Study site

Tamil Nadu is one of the most industrialised states in India. The primary language is Tamil. Tamil Nadu has better education, health and development indices than most other Indian states: 73% of the population is literate compared with 65% for the whole country. ¹¹ Vellore district (area 4314 km², population 3 026 432) is situated in the north central part of the state. Kaniyambadi Block, one of 12 administrative blocks in Vellore district, was the study site. The study took place in the Department of Psychiatry of the Christian Medical College, Vellore. The 100-bed hospital adult psychiatry unit provides short-term care for patients with different organic disorders, substance use, psychoses, neuroses and adjustment problems. The emphasis is on a multidisciplinary approach and eclectic care using a wide variety of therapies. The hospital has a daily out-patient clinic serving 200–250 patients.

Sample and procedure

The study group consisted of patients living within a 100 km radius of the study site. Patients were carefully screened for a diagnosis of schizophrenia and then interviewed by a research psychiatrist (B.S.) at intake using the Structured Clinical Interview for DSM–III–R – Patient version (SCID–P) to confirm the diagnosis. ^{Reference Spitzer, Williams and Gibbon12} Patients with a primary diagnosis of substance use disorder, mood disorder or organic mental disorder were excluded (for details see reference Reference Saravanan, Jacob, Johnson, Prince, Bhugra and David5). Patients meeting inclusion criteria and providing written consent were interviewed as soon as possible, with all patients assessed within a week of starting treatment for psychosis. All patients were interviewed again 6 months and 12 months after the initial assessment. Patients were informed that the purpose of the study was to assess their level of awareness about their illness, but they were unaware of the study's specific hypotheses.

Assessment measures

Social and demographic variables were recorded, including years of education. Patients were interviewed by trained raters using the Tamil version of SCID–P. At baseline we used the Lifetime version of the interview. The Schedule for the Assessment of Insight – Expanded version (SAI–E) with its three dimensions (awareness of illness, relabelling and compliance), the Brief Psychiatric Rating Scale (BPRS) and the Global Assessment of Functioning (GAF) were used at baseline and at each follow-up wave. ^{Reference David, van Os, Jones, Harvey, Foerster and Fahy8,Reference David, Buchanan, Reed and Almeida13–Reference Ventura, Lukoff, Neuchterlein, Liberman, Green and Shaner15} As before, we used the ‘guilt’, ‘low mood’ and ‘suicidality’ items of the BPRS added together as a proxy measure of depression. Duration of untreated psychosis was defined as the interval from the occurrence of prominent psychotic symptoms to first contact with psychiatric services. This onset date was synthesised from the SCID–P interview, which included chronology of psychotic symptoms and significant other collateral information. Finally, whether or not the patient consulted a traditional healer prior to coming to the hospital was noted as a potential indicator of early illness awareness.

We planned to use clinically defined ‘broad’ psychosis outcomes (full remission, remission with deficits, and relapse) as the primary outcome measure, in addition to the repeated measures of GAF (baseline, 6-month and 12-month assessments). Remission is a state following a psychotic episode in which there is no positive or negative symptom for at least 30 days; relapse is the occurrence of at least one positive symptom or two negative symptoms after a period of remission; remission with deficits is the presence of one negative symptom or of social or occupational deficits after a period of remission. Two of the researchers (B.S. and K.S.J.) agreed on the ratings of remission and relapse, basing these on criteria used in the Study of Factors Associated with the Course and Outcome of Schizophrenia (SOFACOS), ^{16,Reference Verghese, John, Rajkumar, Richard, Sethi and Trivedi17} and the Determinants of Outcome of Severe Mental Disorders (DOSMeD), ^{Reference Jablensky, Sartorius, Emberg, Anker, Korten and Cooper18} applied to information extracted from clinical case notes. We used the broad psychosis outcome to tabulate the regression models to make this study more relevant to clinical practice in low- and middle-income countries.

