In the present study, we identified four distinct multimorbidity latent classes among a large sample of community-dwelling adults aged 18 and older and determined how these latent classes are associated with 11-year all-cause mortality, risk factors and HRQoL. To our knowledge, this is the first multimorbidity latent class cohort study of the general population in Malaysia and South East Asia. We used LCA to reduce the complexity of data and identify the multimorbidity subgroups, with a rigorous statistical approach that employs probability-based classification methods, enabling the selection of the most optimal number of classes of chronic conditions based on various diagnostic tests [28] and clinical meaning of the resulting multimorbidity patterns. There are several important learning points highlighted by our study findings. First, the multimorbidity subgroups identified are distinct. The multimorbidity patterns identified were musculoskeletal, mobility and sensory disorders; cardiometabolic diseases; and complex multimorbidity, respectively, in this population. Secondly, there is clear variation in socio-demographic characteristics and the associated adverse health outcomes between the multimorbidity classes, indicating which group of individuals were more vulnerable to multimorbidity and the consequences. We found that age, sex, ethnicity, education level and employment status were associated with distinct multimorbidity patterns. Those in the complex multimorbidity group had a significantly lower HRQoL than other groups across all aspects studied, including physical health, psychological, social relationship and environment domains. Thirdly, the mortality risk differed greatly across the multimorbidity patterns. Compared with the relatively healthy group, the complex multimorbidity group exhibited the highest mortality risk even after adjustment of age and other confounders in the analysis, followed closely by the cardiometabolic group and the musculoskeletal, mobility and sensory disorders group. This also implies that the combined effects of the number of chronic conditions, along with the type and severity of those conditions experienced might be proportional to the mortality risk. Fourthly, our study contributes rigorous evidence about multimorbidity in a middle-income country, aligning with findings from previous HIC studies.

The four multimorbidity classes identified in our study reflected the complexity of multimorbidity measurement and conceptualization. It is difficult to compare our findings directly with previous studies given the considerable methodological differences including study settings, disease types and counts, socio-demographic factors and the statistical methods adopted. Nonetheless, the multimorbidity patterns we identified generally agree with prior work, although limited relevant studies are available in LMICs. Such similarities may indicate that chronic conditions aggregate as they share common aetiology and underlying risk factors. For instance, we found that cardiovascular diseases (i.e. heart disease and stroke in our study) and metabolic diseases (i.e. hypertension, diabetes and obesity in this study) were very likely to co-occur, in agreement with the clinical nature of the chronic conditions, coupled with evidence from other studies [29, 30]. Although the mechanisms behind the combination of arthritis, chronic pain, and physical mobility, hearing and vision problems in this study remain unclear, a similar trend appears among these impairments—all related to physical disability and mobility limitations, which deserve more attention in future studies. Previous studies conducting LCA to describe multimorbidity patterns [31,32,33,34] have yielded similar patterns as our findings. All studies reported a ‘relatively healthy’ latent class (or a ‘minimal disease’ class), which consisted primarily of individuals with ≤ 1 medical condition. Likewise, these studies also reported a complex multimorbidity class with high prevalence in chronic conditions of multiple organ systems. Of the patterns reported, cardiovascular, metabolic, musculoskeletal and respiratory conditions were present in most studies in various combinations (e.g. cardiometabolic class consistent with the pattern we obtained in the present study). Despite using varied datasets and including different numbers of chronic conditions, LCA produced similar results for many studies, showing that some conditions may be more likely than others to co-occur and cluster together. These clusters importantly suggest practical implications for a shift from individual-disease approaches to a more integrated and holistic approach to health service delivery, focusing on clusters of multiple long-term conditions, to potentially improve outcomes and reduce the burden on healthcare systems. Based on our multimorbidity patterns identified, further research to better understand how these diseases interact may be the next helpful step. On the other hand, it is notable that the depression symptom levels did not vary much across the multimorbidity classes. This may be due to the measurement of depressive symptoms using the DASS-21 instrument based on a short recall period of 1 week, a methodological shortcoming that potentially limits the robustness of this result yet an assessment unavoidable in large HDSS data collection.

