[Geographical bias in the inhospital mortality estimated by the Hospital Morbidity Survey].

Background: The Hospital Morbidity Survey (EMH) includes, at the moment, 85% of hospitals and 90% of discharges, and is the only national data source that allows to deepen with information about diagnostics, gender or age, in the study of hospitals as place of death. This work aims to analyze the presence of inhospital mortality geographical biases in the EMH in relation to the sample universe represented by the Statistics of Health Establishments with Inpatient Regime (EESCRI).

Methods: We compared, for each province in 2004, the EMH estimations for discharges, deaths and the percent of mortality with the data from the EESCRI, and adjusting one linear regression model for the number of deaths and a second model for the percent of mortality.

Results: The EMH infraestimated the volume of discharges and deaths (-8.6% and -11.4%), but not the inhospital mortality rate (3.55% vs. 3.45%). In spite of the excellent correlation in the number of deaths, figures in the EMH are inferior to the EESCRI figures in most provinces, and in 13 provinces the discrepancies overcome the 20%. The percent of mortality showed discrepancies overcoming 20% in 9 provinces.

Conclusion: In 2004, the EMH infraestimates the discharges and deaths figures but, except for 9 provinces, there are not evidence of important biases in the percent of mortality.