WITHDRAWN: Practice characteristics influencing general practice mortality: a national longitudinal study 2013 -2017

Problem

Previous studies reporting an association between primary care investment and practice-level mortality have relied on mortality estimates . We investigate the relationship between general practice characteristics and actual mortality rates in England.

Approach

We undertook a retrospective longitudinal study of practice-level data for the financial years 2013/14 - 2017/18. We combined seven aggregate practice level datasets for all English practices (n=7310):

i. General and Personal Medical Services database, providing workforce and patient data;

ii. NHS payments to General Practice, recording payments to practices;

iii. Quality and Outcomes Framework describing performance on clinical achievement indicators;

iv. deprivation data;

v. neighbourhood ethnicity;

vi. patient experience scores from the General Practice Patient Survey;

vii. practice-level mortality

We fitted Poisson models allowing for unobserved practice effects with robust standard errors to allow for overdispersion of errors. It is possible that deaths reported for a given year depends on funding and other practice characteristics in previous years, therefore we included lags of up to three years.

Findings

Mean total funding per patient at 2017 prices was £153.16 (standard deviation £43.74). Practice mortality (per 1000 patients) increased from 8.29 (SD 6.36) in 2013/14 to 8.34 (SD 6.80) in 2017/18. Premature mortality (deaths in those <75 years) decreased from 2.67 (SD 1.31) in 2013/14 to 2.65 (SD 1.28) in 2017/18.

Higher mortality in a given year was significantly associated with lower mortality in the following year (coefficient -0.003; 95%CI -0.003, -0.002). Practice characteristics significantly associated with lower mortality included higher patient satisfaction (coefficient -0.02; 95%CI -0.03, -0.01), higher QOF achievement (coefficient -0.002; 95%CI -0.003, -0.0002), higher practice list growth (coefficient -0.06; 95%CI -0.09, -0.03) and lower deprivation scores (coefficient 0.02; 95%CI 0.01, 0.02).

The prevalence of different diseases impacted on practice mortality differently. For example, higher prevalence of cancer (coefficient 0.04; 95%CI 0.02, 0.07) and dementia (coefficient 0.12; 95%CI 0.10, 0.17) were significantly associated with higher practice mortality. However, a higher prevalence of conditions linked to intermediate outcomes such as blood pressure control in hypertension (coefficient -0.02; 95%CI -0.04, -0.01) was negatively associated with practice mortality.

Higher practice funding was associated with lower mortality, although not significant (coefficient -0.002; 95%CI -0.005, 0.0007). Analysis of individual funding components found that higher expenditure on pension contributions (coefficient -0.002, 95%CI -0.002, -0.001) was significantly associated with lower mortality.

Consequences

Higher mortality in a given year was associated with subsequent, compensatory reduction in mortality in the following year suggesting that in-year random shocks to mortality may affect the most vulnerable. Higher pension contributions, an indicator of GP and practice staff pensionable income, was significantly associated with lower mortality. Lower mortality was associated with higher reported patient satisfaction and clinical achievement implying that these dimensions of healthcare quality translate into reduced mortality.