Tackling social determinants of multimorbidity is necessary to redress the rising burden of multimorbidity in disadvantaged populations. We found significant variations in socio-demographic composition between the multimorbidity groups. Our findings add to existing knowledge about age as a consistent and strong correlate of multimorbidity, across all multimorbidity patterns. Being female, having a lower educational level and not working were also more likely to belong to most of the multimorbidity groups than the relatively healthy group, albeit the significance varied across classes. Individuals with lower education levels were often associated with physically demanding blue-collar jobs, likely leading to higher rates of musculoskeletal problems. Furthermore, lower education levels, frequently linked to lower socioeconomic status, can result in poorer health outcomes and limited access to healthcare, health information and preventive care, thereby increased the risk of complex multimorbidity. Although this may seem counterintuitive, those with higher education levels were likely to engage in sedentary white-collar jobs, which may explain the increased risk of cardiometabolic diseases. Employment provides structured routines, social interactions and better access to healthcare through employment benefits, contributing to better health outcomes among those engaged in employment or non-retired activities, which may explain the consistent findings across latent classes among the study population. In our study, marital status appears to impact participants’ health through social support mechanisms. Married individuals might have better social support, leading to better health outcomes and access to care and, hence, a lower risk of musculoskeletal, mobility and sensory disorders. Single individuals might be younger or experience less stress related to marital responsibilities, reducing the risk of cardiometabolic conditions. The identified risk factors may support better integration of comprehensive management of multimorbidity in interventions and resources for optimal population health and addressing health disparities, targeting older adults, those with lower education levels and lower-income groups. Future research warrants a focus on the underlying pathogenesis connecting these chronic health conditions and the shared risk factors.

Given the potential of multimorbidity to erode financial security and compromise self-care capacity through the burden and complexity of managing multiple long-term conditions, the impact of multimorbidity patterns on quality-of-life outcomes is important. It comes as no surprise that the complex multimorbidity group has the poorest HRQoL compared to other groups across all aspects studied, including physical health, psychological, social relationship and environmental domains. This highlights the substantial impact of the most complex combination of chronic conditions on various aspects of well-being—as diseases clustered and increased in severity, so did the reduced quality of life, functional impairment and deteriorated psychological well-being, social relationships and the environment, in line with other studies [35]. These findings shed light on useful implementation of integrated mental and physical healthcare, and the importance of facilitating development of tailored preventive interventions and treatment for multimorbidity, particularly targeting the earlier stages of diseases.

In the Cox proportional hazards regression models, mortality risk was significantly higher across the multimorbidity groups in reference to the relatively healthy group over 11 years. Interestingly, although individuals in the musculoskeletal, mobility and sensory disorders group had a greater prevalence of chronic conditions (arthritis, chronic pain, physical mobility problems, hearing impairment and vision impairment) than the relatively healthy group, they had the lowest mortality risk among the three disease groups despite adjusted for confounders. This is in concordance with previous studies carried out in a cohort of 7197 community-dwelling adults aged 65 years and older in the USA [31], which found comparable mortality risk between their osteoarticular group (comprised of individuals with arthritis and osteoporosis) and minimal disease group, when stratified by their participants’ frailty status. The complex multimorbidity group in the present study has the strongest magnitude of association with mortality. A plausible explanation is the interplay between age and the most complex treatment and self-management alongside poorer health status in the complex multimorbidity group. Despite the inclusion of younger participants (≥ 18 years old), our sample consisted mostly of older individuals among those with multimorbidity, and the complex multimorbidity group had the highest overall age profile. This underscores the importance of addressing and prioritizing multimorbidity in clinical care to mitigate mortality risk, prioritizing the complex multimorbidity cohort.

