Statistical Analysis

The main variable of interest was whether the cohort member died between March 2005 and December 2019. A stratified analysis has been performed for gender and age (Table 3). We described the distribution of deaths by individual characteristics of cohort members: participants, multimorbidity, number of chronic conditions (5 categories), age groups (3 categories), marital status (3 categories), regions area of residence (geographic regions, 6 categories), life course residence (urban or rural resident at age 10 to 12 and in 2005), education level (4 categories) and monthly income (4 categories, baht, US$1 was equal to 25 baht in 2005). We used Stata 16.0 to tackle the analysis.

Multivariate adjusted hazard ratios according to the interaction between multimorbidity and risk factors (Thailand, 2005 for baseline characteristics and 2019 for all-cause mortality).

AHRs*: Multivariate adjusted Hazard ratios from Cox proportional hazard models controlling for other socio-demographic factors and personal lifestyle elements (variables in Supplementary Table S2 in Supplementary File S1). AHRs bolding: The results of Multivariate hazard ratios (HR) were statistical significance (p < 0.05). p-trend bolding: The results of p-trend were statistical significance (p for trend <0.05).

Kaplan Meier survival curves display explain multimorbidity or not/different number of chronic conditions by baseline in 2005. End-point events were all-cause deaths until 2019. We used the log-rank test to test the statistical probability of observed difference in survival patterns according to having multimorbidity or not/different number of chronic conditions. Univariate and multivariate Cox proportional hazards models were used to examine the association between risk factors (including multimorbidity and the number of chronic conditions) and survival [23]. Multivariate Cox proportional hazards models were also used to evaluate the association between the interaction between multimorbidity and these independent variables, and survival. Hazard ratios (HR) and 95% confidence intervals (95% CI) were generated as the results [23]. For each set of associations, we adjusted other factors in Table 2 for confounders. For any missing data, our study assumed that they were missing at random. For all analyses, p for trend (p-trend) were used to test the linear trend and p values <0.05 were considered statistically significant. All statistical analyses were performed using Stata 16.0.

Hazard ratios according to number of multimorbidity or chronic conditions, and age (Thailand, 2005 for baseline characteristics and 2019 for all-cause mortality).

NA: Not available. AHRs*: Multivariate adjusted Hazard ratios from Cox proportional hazard models controlling for other socio-demographic factors and personal lifestyle elements (variables in Supplementary Table S2 in Supplementary File S1). AHRs bolding: The results of Multivariate hazard ratios (HR) were statistical significance (p < 0.05). p-trend bolding: The results of p-trend were statistical significance (p for trend <0.05).