Cross-sectional study on obesity classification and non-communicable disease (NCDs)

This dataset relates to a community-based cross-sectional study investigating obesity classification and selected non-communicable diseases (NCDs) among adults in Ashanti Mampong Municipality, Ghana. The study aimed to describe the distribution of body mass index (BMI) classifications and reported NCD diagnoses, examine their associations with sociodemographic and behavioural factors, and develop and compare statistical and machine-learning models for predicting reported NCD status. The study involved 421 adults recruited through non-probability sampling at community screening points across 18 communities in Ashanti Mampong Municipality. Data were collected using structured questionnaires and anthropometric measurements. Height and weight were measured to calculate BMI, while questionnaire responses provided information on sociodemographic characteristics, behavioural indicators, family history of NCDs, and previous health-professional diagnoses of hypertension, diabetes and heart disease. BMI classifications included normal weight, overweight, and obesity Classes I, II and III. The primary outcome was the reported diagnosis of at least one of the three selected NCDs. The research employed descriptive statistics, multivariable logistic regression and predictive modelling techniques. Three predictive models—logistic regression, least absolute shrinkage and selection operator (LASSO) logistic regression, and random forest—were developed and evaluated using a 70:30 training–test split and cross-validation. The study found that 61.5% of participants were overweight or obese, including 34.4% classified as obese. At least one selected NCD was reported by 34.2% of participants. Logistic regression and LASSO demonstrated comparable predictive discrimination and performed better than random forest on the held-out test dataset. The data provide evidence for investigating obesity classifications, reported NCD status, associated factors and predictive modelling approaches in a Ghanaian community setting. They may support further epidemiological research, methodological comparisons and independent validation studies. Because the study used non-probability sampling and a cross-sectional design, the findings should not be interpreted as representative prevalence estimates for the entire municipality or as evidence of causality. The predictive models estimate reported existing NCD status rather than future disease occurrence.