Study site

This study was carried out in the Department of Psychiatry, Christian Medical College, Vellore, which is in the north central part of Tamil Nadu. The total area of the Vellore district is 4314.29 km² and is divided into 12 blocks with a total population of 3 026 432. The Department of Psychiatry and Community Health has worked within the Kaniyambadi Block for the past 40 years. The town of Vellore (10.54 km²) has a population of 175 061. The 100-bed hospital provides short-term care for patients with all types of psychiatric diagnoses from the town of Vellore and a much wider rural area beyond. The emphasis is on a multidisciplinary approach and eclectic care using a wide variety of pharmacological and psychological therapies. The hospital has a daily out-patient clinic in which 200–250 patients are seen. The study was approved by the ethics committees of the Christian Medical College, Vellore, and Institute of Psychiatry, London.

Sample

The study group consisted of patients with schizophrenia having their first contact with mental health services and living within a 100 km radius of the study site. Patients were carefully screened for a DSM–IV diagnosis of schizophrenia (American Psychiatric Association, 1994) and then interviewed at intake using the Structured Clinical Interview for DSM–III–R–Patient Version (SCID–P; Reference Spitzer, Williams and GibbonSpitzer et al, 1990) to confirm the diagnosis. Patients with a primary diagnosis of substance use disorder, mood disorder or organic mental disorder were excluded. Patients meeting inclusion criteria and providing written consent were interviewed as soon as possible after the start of treatment for psychosis, with all patients assessed within a week of the onset of treatment. Patients were informed that the purpose of the study was to assess their level of awareness about their illness, but were unaware of the specific hypotheses. A semi-structured interview was used to elicit data regarding demographic characteristics (age, gender, marital status, religion, education, employment, duration of illness and economic data).

Assessments

All assessment instruments were independently translated into the local language (Tamil) by two health professionals. The vernacular version thus obtained was then back-translated into English by two different bilingual professionals. The four translators then arrived at a consensus on the final vernacular version. Content, semantic, technical and conceptual equivalence of the Tamil version of the instruments was examined regularly during the process of translation. Assessments were administered by a local research psychiatrist trained in their use.

Insight

The expanded version of the Schedule of Assessment of Insight (SAI–E; Reference Kemp, David and BlackwellKemp & David, 1997; Reference Sanz, Constable and Lopez-IborSanz et al, 1998) was used for assessment of insight. This has been applied widely in Western and non-Western countries (Reference Kulhara, Chakrabarti and BasuKulhara et al, 1992; Reference Aga, Agarwal and GuptaAga et al, 1995) and comprises questions to assess three dimensions of insight: awareness, relabelling of symptoms and adherence, plus a ‘hypothetical contradiction’ item added to evaluate the person's capacity to consider another's perspective. Each dimension comprises two or three questions which are scored on a 3-point scale from 0 (no insight) to 2 (good insight), with a maximum total score of 24. The supplementary question is scored from 0 to 4 and this is added to the total score. This expanded version also includes items on awareness of change, difficulties resulting from the mental condition and insight into key symptoms.

Explanatory model interview

The Tamil version of the Short Explanatory Model Interview (SEMI; Reference Lloyd, Jacob and PatelLloyd et al, 1998; Reference Joel, Sathyaseelan and JayakaranJoel et al, 2003) was used to elicit patients' attributions of their presenting complaints; their previous help-seeking behaviour (including visiting a temple, a shamam/mantrawadi, a traditional healer, or a doctor); their causal models (e.g. previous deeds/karma, evil spirits, punishment by god, black magic, or disease); perceived consequences (change in the body or mind); and their expectations regarding the index consultation. The SEMI, which combines open-ended questions and a case vignette with a structured coding frame, has been used successfully in a variety of countries and cultures, including India (Reference Manoharam, John and JosephManoharam et al, 2001). The vignette described a young man with florid psychotic symptoms (see Appendix). Participants were asked the same range of questions regarding illness attribution for the vignette and the help-seeking they would advise for such a person.

