The Effect of Coding Practice on Hospital Standardised Mortality Ratios

Abstract

Administrative data is increasingly used for the production of official statistics. However, administrative data is affected by quality issues such as differences in coding practice: parties that contribute to an administrative source register the same case in different ways. The aim of this paper is to determine how we can correct administrative data for differences in coding practice afterwards, to assess whether and how the outcomes are affected by these differences and to estimate the size of this effect. We will apply this method to the calculation of the Hospital Standardised Mortality Ratio (HSMR) which is affected by coding differences in urgency of the admission and comorbidities. A method is presented to adjust the administrative data for those differences, which can subsequently be used to compute an adjusted HSMR. This makes it possible for SN to assess whether a yearly change in the HSMR of a certain hospital is due to a true change in quality of care or due to a change in coding practice. A comparison of the original and the adjusted HSMRs shows that the variation in HSMR is smaller after correction for differences in coding practices.