OBJECTIVE: To evaluate the utility of capture-recapture methods using multiple, routinely collected, computerized data sources to estimate the numbers and prevalence of diabetes. Methods employed for regional and national monitoring of diabetes have been too inaccurate or too expensive. RESEARCH DESIGN AND METHODS: A survey was undertaken that used four sources of ascertainment to identify prevalent cases of known diabetes in community of Northern Italy: diabetic clinic and family physicians, hospital discharges, prescriptions, and reagent strips and insulin syringes. Capture-recapture methods were employed to estimate the number of missing cases and to adjust for undercount to accurately estimate the number of people who had diabetes. RESULTS: We identified 2,069 unique prevalent cases of known diabetes with the intensive case-finding procedure. The diabetic clinic and family physicians data source identified the largest number of cases. The evaluation of the two sample capture-recapture estimates showed that they were all biased downward because of dependencies between sources. Log-linear modeling was employed to take into account the dependence among all data sources and the heterogeneity of diabetic patients. This method estimated that 2,586 cases existed, resulting in an ascertainment-adjusted prevalence of 2.77% (95% confidence interval, 2.44-3.10). Thus, despite the active case identification, approximately 20% could not be identified. However, the number of cases and rates could easily be adjusted using capture-recapture. CONCLUSIONS: The study shows that a two-sample capture-recapture estimate could be very biased if the investigator is not assured that the sources are independent. However, if at least three data sources are employed, log-linear models allow estimation of the number and prevalence rate adjusted for the degree of undercount (in spite of both the dependence of data sources and the heterogeneity of the diabetic population). The critical factor, however, is that the application of multiple sources with capture-recapture methods could be applied across broad geographical areas and across time to have cost-effective monitoring of diabetes at local and national level.
OBJECTIVE: To evaluate the utility of capture-recapture methods using multiple, routinely collected, computerized data sources to estimate the numbers and prevalence of diabetes. Methods employed for regional and national monitoring of diabetes have been too inaccurate or too expensive. RESEARCH DESIGN AND METHODS: A survey was undertaken that used four sources of ascertainment to identify prevalent cases of known diabetes in community of Northern Italy: diabetic clinic and family physicians, hospital discharges, prescriptions, and reagent strips and insulin syringes. Capture-recapture methods were employed to estimate the number of missing cases and to adjust for undercount to accurately estimate the number of people who had diabetes. RESULTS: We identified 2,069 unique prevalent cases of known diabetes with the intensive case-finding procedure. The diabetic clinic and family physicians data source identified the largest number of cases. The evaluation of the two sample capture-recapture estimates showed that they were all biased downward because of dependencies between sources. Log-linear modeling was employed to take into account the dependence among all data sources and the heterogeneity of diabeticpatients. This method estimated that 2,586 cases existed, resulting in an ascertainment-adjusted prevalence of 2.77% (95% confidence interval, 2.44-3.10). Thus, despite the active case identification, approximately 20% could not be identified. However, the number of cases and rates could easily be adjusted using capture-recapture. CONCLUSIONS: The study shows that a two-sample capture-recapture estimate could be very biased if the investigator is not assured that the sources are independent. However, if at least three data sources are employed, log-linear models allow estimation of the number and prevalence rate adjusted for the degree of undercount (in spite of both the dependence of data sources and the heterogeneity of the diabetic population). The critical factor, however, is that the application of multiple sources with capture-recapture methods could be applied across broad geographical areas and across time to have cost-effective monitoring of diabetes at local and national level.
Authors: G Bruno; F Merletti; G Bargero; G Novelli; D Melis; A Soddu; M Perotto; G Pagano; P Cavallo-Perin Journal: Diabetologia Date: 2007-03-02 Impact factor: 10.122
Authors: G Bruno; F Merletti; A Biggeri; G Bargero; S Prina-Cerai; G Pagano; P Cavallo-Perin Journal: Diabetologia Date: 2006-03-09 Impact factor: 10.122
Authors: Terri H Lipman; Lorraine E Levitt Katz; Sarah J Ratcliffe; Kathryn M Murphy; Alexandra Aguilar; Iraj Rezvani; Carol J Howe; Shruti Fadia; Elizabeth Suarez Journal: Diabetes Care Date: 2013-01-22 Impact factor: 19.112
Authors: Geoffrey V Gill; Aziz A Ismail; Nicholas J Beeching; Sarah B J Macfarlane; Mark A Bellis Journal: J R Soc Med Date: 2003-07 Impact factor: 18.000