Bivariate Associations | Domain Models | Final Model | ||||
---|---|---|---|---|---|---|

Common OR | 95% CI | Adjusted OR | 95% CI | Adjusted OR | 95% CI | |

Model 1: Demographics and social environment | ||||||

Age above median (>44 y) | 0.88 | (0.64–1.19) | ||||

Female | 1.43 | (1.04–1.97)^{*} | 1.57 | (1.13–2.17)^{**} | 1.45 | (1.05–2.01)^{*} |

Nonwhite | 1.45 | (1.05–2.01)^{*} | 1.58 | (1.13–2.21)^{**} | 1.51 | (1.08–2.11)^{*} |

Cohabitation | 0.99 | (0.66–1.50) | ||||

Some college education | 0.54 | (0.31–0.95)^{*} | 0.57 | (0.33–0.96)^{*} | ||

Housing categories | ||||||

Independent living [comparison] | ||||||

Homeless | 0.87 | (0.52–1.45) | ||||

Living with family | 0.92 | (0.65–1.30) | ||||

Staffed residence | 0.81 | (0.35–1.84) | ||||

Income above median (>$700) | 0.90 | (0.65–1.25) | ||||

Social support | 1.12 | (0.78–1.60) | ||||

Social support for treatment | 1.22 | (0.89–1.67) | ||||

Model significance | LR = 19.41, df = 3^{****} | |||||

Model 2: Clinical variable model | ||||||

Psychotic disorder | 0.84 | (0.61–1.14) | ||||

Substance abuse | 1.30 | (0.87–1.95) | ||||

BPRS score above median (>30) | 1.28 | (0.94–1.75) | ||||

GAF score above median (>47) | 1.00 | (0.71–1.41) | ||||

Insight score above median (>18) | 1.21 | (0.89–1.65) | ||||

Self harm | 1.77 | (1.25–2.54)^{**} | 1.77 | (1.26–2.49)^{**} | 1.67 | (1.17–2.39)^{**} |

Model significance | LR = 11.29, df = 1^{***} | |||||

Model 3: Treatment and systems utilization | ||||||

Past hospitalizations above median (>3) | 1.05 | (0.77–1.43) | ||||

Time in treatment above median (>20 years) | 0.88 | (0.64–1.21) | ||||

Picked up by the police in past 6 mo (treatment) | 1.80 | (0.81–4.51) | ||||

Picked up by the police in past 6 mo (arrest) | 4.10 | (1.27–21.06)^{*} | 3.58 | (1.08–11.87)^{*} | 3.39 | (1.00–11.44)^{*} |

Outpatient visits above median (>2 monthly) | 1.52 | (1.06–2.19)^{*} | 1.45 | (1.03–2.06)^{*} | 1.39 | (0.97–1.98)^{†} |

Involuntary commitment | — | (0.41–2.28)^{‡} | ||||

Treatment satisfaction above median (>52) | 1.12 | (0.82–1.53) | ||||

Drug Attitudes Inventory above median (>8) | 0.94 | (0.69–1.29) | ||||

Medication compliance above median (>3) | 1.48 | (1.07–2.06)^{*} | 1.41 | (1.02–1.96)^{*} | ||

Appointment compliance above median (>3) | 1.34 | (0.95–1.91)^{†} | ||||

Model significance | LR = 116.57, df = 3^{***} | |||||

Model 4: Leverage Model | ||||||

Outpatient commitment | 1.29 | (0.81–2.09) | ||||

Criminal justice leverage | 0.83 | (0.58–1.19) | ||||

Representative payee | 1.17 | (0.70–2.02) | ||||

Housing | 1.31 | (0.93–1.85) | ||||

Pressure to take medications above median (>1) | 1.81 | (1.31–2.52)^{***} | 1.73 | (1.26–2.38)^{***} | 1.59 | (1.14–2.23)^{**} |

Pressure to keep appointments above median (>1) | 1.66 | (1.19–2.35)^{**} | ||||

Perceived coercion above median (>7.23) | 0.96 | (0.70–1.32) | ||||

Autonomy above median (>24) | 0.69 | (0.51–0.95)^{*} | 0.76 | (0.56–1.03)^{†} | 0.71 | (0.51–0.98)^{*} |

Model significance | LR = 17.55, df = 2^{****} | LR = 46.25, df = 7^{****} |

A common odds ratio with cluster-corrected confidence interval is given only if all five sites’ odds ratios were determined by Zelen’s test (

*p*< 0.05) to represent a sampling distribution from a common population. Results were produced by logistic regression. Where there is no common odds ratio, the confidence interval gives the range of odds ratios across sites.Statistical Significance:

†

*p*< 0.10 (trend);*

*p*< .05;**

*p*< .01;***

*p*< .001;****

*p*< .0001.‡ Significantly positive at one site, significantly negative at one site.