Hospital mortality rates and place of death

Background: Hospital standardized mortality rates (HSMRs) are considered to identify unsafe hospital care, but are criticized for disregarding unmeasured variations of both coding practice and outside of hospital palliative provision. Methods: The validity of HSMR methods is explored by examining whether the additional deaths implied by a Primary Care Trust (PCT) HSMR, as supplied by Dr Foster, are (i) detectable in English PCT mortality data and (ii) correlated with the percentage of PCT deaths out of hospital. Results: Higher PCT HSMRs were positively associated with higher area mortality, although this effect was significantly smaller than the HSMR model implies. There was a significant negative association between PCT HSMR and the percentage of deaths outside of hospital. Conclusions: This evidence suggests that the PCT HSMR is associated with all-cause area mortality, but it overstates the in-hospital mortality attributed to unsafe care. It also suggests that this overstatement is partially explained by the PCT HSMR reflecting the adverse case mix associated with a high share of deaths in hospital, i.e. the HSMR does not only reflect quality failings. Policy analyses that compare HSMRs to measure quality change may be biased by unbalanced samples of hospital patients.