Quality of case ascertainment in cancer registries: a proposal for a virtual three-source capture-recapture technique.

BACKGROUND: The ability and behaviour of the capture-recapture method using a virtual three-source model for evaluation of the level of completeness of case ascertainment requires exploration. METHODS: Cancer cases obtained from 9 population-based cancer registries in Thailand during 2003 to 2007 were applied for capture-recapture using a model based on clinical, pathological and mortality data. These three virtual sources were derived from three actual items common to all cancer registries: the basis of diagnosis, ICD-O morphology code, and last known patient status. Poisson regression models were fit to the data to estimate parameters which were then transformed into demographic values. A linear model was used to determine the predictors and estimated percentage of completeness (EPC) in case ascertainment among the cancer registries. RESULTS: The EPC was greater than 97% in 5 and less than 90% in 4 registries. The worst had an EPC of 70%. The percentage death certificate only (%DCO) and the interaction between %DCO and morphological verification (MV) were significantly associated with EPC. Other factors intrinsic to registries also exerted influence on the EPC. CONCLUSIONS: In addition to other standard indicators to monitor completeness of cancer registries, the present virtual three-source capture-recapture model can be routinely used to estimate the level of completeness of case ascertainment in cancer registries.

Appropriateness of the standard mortality/incidence ratio in evaluation of completeness of population-based cancer registry data.

BACKGROUND: The magnitude of differences in mortality incidence (M:I) ratios derived from the national mortality source and those derived from cancer registry (CR) databases may be used to determine associated factors. METHODS: All information on cancer incidence cases and mortality cases from January 1, 2003 to December 31, 2007 were retrieved from 5 population-based cancer registries in four regions of Thailand. Two sources of mortality were used: death cases within the cancer registries and mortality statistics obtained from the Ministry of Public Health (MOPH). Plots of percentage M:I ratios from cancer registry databases and from national mortality sources against 1 minus 5 years relative survival (1-5yrRS) were used to visualize the correlation between the two mortality sources. A Poisson regression model was used to determine the influence of cancer sites and registries on the M:I ratio/[1-5yrRS]. RESULTS: There was high variability between the standard M:I ratio derived from national mortality compared with 1-5 year RS. The factors affecting M:I ratios are sources of mortality data and misclassification of topographic site as the cause of death. CONCLUSIONS: Use of the M:I ratio is not recommended to evaluate completeness of cancer registry data when the quality of mortality data is poor.