Comparisons of provider performance using patient reported outcome measures should account for missing data
Failure to account for missing data undermines judgments about which providers are best at improving the health of their patients. The latest CHE research paper demonstrates why missing data matter and what can be done about it.
The English Patient Reported Outcome Measures (PROMs) programme asks patients about their health status before they have surgery and again several months after. But not all patients complete the questionnaires.
Missing data creates two problems. First, reduced statistical power means that it is more difficult to identify those providers that perform well or poorly. Second, patients that complete their PROM questionnaires may be different from those that do not. If these differences are related to the effectiveness of treatment, performance assessments that use only completed questionnaires may not reflect the true quality of hospital care.
We address these problems using a statistical approach known as multiple imputation. This involves filling in the missing questionnaire with the most likely responses, which are predicted – or imputed – on the basis of the observed characteristics of the patient and provider. Completed and imputed responses are then analysed together to compare providers.
We find that analysis based solely on completed questionnaires understates the variation in health outcomes experienced by patients treated in English hospitals. By applying our imputation method we can draw more accurate conclusions about the relative performance of providers in improving health outcomes.
Full details of the report are available here: CHE Research Paper 101 (PDF 
, 1,619kb)