of malaria,<br />
Legionnaires’ disease and tuberculosis patients <strong>in</strong> the Netherlands.<br />
Absence of specific <strong>in</strong>terdependencies assumption<br />
For all three <strong>in</strong>fectious diseases studied, co-operation between the conventional registers<br />
used (notification, laboratory and hospital) is expected, result<strong>in</strong>g <strong>in</strong> positive dependence<br />
and underestimation of the number of cases <strong>in</strong> two-source capture-<strong>recapture</strong> models.<br />
Therefore three-source log-l<strong>in</strong>ear capture-<strong>recapture</strong> analysis, <strong>in</strong>corporat<strong>in</strong>g possible pairwise<br />
dependencies, was selected <strong>in</strong> the study design to reduce bias.<br />
For malaria, significant <strong>in</strong>teraction between the laboratory and notification registers is not<br />
identified and could <strong>in</strong>dicate absent pre-notification of laboratory results to the Public<br />
Health Services at the time of this study. The explicable <strong>in</strong>teractions of notification by<br />
161