There has been little study of the relationship between resource utilisation, clinical risks and hospital costs in acute medicine with the question remaining as to whether current funding models reflect patient acuity. We examined the relationship between resource use for investigations/allied professional and patient episode costs in all emergency medical admissions admitted to our institution during 2008-2013. Univariate estimates were compared with a multivariate model adjusted for major cost predictors. Interestingly, the model adjusted cost estimates changed considerably when compared with univariate analysis. We used both linear and non-linear (quantile regression) methods due to the highly skewed nature of hospital costs. The data suggested that hospital episode costs were predictable and driven by objective measures of clinical complexity. The use of expensive investigations and healthcare professional time was secondary to the clinical acuity. Thus, cost was heavily weighted towards higher complexity, and lower resource utilisation associated with lower risk patient groups. However, the non-linear nature of the costings would caution against simple predictor models and non-linear techniques such as quantile regression may, as we have demonstrated, prove superior in defining the underlying relationships.
Computer-aided National Early Warning Score to predict the risk of sepsis following emergency medical admission to hospital: a model development and external validation study
