Centre variation in mortality following post-hospitalization acute kidney injury: analysis of a large national cohort.

BACKGROUND: Routine monitoring of outcomes for patients with acute kidney injury (AKI) is important to drive ongoing quality improvement in patient care. In this study we describe the development of a case mix-adjusted 30-day mortality indicator for patients with post-hospitalization AKI (PH-AKI) across England to facilitate identification of any unwarranted centre variation in outcomes. METHODS: We utilized a routinely collected national dataset of biochemically detected AKI cases linked with national hospitals administrative and mortality data. A total of 250 504 PH-AKI episodes were studied across 103 National Health Service hospital trusts between January 2017 and December 2018. Standardized mortality ratios (SMRs) were calculated for each trust using logistic regression, adjusting for age, sex, primary diagnosis, comorbidity score, AKI severity, month of AKI and admission method. RESULTS: The mean 30-day mortality rate was high, at 28.6%. SMRs for 23/103 trusts were classed as outliers, 12 above and 11 below the 95% confidence limits. Patients with PH-AKI had mortality rates >5 times higher than the overall hospitalized population in 90/136 diagnosis groups and >10 times higher in 60/136 groups. Presentation at trusts with a co-located specialist nephrology service was associated with a lower mortality risk, as was South Asian or Black ethnicity. Deprivation, however, was associated with higher mortality. CONCLUSIONS: This is the largest multicentre analysis of mortality for patients with biochemically ascertained PH-AKI to date, demonstrating once again the considerable risk associated with developing even mild elevations in serum creatinine. Mortality rates varied considerably across centres and those identified as outliers will now need to carefully interrogate local care pathways to understand and address the reasons for this, with national policy required to tackle the identified health disparities.