Hospital standardised mortality ratio needs improvement
Prior to implementing the hospital standardised mortality ratio (HSMR) for hospital comparisons and improvement of quality of care, some adaptations need to be made. HSMR is a promising instrument, however, the current measurement is too distorted by factors other than quality of care. This conclusion was drawn by Wim van den Bosch in his thesis defended on November 2 2011 at the VUmc in Amsterdam, based on a study at Santeon and supervised at NIVEL.
In 2010 ‘crude’ mortality data for 2009 were published in the Netherlands. This publication received a lot of attention from politics, inspectorate, media and patients. However, it is not fair to compare hospitals based on these figures, since differences in diseases are not taken into account. For instance a hospital with exclusively patients suffering from cancer will yield higher mortality rates compared to a regular hospital. Other patient related factors like age and sex may also result into different hospital mortality rates. Clearly crude mortality rates are not well suited to measure differences in quality of care at the hospital level. An adjustment for disease, age, etc. is necessary in order to make mortality rates better comparable.
Standardised measure
Such an adjusted measure has been developed in England: the ‘hospital standardised mortality ratio’ (HSMR). The benchmark is the national average of all the hospitals and is set at 100. The adjusted measure has been applied to admissions in 2010 for most of the Dutch hospitals. The outcomes will be published before the end of the year 2011. In addition to hospital comparisons this instrument can also be used to compare patient populations within identical diagnostic groups.
Measurement errors
Some years ago six large non-university teaching hospitals under the name of ‘Santeon’ started using the HSMR as a benchmarking tool with the intention to improve their quality of care. Gradually, however, the doubt about the validity was growing. Measurement errors seemed to have more influence on the HSMR scores than quality of care did. They therefore investigated the nature of the measurement errors and how they potentially could affect the scores. Significant differences in HSMR scores were found to be caused by variations in: admission registrations, re-admission frequencies, the severity of the casemix and special procedures that are restricted to a limited number of hospitals only, such as open heart surgery. For all these distorting factors, not related to quality of care, no adjustment is being made.
Erroneously high
The adjustment by the HSMR model that was used in 2010 is insufficient. There is a fair chance that various good and safe hospitals could score an unduly high HSMR. Only after improvement of the HSMR instrument, by applying the findings of the thesis, can the quality of the instrument be sufficiently guaranteed so as not prohibit disclosure.
Wim van den Bosch is Senior Advisor Quality and Accountability in the St. Antonius Hospital in Utrecht/Nieuwegein, the Netherlands.
The study was initiated, facilitated and financially supported by Santeon, a group of six cooperating hospitals in the Netherlands.