Bayesian multistate modelling of prefrailty and frailty: a microsimulation study with weight management interventions for prevention in Asian settings

BackgroundPre-frailty and frailty prevalence is increasing in the Western Pacific, resulting in substantial morbidity and mortality in older populations. Modelling frameworks are required to estimate the prevalence of frailty and potential impacts of ongoing population-level nutritional interventions. MethodsUsing a microsimulation sociodemographic model of 3,353,032 individuals from 1990 to 2050, and data from the Singapore Longitudinal Ageing Study 2 of 3,270 participants, we developed a Bayesian multistate model of robust, pre-frailty and frailty stages with estimated transition probabilities by age, ethnicity, and gender for each body-mass index (BMI) category. We then explored four scenarios where weight management interventions are applied that modify the annual distribution of underweight, normal weight, overweight, obese I and obese II individuals. FindingsBetween 2011 and 2050, the overall prevalence of pre-frailty and frailty increased from 44{middle dot}2% to 46.9%, and from 3{middle dot}2% to 11{middle dot}3% respectively. Reductions of 811 pre-frail individuals (95% CrI: 624-1,127) and 36,202 frail individuals (22,109- 41,124) were estimated when underweights shifted to normal weights, 5,787 (3,670- 7,707) and 55,777 (32,683-80,941) when obese II moved to obese I, and 22,045 (18,430-23,487) and 62,847 (40,165-91,517) when both groups shifted respectively. Total healthcare utilization decreases by 6{middle dot}9% (4.3%-8{middle dot}1%) with the latter intervention. InterpretationFrailty prevalence is projected to substantially increase by 2050 where large-scale weight management interventions could be utilised to avert cases of both pre-frailty and frailty in older individuals. FundingThis research was supported by the Population Health Metrics and Analytics project, the Ministry of Health and National Innovation Challenge (NIC Ageing), Healthy Longevity Catalyst Awards (HLCA) MOH-HLCA22Feb-0007.