The physical health of people with severe mental illness in Bangladesh and Pakistan: A cross-sectional survey

Gerardo Antonio Zavala (  g.zavala@york.ac.uk ) ARK Foundation Bangladesh https://orcid.org/0000-0002-9825-8725 Asiful Haidar ARK Foundation Bangladesh Krishna Prasad-Muliyala NIMHANS: National Institute of Mental Health and Neuro Sciences Faiza Aslam IOP: Institute of Psychiatry Rumana Huque Arkin Foundation Bangladesh Humaira Khalid IOP: Institute of Psychiatry Pratima Murthy NIMHANS: National Institute of Mental Health and Neuro Sciences Asad T Nizami IOP: Institute of Psychiatry Sukanya Rajan NIMHANS: National Institute of Mental Health and Neuro Sciences David Shiers University of Manchester Najma Siddiqi University of York Kamran Siddiqi University of York Jan R Boehnke University of Dundee


Introduction
Severe mental illnesses (SMI) are conditions such as schizophrenia and bipolar disorder that are debilitating, persistent and associated with serious functional impairment. People with SMI die on average 10-20 years earlier than the general population, and this 'mortality gap' is widening. [1] Although suicide accounts for 15% of deaths, an estimated 80% of the observed premature mortality is attributable to physical disorders (physical multimorbidity), most commonly due to non-communicable diseases (NCDs). [2] The excess disease burden from physical multimorbidity in people with SMI may be explained by a combination of factors associated with these mental disorders including clustering of, and predisposition to health-risk behaviours (e.g. tobacco and alcohol use, lack of physical activity, and poor diet), side effects of medication, social determinants of poor health (e.g. stigma, and poverty) and barriers to accessing healthcare. [3] Our current understanding of the distribution and determinants of physical multimorbidity in people with SMI is based mostly on evidence from high-income countries. A few small studies from low-and middle-income countries (LMICs) show similar patterns, but with an even shorter life expectancy and higher mortality for people with SMI. [4,5] These studies indicate that physical multimorbidity in SMI may be at least as much of a challenge in LMICs as in high-income countries. [1] In South Asia, the prevalence of both mental disorders and NCDs has been increasing rapidly. [6,7] This increase is coupled with limited access to essential health services and a widespread neglect of the physical health needs of people with SMI by policy makers and healthcare services.
[8] The overall burden of disease due to physical multimorbidity in this population is, therefore, likely to be high and is set to rise further, with a corresponding increase in within-country and global health inequalities. Despite these concerns, there is a lack of empirical studies originating in South Asia on the distribution and determinants of physical multimorbidity in people with SMI. [9] Addressing multimorbidity in LMICs is a global priority, recognised in global policies to help achieve the Sustainable Development Goals. [10] A detailed understanding of the prevalence of physical multimorbidity and current access to health advice and treatments for physical disorders in people with SMI in LMICs can inform appropriate service provision and contribute to achieving these goals.

Methods
Aims 1) estimate the prevalence of physical health conditions and health risk behaviours, 2) assess access to physical healthcare and health-risk modi cation advice, and 3) compare ndings with those of the general population. for people with SMI in Bangladesh, and Pakistan.

Design and settings
We conducted a cross-sectional survey of patients with a clinical diagnosis of SMI recruited at two national specialist mental health institutions in South Asia, the National Institute of Mental Health (NIMH) in Dhaka, Bangladesh; and Institute of Psychiatry (IOP) Rawalpindi, Pakistan. We originally planned to include a third site-the National Institute of Mental Health and Neurosciences, Bangalore, India, However, due to a delay in obtaining all requisite approvals, data collection could not be initiated before COVID-19 pandemic. Further details of the methods are reported in the published protocol, [11] and summarised below.

Sample size
We aimed to build as large a sample as possible within the resources available over the study period, with an initial target of 1,500 participants at each site. As an indicative example of precision to address some of the key research questions, we used the example of diabetes. For investigating the prevalence of type-2 diabetes, assuming a prevalence estimate of 10%, 857 participants per country would provide a precision of ± 2% (95% con dence interval).

Eligibility
Consenting adults (18 years and over) with a clinical diagnosis of SMI de ned by the International Classi cation of Disease 10th revision (ICD-10) as schizophrenia, schizotypal and delusional disorders (F20-F29); bipolar affective disorder (F30, F31); and severe depression with psychotic symptoms (F32.3, F33.3), and able to provide informed consent, as assessed by the treating clinician, were eligible.

