The development and validation of a marginalization index for inpatient psychiatry

Marginalization is a multidimensional social construct that influences the mental health status of individuals and their use of psychiatric services. However, its conceptualization and measurement are challenging due to inconsistencies in definitions, and the lack of standard data sources to measure this construct.

To create an index for screening marginalization based on an existing comprehensive assessment system used in inpatient psychiatry.

Items anticipated to be indicative of marginalization were identified from the Resident Assessment Instrument-Mental Health (RAI-MH) that is used in all inpatient mental health beds in Ontario, Canada. Principal Component Analysis (PCA) and cluster analysis of these items was performed on a sample of 81,232 patients admitted into psychiatric care in Ontario between 1 January 2011 and 31 December 2016 to identify dimensions being measured. Various weights and scoring methods were tested to assess convergent validity on multiple outcomes of marginalization. Receiver Operating Characteristic (ROC) curve analysis was utilized to determine optimal cut-offs for the index by modeling the likelihood of different marginalization outcomes, including homelessness.

Fifteen items were identified for the development of the Marginalization Index (MI). PCA and cluster analysis identified that the items measured five dimensions. ROC curve analysis among homeless individuals identified an Area Under the Curve of 0.76 and an optimal cut-off of five on the MI. Frequency analysis of the index by different characteristics identified homeless individuals, frequent mental health service users, persons with a history of violence and police intervention, and persons with addictions issues, as groups with the highest scores, confirming the convergent validity of the index.

The MI is a valid measure of marginalization and is strong predictor of risk of homelessness among psychiatric inpatients. MI provides a resource that can be used for social and health policy, decision-support and evaluation.