Background <p>Alteration in serum lipids is a well-recognized metabolic accompaniment of chronic hepatitis B (HBV) and hepatitis C (HCV) viral infections, and may reflect a combination of disease severity and metabolic dysregulation. Despite this established association, lipid profile monitoring is not standard practice in these patients, and lipid abnormalities may remain unrecognized until they influence long-term cardiometabolic outcomes or complicate overall risk stratification. This gap in routine practice provided the rationale for the present study.</p> Aims <p>This study aimed to evaluate whether routinely available liver function tests (LFTs) can be leveraged to predict hypolipidemia, and translate these findings into clinically actionable risk prediction tools by using machine learning (ML) prediction models.</p> Methods <p>The study included a total of 2,440 participants divided into cases (<i>n</i> = 1104) and controls (<i>n</i> = 1336). Relevant demographic and laboratory data were collected, including viral serology, liver function test (LFT) parameters, and lipid profiles and structured in an ontological framework for ease of accessibility and interoperability. Potential predictors of hypolipidemia were screened using univariate analysis, followed by assessment of multicollinearity using the variance inflation factor. Predictors retaining statistical significance were subsequently entered in a multivariable logistic regression model to determine independent predictors of hypolipidemia. Furthermore, different ML algorithms were used to develop a predictive model for hypolipidemia in these patients, and their discriminatory performance was assessed using multiple evaluation metrics, including accuracy, area under the curve (AUC), sensitivity, specificity, recall, Matthews Correlation Coefficient (MCC), balanced accuracy, false positive rate (FPR), false negative rate (FNR), and Brier score. Decision curve analysis (DCA) were performed to assess clinical utility.</p> Results <p>Significant differences were observed between cases and controls across all lipid parameters, including total cholesterol, triglycerides, high-density lipoprotein (HDL), low-density lipoprotein (LDL), and very low-density lipoprotein (VLDL), all of which were lower in cases (<i>p</i> &lt; 0.05). A markedly higher proportion of cases exhibited hypolipidemia (335/1104; 30.3%) compared to hyperlipidemia (62/1104; 5.6%). LFT parameters showed modest ability to predict hypolipidemia on multivariate logistic regression. The ML algorithms outperformed the conventional statistical method of multivariate logistic regression in predicting hypolipidemia in infected individuals. Among the ML algorithms, ensemble models demonstrated superior discriminative performance. The Random Forest (RF) model achieved the best overall predictive performance, with an AUC of 0.926, accuracy of 0.910, sensitivity of 0.88, specificity of 0.91, balanced accuracy of 0.901, and an MCC of 0.82. The model further exhibited favorable calibration with a low Brier score (0.060) and reduced misclassification errors, reflected by a low FPR (0.087) and FNR (0.111). DCA demonstrated net clinical benefit of ML models over test-all, test-none strategy.</p> Conclusions <p>Routine LFT parameters can predict hypolipidemia in chronic HBV and HCV. ML models improve discrimination and yield clinically meaningful net benefit, supporting risk-guided lipid profiling using routinely available LFTs. This can offer a simple screening and triage tool that converts routinely available LFTs into an estimated probability of hypolipidemia.</p>