Skip to main content

Table 3 Results from a binominal logistic regression (full model) with a stepwise forward selection method (likelihood ratio). Suicide cases, n = 154; Controls, n = 154

From: Are quantity and content of psychiatric interventions associated with suicide? A case-control study of a Swedish sample

Full model* B (SE) Wald Sig. OR (CI) 95% CI for OR
     Lower Upper
Number of outpatient visits −0.012 (0.004) 7.908 .005 0.988 0.980 0.996
Serious suicide attempt (yes)1 1.043 (0.364) 8.233 .004 2.839 1.392 5.790
Other structured psychotherapy −0.970 (0.509) 3.641 .056 0.379 0.140 1.027
  1. Note: The dependent outcome variable ‘Suicide’ was dichotomized (Yes = 1, No = 0)
  2. Abbreviations: B: beta coefficient, SE: standard error, OR: odds ratio / expected beta coefficient, CI: confidence interval for OR. For each step, the entry testing was based on the significance of the score statistic, and removal testing was based on the probability of a likelihood-ratio statistic. Degrees of freedom (df): 1 in all steps, except for ‘Intervention form’ (df = 3)
  3. 1 Somatic specialist medical inpatient treatment due to serious suicide attempt; for diagnostic ICD-10 codes, see Appendix
  4. *Full model: The six independent predictor variables/covariates entered in the stepwise forward analysis were: ‘Number of visits’ (continuous variable), ‘Intervention form’ (categorical variable: Combination therapy, Psychotherapy, Psychotropics (reference), Other), ‘CBT’ (incl. DBT), and ‘Other structured psychotherapy’, ‘Serious suicide attempt’, and ‘Somatic comorbidity’ (dummy coded [Yes = 1, No = 0] variables
  5. Full model fit: R2 = 0.099 (Nagelkerke). Model χ2 = 23.818; p = .001