Con rmation of SMI diagnosis
To increase standardisation across sites and alignment with other studies, each SMI diagnosis was con rmed by trained researchers using the Mini-international neuropsychiatric interview (MINI) version 6.0. [12] The MINI is a short diagnostic structured interview for mental disorders, designed to allow administration by non-specialists.

Recruitment of participants
We used strati ed random sampling to recruit a sample comprising 80% outpatients and 20% inpatients, re ecting the service case mix. [11] Patient and public involvement A community panel comprising patients, caregivers and advocacy group members ensured community, patient, and public involvement. The panel reviewed and piloted the planned survey questionnaire and advised on its feasibility.

Data collection
We conducted a face-to-face survey using tablets (Qualtrics, Provo, UT) to collect information about physical disorders, mental health, health risk behaviours, health-related quality of life, health-risk behaviour advice and healthcare utilisation using, wherever available, validated instruments as described below. The survey was translated into Urdu and Bangla. Interviewers (including males and females to accommodate participant preference) used regional dialects where required, consistent with usual clinical practice in these settings. Data were collected between July 2019 and March 2020, when recruitment was halted due to the COVID-19 outbreak.

STEPwise approach to surveillance (STEPS)
We used the WHO STEPwise approach to Surveillance of NCDs (STEPS) instrument Version 3.2 to collect information about NCDs, associated risk factors and behaviours, access to physical healthcare and health-risk modi cation advice. [13] STEPS is an international standardised tool that has already been translated, used and validated in the general population in Bangladesh and Pakistan, and therefore allows comparisons with the general population within and between countries. [14,15] The STEPS survey includes the use of show-cards with culturally relevant examples used to aid respondents in classifying health-risk behaviours. Categorisation of health conditions and risk behaviours followed the WHO guidelines. [16] The STEPS module for NCDs was used to ask participants about medically-diagnosed type-2 diabetes, raised blood pressure, heart disease, and hypercholesterolemia, and treatments advised by a health worker for these conditions (such as medication and dietary, weight management, smoking cessation, or physical activity advice). Questions about lung disease, hepatitis B, C, syphilis, tuberculosis (TB) and human immunode ciency virus (HIV) (which are not part of the STEPS survey) were asked in the same format as for the other chronic physical conditions.

Health-risk behaviours
Current or past use of smoked or smokeless tobacco was recorded.
[16] The alcohol module was used to categorise participants into lifetime abstainers, abstainers in the past 12 months and current users of alcohol;[16] and the diet module was used to record the number of days that respondents consumed fruit and vegetables in a typical week, the number of servings consumed on average per day, and adherence to the WHO recommendations of at least 5 fruits and vegetables per day. [17] The physical activity module was used to record activity for transport purposes (e.g. walking, cycling), vigorous and moderate activity at work, and vigorous, moderate activity in leisure time, time spent sitting and adherence to the WHO recommendations of < 600 metabolic equivalents minutes/week. [18] In addition, risk behaviours related to sexually transmitted diseases, including multiple sexual partners, unprotected sexual contact, and use of injectable drugs, were assessed using three questions adapted from the 10-item HIV risk Screening Instrument. [19,20] Physical measurements Blood pressure (BP) was taken using an automated blood pressure measuring instrument (OMRON®) following instructions in the WHO STEPs surveillance manual); the average of the second and third readings was used for analysis.
[16] High blood pressure was de ned as a measurement of >140/90 mmHg. [16] Height, weight, and waist circumference were measured for all participants except pregnant women. All measurements were taken in duplicate and the average of the two values was calculated, following the protocols set out in the WHO STEPS surveillance manual.
[16] We calculated the Body Mass Index (BMI) and classi ed participants using the WHO classi cation, namely underweight (BMI<18.49), normal weight (BMI= 18.5-24.9), overweight (BMI= 25-29.9), obesity (BMI≥30). Abdominal obesity was de ned as a waist circumference of (≥ 94 cm) for males and (≥ 80 cm) for females. [16] Mental health In addition to administering the MINI, we collected information relevant to the SMI diagnosis, including duration of illness and type and duration of treatments. The Patient Health Questionnaire (PHQ-9) was used to measure the severity of depressive symptoms, [21] and the Generalized Anxiety Disorder-7 (GAD-7) for severity of anxiety symptoms. [22] Health-Related Quality of life The EQ-5D-5L was used to measure health-related quality of life. [23] We used the Urdu and Bangla validated versions, provided by EuroQol.

