Mutambayi, M.Lala, M.Mbhele, Nhlonipho Errol2026-09-012026-09-012024-11Mbhele, N.E. (2024) Statistical comparative performance of logistic model and proc survey in the determination of risk factors related to high blood pressure in South Africa. Master of Biostatistics mini dissertation. Alice, South Africa: University of Fort Hare.http://hdl.handle.net/20.500.11837/5083Mini dissertationBackground: The study aims to investigate the most appropriate model for data analysis and presentation within the context of the healthcare sector in South Africa, with specific reference to hypertension. This study sets out to compare logistic regression analysis and the survey logistical model with the aim of identifying and presenting the most appropriate hypertension risk factors in the South African context. Methods: This study provides an analysis of secondary survey data, collected through the administration of the 2016 South African Demographic and Health Survey (SADHS). The survey was implemented by Statistics South Africa (Stats SA) with the South African Medical Research Council (SAMRC). As the nature of this study is comparative, it is appropriate to apply the identified models of healthcare analytics to a single set of data. As such, data collection comprises the acquisition of the SADHS results. The data analysis in this study comprises the application of the Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC) to the logistics regression model results, as well as the survey logistic model results, based on the models’ application to the SADHS data. This study undertook a full LOGISTIC regression application, using all potential predictors of hypertension. Following the identification of potential predictors, a Random Forest was then used to determine the reduced dimension of the data. Using the relevant variables that emanated from the Random Forest algorithm, the LOGISTIC and SURVEYLOGISTIC procedures were performed to assess factors associated with hypertension and compare the model fit. Results: Current age showed a modest but statistically significant positive association with hypertension (AOR = 1.05, 95% CI: 1.04; 1.06). Geographical location also played a crucial role, with individuals residing in the Western Cape (AOR = 3.27, 95% CI: 1.90; 5.65), Northern Cape (AOR = 2.57, 95% CI: 1.45; 4.56), Free State (AOR = 2.66, 95% CI: 1.53; 4.63), and Kwazulu Natal (AOR = 2.07, 95% CI: 1.26; 3.42) exhibiting significantly higher odds of hypertension compared to those in Limpopo. Moreover, socioeconomic status was a significant determinant, as individuals classified as poor (AOR = 0.43, 95% CI: 0.30; 0.61) or middle class (AOR = 0.63, 95% CI: 0.46; 0.87) had lower odds of hypertension compared to those classified as rich. Lifestyle factors also emerged as significant contributors, with non-smokers exhibiting lower odds of hypertension compared to smokers (AOR = 1.33, 95% CI: 1.05; 1.70), and individuals perceiving their own health as poor having substantially higher odds of hypertension (AOR = 2.90, 95% CI: 1.76; 4.77). Notably, intention to reduce salt consumption was associated with decreased odds of hypertension (AOR = 0.61, 95% CI: 0.47; 0.78). Additionally, a history of CVA (AOR = 0.43, 95% CI: 0.19; 0.98) or diabetes (AOR = 0.23, 95% CI: 0.15; 0.35) were associated with lower odds of hypertension. These findings underscore the multifactorial nature of hypertension, with implications for targeted interventions addressing geographical, socioeconomic, and lifestyle-related risk factors. The comparison between the results obtained from the LOGISTIC and SURVEYLOGISTIC procedures for factors associated with hypertension revealed several significant findings. Firstly, current age remained consistently and significantly associated with hypertension in both procedures, with each unit increase in age corresponding to a 6% increase in the odds of hypertension (LOGISTIC: AOR = 1.06, 95% CI 1.05; 1.07; SURVEYLOGISTIC: AOR 1.06, 95% CI: 1.05; 1.07). Additionally, the absence of a diabetes diagnosis was significantly associated with lower odds of hypertension in both procedures, with similar effect sizes and confidence intervals (LOGISTIC: AOR = 0.17, 95% CI: 0.11; 0.25; SURVEYLOGISTIC: AOR = 0.16, 95% CI: 0.11; 0.24). Moreover, individuals who had not been diagnosed with a CVA exhibited significantly lower odds of hypertension in both procedures, with comparable effect sizes and confidence intervals (LOGISTIC: AOR = 0.37, 95% CI: 0.17; 0.81; SURVEYLOGISTIC: AOR = 0.37, 95% CI: 0.15; 0.91). However, marital status yielded less consistent results between the two procedures. While there were some similarities in the associations between marital status categories and hypertension, the confidence intervals were wider in the SURVEYLOGISTIC procedure, indicating greater uncertainty. The Akaike Information Criterion (AIC) values were lower for the LOGISTIC procedure (AIC = 2291) compared to the SURVEYLOGISTIC procedure (AIC = 2473), suggesting that the logistic regression model may provide a better fit to the data. Conclusion: This study proved that the logistic regression model is more appropriate to survey data analysis in terms of the established dataset of the 2016 South African Demographic and Health Survey, with a lower AIC, and narrower confidence intervals concerning the survey regression model.enHypertension -- South AfricaLogistic regression analysisMultiple Logistic RegressionThesis