Psychopathology

The Brief Psychiatric Rating Scale (BPRS; Reference Ventura, Green and ShanerVentura et al, 1993) and the Global Assessment of Functioning (GAF; Reference Endicott, Spitzer and FleissEndicott et al, 1976) were used to assess psychopathology. The BPRS total score and a depression subscale (investigating guilt, low mood and suicidality) were utilised.

Insight and psychopathology

There was no significant correlation between BPRS total score and SAI–E total score (Pearson's r=–0.13, P=0.1), general awareness (r=–0.05) or treatment adherence sub-scales (r=–0.15). However, a significant negative correlation was observed between the relabelling dimension of the SAI–E and BPRS total score (r=–0.2, P=0.04). Of the individual items on the BPRS, anxiety showed the strongest (positive) correlation with SAI–E total score and with the illness awareness items (r=0.25 and 0.28 respectively, P<0.01).

Explanatory models

Table 2 shows the relationship between insight and responses to the SEMI summarised according to major theme. The 18 patients who endorsed a disease explanation had significantly higher scores on the SAI–E, indicating greater insight. Most patients acknowledged social (but not occupational) dysfunction and that their well-being had been affected; and these had higher insight scores than those who did not. There was a significant inverse correlation between non-biomedical beliefs and insight scores. Patients who held such beliefs, most commonly regarding black magic (n=94), but also those who attributed their problems to evil spirits, previous deeds and punishment by god, tended to have lower insight scores. Patients were frequently unsure or unable to give a definite attribution of their illness and this tended to be associated with low insight scores.

Table 2 Relationship between insight and explanatory models variables

SEMI questions	Total number of responses	SAI–E total score Mean (s.d.)	F (P) 1
Black magic
Yes	94	4.1 (4.0)
No	26	7.7 (5.6)	9.0 (0.001)
Not sure	11	2.5 (2.6)
Previous deeds
Yes	12	5.7 (4.8)
No	78	5.8 (5.0)	9.8 (0.001)
Not sure	41	2.2 (1.9)
Evil spirits
Yes	22	3.6 (4.3)
No	80	5.7 (4.9)	6.7 (0.002)
Not sure	29	2.5 (2.0)
Punishment by god
Yes	13	4.0 (2.7)
No	71	5.8 (4.9)	6.3 (0.003)
Not sure	47	3.0 (3.8)
Disease
Yes	18	10.4 (6.1)
No	64	3.7 (3.4)	21.4 (0.001)
Not sure	49	3.8 (3.8)
Social dysfunction
Yes	110	5.1 (4.8)
No	13	2.9 (2.5)	3.1 (0.05)
Not sure	8	1.7 (1.4)
Occupational dysfunction
Yes	124	4.8 (4.6)
No	5	2.1 (2.5)	1.2 (0.3)
Not sure	2	2.0 (2.8)
Well-being affected
Yes	103	5.4 (4.7)
No	8	2.1 (3.0)	6.1 (0.003)
Not sure	20	2.1 (2.5)

SEMI, Short Explanatory Model Interview; SAI–E, Schedule for Assessment of Insight – Expanded version

1. One-way analysis of variance (ANOVA)

Awareness of illness – case vignette

Patients were more likely to judge the person described in the vignette as having an illness or disease (n=63) than they were to make this attribution about themselves (n=18; χ²=13.9, d.f.=1, P<0.001), excluding those making attributions to neither (n=66). Slightly fewer patients judged the person described in the vignette to be suffering from the effects of black magic (n=81, 61.8%) than they themselves (n=94, 71.8%; 95% CI–21.3 to 1.4, NS).

Factors associated with insight

A stepwise conditional logistic regression analysis was performed with total insight score as the dependent variable, entering psychopathology, endorsed explanatory models, pathways to care and other clinical variables, controlling for age, gender and education and other demographic factors. This identified causal attribution (disease as a positive predictor and black magic as a negative predictor), the perceived consequences item of the SEMI and presence of help-seeking as factors significantly associated with insight (Table 3).

