Hospital financing of ischaemic stroke: determinants of funding and usefulness of DRG subcategories based on severity of illness.

Abstract

Several Western and Arab countries, as well as over 30 States in the US are using the "All-Patient Refined Diagnosis-Related Groups" (APR-DRGs) with four severity-of-illness (SOI) subcategories as a model for hospital funding. The aim of this study is to verify whether this is an adequate model for funding stroke hospital admissions, and to explore which risk factors and complications may influence the amount of funding. A bottom-up analysis of 2496 ischaemic stroke admissions in Belgium compares detailed in-hospital resource use (including length of stay, imaging, lab tests, visits and drugs) per SOI category and calculates total hospitalisation costs. A second analysis examines the relationship between the type and location of the index stroke, medical risk factors, patient characteristics, comorbidities and in-hospital complications on the one hand, and the funding level received by the hospital on the other hand. This dataset included 2513 hospitalisations reporting on 35,195 secondary diagnosis codes, all medically coded with the International Classification of Disease (ICD-9). Total costs per admission increased by SOI (€3710-€16,735), with severe patients costing proportionally more in bed days (86%), and milder patients costing more in medical imaging (24%). In all resource categories (bed days, medications, visits and imaging and laboratory tests), the absolute utilisation rate was higher among severe patients, but also showed more variability. SOI 1-2 was associated with vague, non-specific stroke-related ICD-9 codes as primary diagnosis (71-81% of hospitalisations). 24% hospitalisations had, in addition to the primary diagnosis, other stroke-related codes as secondary diagnoses. Presence of lung infections, intracranial bleeding, severe kidney disease, and do-not-resuscitate status were each associated with extreme SOI (p < 0.0001). APR-DRG with SOI subclassification is a useful funding model as it clusters stroke patients in homogenous groups in terms of resource use. The data on medical care utilisation can be used with unit costs from other countries with similar healthcare set-ups to 1) assess stroke-related hospital funding versus actual costs; 2) inform economic models on stroke prevention and treatment. The data on diagnosis codes can be used to 3) understand which factors influence hospital funding; 4) raise awareness about medical coding practices.