Errors in Survey Based Quality Evaluation Variables in Efficiency Models of Primary Care Physicians
Link to article:
Kittelsen, S.A.C., Kjæserud, G.G., & O.J.Kvamme
Number in series: 12
This report focuses on efficiency analyses in the health care sector which are often criticized for not incorporating quality variables. The report is mainly a theoretical contribution to this critique. The authors purpose is to analyze by means of statistical tests if the patients perception of health care services have an effect on efficiency results. This is done to judge if such patient related measures have a significant impact on the use of resources in various Data Envelopment Analysis (DEA) models. As the use of survey data implies that the quality variables are measured with error, the assumptions underlying a DEA model are not strictly fulfilled. The report therefore establishes ways of correcting for biases that might result from the violation of selected assumptions. Firstly, any selection bias in the patient mix of each physician is controlled for by regressing the patient evaluation responses on the patient characteristics. The corrected quality evaluation variables are then entered as outputs in the DEA model, and model specification tests indicate that out of 25 different quality variables, only waiting time has a systematic impact on the efficiency results. Secondly, the effect on the efficiency estimates of the remaining sampling error in the patient sample for each physician is accounted for by constructing confidence intervals based on resampling. Finally, as an alternative approach to including the quality variables in the DEA model, a regression model finds different variables significant, but not always with a trade-of between quality and quantity.
C61, D24, I12
DEA, Health economics, Quality, Patient evaluation, Efficiency, Errors in variables, Resampling, Bootstrap, Selection bias, Sampling error.
Project:Oppdragsgiver: Norges forskningsråd
Frisch prosjekt: 4101 - Economics of health
Norwegian Research Council through HERO and the project Quality and efficiency analysis in the public sector