Evaluation of care and surveillance of cardiovascular disease: can we trust medico-administrative hospital data?

BACKGROUND

The evaluation of care and the surveillance of disease are important in respect to cardiovascular disease because it is prevalent and costly. In Canada, medico-administrative hospital data are readily available, continuously updated, and offer comprehensive coverage of the patient population. However, there is concern about the quality of the information.

METHODS

The reliability and predictive capability of comorbidity data contained within Québec's hospital discharge database were assessed in comparison with data collected by clinical medical record reabstraction in a sample of 1989 patients hospitalized from 2002 to 2006 in a mix of 13 hospitals. Patients either had a principal diagnosis of myocardial infarction or underwent angioplasty or bypass surgery. Twenty-one comorbidities included in the Charlson comorbidity index or known to be associated with mortality were validated via medical record reabstraction.

RESULTS

Of 14 comorbidities with > 2% prevalence, 8 had excellent agreement with medical record review (κ > 0.8) while 6 had substantial agreement (κ > 0.6). In general, positive predictive values were high, while measures of sensitivity were more variable. Univariate associations between comorbidities and 30-day and 1-year mortality were generally similar in the 2 data sources. Comorbidities retained in the final multivariate stepwise regression models from each data source were almost identical, as were the 2 models' abilities to predict mortality.

CONCLUSIONS

Hospital discharge data in Québec are, in general, reliably coded and compare favourably with clinical medical record review in their ability to predict mortality. It appears sufficiently reliable to provide useful information about clinical outcomes of cardiac care and to identify problems that warrant investigation.