BACKGROUND: Linked hospital morbidity data can be used to estimate the incidence of serious chronic disease. However, incidence rates calculated from first-time hospital admissions tend to be overestimated as a result of the erroneous inclusion of prevalent cases that have had previous hospital admissions prior to the study observation period. To address this problem, we have developed the backcasting method. METHOD: A retrograde survival model was implemented to calculate the level of over-ascertainment of incidence according to the number of years of linked data on which the estimates were based and corresponding correction factors were calculated. The method is illustrated using the example of linked hospital morbidity data on diabetes mellitus and then acute myocardial infarction, which was validated against the Perth MONICA database for cardiovascular disease. RESULTS: Corrected estimates of the incidence of diabetes and acute myocardial infarction were produced. The incidence of diabetes was shown to be lower than in North America in accordance with prevalence estimates, whereas the incidence of acute myocardial infarction was overestimated by approximately 10%. CONCLUSION: A new method is presented for estimating incidence trends in disease from linked hospital morbidity data. The advantages of this method are its ease of use with routinely collected data and the relatively low cost of applying it in comparison with community surveys or maintaining formal disease registers. The method has other applications using linked data, such as the study of trends in first-time health care procedures and pharmaceutical prescriptions.
BACKGROUND: Linked hospital morbidity data can be used to estimate the incidence of serious chronic disease. However, incidence rates calculated from first-time hospital admissions tend to be overestimated as a result of the erroneous inclusion of prevalent cases that have had previous hospital admissions prior to the study observation period. To address this problem, we have developed the backcasting method. METHOD: A retrograde survival model was implemented to calculate the level of over-ascertainment of incidence according to the number of years of linked data on which the estimates were based and corresponding correction factors were calculated. The method is illustrated using the example of linked hospital morbidity data on diabetes mellitus and then acute myocardial infarction, which was validated against the Perth MONICA database for cardiovascular disease. RESULTS: Corrected estimates of the incidence of diabetes and acute myocardial infarction were produced. The incidence of diabetes was shown to be lower than in North America in accordance with prevalence estimates, whereas the incidence of acute myocardial infarction was overestimated by approximately 10%. CONCLUSION: A new method is presented for estimating incidence trends in disease from linked hospital morbidity data. The advantages of this method are its ease of use with routinely collected data and the relatively low cost of applying it in comparison with community surveys or maintaining formal disease registers. The method has other applications using linked data, such as the study of trends in first-time health care procedures and pharmaceutical prescriptions.
Authors: Yariv Gerber; Susan A Weston; Margaret M Redfield; Alanna M Chamberlain; Sheila M Manemann; Ruoxiang Jiang; Jill M Killian; Véronique L Roger Journal: JAMA Intern Med Date: 2015-06 Impact factor: 21.873
Authors: Clarabelle Pham; Orla Caffrey; David Ben-Tovim; Paul Hakendorf; Maria Crotty; Jonathan Karnon Journal: BMC Health Serv Res Date: 2012-08-21 Impact factor: 2.655
Authors: Huberdina L Koek; Jan W P F Kardaun; Evelien Gevers; Agnes de Bruin; Joannes B Reitsma; Diederick E Grobbee; Michiel L Bots Journal: Eur J Epidemiol Date: 2007-09-08 Impact factor: 8.082