Variations in Dutch National Medical Registration hardly affect the hospital standardised mortality rate (HSMR)

To analyse the variation in the registration of hospital admissions across Dutch hospitals and determine how this variation affects the Hospital Standardised Mortality Rate (HSMR).

Retrospective, descriptive.

We used data from the National Medical Registration (LMR), covering the records of all hospital admissions in 2005 in Dutch hospitals, to analyse the variation between hospitals in 3 variables: the number of secondary diagnoses, the percentage of unplanned admissions, and the percentage of non-specified diagnoses (‘other diagnoses’). The impact of this variation on the HSMR was analysed by calculating the correlation between the HSMR and each of the variables. The correlation between the original HSMR and the HSMR without adjustment for these variables was also calculated.

The variation in the percentages of unplanned admissions and admissions with a non-specified diagnosis was low. The variation in these two variables had a small or no effect on the HSMR. There was a considerable variation in the mean number of secondary diagnoses per hospital. This variation had a limited but statistically significant effect on the HSMR. The HSMR calculated without adjustments for secondary diagnoses correlated strongly with the original HSMR.

This analysis does not support the view that the HSMR is strongly affected by variation in the registration of hospital admissions and is therefore not reliable. Therefore, there is no need for restraint with regard to publication of the Dutch HSMR.

Conflict of interest: D. Pieter and R.B. Kool are both employees of Kiwa Prismant, the organisation appointed to calculate the hospital standardised mortality rates of the Dutch hospitals. Financial support: none declared.