Table 3 Factors associated with level of insight in schizophrenia ¹

Variable	Unadjusted odds ratio (95% CI)	Adjusted odds ratio 2 (95% CI)	P
Help-seeking (0=no, 1=yes)	5.64 (1.6–16.8)	4.9 (1.97–12.2)	0.001
Attribution to disease (0=no, 1=yes)	10.2 (3.4–30.3)	9.4 (3.1–29.0)	0.005
Attribution to black magic (0=no, 1=yes)	0.3 (0.12–0.73)	0.36 (0.12–0.94)	0.04
Awareness of change in body or mind (0=no, 1=yes)	8.8 (1.1–67.8)	9.7 (1.2–75.9)	0.03

SAI–E, Schedule for the Assessment of Insight – Expanded version

1. SAI–E total scores were dichotomised into high and low groups (with the median as the cut-off point)

2. Adjusted using conditional logistic regression with gender, age, years of schooling, location of residence (rural v. urban) and marital status (currently married v. not currently married)

Psychopathology

There was some evidence to support the hypothesis that insight is related to the psychopathology of schizophrenia. We found a weak but significant inverse correlation between total BPRS score and the relabelling dimension of insight, and a positive relationship between insight total score and indices of anxiety and worry. Given that correlations were in opposite directions, it is not surprising that the overall correlation was non-significant, although the weight of psychosis items has usually outweighed the mood/anxiety items in previous studies (for review see Reference David, Amador and DavidDavid, 2004). Nevertheless this pattern is in line with studies carried out in both high- and low- and middle-income countries (Reference Kulhara, Chakrabarti and BasuKulhara et al, 1992; Reference Aga, Agarwal and GuptaAga et al, 1995; Reference Mintz, Dobson and RomneyMintz et al, 2003; Reference David, Amador and DavidDavid, 2004). The strength of the correlations was modest, bolstering the argument that insight and psychopathology are not merely two sides of the same coin. Lower insight scores were seen in patients brought to the clinic involuntarily, as expected from previous studies (Reference McEvoy, Applebaum and AppersonMcEvoy et al, 1989; Reference David, Buchanan and ReedDavid et al, 1992). The lack of correlation with scores of social functioning is again in line with some (Reference Bartko, Herczeg and ZadorBartkó et al, 1988; Reference SchwartzSchwartz, 1998; Reference Simon, Berger and GiacominiSimon et al, 2004) but not all previous studies (Reference Amador, Flaum and AndreasenAmador et al, 1994; Reference Trauer and SacksTrauer & Sacks, 2002), and might reflect the particularly low levels of social functioning (and small variance) in this patient sample. As in previous studies, duration of illness and gender were not determinants of insight.

The association between insight and symptoms of anxiety – assessed somewhat crudely using the BPRS – may be predicted on the basis of previous work (Reference Amador, Friedman and KasapisAmador et al, 1996; Reference Mintz, Dobson and RomneyMintz et al,2003; Reference Rathod, Kingdon and SmithRathod et al, 2005), but the direction of causality is by no means clear. The received wisdom is that anxiety and lowered mood are a reaction to illness awareness, but it is equally plausible that anxiety leads to a more self-critical attitude, akin to ‘depressive realism’. Whatever the explanation, this pattern of associations appears not to be culturally specific.

