Background Adjustment for morbidity level is important in ensuring fair comparison of outcomes between patient groups and healthcare providers. The Quality and Outcomes Framework (QOF) in UK primary care, which records numerous diseases systematically, offers potential for developing a standardised morbidity score that can be easily applied in research and service settings.

Methods Using The Health Improvement Network (THIN), a large primary care database of 375 UK general practices in 2008–9, half the practices were randomly selected as a training set to derive a morbidity score based on chronic conditions recorded in QOF, and the other practices formed a validation set to assess predictive performance. A total of 653,780 patients aged 60 and over registered in 2008 were included, and mortality at one year was assessed.

Results Nine QOF conditions were identified as robust co-predictors (Hazard Ratio ≥1.2) of mortality independent of age and sex, and were assigned integer score weights based on the strength of their association with mortality. Cancer (HR=3.4) and Dementia (HR=2.8) were the strongest predictors. In a Cox model with age and sex included, the addition of the QOF score improved model discrimination in predicting mortality (c-statistic=0.82 vs. 0.78), performing similarly to the Charlson index, an established morbidity index. In a multilevel logistic model, an individual’s QOF score explained more of the variation in mortality between practices than the Charlson index (46% compared to 32%). At practice level, the mean QOF score per patient was strongly correlated with practice standardised mortality ratios (r=0.64) and explained more variation in practice death rates than the Charlson index.

Conclusion A simple score derived from routine QOF recording provides a morbidity index which is highly predictive of one year mortality in older UK Primary Care patients, is simpler to implement than existing morbidity scores, and explains practice level variations in mortality. This new score has potential utility in research and healthcare outcome monitoring and could be easily implemented nationally through existing mechanisms for anonymised collection of QOF data from practices.