Less Than Half of Major CVD Hospitalizations Not Recorded in Primary Care Morbidity Data

Data on drug prescriptions, diagnoses, and events are increasingly being used for epidemiologic and pharmacologic studies and for the evaluation of quality of care, but questions arise about how complete and accurate the diagnoses and events in these databases are.

Therefore, Dutch researchers evaluated how well patients with major cardiovascular disease (CVD) hospitalizations can be identified from primary care data on morbidity and/or drug prescriptions in a regional database of patients with type 2 diabetes. The database included prescriptions, morbidity, lab test results, and physical exams. The researchers also evaluated the accuracy of using different drugs (platelet aggregation inhibitors, vitamin K antagonists, or nitrates) as proxies to identify patients with a history of major CVD and to identify a first major CVD event in patients without a history of CVD. The study, by Koen Bernardus Pouwels and colleagues from the University of Groningen, The Netherlands, was published in BMC Health Services Research.

The study used data from 17,230 diabetes patients from the GIANTT databases and Dutch Hospital Data register. At baseline, the mean age of the study population was 66 years; 48% were men, median diabetes duration was 6 years, and the prevalence of at least 1 major CVD diagnosis recorded by a general practitioner was 16%. The study period was January 1, 2008, to December 31, 2009. Only patients surviving the whole study period were included; the study did not assess fatal events.

Only 43% of major CVD hospitalizations could be identified using diagnoses recorded in primary care. Adding drug prescriptions to the search increased the sensitivity up to 94%. Using at least 1 prescription of either a platelet aggregation inhibitor, vitamin K antagonist, or nitrate as a proxy could identify 85% of patients with a history of major CVD recorded in primary care.

“Our finding that less than half of the hospitalizations for a major CVD were identified using primary care morbidity records indicates that even for major events such records are incomplete,” the authors concluded.

The finding that a large proportion of major CVD hospitalizations is not recorded in primary care morbidity data is particularly relevant in the context of upcoming regional and nationwide initiatives to improve electronic healthcare databases, they noted. Their findings stress the importance of combining different types of data and linking different data sources to identify patients with disease and using drug prescriptions as proxies to help identify patients with major CVD.