: Variation in healthcare utilization has been discussed extensively, with many studies showing that variation exists, but fewer studies investigating the underlying factors. In our study, we used a logistic multilevel-model at the patient, hospital, and regional levels to investigate (i) the levels to which variation could be attributed and (ii) the hospital and regional factors associated with treatment decisions. To do so, we used hospital discharge records for the years 2012-2016 in Germany and Italy and for 2014-2016 in the Netherlands combined with hospital and regional characteristics in nine case studies. We used a theoretical framework to categorize these case studies into effective, preference-sensitive, and supply-sensitive care. Our results suggest that most variation in the treatment decision can be attributed to the hospital level (e.g., case volume), whereas only a minor part is explained by regional characteristics. Italy had the highest share attributable to the regional level, whereas the Netherlands had the lowest. We observed less variation for procedures in the effective-care category compared to the preference- and supply-sensitive categories. Although our results were heterogeneous, we identified patterns in line with the theoretical framework for treatment categories, underlining the need to address variation differently depending on the category in question.