Prevalence and patterns of multimorbidity in the Jamaican population: A comparative analysis of latent variable models

Abstract

Background Evidence suggests that the single-disease paradigm does not accurately reflect the individual experience, with increasing prevalence of chronic disease multimorbidity, and subtle yet important differences in types of co-occurring diseases. Knowledge of multimorbidity patterns can aid clarification of individual-level burden and needs, to inform prevention and treatment strategies. This study aimed to estimate the prevalence of multimorbidity in Jamaica, identify population subgroups with similar and distinct disease profiles, and examine consistency in patterns identified across statistical techniques. Methods Latent class analysis (LCA) was used to examine multimorbidity patterns in a sample of 2,551 respondents aged 15-74 years, based on data from the nationally representative Jamaica Health and Lifestyle Survey 2007/2008 and self-reported presence/absence of 11 chronic conditions. Secondary analyses compared results with patterns identified using exploratory factor analysis (EFA). Results Nearly one-quarter of the sample (24.1%) were multimorbid (i.e. had �2 diseases), with significantly higher burden in females compared to males (31.6% vs. 16.1%; p<0.001). LCA revealed four distinct classes, including a predominant Relatively Healthy class, comprising 52.7% of the sample, with little to no morbidity. The remaining three classes were characterized by varying degrees and patterns of multimorbidity and labelled Metabolic (30.9%), Vascular-Inflammatory (12.2%), and Respiratory (4.2%). Four diseases determined using physical assessments (obesity, hypertension, diabetes, hypercholesterolemia) were

Social determinants of multimorbidity in Jamaica: application of latent class analysis in a cross-sectional study

“…LCA results have been previously described [ 16 ]. To summarize, indices and probability plots collectively supported the four-class model as the optimal baseline model, based on identification (i.e., the model converged on the same solution the majority of the time), entropy (i.e., distinctiveness of the latent classes) and meaningful interpretation.…”

“…This is the first study to examine the social patterning of multimorbidity in Jamaica. Building on identified latent classes [ 16 ], this study shows that beyond differences in the type and number of diseases comprising multimorbidity patterns, there are also important social differences associated with class membership. Specifically, the study found that advancing age and recent health service use were significantly associated with all multimorbidity patterns.…”

“…Indicators of multimorbidity class membership are all fully described elsewhere [ 16 ]. Briefly, selection of indicators for multimorbidity measurement was guided by criteria outlined in a 2011 systematic review [ 18 ], being limited to NCDs with the highest prevalence in this population (i.e., at least 1% in each sex) and grouping related conditions such as cardiovascular disease (i.e., heart disease, myocardial infarction, and circulation problems) and mental health disorders (i.e., depression, anxiety, psychosis, and other mental health problems) to enhance data quality.…”

“…The final list of conditions included hypertension, obesity, hypercholesterolemia, diabetes, asthma, arthritis, cardiovascular disease, mental health disorders, COPD, stroke, and glaucoma. Of note, obesity was included as a chronic condition given its recognition as a chronic disease requiring intervention [ 19 ], its inclusion in previous studies of multimorbidity indices and patterns [ 7 , 18 ] and ability to corroborate findings in baseline LCA models with and without obesity, in our previous work [ 16 ].…”

“…However, no investigation of the social determinants of specific patterns or combinations of diseases has been undertaken to date in Jamaica. Recent work using latent class analysis (LCA) points to a high prevalence of multimorbidity in the Jamaican population, that is disproportionately borne by women, and comprises three distinct multimorbidity patterns (labelled Metabolic , Vascular-Inflammatory , and Respiratory ), in addition to a Relatively Healthy class characterized by little to no morbidity [ 16 ]. LCA provides a novel approach to reducing data complexity and has been widely used in prevention research to identify underlying subgroups of individuals characterized by multiple dimensions (e.g., with similar disease probability profiles) [ 17 ].…”