Hospital Standardized Mortality Ratios, an unfairly judgement of delivered quality of acute care

Author: Sanne J. den Hartog, M.D. PhD candidate
Patient in hospital

For quality purposes, each year Statistics Netherlands calculated the Standardized Mortality Ratios (SMRs). In the current SMR calculation regionalization of acute care is not taken into account. The SMR calculated on a regional level instead of hospital level might provide a more valid quality indicator.

In a study of our team (Department of Neurology and Public health Erasmus MC) we compared hospital SMRs, specifically for non-specialized and specialized hospitals, to regional SMRs (RSMRs) for acute cerebrovascular disease, acute myocardial infarction, and intracranial injury. 

Standardized Mortality Ratios

SMRs are ratios of observed and expected numbers of deaths per hospital. The observed in-hospital mortality is obtained from the administrative data from the hospitals. The expected in-hospital mortality is calculated with data about age, sex, socioeconomic status, severity of main diagnosis, urgency of admission, Charlson comorbidity index, place of residence before admission, month/year of admission. When the SMR is below 1, the in-hospital mortality number is lower than expected and above 1 the in-hospital mortality is worse than expected. The SMRs can be used to compare hospital performance.

Regionalization of acute care

For the comparison of hospital performance, adjustment for differences in patient characteristics (case-mix) between hospitals is needed. Within regions there is agreement between specialized and non-specialized hospitals about the treatment location of patients with a specific disease. Patient with a more severe disease are more often treated in specialized hospitals. These patients have a higher risk of in-hospital mortality. Un insufficient case-mix adjustment will lead to higher SMRs of specialized hospitals, with an unfairly judgment about the delivered quality of care. We hypothesized that the current case-mix adjustment does not capture the differences between non-specialized and specialized hospitals. An SMR calculated on a regional level is more reasonable and less influenced by differences in case-mix between hospitals.

Our study

We used data from the Dutch National Basic Registration of Hospital Care. This database contains hospital admissions from all hospitals in the Netherlands. The hospital SMRs were calculated using the current approach of Statistics Netherlands. The RSMRs were calculated with weighted SMRs per hospital.


We found that the range in hospital SMRs of specialized hospitals was higher than for non-specialized hospitals in all three acute diseases. The range in RSMRs was small. Additional adjustment with proxies of disease severity, by for example a specialized treatment, did not change the range in hospital SMRs.


We conclude that the SMRs of three acute, regionally organized diseases vary substantially between hospitals. Instead of showing differences in quality of care between hospitals, the between-hospital difference in SMRs partly represent differences in case-mix between specialized and non-specialized hospitals. An SMR measured on a regional level might provide a more valid quality indicator and is more in line with the current regionalization of acute care.

PhD student
Sanne den Hartog
Sanne den Hartog is a M.D. PhD candidate at the department of Neurology, Radiology and Nuclear Medicine, and Public Health at Erasmus MC. Her main focus is on quality of stroke care. You can contact her via email:

Compare @count study programme

  • @title

    • Duration: @duration
Compare study programmes