Meaning Behind the Metrics of Misery: Understanding Prevalence Estimates of Poor Mental Health in Two Samples of Older Rural Indonesians

Abstract

Background: Late life is typically accompanied by unique physical and mental health challenges. Fewer older people are diagnosed with mood or anxiety-specific disorders than their younger counterparts. However, older people score more highly than younger people on symptom screens indicating high levels of clinically relevant depressive, anxiety, and nonspecific psychological distress symptoms which cause high morbidity, mortality, disability, and poor quality of life. The unique presentation of late life psychiatric syndromes, such as depression and anxiety, remain largely unaddressed in existing psychiatric nosology and measurement techniques, as do depictions of depression and anxiety across diverse cultural contexts. Very few studies exist investigating either the descriptive epidemiology of depression and anxiety among older adults living in low-middle income countries (LMIC) or the unique challenges of mental health measurement in LMIC contexts. This dissertation contributes to this developing evidence base by providing a critical analysis of point prevalence estimates of depression, anxiety, and nonspecific psychological distress (distress) symptoms in two samples of Indonesian rural older persons. Methods: We enumerated greater than or equal to 60-year-olds in 12 Indonesian rural villages as part of the Ageing in Rural Indonesia Study in 2015/16 (N=2526; sample 1). We re-enumerated two of the 12 villages surveyed in 2015 in 2017 (N=536; sample 2). Depressive and distress symptoms were each measured using three scales: PHQ-8/9, CES-D, GDS, and K6, DQ5 and SRQ-20 respectively. Anxiety symptoms were evaluated with the GAD-7. Classical Test Theory and Item Response Theory were used to investigate the psychometric properties of symptom screens. We also undertook mixed effects modelling and Moderated Nonlinear Factor Analysis to identify sources of variability in prevalence estimates. Results: Commonly used cut points of short symptom screens used to approximate diagnostic depressive disorders produced estimates that typically lacked comparability (e.g., sample 2 point-prevalence 3.2%-39.9%). Psychometric analysis further identified mental health scales with better (PHQ-8/9, GAD-7, K6, DQ5) and poorer (GDS, SRQ) construct validity. Sources of variability in point prevalence estimates of depression, anxiety and distress symptoms were identified, and related to study design, cognitive ability, marital status, financial means, level of social support, lifestyle, and health related status. Pervasive non-invariance was identified in participant responses to scale items related to gender, literacy, and ethnicity. However, when modelled, measurement non-invariance did not substantially modify means. Females, respondents with lower literacy levels, and Batak and Sundanese sample villages had significantly higher levels of depression, anxiety, and distress symptoms. Conclusion: The practice of using existing mental health symptom screens combined with commonly used cut points as proxies for depression and anxiety in older rural Indonesians and other diverse populations should be avoided. Rigorous psychometric and diagnostic validation evidence should be ascertained. In the interim, better performing symptom screening tools (i.e., PHQ-8/9, GAD-7, K6, DQ5) may be used as measures of continuous symptom severity. Future research should focus on evaluating the distinctive and overlapping features of mental ill-health in specific subpopulations of Indonesians.