Our study shows that people with complex multimorbidity, characterized by multiple severe chronic conditions, specifically overlapping musculoskeletal, mobility and sensory disorders, and cardiometabolic diseases, experienced the poorest HRQoL and the strongest mortality risk. This is particularly novel from a clinical perspective as it highlights the compounded detrimental effects of these chronic conditions, emphasizing the need for targeted care plans and resource allocation. Clinicians can use these insights to develop interventions that address the unique needs of those people with the highest number and severity of these chronic conditions. Our findings can also guide healthcare policymakers in optimizing resource distribution to high-risk groups, particularly in LMICs, where resources are often limited and need to be used strategically. Our findings advocate for integrated care models tailored to the needs of those with complex multimorbidity that ensure early detection and preventive measures, such as lifestyle modifications, regular screenings and proactive management to prevent the progression to more severe health states, for effective care and improved quality of life. This study adds to the body of literature by elucidating latent classes of multimorbidity and the consequences in Malaysia and South East Asia, and providing practical insights for improving patient outcomes and informing healthcare strategies, particularly beneficial for those with the most relevant complex and severe disease profiles.

There are some important limitations. First, we used 13 self-reported physician-diagnosed chronic conditions commonly available in epidemiological and clinical studies for LCA. On the ground that specific multimorbidity patterns may be sensitive to the number of chronic conditions [36], chronic conditions not evaluated in this study and our inclusion of additional chronic conditions might have yielded different multimorbidity patterns. In our study, information about cancer was not available. Conversely, other highly prevalent conditions such as physical mobility problems, obesity and depressive symptoms were considered in our study, although they are commonly omitted in other research. Second, although misclassification of individuals assigned to each latent class is reasonable because multimorbidity class membership is determined based on predictive probability calculated from LCA, there may still be some degree of uncertainty associated with latent class membership in some cases. Even though each class seemed to show clinically distinct patterns of chronic conditions, results should be interpreted with caution. We have shared clinical interpretations for these latent class memberships; future research may help to provide further certainty about the nature of these latent classes. One possibility is to extend the analysis to populations in other South East Asian countries or beyond. Third, our data on chronic conditions did not allow us to assess changes in the patterns of multimorbidity over time. It was also not possible to draw temporal associations with socio-demographic factors due to the limitation of datasets. Lifestyle risk factors which may be associated with the development of multimorbidity could not be included; future investigation may enhance understanding and offer opportunities for intervention. Lastly, it should be acknowledged that the depressive symptoms assessed by the DASS-21 instrument are not clinical diagnoses of depression. This may explain the lack of variation in depressive symptoms across the LCA groups and potentially their limited informativeness for constructing multimorbidity.

The main strengths of our study included a relatively large sample size provided enough numbers for LCA (thus leaving out bias towards conditions of specific prevalence or problems with low count rates), diverse and multi-ethnic representation, and a wide range of diseases and socio-demographic covariates studied. Another key strength of the present study from the modelling perspective is the use of a more objective LCA algorithm based on a rigorous statistical basis. We were able to identify distinct latent classes of the chronic health conditions in a robust manner beyond chance that are strongly associated with multimorbidity. Furthermore, the use of our health and demographic surveillance system data drawn from community-dwelling individuals allowed capturing a large number of participants who are not restricted to certain pre-requisites, such as belonging to a specific facility (e.g. solely hospitalized patients) or a specific health insurance company. Moreover, our databases with extensive and regular follow-up and up-to-date longitudinal death data make it a powerful surveillance tool for studying mortality in relation to multimorbidity. Incorporating this temporal dimension, spanning up to 11 years of mortality data, enriched our analysis by allowing associations to be directional, thereby provided stronger evidence for causality. We have also explicitly included relevant factors particularly the socio-demographic characteristics, thus, allowed the isolation of the effect of the survivorship bias. Most previous studies were conducted among the elderly, where multimorbidity is more likely to occur; our study bridged this gap by providing crucial evidence on population-based multimorbidity patterns across a broad age span comprising young, middle-aged and older persons. Our findings support proposals that interventions to improve outcomes in multimorbidity may be more appropriately targeted on the specific clusters uncovered. Our findings also underline the need for tailored approaches for higher-risk patient groups with important differences in socio-demographic characteristics and mortality risk.