TABLE 2.
Logistic regression models of effects of placement on a community treatment order (CTO) in Victoria, Australia, from 2000 to 2010a
| Variable | Model χ2b | B | SE | Exp(b) | 95% Cl |
|---|---|---|---|---|---|
| All deaths | 4,336.84 | ||||
| CTO cohort | −.09 | .06 | .91 | .79–.99 | |
| Treatment days per community care episode | .00 | .00 | 1.00 | 1.00–1.01 | |
| Interaction CTO × treatment days per community care episode | .00 | .00 | 1.00 | 1.00–1.01 | |
| Deaths excluding assault, self-harm, and undetermined intent | 4,983.54 | ||||
| CTO cohort | −.18 | .07 | .83 | .75–.93 | |
| Treatment days per community care episode | .00 | .00 | .99 | .99–1.01 | |
| Interaction CTO × treatment days per community care episode | .00 | .00 | .99 | .99–1.01 | |
| Deaths by external causes, assault, self-harm, and undetermined intent | 298.45 | ||||
| CTO cohort | .23 | .11 | 1.26 | 1.01–1.58 | |
| Treatment days per community care episode | .01 | .01 | .99 | .98–1.01 | |
| Interaction CTO × treatment days per community care episode | .01 | .01 | 1.01 | .99–1.02 | |
| Deaths by self-harm | 298.45 | ||||
| CTO cohort | .28 | .12 | 1.32 | 1.04–1.69 | |
| Treatment days per community care episode | .01 | .01 | .99 | .98–1.01 | |
| Interaction CTO × treatment days per community care episode | .01 | .01 | 1.00 | .99–1.02 | |
| Deaths by external causes, assault, and undetermined intent | 220.21 | ||||
| CTO cohort | −.04 | .13 | .96 | .74–1.24 | |
| Treatment days per community care episode | .00 | .00 | 1.00 | 1.00–1.01 | |
| Interaction CTO × treatment days per community care episode | .00 | .00 | 1.00 | 1.00–1.01 |
All models in addition to the primary explanatory variables regressed the following explanatory variables on the model criterion variable: propensity of a patient to be selected into the CTO sample from among hospitalized patients, age, gender, socioeconomic status or vocational challenge (that is, >11th grade education, unemployment, and the lowest Socio-Economic Indexes for Areas rank of a neighborhood in which the patient lived), risk period associated with study and institutional involvement (that is, age at first date known to the mental health system, total time known to the mental health system, and lifetime total inpatient days), diagnoses (major affective disorder, dementia, schizophrenia, and paranoia or other psychosis), imprisonment or police custody during the study, aboriginal or Torres Strait Islander status, language interpreter required at the mental health tribunal, and psychosocial profile reflected in the 12 Health of the Nation Outcome Scale scores at both inpatient admission and release [see online supplement].
N for regression=27,585; df=43 and 27,542. All models were significant (p<.001).