Thirty-day Readmission Rates and Associated Factors: A Multilevel Analysis of Practice Variations in French Public Psychiatry.

BACKGROUND: Inpatient psychiatric readmissions are often used as an indicator of the quality of care and their reduction is in line with international recommendations for mental health care. Research on variations in inpatient readmission rates among mental health care providers is therefore of key importance as these variations can impact equity, quality and efficiency of care when they do not result from differences in patients' needs.

AIMS OF THE STUDY: Our objectives were first to describe variations in inpatient readmission rates between public mental health care providers in France on a nationwide scale, and second, to identify their association with patient, health care providers and environment characteristics.

METHODS: We carried out a study for the year 2012 using data from ten administrative national databases. 30-day readmissions in inpatient care were identified in the French national psychiatric discharge database. Variations were described numerically and graphically between French psychiatric sectors and factors associated with these variations were identified by carrying out a multi-level logistic regression accounting for the hierarchical structure of the data.

RESULTS: Significant practice variations in 30-day inpatient readmission rates were observed with a coefficient of variation above 50%. While a majority of those variations was related to differences within sectors, individual patient characteristics explained a lower part of the variations resulting from differences between sectors than the characteristics of sectors and of their environment. In particular, an increase in the mortality rate and in the acute admission rate for somatic disorders in sectors' catchment area was associated with a decrease in the probability of 30-day readmission. Similarly, an increase in the number of psychiatric inpatient beds in private for-profit hospitals per 1,000 inhabitants in sectors' catchment area was associated with a decrease in this probability, which also varied with overall sectors' case-mix characteristics and with the level of urbanisation of the area.

DISCUSSION: The extent of the variations and the factors associated with it question the adequacy of care and suggest that some of them may be unwarranted. Our findings should however be interpreted in consideration of several limits inherent to data quality and availability as we relied on information from administrative databases. While we considered a wide range of factors potentially associated with variations in 30-day readmissions, our model indeed only explained a limited part of the variations resulting from differences between sectors.

IMPLICATIONS FOR HEALTH POLICIES: Our findings underscored that practice variations in psychiatry are a reality that merits the full attention of decision makers as they can impact the quality, equity and efficiency of care. A specific data system should be established to monitor practice variations in routine to promote transparency and accountability.

IMPLICATIONS FOR FURTHER RESEARCH: Few associations were found between variations in 30-day inpatient readmissions and the supply of care. The routine collection of detailed organizational characteristics of health care providers at a national level should be supported to facilitate additional research work, both in France and in other contexts.