A scoping review of the statistical methods and risk-adjustment approaches used to compare cardiovascular disease services using Australian health system data

Abstract

Routinely collected administrative hospital data is employed in comparative analyses of patient outcomes across different hospitals and time periods to measure performance or evaluate interventions. Due to the nonrandomised nature of these data, analyses depend on robust statistical risk-adjustment techniques to control for differences in the case mix of patient cohorts. This scoping review describes the alternative statistical methods used to balance groups of cardiology patients in comparative analyses of hospital outcomes dependent on Australian routinely collected data. Databases, including PubMed, Medline, Embase, Cinahl and Econlit were searched for papers published in English from the period 1st January 2012 to 31st December 2021, which relied on routinely collected Australian hospital data and used statistical methods to adjust for difference in groups of hospital cardiology patients across time or between institutions. 22 papers met the eligibility criteria for inclusion, 6 papers reported evaluations of service changes and 16 papers reported on comparisons of outcome between different groups of patients. Most studies used conventional multivariable methods to balance confounders and the choice of covariates was highly heterogenous. Modern statistical approaches for observational data are not yet commonly applied to analyses using routinely collected Australian hospital data. More studies comparing the sensitivity of analyses results to methods would be helpful to assess the value of these newer methods.

Keywords

• Risk-adjustment
• Hospital data
• Statistical methods
• Routine data
• Cardiology