Study design
We interviewed every third person committed to a young offender institution over a twelve month period. Those interviewed were seen within seven days of reception. We used a semi-structured interview for background and demographic details, a semi-structured interview for substance misuse [
26,
27], and the CAARMS structured interview for ultra-high risk states of psychosis [
21,
22]. The study was approved as a clinical and service audit project by the audit, effectiveness and research ethics committee for the National Forensic Mental Health Service. All participants gave informed consent for participation.
Setting
St Patrick's Institution is a young offender institution in Ireland with a bed capacity of 217. At the time of this study it was the only prison accepting males aged 16 to 20 in the state (population 4.6 m). All committals were screened by a prison nurse within 6 hours of reception at the prison and all were seen by a general practitioner within 24 hours who carried out an unstructured general health assessment as for any new patient presenting to primary care. One in three from a list of chronological receptions was selected for a more detailed assessment by the visiting psychiatrists (DF and DS, post-membership psychiatric registrars, equivalent to US fellows or UK ST4). If the person selected was not eligible because it was their second or subsequent committal, the next on the list was selected.
Participants
Of the 836 committals in the study period, 78 were not eligible because they were second or subsequent committals leaving 758 eligible new committals. Every third committal eligible (n = 278) was selected for interview and 480 were not selected. Of the 278 selected, 107 were either released before they could be interviewed, declined the interview, were absent from the prison at court, released from custody or transferred to another prison and the remaining 171 were interviewed.
Instruments
The Comprehensive Assessment of At-Risk Mental States (CAARMS) is a semi structured research diagnostic interview schedule which was developed to reliably detect the prodrome of first episode psychosis prospectively [
21,
22]. It has seven subscales (positive symptoms, cognitive change, emotional disturbance, negative symptoms, behavioural change, motor/physical change, and general psychopathology), each of which are scored from 0 (absent) to 6 (psychotic and severe / extreme). Each subscale scores for threshold, frequency and duration of symptoms. It takes account of the relationship between symptoms and substance use and also measures subjective level of distress caused by symptoms (0–100 scale).
To meet the symptom component of the criteria for an UHR state, only symptoms in the positive symptom subscale require assessment. This subscale is comprised of four domains: Unusual Thought Content (e.g. delusional mood and perplexity, ideas of reference, bizarre ideas), Non-Bizarre Ideas (e.g. suspiciousness, grandiose ideas, somatic ideas, nihilistic ideas, religious ideas), Perceptual Abnormalities (e.g. distortions, illusions, hallucinations), and Disorganised Speech.
The CAARMS defines three groups which outline criteria required to be diagnosed with the ultra-high risk state - 1) A “vulnerability” state in which there is a family history of psychosis in first degree relative or schizotypal personality disorder in the identified patient is present 2) “attenuated psychosis” , a pattern of psychotic symptoms which are sub-threshold in intensity or frequency for a diagnosis of psychotic disorder 3) Brief limited intermittent psychotic symptoms (BLIPS) describes a recent history of frank psychotic symptoms that resolved spontaneously without anti-psychotic medication within one week 4) a fourth category of psychotic disorder is also included in the CAARMS. This is defined as symptoms rated above threshold as described in the CAARMS and present for at least one week.
To fulfil the criteria for each of the UHR groups, quantitative evidence of functional impairment is also required as measured by a 30% drop in score on the Social and Occupational Functioning Scale (SOFAS) [
28]. The SOFAS is a tethered rating scale ranging from 0 'grossly impaired' to 100 'superior functioning' and differs from the Global Assessment of Function scale (GAF) [
29] in that it does not include ratings for severity of symptoms.
Inter-rater reliability was tested by jointly interviewing and independently scoring 13 individuals, Spearman rank correlation coefficient r = 0.962, p < 0.001.
Alcohol and substance use disorders were diagnosed using the short form of the Developmental Understanding of Drug Misuse and Dependence (DUNDRUM-DS) [
26,
27]. This instrument was developed by the authors to elicit DSM-IV diagnostic criteria for substance use disorders using a brief yet comprehensive interview. We chose not to use the alcohol and substance use modules of the Structured Clinical Interview for DSM-IV [
30] for a number of reasons including the longer duration of interview required and its omission of newer substances like Mephedrone. The DUNDRUM-DS assesses the severity of use for a range of intoxicants on four levels: never used, ever used, abuse and dependence. Its validity is currently being tested in a forensic inpatient population. Joint interviewing (DS & DF) and separate scoring of ten individuals revealed good to excellent inter-rater reliability (cannabis Cohen's Kappa = 0.855 p < 0.001, mephedrone k = 0.863 p < 0.001, alcohol, ecstasy, amphetamine, cocaine, benzodiazepines and heroin all k = 1, p < 0.001).
Study size
There is no previous research concerning the prevalence of ultra-high risk states in prisons. It is known that 3% of committals to adult prisons in Ireland have a current psychosis [
31], and community help-seeking samples have shown that between 20% and 40% of ultra-high risk cases identified in community help seeking samples go on to develop psychosis [
13‐
15,
19,
21,
23,
32,
33] We inferred that with a 20% conversion rate, 15% of committals to a young offenders group might be expected to fulfil the ultra-high risk criteria. Using the Confidence Interval Analysis statistics programme [
34], we calculated that a sample of 100 would detect a 15% prevalence with a 95% confidence interval of 9.3% to 23.3%.
Statistical methods
All data were stored anonymously and entered in SPSS-18 [
36]. Where needed, added statistics were calculated using Confidence Interval Analysis [
34] to derive relative risks and their confidence intervals. Age (mean 18.2, SD 1.3, range 16 to 20), SOFAS score (mean 50.4, SD 12.9, range 20 to 86) and number of substance misuse problems (mean 2.3, SD 1.8, range 0 to 7) all had the characteristics of normally distributed variables, with no outlier observations, and mean, median and mode close in each case. However number of substance misuse problems was treated as an ordinal.
Chi-squared (χ²) and relative risk statistics were calculated for associations between ordinal or categorical variables. Analysis of variance was used for normally distributed variables and categorical variables such as UHR state. Binary logistic regression using likelihood ratio was also used to examine the relationship between UHR state and co-variates.