Objective To identify the main drivers of inpatient stroke care resource use, estimate the influence of stroke teams on the length of stay (LoS) of its patients and analyse the variation in relative performance across teams.

Design For each of four types of stroke care teams, a two-level count data model describing the variation in LoS and identifying the team influence on LoS purged of patient and treatment characteristics was estimated. Each team effect was interpreted as a measure of stroke care relative performance and its variation was analysed.

Setting This study used data from 145 396 admissions in 256 inpatient stroke care teams between June 2013 and July 2015 included in the national stroke register of England, Wales and Northern Ireland—Sentinel Stroke National Audit Programme.

Results The main driver of LoS, and thus resource use, was the need for stroke therapy even after stroke severity was taken into account. Conditional on needing the therapy in question, an increase in the average amount of therapy received per inpatient day was associated with shorter LoS. Important variations in stroke care performance were found within each team category.

Conclusions Resource use was strongly associated with stroke severity, the need for therapy and the amount of therapy received. The variations in stroke care performance were not explained by measurable patient or team characteristics. Further operational and financial analyses are needed to unmask the causes of this unexplained variation.