OBJECTIVES Hopital's attractiveness can impact the number of inpatients, and so health care expenditures. The objective is to identify parameters related to patients or drugs, wich influence the proportion of inpatients not residing in the hospital's region, in order to better understand expensive drug expenditures. METHODS A database on expensive inpatient drugs listed on a national formulary was set up, with information related to drugs -from National and European Health authorities- and patients -from our university hospital centre’s database (UHC) of the Paris region. A censored regression model, Tobit model, was developed in which the dependent variable was, for each drug, the ratio of the number of inpatients not residing in the Paris region, on the total number of inpatients receiving this drug. Explanatory variables were related to drugs and to patients. All statistical tests were run in Stata/IC13®. RESULTS During 2012, 526 091 doses -of the 113 drugs list- are administered to 30 499 UHC inpatients. 13% of inpatients come from off the Paris region. According to the Tobit model, three variables positively impact the ratio (innovative status, percentage of women and Diagnosis Related Group coded as severe) and three negatively (orphan designation, percentage of patients over 65 years old, biological origin). Therapeutic classes and Major Diagnostic Categories also impact significantly the ratio according to the statistic class. For instance, innovative status, defined by a market authorization granted in the five last years and a high level of improvement in actual benefit, improves the ratio of 8.6 points. To benefit from the most innovative treatments, patients tend to be cared in UHC's hospitals. CONCLUSIONS This study represents the first step of modelling significant determinants of hospital's attractiveness for patients. It focuses, quite uniquely, on patients and drugs factors, and could be extended to other variables.