Blood Tests
A blood sample was taken from consenting participants for: haemoglobin, glycated haemoglobin (HbA1c), lipid pro le, thyroid function tests, liver function tests and creatinine. The cut off for high HbA1c was according to the WHO de nition ≥6.5%. [24] The prevalence of high total triglycerides was de ned as ≥ 180 mg/dl, [25] high serum cholesterol was de ned as LDL≥100 mg/dl. [25] Anaemia was de ned according as haemoglobin ≤13 g/dl for males and ≤12 g/dl for females. [26] Statistical analysis Quantitative data were summarised using descriptive statistics with means, standard deviation and 95% con dence interval for continuous data and counts, percentages and 95% con dence interval for categorical data. Overall, and by site we described the: 1) prevalence of chronic physical conditions.
2) prevalence of risk factors (obesity, hypertension, hypercholesterolemia) and risk behaviours (poor diet, physical inactivity, tobacco and alcohol use).
3) severity of common mental disorder symptoms (anxiety, depression) and health-related quality of life.

4) access to treatment for physical conditions and health risk modi cation advice
To compare our ndings with those in the latest STEPS reports from Bangladesh [15] and Pakistan, [14] we calculated weights by comparing the sex and age distribution reported in these STEPS surveys with the distribution in our data. Weights were applied using the complex sample module in SPSS 26.0 and we calculated the odds of people with SMI of having a NCD, related risk factors, engaging in health risk behaviours and being screened, treated and receiving risk modi cation advice compared with the STEPS survey participants in Bangladesh and Pakistan. [14,15] Results were presented as odds ratios from cross-tabulations of STEPS and weighted survey data. Signi cance levels were adjusted via Bonferroni correction for multiple hypothesis testing (adjusted level p < 0.006). [27] Results Between July 2019 and March 2020, 2,757 people with SMI were approached in the two sites and 2,344 (85.0%) participated in the survey (1422 in Bangladesh and 922 in Pakistan). Out of those who did not participate, 368 were ineligible, 36 refused and 9 left before completing the survey (Fig. 1).
Participant characteristics are shown in Table 1. On average, 60.5% of the sample was male, and the mean age was 34.2 years -the  Physical disorders, risk factors, health risk behaviours and healthcare Tables 2 and 3 summarise ndings for the prevalence of physical disorders, risk factors and healthcare use.  .6] § people that self-reported not to have type-2 diabetes, hypertension and hypercholesterolemia or not tested but found positive with the test performed during the IMPACT survey. ‡Data not reported due to low numbers for statistical disclosure control. Con dence intervals were calculated using bootstrap sampling procedure (n = 1000) for binomial and continuous variables, and using Goodman's method for multinomial proportions. ¶Includes type-2 diabetes, hypertension and hypercholesterolemia Eight percent of participants had type-2 diabetes (self report of clinician diagnosis or those with HBA1c ≥ 6.5%), 1.3% had chronic respiratory disorders and 3.6% had cardiovascular diseases, 3.1% were diagnosed with tuberculosis and 2.6% with chronic hepatitis.
Overall, 43.4% participants had overweight or obesity; most women (73.3%) and a high proportion of men (25.6%) had a high waist circumference. Underweight was also prevalent in 8.6% with a similar distribution in both countries.
Almost a quarter (24.7%) either reported a diagnosis of hypertension or had high measured blood pressure (≥ 140/90mmHg), 16.5%  reported tobacco use in Bangladesh and Pakistan, respectively. Around half of participants did not meet the WHO recommendations for physical activity (of 600 metabolic equivalents minute/week); and 84.5% of the participants in Bangladesh, and 91.0% in Pakistan reported not to meet the WHO recommended levels of fruit and vegetable intake (at least 5 servings). Only 1.0% of males and 0.1% of females reported consuming alcohol in the last month. Less than 13% of the sample reported to have more than 2 sexual partners in the last 10 years. Table 3, only 63.7% of the participants had been previously tested for any NCDs or NCD risk factor; 60.2% for hypertension, 27.3% for type-2 diabetes, and 7.3% for hypercholesterolemia. In general, a low proportion of participants in both Bangladesh and Pakistan received treatment for physical conditions or to address risk factors. Of those with self-reported NCD or a NCD risk factor, only 43.5% reported receiving related treatment or health-risk modi cation advice. The provision of relevant treatment was highest in those reporting type-2 diabetes (68.8%, 95%C.I.=58.5 to 77.6), followed by hypertension (41.9%, 95%C.I.=36.6 to 47.6) and hypercholesterolaemia (32.8%, 95%C.I.=21.6 to 41.6). Only 35.0% received any type of advice to modify health-risk behaviours; among those who consumed tobacco, only 25.2% had been advised to quit.

