Prediction of general criminal recidivism estimated by extended Cox-proportional-hazards regression (n = 148)a.

a Two participants had not been assigned a subgroup membership and were therefore excluded from the analysis [4].b 95% CI = 95% Confidence Interval.c The subgroups were defined with reference to the term “triply troubled”, referring to individuals with a combination of mental health problems, substance use problems, and criminal behavior [4,40]. Membership of the subgroup “Working triply troubled” (i.e. the subgroup with low levels of employment problems and medium levels of alcohol-, psychiatric-, and legal problems) was used as a reference category. Each of the remaining subgroups was compared to this particular group (not shown in the table).d Subgroup with low degrees of various problems.e Subgroup with severe legal-, psychiatric-, alcohol-, drug- and family/social problems.f Subgroup with elevated legal-, medical-, psychiatric and drug problems.g Compared to h Compared to no residence in dry housing.Prediction of general criminal recidivism estimated by extended Cox-proportional-hazards regression (n = 148)a.