Aim: to define an algorithm to estimate prevalence of ischemic heart disease from health administrative datasets.
Setting: four Italian areas: Venezia, Treviso, Torino, Firenze.
Participants: resident population in the four areas in the period 2002-2004 (only 2003 for Firenze) for a total of 2,350,000 inhabitants in 2003.
Main outcomes: annual crude and standardized prevalence rate (x100 inhabitants), 95% confidence intervals by area. Quality (comparability and coherence) indicators are also reported
Methods: the algorithm is based on record linkage of hospital discharges (SDO), pharmacological prescriptions (PF), exemptions from health-tax exemptions (ET) and causes of mortality (CM). From SDO we extracted discharges for ICD9-CM codes 410*-414* in all diagnoses in the estimation year and during the four years immediately preceding. We selected from PF subjects with at least two prescriptions of organic nitrates (ATC = C01DA*) in the estimation year. From ET subjects with a new exemption for ischemic heart disease (002.414) or who obtained exemption in the three years preceding, were selected. We also considered all deaths in the year for ischemic heart disease (ICD9 CM 410-414). Cases were defined as ischemic heart disease prevalent cases if they were extracted at least once from one of the datasets and if they were alive on January 1 of the estimation year.
Results: estimated crude prevalence ranges from 2.5 to 4%. The standardized prevalence led to a narrower range of values (2.8-3.3%). Venezia and Firenze show a higher standardized prevalence in both sexes (men 4.7% and 4.4%; women 2.3% and 2.2% respectively); Treviso and Torino present a lower standardized prevalence (men: 3.9%; women: 1.9%). The hospital discharges are the main source to identify prevalent subjects (34-48% of subjects are solely identified by SDO), pharmacological prescriptions are a relevant source in Firenze and Torino (27-28%), while they are less relevant in Venezia and Treviso (13-15%). ET shows a different contribution to prevalent case identification in the four areas: Venezia (8%), Treviso (3.2%), Firenze (1.3%), whereas in Torino this source was not available at all. Subjects classified as prevalent cases only through causes of death are less than 2%. The percentage of subjects simultaneously identified by multiple sources is high in Venezia (43%) and low in Torino (30%).
Conclusions: patterns in use of pharmaceuticals and exemptions from prescription charges appear to be heterogeneous in the different areas under study. These two aspects make a proper comparison between areas difficult. The algorithm could be applied only in areas with a similar use of organic nitrates and with a good comparability of the exemptions dataset.