Hospital Standardized Mortality Ratio

From Clinfowiki
Jump to: navigation, search

Factors Affecting the Hospital Standardized Mortality Ratio


Background

The ongoing investigation and governmental inquiries into the Mid-Staffordshire Trust scandal1 have brought the use of Hospital Standardized Mortality Ratios (HSMR) to the attention of the public. This paper will attempt to define the HSMR and to identify some of the factors that affect its scoring.

Definition of the HSMR

Simply put, the HSMR is; (Number of deaths)/ (Number of expected deaths) X100

For example; if you expected 100 death to occur, and 100 deaths did in fact occur, then: (100)/ (100) X100=100 .This is considered a neutral score or the average death rate.

While simple to calculate, the result can be misleading. A score below 100 may be considered a sign of improved quality of care (hence fewer deaths) while a score above 100 might be considered a sign of poor quality care (more deaths than one would expect.) We will attempt to show why this is not necessarily true.


Factors affecting the HSMR

Number of Expected Deaths

It seems simple, but how do you calculate the number of expected deaths in your hospital over, say, a one year period? Historical data might be of some help, but changes in equipment, case loads, and in the complexity of procedures performed, will render this method meaningless. Also, how do you measure the comparison with other hospitals in your area? There are different methods of calculating the average death rates, but they vary and can produce different HSMRs as a result2.


Data Errors

Errors in data input, especially in the patient demographics, can lead to an improper patient case-mix resulting in a comparison of HSMRs as useful as a comparison of apples to oranges. For example, if one hospital mixed their emergency surgical patients into the same case-mix as their ambulatory patients, they could see a greater amount of deaths and a higher HSMR as a result, compared to a hospital that kept the two categories separate.


Coding Practices

HSMRs are affected by changes in the coding of the primary diagnosis, the rate of admission and readmissions, the number of co-morbidities recorded and the use of special codes, such as the code for palliative care, resulting in manipulation, or “gaming” of the system. Medway Trust of England, was advised it was underusing the palliative code Z51.5. By increasing the use of this code, it was able to dramatically lower its HSMR from a score above 105 to a score below 354. Changes in coding practices may account for the fact that overall in England, the HSMR rate fell faster than the crude death rate which remained relatively unchanged between 2001 and 200811.


Case Mix

Case mix factors include; age, sex, admission type, comorbidity, social deprivation, diagnostic group, source of referral and may include other factors as well3. It is important to make proper comparisons, little can be learned by comparing adult cardiac patients in an Intensive Care Unit (ICU) with children admitted for observation of flu like symptoms.


Admitting/Discharge/Transfer Policies

Another method of manipulating the system is transferring patients to other facilities, or home if that is the patients wish, prior to death. Also, readmitting the same patient over and over will greatly improve the death to admission ratio3.


Conclusion

The HSMR is not a reliable measure, taken alone, to rate quality of hospital care or to rate hospital against hospital, as a range of factors can influence the HSMR results. The public needs a better tool or a selection of tools in order to make informed choices about quality of care in deciding which hospital they chose for treatment.


References

1. Hospital standardized mortality ratios: their uses and abuses; briefing: Nov 2010 Issue 208. Wwwnhsconfed.org/publications.

2. Mohammed M.A., Lilford R., Rudge G., Holder R., Stevens A. The findings of the Mid-Staffordshire Inquiry do not uphold the use of hospital standardized mortality ratios as a screening test for “bad” hospitals. http://qjmed.oxfordjournals.org/content/early/2013/05/07/qjmed.hct101.full. Downloaded on 07-16-2013

3. Bosch W., Spreeuwenberg P., Wagner C. Variations in hospital standardized mortality ratios (HSMR) as a result of frequent readmissions. http://www.biomedcentral.com/1472-6963/12/91. Downloaded on 07-17-2013

4. Taylor P. Rigging the Death Rate. London Review of Books, vol. 35 no.7 April 11, 2013.

Submitted by Ed Evans