Baseline characteristics of the study population
General demographics | Development sample N=37 523 | Validation sample N=15 649 | ||
N | % | N | % | |
Age in years, median (IQR)* | 32.2 | 21.4–48.9 | 32.5 | 21.4–49.1 |
Individuals under 16 years | 2224 | 5.9 | 864 | 5.5 |
Sex, female* | 20 561 | 54.8 | 8685 | 55.5 |
Substance misuse | ||||
Current or lifetime alcohol use disorder, excluding alcohol intoxication* | 7257 | 19.3 | 3025 | 19.3 |
Current or lifetime drug use disorder (including drug intoxication)* | 8384 | 22.3 | 3672 | 23.5 |
Alcohol intoxication at index | 1416 | 3.8 | 815 | 5.2 |
Living status† | ||||
Living with other adults | 14 151 | 37.7 | 5697 | 36.4 |
Living with children | 15 493 | 41.3 | 6494 | 41.5 |
Treatment in the past 3 months | ||||
Any psychotropic medication | 20 888 | 55.7 | 9537 | 60.9 |
Antidepressant treatment | 12 527 | 33.4 | 6009 | 38.4 |
Antipsychotic treatment | 4259 | 11.4 | 1834 | 11.7 |
Mood-stabiliser treatment | 677 | 1.8 | 310 | 2.0 |
Physical health problems | ||||
New cancer diagnosis | 411 | 1.1 | 149 | 1.0 |
History of self-harm | ||||
Method of index self-harm event‡ | ||||
Any psychotropic medication overdose | 1360 | 3.6 | 539 | 3.4 |
Cutting | 4925 | 13.2 | 1614 | 10.3 |
Hanging, strangulation or suffocation | 308 | 0.8 | 150 | 1.0 |
Drowning | 51 | 0.1 | 23 | 0.2 |
Lifetime history of self-harm prior to index* | 11 277 | 30.1 | 5372 | 34.3 |
History of self-harm in the past 12 months prior to index* | 4281 | 11.4 | 1994 | 12.7 |
Number of lifetime prior self-harm episodes, including the index self-harm event | ||||
1–2 episodes | 31 740 | 84.6 | 12 879 | 82.3 |
3+ episodes | 5783 | 15.4 | 2770 | 17.7 |
Overnight admission | 16 991 | 45.3 | 7531 | 48.1 |
Time between episodes ≤1 month | 1929 | 5.1 | 838 | 5.4 |
Mental health in the past 12 months | ||||
Any psychiatric disorder (except substance use disorders) | 16 472 | 43.9 | 7281 | 46.5 |
Serious psychiatric disorder | 3006 | 8.0 | 1305 | 8.3 |
Criminal/violence/legal issues | ||||
Lifetime criminal record for any crime | 13 451 | 35.9 | 5974 | 38.2 |
Criminal record for any crime in past 12 months | 3816 | 10.2 | 1809 | 11.6 |
Lifetime arrest history for any crime | 15 068 | 40.2 | 6506 | 41.6 |
Arrest history for any crime in past 12 months | 6416 | 17.1 | 2818 | 18.0 |
Lifetime criminal record for violent crime | 5729 | 15.3 | 2576 | 16.5 |
Criminal record for violent crime in past 12 months | 897 | 2.4 | 416 | 2.7 |
Lifetime arrest history for violent crime | 5402 | 14.4 | 2266 | 14.5 |
Arrest history for violent crime in the past 12 months | 2141 | 5.7 | 872 | 5.6 |
Family history | ||||
Family history of suicide | 1118 | 3.0 | 530 | 3.4 |
Family history of any psychiatric disorder | 15 112 | 40.3 | 6503 | 41.6 |
Values are all numbers and percentages, except for age for which median years and IQR are reported. Criminal arrest refers to being charged for an offence.
*Core factor, kept in the final model independently of its statistical significance or predictive strength.
†Living status information was missing for 641 (2%) individuals in the development sample and 199 (1%) in the validation sample. Individuals under 16 years old with missing living status were reclassified as ‘living with other adults’ (2049 (92%) of the under 16s in the development sample; 815 (94%) in the validation sample). Of those with missing living status, two individuals died from suicide more than 12 months since index in the development sample, zero in the validation sample. For variables psychiatric treatment, mental health, family history, new cancer diagnosis and self-harm method, when there was no information, it was assumed to indicate the variable did not occur rather than a missing value.
‡It was possible for an individual’s index self-harm event to be coded under more than one of the methods listed. The majority of individuals not included in any of the listed categories had either non-psychotropic medicine overdose or an unspecified method of index self-harm. Drowning was too rare to enter into the multivariable model.