Capture-recapture

Department of Public Health Sciences

Home of the
South LondonIStroke Register SLSR

Capture-recapture models for estimating incidence

Funded by:

MRC Training Fellowship

Study team:

Dr Kate Tilling*

Dr Jonathan Sterne

Prof Charles Wolfe

Background:

Incidence of stroke is often estimated using population-based registers. Any register, however carefully conducted, is likely to miss some cases and cases missed may differ from those captured (for example, in disease severity). Any register not attempting to estimate the number of cases missed will under-estimate the true incidence of disease. Capture-recapture is often used to assess completeness of a register, although the usual two source model relies on assumption of independence of sources and equality of capture probability which are rarely satisfied in epidemiology.

Methods:

We compared traditional capure-recapture with recently-developed capture-recapture models including covariates. The models including covariates were used to estimate incidence of stroke using a population-based stroke register (in South London , UK), allowing for the association of patient characteristics with probability of capture.

Results:

616 strokes were registered in 2 years. Probability of capture by the register was associated with both demographic variables and the stroke-severity indicators. The model without covariates underestimated the number of non-fatal cases. The capture-recapture (including covariates) estimate of the population size showed that the register was approximately 88% complete. Adjusting for the number of cases missed increased the estimated incidence from 1.31 (95% CI 1.21 to 1.42) to 1.49 (95% CI 0.38 to 2.60) per 1000 people. The estimated incidence rate ratio for black compared to white people decreased from 2.18 (95% CI 1.8 to 2.7) to 1.93 (95% CI 1.05 to 3.54) after correction for under-ascertainment.

Conclusion:

Using these methods, data from an incomplete register can be used to calculate both incidence and age-standardised incidence. Stroke registers might be made more cost effective by using the combination of sources and covariates which most accurately estimates population size, instead of aiming for 100% completeness. Comparisons of incidence should only use registers which have attempted to estimate the number and type of cases missed.

References:

Tilling K, Sterne JAC. Capture-recapture models including covariate effects. American Journal of Epidemiology 1999; 149(4): 392-400.

Tilling K, Sterne JAC, Wolfe CDA. Estimation of the incidence of stroke using a capture-recapture model including covariates. International Journal of Epidemiology (in press).