Pros and cons of standardised mortality ratios

Differences in case-mix can influence the comparison of standardised mortality ratios even with optimal risk adjustment: an analysis of data from paediatric intensive care

“…This restriction is not always recognised by the users of SMRs. When it is acknowledged there are those who argue that any bias that may arise when directly comparing two SMRs is likely to be small and would not adversely affect any inferences drawn 22 23Calculation and comparison of the Standardised Mortality Ratio (SMR)

The standardised mortality ratio (SMR) is defined as the ratio of the number of deaths observed for the hospital of interest to the number of deaths that would have been expected if the patients at the hospital have experienced the same death rates at that found in the reference population:

where π i is the probability that a patient in the i - th case-mix stratum in the hospital will die; π Ri is the probability that a patient in the i-th case-mix stratum in the reference population will die; p i is the proportion of patients in the i-th case-mix stratum in the hospital.

For the SMRs for two hospitals (A and B) to take the same value the following would need to be true:

However, even if the stratum-specific event probabilities were identical for both hospitals for all strata (ie, π Ai =π Bi =π ABi for all values of i), and different from the reference population for at least one stratum, the SMRs would only be sure to take the same value if their population structures were also the same:

…”

Incidence, Risk Factors, and Mortality Associated With Second Malignant Neoplasms Among Survivors of Adolescent and Young Adult Cancer

“… 16 , 17 , 18 , 19 , 20 Most of these studies have examined data from large cancer registries such as SEER and used an external comparison group (eg, the general population) for evaluating the excess risk of SMN. However, using external comparison for causal inference is associated with several known limitations, 21 including the inability to address potential incomparability between cancer survivors and the general population, such as insurance status and health care access. 22 Information on therapeutic agents is also not typically available in these cancer registries.…”

Quantifying the Potential Bias when Directly Comparing Standardised Mortality Ratios for In-Unit Neonatal Mortality

“…It has long been recognised that directly comparing the SMRs of two different healthcare providers may be inappropriate as the SMRs for two providers performing equally for each patient type will not necessarily take the same value if the providers' population structures are different [16]–[18]: “[the SMR] is only a ‘ single-pair ’ method , and if it is applied to a number of groups it may only be thanks to the mercy of Providence that it is not grossly misleading ” [19]. On the other hand, there are others who argue that any bias that arises when comparing two SMRs is likely to be small and would not adversely affect any inferences drawn [20], [21]. Although it is unclear how much bias would really occur in practice, measures such as the HSMR and SHMI are increasingly being used to compare hospitals that potentially have patient populations with very different sets of risk factors.…”