Adult Outcomes of Child- and Adolescent-Onset Schizophrenia: Diagnostic Stability and Predictive Validity

A retrospective psychosis screen was applied to the records of all patients under 18 years of age who had attended the Maudsley Hospital in South London between 1973 and 1991. At the Maudsley Hospital Children’s Department, data from clinical assessments are combined through standard coding of specified symptoms, and the symptom codes are recorded on standardized clinical data summaries (“item sheets”) and stored on a computer (21). Item sheets were completed on more than 95% of all patients during the study period, during which the format of the item sheets remained constant. Patients who screened positive for psychosis were identified from the item sheets if at least one of the following criteria were met: 1) an ICD-9 diagnosis of schizophrenia (295.0–295.9), affective psychosis (296.0–296.9), paranoid state (297.0–297.9), or other nonorganic psychosis (298.0–298.9) (equivalent ICD-8 codes were used from 1973 to 1977), or 2) one or more of the symptoms of hallucinations, delusions, or ideas of reference was definitely present. Patients attending the Maudsley Hospital Adult Department were also included in the screened-positive group if they were under age 18 at the baseline assessment and had an ICD-9 psychotic diagnosis (ICD-8 codes were used from 1973 to 1977). A total of 196 patients screened positive for psychosis.

At the second stage of the study, a detailed chart review was conducted for the 196 screened-positive patients. The selection criterion was unequivocal evidence of one or more of the following Research Diagnostic Criteria (RDC) psychotic symptoms (22) occurring in clear consciousness and lasting for at least 2 weeks: prominent hallucinations, delusions, or incoherence/loosening of associations. The records of 23 of the initial 196 patients who screened positive had missing case notes or not enough clinical details to confidently determine the presence or absence of psychotic symptoms, 58 patients were confirmed as nonpsychotic after their case records were examined, and five had a diagnosis of autism in the absence of an RDC psychotic symptom. The remaining 110 patients constituted the baseline adolescent-onset psychosis sample and were eligible for follow-up.

Adequate follow-up data were obtained on 93 (84.5%) of the 110 patients over a 2-year period. These 93 patients will be referred to as the followed-up sample. Subjects were located by using a variety of methods, including searches of the U.K. National Health Service Central Register and records of local U.K. Family Health Service Authorities. Information recorded in the patients’ medical records at the first admission was also used to contact subjects’ parents, family doctors, and consultant psychiatrists. After subjects had been traced, they received a letter introducing the study. Finally, subjects were visited by the author to explain the purposes of the study and to seek their consent. After complete description of the study to subjects, written informed consent was obtained.

Seventeen subjects (15.4%) were excluded from the final followed-up sample because of inadequate data. Seven subjects refused to be interviewed, and sufficient clinical information for these subjects was not available from other sources. Four subjects could not be traced, and nine had died. However, for three of the deceased subjects, there were enough detailed clinical data to include them in the followed-up sample. Seventy-four (80%) of the followed-up sample were assessed in face-to-face interviews. For a further 19 subjects, high-quality information was available from relatives, professionals, and case records. There were no significant differences between the followed-up sample (N=93) and those not followed-up (N=17) in age at onset, gender, ethnicity, place of residence, and DSM-III-R baseline diagnoses.

Baseline information was extracted from patients’ medical records by using a structured coding procedure specifically designed for the study. The quality of case note information recorded by Maudsley Hospital psychiatry trainees was uniformly high and followed the guidelines on obtaining and recording clinical information produced by the Maudsley Hospital and Institute of Psychiatry. To minimize potential bias and to avoid inferential impressions, items were rated only if the case notes contained explicit positive statements concerning the patient’s status. All baseline ratings were made blind to adult outcome status. The mean age of subjects at baseline assessment was 14.9 years (SD=1.5, range=11–17). The mean duration from onset of psychotic symptoms to baseline assessment was 5.3 months (SD=7.1, range=0.2–36).

Data on psychopathological characteristics at baseline from medical records were rated by using the OPCRIT checklist for psychotic illness, version 3.31 (23). OPCRIT comprises a checklist of 90 items constructed from operational criteria for the major psychiatric classifications (ICD, RDC, DSM-III-R, etc.) and a suite of computer programs that allows psychopathological data to be entered and edited and diagnoses to be generated according to each set of diagnostic criteria. OPCRIT has been shown to have good reliability for DSM-III-R diagnoses made by using the 90-item checklist (kappa=0.73) (24). The concurrent validity of OPCRIT DSM-III-R diagnoses has been established in analyses showing good to excellent agreement with consensus best-estimate diagnoses (25).