Explanatory models

Insight was also evaluated indirectly using a vignette that forms part of the SEMI. In line with our second hypothesis, a pattern was seen wherein the person in the vignette was readily labeled as having a mental illness whereas patients rarely used this attribution for themselves. This pattern has been observed in other vignette studies in the UK, USA and Asia (Reference McEvoy, Schooler and FriedmanMcEvoy et al, 1993; Reference Swanson, Freudenreich and McEvoySwanson et al, 1995; Reference Chung, Chen and LamChung et al, 1997; Reference StartupStartup, 1997). The self–other comparison overcomes the criticism levelled at cultural comparisons in psychiatry that the ‘content’ of attributions, delusions, hallucinations etc. may differ but that this is a relatively trivial consideration. Previous vignette studies show that the ‘medical model’ is an available heuristic to people with major mental illness, including those in South India. This is not surprising given the high levels of literacy and rapid cultural transition – first through colonisation and subsequently through globalisation – affecting that part of the world. It is of course highly likely that complex illness attributions of the kind elicited have been around in Asia for centuries, but here we are talking about a narrower Western biomedical model. The pattern of responses in relation to self and others might be described as an example of self-serving bias, a common, perhaps universal, defensive cognitive style (Reference Pronin, Gilovich and RossPronin et al, 2004). It is understandable that a person might be uncomfortable in ascribing to themself a diagnosis of mental illness, given the stigma that this carries and, perhaps, implications of blame. However, the alternative explanations of behaviour such as black magic or past misdeeds are not value free. The meaning and implications of mental illness within cultures are clearly varied, and we are undertaking further detailed qualitative work to explore this.

Finally, we tested the hypothesis that indigenous explanatory models or ‘emic perspectives’ would be independent of insight. The study showed that the expression of culturally appropriate illness attributions (e.g. evil spirits, black magic, etc.) was common. Adherence to such ideas was deemed, at best, not consistent with high levels of insight or, at worst, indicative of poor insight as judged by psychiatrists from the same culture. In the case of black magic, this illness attribution predicted significantly lower levels of insight in the multiple regression analysis after controlling for other factors. However, the same analysis revealed a strong association between awareness of the consequences of illness, early help-seeking and level of insight, elicited separately through the open-ended SEMI in a way that did not rely upon the conventions of a psychiatric interview. This supports the notion that insight or an awareness of illness that is adaptive in terms of appropriate help-seeking might easily be missed if the assessment of insight is undertaken in a rigid culturally insensitive way (Reference Saravanan, Jacob and PrinceSaravanan et al, 2004).

Our results compliment a recent study of second-generation minority ethnic groups in the UK of West African, Caribbean and Bangladeshi origin. This study found that there was an association between attribution of illness to supernatural phenomena – not invoked by the White UK comparison group – and poorer insight as assessed using the SEMI (Reference McCabe and PriebeMcCabe & Priebe, 2004). We also noted a tendency for those who were unsure whether disease or supernatural forces accounted for their state to have low insight scores. This might imply that any attribution is better than none.

Conclusions

We have shown that insight is amenable to study in a non-Western setting using standard assessment instruments. Patients with schizophrenia, wherever they are, appear to vary in their willingness to accept an illness attribution for their state, but are more likely to do so if they are anxious and if considering someone else's predicament rather than their own. However, the study supports a more inclusive conceptual model for assessing insight which emphasises help-seeking. This will be of value not just for clinical and research work in non-Western countries but for a richer understanding of the experience of the individual in their cultural context (Reference White, Bebbington and PearsonWhite et al, 2000; Reference McCabe and PriebeMcCabe & Priebe, 2004). Understanding the relationship between cultural variations of insight and prevailing belief systems will be a step forward in designing interventions to enhance engagement of patients with mental health services in Western and non-Western countries.

Case vignette

Mr M.E. was hospitalised at the age of 18. He was single and lived with his parents. His relatives described that he functioned quite well as a child and that he was well adjusted at school until the age of 11–12. At that age he became preoccupied with strange ideas and for this reason he saw a psychiatrist weekly for about 1 month. This treatment, which did not include any mediction, had a positive effect on his worries, but he became more withdrawn and participated in fewer social activities in the months to come. Before hospitalisation he felt that others could hear his thoughts and he also felt that a ‘satanic’ group living in his native place were persecuting him. In a mysterious way he felt that this group did ‘black magic’ against him and that they could influence his body from a long distance. He could feel this as a pain in his stomach. These symptoms lasted for several months. At the time of hospitalisation he felt that his brain was damaged and ‘empty’, and that ‘someone’ was inserting thoughts into his head. He was withdrawn and preoccupied with the idea that the ‘black magic’ that they had done might have destroyed his brain tissue.