Statistical analysis

The three broad psychosis outcome variables were collapsed to yield two: remission and relapse. Remission was defined as being symptom-free, and relapse was defined as meeting criteria for a schizophrenic episode after meeting criteria for remission in a previous follow-up period. We used general linear modelling (GLM) analysis in SPSS version 15 for Windows, to test whether baseline insight or psychopathology (BPRS score, model 1), or change in insight or psychopathology at 0–6 months (model 2) or at 6–12 months (models 3 and 4) predicted good outcome using GAF score as a continuous outcome measure at 12 months, controlling for sociodemographic and clinical variables (gender, marital status, employment, urbanicity, DUP, traditional healer visit plus baseline insight and psychopathology scores where appropriate). A second set of logistic regressions, each with the same model structure as above, was performed with clinical outcome (complete remission or relapse) as the dependent variable, entering the same independent variables and with the same controls. Interactions between insight and outcome and significant covariates were checked in the modelling. Model fits were checked by examining partial residual plots.

Good and poor outcomes

Table 3 summarises the rates of remission and relapse found in the cohort over the year of follow-up. Of the 115 patients assessed, half (n = 58) had experienced complete remission of symptoms and half (n = 57) experienced remission with deficits by the 12-month follow-up. At the 6-month assessment, 1 patient had died, 15 had withdrawn and 12 could not be traced, and at 1 year there was 1 further death, 4 patients withdrew and a further 11 could not be traced. Fourteen patients (12%) relapsed during the entire 12-month follow-up period. Most of those assessed at 12 months (96%, n = 110) were found to satisfy the study criteria for schizophrenia at any time during the year of follow-up. In addition, 65% (n = 75) were in regular contact for the monthly follow-up appointments.

Table 3

Remission and relapse 1 year after first contact with psychiatric services

Outcome	n (%)
Complete remission	58 (50)
Relapse	2
No relapse	56
Remission with deficits	57 (50)
Relapse	12
No relapse	45

Insight and psychopathology

Pearson correlation coefficients between insight and BPRS scores at baseline and follow-up assessment points were examined. At the baseline assessment there was no significant correlation between total BPRS score and total insight score. However, a significant albeit weak negative correlation between the relabelling dimension of the SAI–E and BPRS was observed (Pearson's r = –0.2, P = 0.04). In addition, the depression subscale score (mean = 6.5, s.d. = 2.3) showed a weak positive correlation with SAI–E total score and illness awareness items (r = 0.17 and r = 0.19 respectively, P≤0.05).

At follow-up the pattern of associations was rather different: there was a moderate significant negative correlation between total BPRS score and total insight score at 1 year (r = –0.3, P = 0.007). Lower BPRS scores were significantly associated with increase in the awareness (r = –0.35, P = 0.01) and treatment dimensions (r = –0.34, P = 0.01) at 6 months and the awareness (r = –0.4, P = 0.01), relabelling (r = –0.5, P = 0.01) and treatment dimensions (r = –0.5, P = 0.01) at 1 year. To find out whether changes in insight might be accounted for by improvement in clinical status, correlations between change scores in BPRS (baseline minus follow-up) and change scores in insight measures were obtained. The correlation coefficient was r = 0.3 (P = 0.01).

Predictors of outcome

Both regression analyses identified duration of untreated psychosis and early changes in insight and psychopathology as predictors of good outcome (Tables 4 and 5). Table 4 shows predictors of GAF and Table 5 shows predictors of clinically classified good outcome at 12 months. Duration of untreated psychosis is a significant predictor in both analyses in all four models, along with changes in psychopathology (BPRS) and insight (SAI–E) scores between baseline and 6 months (model 2). Interestingly, neither baseline insight nor baseline psychopathology was predictive of outcome (model 1). Similarly, changes between 6 months and 12 months were not related to outcome, with or without adjustment for baseline psychopathology/insight (models 3 and 4; Fig. 1).

