Appendix Table 2 – untransformed model coefficients
| Variable | Coefficient |
| constant | -6.1304 |
| Sex – male | 0.6507 |
| Age (per year) | -0.008548 |
| Previous violent crime | -0.2427 |
| Previous drug use | 0.0817 |
| Previous alcohol use | 0.2541 |
| Previous self-harm | 0.9359 |
| Educational level
Upper secondary Post-secondary |
0.2133 0.5217 |
| Parental drug or alcohol use | -0.3549 |
| Parental suicide | 0.5618 |
| Recent treatment – antipsychotic | 0.2533 |
| Recent treatment – antidepressant | 0.5624 |
| Inpatient diagnosis | 1.0829 |
| Length of first inpatient stay > 7 days | 0.2047 |
| Number of episodes > 7 | -0.2562 |
| Benefit receipt | -0.1920 |
| Parental psychiatric hospitalisation | 0.1816 |
| Comorbid depression | 0.2362 |
To calculate predicted probability for an individual:
1) Calculate X. We illustrate the equation for X for the first three variables, however all 17 need to be included using the coefficients from the table above.
X = -6.1304 + 0.6507*Male – 0.008548*Age – 0.2427*Previous violent crime + …
2) Calculate predicted probability = 1 / [1 + exp(-X)]
3) If predicted probability > 1%, individual is classified at high risk; otherwise individual is classified at low risk