Logistic Regression Modelling of Complete Blood Count Parametersfor Predicting Diabetes Mellitus among Sabahan Women

Abstract

According to the National Diabetes Registry, 1.6 million Malaysians have diabetes, with women having a higher prevalence than men (57.1% versus 42.9%). These figures highlight the necessity of customized diabetes care strategies that consider genderspecific characteristics. Therefore, this study investigates the relationship between diabetes in females and complete blood count (CBC) parameters. A dataset containing the CBC test results of 537 female patients obtained from the Clinical Laboratory at the Faculty of Medicine & Health Science at Universiti Malaysia Sabah was analyzed. The univariate analysis showed all variables were significant predictors of diabetes except for ethnicity. At the multivariate analysis, the following parameters were identified as significant: age (OR = 2.77, 95% CI: 1.402–5.486, p = 0.003), hemoglobin (OR = 22.56, 95% CI: 10.556–48.178, p = 0.000), MCV (OR = 26.20, 95% CI: 12.23–56.14, p = 0.000), MCH (OR = 2.29, 95% CI: 1.17–4.51, p = 0.016), MCHC (OR = 2.81, 95% CI: 1.40–5.65, p = 0.004), HCT (OR = 3.47, 95% CI: 1.72–6.70, p = 0.001), WBC (OR = 4.72, 95% CI: 2.39–9.32, p = 0.000), and platelets (OR = 7.95, 95% CI: 3.25–19.46, p = 0.000). The Hosmer-Lemeshow goodness-of-fit test indicates that the model aligns well, accurately reflects the data and satisfies the assumptions of logistic regression. When used and interpreted appropriately, these CBC parameters can be useful indicators to support decision-making for the monitoring and treatment of diabetes mellitus-related issues in females.