Comparison between people with SMI and the general population (STEPS survey)
The results for the comparisons between our data and country STEPS reports are summarized in Table 4.  [15] report and STEPS 2014 Pakistan [14] report, data from the SMI survey were weighted by age and gender according to the distribution of the STEPS report, N/A: Not available, ‡ After Bonferroni correction for multiple testing the p < .05 signi cance level was corrected to p < .006. ‡ Blood glucose ≥ 126mg/dl for the STEPS survey and HbA1c ≥ 6.50% for the SMI survey § People that self-reported not to have the condition or not previously tested but tested positive in assessments performed for the IMPACT survey or the STEPS survey in Bangladesh. ¶I Data not reported due to low numbers for statistical disclosure control.

Discussion
This is the rst multi-country study from South Asia to report on physical multimorbidity, health risk behaviours and access to related healthcare in people with SMI. We found a high prevalence of physical health conditions, primarily NCDs and related risk factors. We also found that people with SMI were more likely to have NCDs and NCD risk factors (overweight/obesity, hypertension, hypercholesterolemia) and engage in some health risk behaviours (tobacco use) but less likely to receive risk modi cation advice than the general population. Many people with SMI in our sample reported they had never been tested or screened for NCDs or NCD risk factors despite the well-established link between SMI and cardiometabolic conditions. [4,5] Moreover a large proportion of people with type-2 diabetes, hypertension and hypercholesterolaemia had not been previously diagnosed and these conditions were only detected on testing during the survey. Most had not received appropriate treatment and risk modi cation advice for their physical health. Therefore, even in the two major specialist mental health institutes included in our survey, most people with SMI failed to receive adequate screening, prevention and management of NCDs and NCDs risk factors.
The nding that people with SMI are more likely to have NCD risk factors compared with the general population extends previous ndings for example for obesity,[28] hypercholesterolemia, [29] and decreased physical activity. [30] Importantly it should be noted that psychotropic medication may contribute to some of these adverse risks. [31] Almost all survey participants were prescribed antipsychotics, which are associated with tiredness and sedation, an increased risk of obesity, and adverse effects on glucose and lipid metabolism.
In Pakistan, we found a higher prevalence of tobacco use in people with SMI compared with the general population. This is consistent with other studies in people with SMI, [32] where tobacco use has been associated with a greater susceptibility to addiction because of a higher subjective experience of reward and an attempt to self-medicate to mitigate anxiety and depressive symptoms. [33,34] Unexpectedly the opposite was found in Bangladesh. This may be because the STEPS survey for Bangladesh reported an unusually high estimate of the prevalence of tobacco use (70% in men). The more reliable Global Adult Tobacco Survey [35] for the same period reported a prevalence of 58% in th e same group that is closer to the gures reported in our study.The low observed prevalence of alcohol use in both men and women is similar to the STEPS survey reports, [14,15] and is likely to be explained on the basis of religious proscription.
Despite the high prevalences of overweight/obesity, hypercholesterolemia, hypertension and tobacco use, health-risk modi cation advice was provided to less than one quarter of people with SMI and we found that the odds of receiving such advice was lower in people with SMI than in the general population in both countries. Similar treatment gaps have been reported in high income countries.
[36] Although psychiatrists are trained in motivational interviewing, there are attitudinal barriers that make mental health professionals reluctant to engage with patients about their tobacco use. [37] Moreover, misconceptions about potential side-effects of tobacco cessation medication, unfounded fears of exacerbating depressive symptoms following quitting and low expectations of patients' motivation or ability to stop smoking are additional barriers.
[38] On the other hand, there is high quality evidence from high income countries about both the effectiveness and cost bene ts of smoking cessation interventions in people with SMI. [39] Such approaches need to be adopted in South Asia, where tobacco use is common.
Similarly, lifestyle interventions have shown promise in reducing weight and improving metabolic risk factors and are recommended as an essential part of the management of SMI in these countries. [36] An important study nding is the high proportion of participants with moderate or severe depressive and anxiety symptoms. In Pakistan this was higher than for Bangladesh, perhaps related to the higher proportion of participants in the depression with psychosis category in the Pakistan sample. Depression and anxiety are independently linked with NCDs, therefore addressing these may help to prevent and better manage NCDs in people with SMI.