Table 4

Predictors of good outcome in first-episode schizophrenia using Global Assessment of Functioning (GAF) score as the primary outcome: general linear model analysis

					Model 1		Model 2		Model 3		Model 4
	Model 1a Beta	Model 2b Beta	Model 3c Beta	Model 4d Beta	F	P	F	P	F	P	F	P
Age	-1.05	6.77	0.23	0.23	0.002	0.96	0.12	0.73	1.30	0.26	1.34	0.25
Gendere	0.57	-1.13	-0.40	-0.77	0.04	0.85	0.19	0.66	0.02	0.88	0.08	0.77
Marital statusf	-1.76	-3.37	-0.99	-0.42	0.32	0.57	1.60	0.21	0.12	0.73	0.02	0.89
DUP	-3.31	-3.45	-3.90	-3.75	9.5	0.003*	11.50	0.001*	13.6	0.001*	12.2	0.001*
Traditional healerg	2.12	1.33	2.72	2.73	0.52	0.47	0.25	0.62	0.98	0.33	0.99	0.32
Education	-0.32	-0.33	-0.30	-0.31	0.86	0.36	1.24	0.27	0.90	0.35	0.95	0.33
BPRS score
Baseline	9.69			4.56	0.12	0.73					0.03	0.86
Change 0-6 months		-0.43					0.66	0.003*
Change 6-12 months			-0.30	-0.30					0.66	0.08	3.10	0.08
SAI—E score
Baseline	0.43			0.24	2.0	0.16					77	0.38
Change 0-6 months		0.70					0.66	0.001*
Change 0-12 months			0.24	0.24					1.66	0.20	1.61	0.21

Table 5

Predictors of good outcome (remission without relapse) in first-episode schizophrenia using clinically defined outcome criteria: logistic regression analysis

					Model 1		Model 2		Model 3		Model 4
	Model 1a CI	Model 2b CI	Model 3c CI	Model 4d CI	OR	P	OR	P	OR	P	OR	P
Age	0.90-1.03	0.88-1.03	0.92-1.06	0.92-1.06	0.97	0.32	0.95	0.23	0.98	0.66	0.98	0.66
Gendere	0.33-2.01	0.23-1.68	0.28-1.88	0.28-1.92	0.81	0.65	0.62	0.34	0.72	0.51	0.74	0.53
Marital statusf	0.24-1.56	0.16-1.28	0.25-1.80	0.23-1.81	0.61	0.30	0.45	0.13	0.67	0.42	0.65	0.41
DUP	0.99-1.00	0.99-1.00	0.99-1.00	0.99-1.00	0.99	0.04*	0.99	0.05*	0.99	0.03*	0.99	0.03*
Traditional healerg	0.31-1.79	0.23-1.76	0.32-2.17	0.32-2.18	0.75	0.51	0.64	0.39	0.83	0.72	0.84	0.72
Education	0.85-1.06	0.84-1.07	0.85-1.07	0.85-1.07	0.95	0.37	0.95	0.23	0.95	0.42	0.95	0.42
BPRS score
Baseline	0.93-1.08	0.90-1.07	0.92-1.08	0.92-1.08	1.00	0.95
Change 0-6 months		0.89-1.0					0.95	0.05*
Change 6-12 months			0.91-1.06	0.91-1.03					0.97	0.32	0.96	0.32
SAI—E score
Baseline	0.94-1.13		0.90-1.09	0.90-1.09	1.03	0.59					0.99	0.85
Change 0-6 months		1.04-1.23					1.13	0.006*
Change 0-12 months			0.96-1.09	0.96-1.09					0.84	1.03	1.03	0.42

Components of insight

We did not have sufficient statistical power to examine separately specific components of insight or symptom clusters. Nevertheless, exploratory analyses using GAF as the outcome measure indicated that at 6 months (model 2) the illness awareness dimension of the SAI–E was a significant predictor (F = 21.6, P<0.001; adjusted R ² = 0.330). Significant effects also emerged in model 3 with the illness awareness (F = 7.96, P = 0.006) and treatment compliance (F = 12.2, P = 0.001) dimensions (but not relabelling).