Of the 93 subjects in the followed-up sample, 51 (55%) had a DSM-III-R diagnosis (OPCRIT derived) of schizophrenia at baseline, 12 (13%) had a schizoaffective psychosis, 23 (25%) had an affective psychosis (unipolar major depressive and bipolar psychoses), and seven (8%) had an atypical psychosis (unspecified functional psychosis). Data for all patients with nonschizophrenic psychoses (N=42) were combined for further analysis. Table 1 shows the characteristics of the 51 patients with schizophrenia and the 42 patients with nonschizophrenic psychoses. Both groups were similar in age at onset, duration of follow-up, gender ratio, referral source (local area versus elsewhere), socioeconomic status, and ethnicity.

All assessments were done by raters who were blind to baseline diagnosis, as the OPCRIT data were not analyzed until all follow-up interviews had been completed. Although the main emphasis in the adult follow-up assessment was on direct interviewing of subjects, where possible multiple sources of information (medical records and interviews with health professionals and relatives) were used to corroborate or supplement information from subject interviews. A median of two sources of information was obtained per subject. After all available sources of information were combined, adult best-estimate DSM-III-R diagnoses were made. The mean length of follow-up (time from baseline) for the 93 subjects in the followed-up sample was 11.5 years (SD=5.2, range=4–22). The mean age at follow-up assessment for the sample was 27.2 years (SD=5.2, range=17–39).

Interviews to determine adult lifetime diagnosis were conducted with a modified version of the Schedule for Affective Disorders and Schizophrenia—Lifetime Version (SADS-L) (26). Coverage was broadened to include diagnoses not covered in the original schedule (27) and some additional questions specifying precise DSM-III-R diagnostic criteria for psychoses. The author, who conducted the SADS-L interviews, obtained training from experienced interviewers at the Institute of Psychiatry.

Although no “gold standard” exists to measure lifetime psychiatric diagnosis, the currently accepted standard procedure involves consensus best-estimate diagnosis on the basis of all available data by two or more clinicians (28). Diagnostic data were extracted and rated independently on a random sample of 25 patients by a second experienced psychiatrist who trained at the Maudsley Hospital. There was a 100% concordance for DSM-III-R diagnoses for the 25 patients rated by two psychiatrists.

Cross-sectional symptomatic and social outcome were measured with several instruments. Cross-sectional measures of psychopathology were obtained by using the Scale for the Assessment of Positive Symptoms (SAPS) (29), the Scale for the Assessment of Negative Symptoms (SANS) (30), and the Global Assessment of Functioning symptom scale (31). The total SAPS score represented the sum of scores on all 35 items (maximum score=170). The total SANS score represented the sum of scores on all 24 items (maximum score=120). The Global Assessment of Functioning symptom scale rates severity of symptoms in the past month on an ordinal scale from 1 (most severe) to 90 (no symptoms). Cross-sectional social outcome/disability was measured with the Global Assessment of Functioning disability scale (31) and the World Health Organization (WHO) Disability Assessment Schedule (32). The Global Assessment of Functioning disability scale rates severity of disability in the past month on an ordinal scale from 1 (most severe) to 90 (no disability).

Course of illness and disability was assessed by using a slightly modified version of the Life Chart instrument from the WHO Multi-Centre Study on the Course and Outcome of Schizophrenia (31). The Life Chart assesses course of illness according to quality of remission, course type, and severity of symptoms, each with clear definitions. Two varieties of remission are distinguished: 1) complete remission, in which the subject is virtually symptom-free and shows his/her usual premorbid personality, and 2) incomplete remission, in which the subject is no longer psychotic but shows any combination of residual symptoms, nonpsychotic symptoms, or personality change. Course type characterizes the entire follow-up period in one of four ways: 1) continuous course, in which the patient was psychotic over most of the period and any remissions that occurred were brief (none longer than 6 months), 2) episodic course, in which discrete psychotic episodes (none longer than 6 months) occurred with clear periods of remission between episodes and at least one period of remission lasting 6 months or more, 3) mixed course, which had characteristics other than those of an episodic or continuous course (e.g., the longest psychotic episode lasted 12 months, and longest remission lasted 9 months), and 4) not psychotic, in which the patient was never actively psychotic during follow-up. Severity of symptoms during the majority of the follow-up period was characterized by four levels: 1) severe, in which the patient could not carry on a coherent conversation, could not work without close supervision, or required constant care; 2) moderate, in which the patient consistently showed signs of health despite marked psychotic symptoms, e.g., could speak with reasonable clarity on some occasions or could work or care for others to some extent; 3) mild, in which the patient’s behavior was generally normal and illness was not immediately obvious in conversation, although definite psychotic symptoms persisted; and 4) recovered, in which the patient experienced no psychotic symptoms or only residual symptoms.

The Global Assessment of Functioning symptom scale range expressed the variability in levels of symptoms during the follow-up period and was defined as the difference between the highest and lowest Global Assessment of Functioning symptom scale score for each subject. Variables measuring time were expressed as the proportion (percentage) of the individual follow-up period. The definition of employment included any time spent in full-time education or vocational training (or full-time child care). Independent adult living was defined as living either alone or with friends/partner or with parents and receiving no more support than would be received by the majority of individuals at a similar age.

