Article: Hospital standardised mortality ratio: A reliable indicator of quality of care? (abstract) - March 2018 - NJM
Issue: 2018 > March > original article

Hospital standardised mortality ratio: A reliable indicator of quality of care?



ORIGINAL ARTICLE
J.A. van Erven, L.S. van Galen, A.A. Hettinga-Roest, E.P.J. Claessens, J.C. Roos, M.H.H. Kramer, P.W.B. Nanayakkara
AbstractFull textPDF

Abstract

Background: This study investigates (1) whether the hospital standardised mortality ratio (HSMR) model underestimates or overestimates disease severity and (2) the completeness of the data collected by administrators to calculate HSMR in a cohort of deceased patients with the diagnosis of pneumonia.
Methods: In this retrospective cohort study Pneumonia Severity Index (PSI) and Abbreviated Mortality in Emergency Department Sepsis (abbMEDS) scores and associated mortality probabilities were obtained from 32 deceased pneumonia patients over the year 2014 in the VU University Medical Centre. These were compared with mortality probabilities of the Central Bureau for Statistics (CBS) calculated for every patient using the HSMR model. Clinical charts were examined to extract relevant comorbidities to determine the reliability of data sent to the national registration of hospital care.
Results: Risk categories determined by using the PSI and the abbMEDS were significantly higher compared with the risk categories according to HSMR (p = 0.001, respectively p = 0.000). The mean difference between the number of comorbidities in our registration and the coders’ registration was 1.97 (p = 0.00). The mean difference was 0.97 (p = 0.000) for the number of comorbidities of influence on the Charlson Comorbidity Index (CCI) and 1.25 (p = 0.001) for the calculated CCI.
Conclusion: The results of this study suggest that the mortality probabilities as calculated by the CBS are an underestimation of the risk of dying for each patient. Our study also showed that the registration of data sent to the CBS underestimated the actual comorbidities of the patient, and could possibly influence the HSMR.