Implications of incorporating morbidity into primary care workload models for NHS funding allocations: a retrospective observational study in England
Lyvia de Dumast, Patrick Moore, Kym I E Snell, Tom MarshallBackground
Weighted capitation formulas are used in many countries to allocate primary care funding. Most rely heavily on demographic and area-level factors to weight payments, with limited direct adjustment for morbidity. This may disadvantage practices serving populations with greater morbidity or earlier onset of disease.
Objectives
To assess how well the current National Health Service (NHS) weighting formula reflects morbidity-related workload and how incorporating morbidity would alter national funding allocations.
Design
Retrospective observational study using patient-level electronic health records and national practice-level administrative data. Analyses comprised: (1) practice-level modelling of the NHS practice index (the ratio of formula-adjusted to registered patients); (2) patient-level fixed-effects regression of consultation workload and (3) national simulations applying coefficients from the demographic-only and morbidity-inclusive models to all practices to generate alternative practice weights for comparison.
Setting
Primary care in England (4440 general practices) for practice-level analyses and 627 UK general practices contributing over four million patients to a national electronic health record database for patient-level modelling.
Primary outcome
Annual primary care consultation workload (minutes per patient-year), estimated at patient level and applied in national simulations.
Secondary outcomes
Practice-level predicted workload, morbidity-based practice indices and proportional redistribution under alternative weighting models.
Results
In practice-level analyses of the NHS practice index, the multivariable model explained 77% of variation (R²=0.77). Deprivation, age structure and region accounted for most of this, whereas recorded morbidity contributed relatively little, indicating that the current weighting formula reflects demographic and area characteristics more strongly than morbidity burden. Mean consultation workload was 64.5 min per patient-year. In patient-level models, adding morbidity indicators to the demographic-only specification increased the proportion of variation in workload within practices explained from 16% to 26%. Morbidity was a strong independent predictor of workload and substantially reduced age and sex differences in predicted workload, although socioeconomic differences remained after adjustment. When coefficients from the morbidity-inclusive model were applied nationally to generate alternative practice weights and compared with the demographic-only specification, overall redistribution was modest: practice weights changed by about 2.5% on average. Across practices, 86.7% experienced changes within±5%, while 6.7% gained at least 5% and 6.5% lost at least 5%. Practices serving populations with higher morbidity tended to gain, whereas those serving older populations tended to lose. Practices in more deprived quintiles were significantly less likely to gain and more likely to lose.
Conclusions
The current demographic-based weighting formula captures age and regional variation but only weakly reflects recorded morbidity. Incorporating morbidity improves prediction of workload and produces modest redistribution towards populations with higher disease burden, although deprivation-related differences remain.