Table 3 shows the results of the comparison between the clinical course of adolescent-onset schizophrenia and other psychoses on measures adapted from the WHO Life Chart (31). The results indicate that an adolescent diagnosis of schizophrenia predicted a greater likelihood of a severe, unremitting illness course. Only 12% (N=6) of patients given a diagnosis of schizophrenia in adolescence were in complete remission for most of the follow-up, compared to 52% (N=22) of those with nonschizophrenic psychoses. Subjects with a diagnosis of adolescent-onset schizophrenia were psychotic for a greater proportion of the follow-up (N=46; median 84.0% of follow-up, interquartile range [IQR]=33.7–100) than were those with other psychoses (N=42; median 15.5%, IQR=8.3–39.0) (Mann-Whitney U, z=–4.49, p<0.0001).

Table 4 shows that subjects with adolescent-onset schizophrenia were significantly more impaired at outcome than those with other psychoses. Compared to subjects with adolescent-onset schizophrenia, subjects with nonschizophrenic psychoses had greater symptom variability over the course of follow-up. At outcome, subjects with adolescent-onset schizophrenia had more negative symptoms (N=37, median total SANS score=30 of the maximum score of 120, IQR=13–49) than subjects with other psychoses (N=39, median total SANS score=4 of the 120 maximum score, IQR=0–25) (Mann-Whitney U, z=–4.14, p<0.0001) and more positive symptoms (median total SAPS score=14 of the maximum score of 170, IQR=3–21 for subjects with schizophrenia [N=48] versus median total SAPS score=0 of the 170 maximum score, IQR=0–12 for subjects with other psychoses [N=42]) (Mann-Whitney U, z=–3.12, p=0.002).

Subjects with adolescent-onset schizophrenia spent less time living independently (N=50, median time independent <1 year, <1% of follow-up period) than subjects with nonschizophrenic psychoses (N=42, median time independent=5.9 years, 74% of follow-up period) (Mann-Whitney U, z=–4.5, p<0.0005). Twenty percent (N=10) of the subjects with adolescent-onset schizophrenia were living independently (nine of these 10 were still living with parents), compared to 50% (N=21) of subjects with other psychoses (10 of 21 were living with friends or partners). Thirty-two percent (N=16) of the subjects with schizophrenia had spent the majority of the follow-up period in long-stay residential or psychiatric hospital care, compared to only 17% (N=7) of subjects with other psychoses. There was a significant linear association between adolescent-onset schizophrenia and nonindependent adult living (χ²=10.8, df=5, p=0.001).

Sixty percent (N=30 of 50) of the subjects with adolescent-onset schizophrenia left school with no formal qualifications and undertook no further postschool training. Despite this overall low level of educational achievement, it is notable that 22% (N=11) achieved passes on the “GCSE” (General Certificate of Standard Education, the U.K. national scholastic examination for 16-year-olds) or the “A level” (U.K. national scholastic examination for 18-year-olds) after the onset of psychosis. Educational achievement was significantly better for those with nonschizophrenic psychoses, with only 33% (N=14) without any formal qualification or training, while 38% (N=16) achieved GCSE or higher qualifications. There was a significant linear association between the diagnosis of schizophrenia and low educational achievement (χ²=6.9, df=5, p<0.01).

Subjects with an adolescent diagnosis of schizophrenia spent significantly less time in employment over the follow-up period (N=50, median=0%, IQR=0%–18.0%) than subjects with nonschizophrenic psychoses (N=42, median=32.4%, IQR=5.0%–63.6%) (Mann-Whitney U, z=–3.73, p=0.0002).

Table 5 indicates that adolescent-onset schizophrenia predicts impoverished social relationships in adult life. Almost half of subjects with adolescent-onset schizophrenia (N=22, 44%) had no social contacts other than with professionals or close relatives. Only 14% socialized with a friend or friends, only one subject had had a reciprocal love relationship, and none had married or had a stable partner. Subjects with an adolescent diagnosis of nonschizophrenic psychoses were much less socially impaired, with more than half (N=24, 57%) having either a regular friendship or love relationship.

To estimate the magnitude of confounding in the association between diagnosis and symptomatic outcome, the Global Assessment of Functioning symptom scale score was regressed on DSM-III-R adolescent-onset diagnostic category (schizophrenia versus nonschizophrenic psychosis) and the following variables were then added to the model: gender, social class, location (area of residence), and ethnicity. The unadjusted regression coefficient (Beta) for diagnosis was 14.9, and the fully adjusted regression coefficient (Beta) was 13.4. Overall, adjusting for demographic confounding had a minor effect (about 10%) on the association between diagnosis and symptomatic outcome.