Quality of outcome

Our second finding was that during the follow-up period remission was common and relapses rare. Almost half of the patients had full remission of symptoms over the course of the 12-month follow-up. This rate of remission (50%) is midway between rates reported from urban and rural Chandigarh as part of the DOSMeD project but lower than the 69% recorded in Vellore in SOFACOS in the early 1980s. ^{16–Reference Jablensky, Sartorius, Emberg, Anker, Korten and Cooper18} This is consistent with the ‘favourable outcome’ hypothesis in low- and middle-income countries, ^{Reference Jablensky, Sartorius, Emberg, Anker, Korten and Cooper18,Reference Lin and Kleinman32–Reference Hooper, Harrison, Janca and Sartorius35} but see Cohen et al ^{Reference Cohen, Patel, Thara and Gureje36} and Leff ^{Reference Leff37} for a debate on this question.

Duration of untreated psychosis

Our third finding is the significant association between DUP and both outcome measures, even after changes in insight and psychopathology were controlled for (see Drake et al). ^{Reference Drake, Haley, Akhtar and Lewis38} Patients with longer DUP were less likely to achieve full remission. This is in line with many published studies in various settings including low- and middle-income countries. ^{Reference Marshall, Lewis, Lockwood, Drake, Jones and Croudace39–Reference Farooq, Large, Nielssen and Waheed41} In our study the good outcome and poor outcome groups had a mean DUP of 67.1 weeks and 122.5 weeks respectively. It is not certain that reducing this period would necessarily improve outcome, but this is surely a worthwhile issue for research and clinical development. We did not find any gender effects on outcome in this study. Superior outcome in women was noted both in DOSMeD and other studies from India and in rural China, although not in Nigeria or Colombia. ^{Reference Hooper, Harrison, Janca and Sartorius35,Reference Ran, Xiang, Huang and Shan40,Reference Thara and Rajkumar42–Reference Ohaeri44}

Insight, psychopathology and culture

We found fairly consistent associations between insight and psychopathology. Our baseline findings using the expanded BPRS and SAI–E showed weak but significant inverse correlations between total BPRS score and the relabelling dimension. ^{Reference Saravanan, Jacob, Johnson, Prince, Bhugra and David5,Reference Drake, Pickles, Bentall, Kinderman, Haddock and Tarrier45} However, at follow-up all three dimensions of insight (awareness, compliance and relabelling) and total insight scores showed predictably significant although weak relationships with psychopathology. This is in line with most studies in a variety of settings, including India, ^{Reference David, Amador and David6,Reference Saeedi, Addington and Addington29,Reference Kulhara, Chakrabarti and Basu46–Reference Mintz, Dobson and Romney48} but see Drake et al who found that paranoia and delusions specifically were unrelated to global insight. ^{Reference Drake, Pickles, Bentall, Kinderman, Haddock and Tarrier45} Studies in our cohort using quantitative and qualitative methods suggest that early change in insight or remaining within the existing ‘collaborative social world’ may have benefits. ^{Reference Saravanan, Jacob, Deepak, Bhugra, Prince and David19} One clinical implication is that efforts to improve insight taking account of local cultural factors could yield benefits in outcomes and might relate to the supposedly superior prognosis of psychosis in settings such as those in south India.

Methodological issues

The participants in this study were located in a mental health service centred on a large hospital. This may represent a selection bias. However, this is likely to have resulted in a more severely affected patient cohort than one derived from community-based case ascertainment. Initially, 29% had to be excluded, mostly owing to florid psychotic disturbance rendering formal assessment impossible. Withdrawal rates were low, reflecting in large part the influence of intact family structures which support patients and value medical treatment. Outcome was categorised relatively crudely to assist comparison with other studies, yet the continuous and categorical outcomes yielded highly consistent findings. The clinically defined categorical outcome was made in part by the psychiatric researcher (B.S.), who also participated in the initial and follow-up ratings of psychopathology (including insight) and measurement of DUP, so may have been biased. However, the research nurse (S.J.) who rated GAF outcome was unaware of DUP status. This method ensured that at least one of the outcome measures is unlikely to have been biased by knowledge of DUP or other baseline factors. Again, the consistency in the findings regarding predictors of outcome provides reassurance that such observer bias was minimal. Finally, we were unable to measure the extent to which aspects of culture shape insight and hence outcome of schizophrenia in this study in the absence of contemporaneous control data from other countries or settings.