The use of league tables during the last decade has frequently been employed to assess quality in health care. However, few studies have attempted to assess quality by quantifying the variability across the organizational units or attempted to investigate whether the units are the correct context that really influences the outcome under study.
To quantify the variation between different organizational units regarding 5 different Swedish national pharmacological performance indicators and to examine whether the organizational units under study are a valid construct of the context that influences the specific outcome.
A multilevel model with patients nested within health care units that in turn were nested within County councils was used. By using measures of variance (intraclass correlation [ICC]), we quantified the extent to which the 5 indicators of health care quality were conditioned by the specified units.
For all 5 studied indicators, the variation between county councils was small (ICC ranged from 2% to 7%), whereas the variation among health care units seemed to be more important (ICC ranged from 20% to 40%).
As the variation between county councils was small, using league tables for performance evaluation seems to be inappropriate. If league tables are to be presented, the relative size of the variation at the higher levels and an analysis regarding the possible influence of the context for the specific outcome should be included. This approach provides useful information for identifying relevant contexts to capture health care variation.
From the *Department of Clinical Sciences, Social Epidemiology, Lund University, Malmö, Sweden; †Department of Clinical Science, Center for Primary Health Care Research, Lund University, Malmö, Sweden; and ‡Centro Superior de Investigación en Salud Pública (CSISP), Valencia, España.
Reprints: Henrik Ohlsson, PhD, Center for Primary Health Care Research CRC, Building 28, Floor 11, Entrance 72, Malmö University Hospital, S-205 02 Malmö, Sweden. E-mail: email@example.com.