Table 1.

Standardised ‘whole hospital’ mortality measures

Standardised mortality measures, such as the hospital standardised mortality ratio (HSMR) or the summary hospital mortality indicator (SHMI), are presented as a ratio of actual to expected mortality. An expected mortality rate is calculated for each hospital using data derived from discharge coding: using a statistical model to forecast the number of deaths that a hospital would be expected to have, based on the characteristics of the admitted patients. Because expected mortality rates are based on discharge coding, it is important for clinicians to support accurate coding (for example, by avoiding the use of symptom diagnoses such as ‘chest pain’). The SHMI and HSMR have a number of important differences:
SHMIHSMR
Data sourceDischarge codesDischarge codes
MethodologyStatistical model to forecast expected number of deathsStatistical model to forecast expected number of deaths
Mortality measureInpatient deaths or death within 30 days of dischargeInpatient deaths
Included patientsAll deathsExclusions for certain diagnoses
Standardised measures are presented using statistical process control (SPC) methodology with 100 as the reference value and upper and lower control limits calculated (analogous to confidence intervals). Control limits are usually set so that the probability of a value lying outside them by chance is less than 2 per 1,000. Rates are reported as abnormal if they are outside the control limits. Although HSMR and SHMI are often presented as a single monthly or quarterly figure, this is unhelpful without knowledge of the control limits and the pattern over time. Observing changes in mortality data over time is generally a much more useful guide to quality than a one-off measurement.
Common problems with hospital SMRs:

  • reliance on a single figure without reference to data over time

  • the use of mortality measures on their own as an indicator of quality

  • dismissing high readings as being due to coding defects

  • taking reassurance from low readings without further understanding the data

  • use for ranking hospitals

  • no limit to the length of follow up in England

  • do not account for transfers of patients.

  • Adapted and reproduced from RCP’s Acute care toolkit 11.15