The prevalence of TB was three times higher than in the general population. [40] This is consistent with previous ndings and clustering of TB risk factors reported in people with SMI. [41] In contrast, the prevalence of HIV [42] and Hepatitis B and C [43] were similar to those reported in the general population, a surprising nding considering the several risk factors for blood-borne viruses that have been reported to cluster in people with SMI. [44] While most of the comparisons between people with SMI and the general population are in line with clinical expectations and previous ndings, [28][29][30] there were some anomalous results. These include the lower odds of people with SMI with a self-reported clinical diagnosis of type-2 diabetes, hypertension and hypercholesterolemia in Bangladesh. This may be due to 'diagnostic overshadowing' where the presence of a mental disorder means clinicians do not look for physical health problems, or failure to recall such diagnosis by patients. The lower education and socioeconomic levels for participants from Bangladesh (compared with Pakistan) may have contributed to the latter. [45,46] We report ndings from the rst large-scale effort to document physical multimorbidity in people with SMI attending specialist services in two South Asian countries. We used standardised tools for data collection (i.e. STEPS, EQ-5D, PHQ9, GAD7) that allowed us to compare our ndings with those in the general population. Data were collected by trained researchers having experience of working with this population.
Finally, we gathered objective data on physical conditions (including blood tests) and report on both previously diagnosed and undiagnosed conditions.
Of the several limitations that need to be mentioned, the rst is that due to COVID-19 the participant sample was smaller than the originally planned sample of 1500 per site. Therefore, our ndings need to be interpreted with caution. Second, while we have used ndings from studies in the general population to compare and discuss our ndings, caution needs to be exercised in such comparisons since the analyses were only adjusted by sex and age. Moreover, we need to be mindful of the time lag between these studies, during which a number of parameters of interest might have changed.
Third, we relied on blood results from each mental health institution's laboratory but we did not standardise these tests between laboratories.
Fourth, since the sample was drawn from tertiary care, the ndings may not be representative of the total SMI population in each country.
However, unlike mental health services in high-income countries, tertiary care services in South Asian countries accept self-referral without the need for primary or secondary care referral and often function as 'the rst port of call' for people with severe mental illness. They also attract patients from both urban and rural areas. Therefore, the study population is likely to be similar to the overall population of people with SMI in these countries.
The high prevalence of physical health conditions and health risk behaviours in SMI compared with the general population and their underdetection even in specialist centres merits attention to improve early identi cation, prevention and management, in line with international recommendations and guidance. There is a need to periodically audit and improve adherence to these recommendations. In view of challenging resource limitations, interventions to address health risk behaviours that are brief, and delivered by non-specialist personnel need to be tested in these settings. Integration of physical health care with mental health care that has been envisioned at all levels of mental health care delivery needs to be actioned and scaled up. [9] Representative community-based studies may further answer questions related to regional differences in physical health conditions and health risk behaviours.

Conclusion
People with SMI in South Asia have a high prevalence of NCDs, which may be due to the associated clustering of several health risk factors and behaviours in this population. There is an unmet need to address physical multimorbidity in people with SMI in South Asia. Policy makers and healthcare professionals working with people with SMI need to recognise the extent and importance of physical multimorbidity in this vulnerable group and prioritise prevention, screening and treatment of NCDs in people with SMI. Trained researchers provided verbal and written study information to patients and their relatives or carergivers, highlighting that participation was voluntary, the decision would not affect care, and consent could be withdrawn at any stage without providing a reason. Written consent was obtained (a thumbprint was accepted where a signature could not be provided). No assessments were conducted where the patient appeared reluctant, even if consent had previously been obtained.

Abbreviations
The study was approved by the ethics committees of the Department of Health Sciences, University of York, UK; Centre for Injury Prevention and Research, Bangladesh; and the National Bioethics Committee, Pakistan. All study procedures complied with legislation and guidance for good practice governing the participation in research of people who may lack capacity (Mental Capacity Act (UK) 2005). Participants did not receive nancial inducements to participate, but results of physical health measurements and blood tests were shared with them and with the treating clinician.

Consent for publication
Not applicable Availability of data and materials The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.