Identifying dietary consumption patterns from survey data: a Bayesian nonparametric latent class model.

Dietary assessments provide the snapshots of population-based dietary habits. Questions remain about how generalisable those snapshots are in national survey data, where certain subgroups are sampled disproportionately. We propose a Bayesian overfitted latent class model to derive dietary patterns, accounting for survey design and sampling variability. Compared to standard approaches, our model showed improved identifiability of the true population pattern and prevalence in simulation. We focus application of this model to identify the intake patterns of adults living at or below the 130% poverty income level. Five dietary patterns were identified and characterised by reproducible code/data made available to encourage further research.

Bayesian estimation methods for survey data with potential applications to health disparities research.

Understanding how and why certain communities bear a disproportionate burden of disease is challenging due to the scarcity of data on these communities. Surveys provide a useful avenue for accessing hard-to-reach populations, as many surveys specifically oversample understudied and vulnerable populations. When survey data is used for analysis, it is important to account for the complex survey design that gave rise to the data, in order to avoid biased conclusions. The field of Bayesian survey statistics aims to account for such survey design while leveraging the advantages of Bayesian models, which can flexibly handle sparsity through borrowing of information and provide a coherent inferential framework to easily obtain variances for complex models and data types. For these reasons, Bayesian survey methods seem uniquely well-poised for health disparities research, where heterogeneity and sparsity are frequent considerations. This review discusses three main approaches found in the Bayesian survey methodology literature: 1) multilevel regression and post-stratification, 2) weighted pseudolikelihood-based methods, and 3) synthetic population generation. We discuss advantages and disadvantages of each approach, examine recent applications and extensions, and consider how these approaches may be leveraged to improve research in population health equity.

Higher Imported Food Patterns Are Associated with Obesity and Severe Obesity in Tuvalu: A Latent Class Analysis.

Background: Tuvalu is a Pacific Island country within the small island developing states that has observed a significant and alarming increase in obesity rates over the past 40 years, affecting ∼60 %-70 % of the current population. Objectives: This study aimed to investigate the association between food patterns and the proportion of obesity in a Pacific Island country. Methods: The 2022 COMmunity-based Behavior and Attitude survey in Tuvalu (COMBAT) included 985 adults with complete data on sociodemographic information and the frequency of consumption of 25 common foods. A latent class analysis determined 4 food patterns. Bayesian multilevel logistic and linear regression models estimated the association between food patterns and the proportion of obesity [body mass index (BMI) ≥30 kg/m2], severe obesity (BMI ≥40 kg/m2), and weight (kg), adjusting for potential confounders and accounting for clustering by region. Results: The latent class analysis revealed 4 food patterns with an entropy of 0.94 and an average posterior probability of class assignment for each individual of 0.97, described as follows: 1) local: locally produced foods with moderate food diversity (proportion of individuals = 28 %); 2) diverse-local: local with greater food diversity (17 %); 3) restricted-imported: more imported with restricted diversity (29 %); and 4) imported: heavily imported with high diversity (26 %). Compared to those following the diverse-local pattern, the odds of having obesity were greater for those classified with the imported pattern [odds ratio (OR): 2.52; 95 % credible interval (CrI): 1.59, 3.99], restricted-imported pattern (OR: 1.89; 95 % CrI: 1.59, 3.99), and local pattern (OR: 1.54; 95 % CrI: 0.94, 2.50). Similar trends were observed for severe obesity while body weight was positively associated with both restricted-imported and imported food patterns. Conclusions: The high consumption of imported foods, together with the low consumption of plant-based foods and protein-rich foods, could be a relevant modifiable lifestyle factor explaining the high levels of obesity and severe obesity in Tuvalu, a Pacific Island country.

Exploring the link between home garden use and severe obesity: Insights from a nationwide survey in Tuvalu.

Background: Obesity is prevalent and increasing but understudied across Pacific Islanders. Tuvalu is a South Pacific country with a high obesity rate and faces multiple threats of food insecurity. Home garden serves as a sustainable food source and can be a possible intervention for the obesity pandemic in Tuvalu. This study investigated Tuvaluans' home garden use and obesity, and explored factors associated with increased use of home gardens. Methods: We conducted a nationwide, cross-sectional study in Tuvalu during 2022. Structured questionnaires were administered during the in-person interviews, and trained interviewers measured the height and weight of each participant. The association between home garden use, obesity and severe obesity were tested with univariate and multivariable logistic regression. We also applied overlapping weights to balance the distribution of baseline demographic factors. Results: The average body mass index was 34.87 kilogrammes (kg) / square metre (m2) among the study population of 1024 adults (630 from Funafuti and 394 from other islands in Tuvalu). Overall, people having home gardens was associated lower odds for severe obesity compared to those without a home garden in overlap weighting models (odds ratio (OR) = 0.946, 95% CI = 0.897-0.997, P = 0.039) and the association was stronger in Funafuti (OR = 0.927, 95% CI = 0.866-0.991, P = 0.027) than in the outlying islands (OR = 0.967, 95% CI = 0.889-1.052, P = 0.435). Furthermore, increased age was positively associated with having a home garden in Funafuti, and smoking showed an inverse association. Conclusions: Having a home garden is associated with lower odds of severe obesity in Tuvalu, and the association is stronger in Funafuti. Smokers are less likely to have home gardens, and increased age is positively associated with having home gardens. These findings promote more home garden utilisation and provide evidence for targeted interventions in Tuvalu.

Modeling the relative risk of incidence and mortality of select vaccine-preventable diseases by wealth group and geographic region in Ethiopia.

Immunization is one of the most effective public health interventions, saving millions of lives every year. Ethiopia has seen gradual improvements in immunization coverage and access to child health care services; however, inequalities in child mortality across wealth quintiles and regions remain persistent. We model the relative distributional incidence and mortality of four vaccine-preventable diseases (VPDs) (rotavirus diarrhea, human papillomavirus, measles, and pneumonia) by wealth quintile and geographic region in Ethiopia. Our approach significantly extends an earlier methodology, which utilizes the population attributable fraction and differences in the prevalence of risk and prognostic factors by population subgroup to estimate the relative distribution of VPD incidence and mortality. We use a linear system of equations to estimate the joint distribution of risk and prognostic factors in population subgroups, treating each possible combination of risk or prognostic factors as computationally distinct, thereby allowing us to account for individuals with multiple risk factors. Across all modeling scenarios, our analysis found that the poor and those living in rural and primarily pastoralist or agrarian regions have a greater risk than the rich and those living in urban regions of becoming infected with or dying from a VPD. While in absolute terms all population subgroups benefit from health interventions (e.g., vaccination and treatment), current unequal levels and pro-rich gradients of vaccination and treatment-seeking patterns should be redressed so to significantly improve health equity across wealth quintiles and geographic regions in Ethiopia.