Related Experiment Video
Updated: Dec 26, 2025
An In Vitro Model for Measuring Immune Responses to Malaria in the Context of HIV Co-infection
Published on: October 6, 2015
Comparison of predictive models for hepatitis C co-infection among HIV patients in Cambodia
Jozefien Buyze1, Anja De Weggheleire2, Johan van Griensven2
1Department of Clinical Sciences, Institute of Tropical Medicine, Nationalestraat 155, Antwerpen, 2000, Belgium. jbuyze@itg.be.
Insights
Targeted Hepatitis C virus (HCV) screening in HIV cohorts is crucial for resource-limited settings. Logistic regression identified fewer missed HCV co-infections but referred more patients, while Spiegelhalter-Knill-Jones offered a better balance for efficient screening.
Area of Science:
- Epidemiology
- Biostatistics
- Public Health
Background:
- Hepatitis C virus (HCV) infection poses a significant global health challenge.
- World Health Organization guidelines advocate for HCV screening in all individuals with HIV.
- Targeted HCV screening is a practical approach for HIV cohorts in resource-limited settings, necessitating effective predictive tools.
Purpose of the Study:
- To compare the performance of logistic regression, Spiegelhalter-Knill-Jones (SKJ), and Classification and Regression Trees (CART) for predicting HCV co-infection in an HIV cohort.
- To identify clinician-friendly tools for prioritizing HCV testing among specific subgroups of HIV patients.
Main Methods:
- A cross-sectional study was conducted in an HIV cohort in Phnom Penh, Cambodia.
- The predictive performance of logistic regression, SKJ, and CART models was evaluated.
- Leave-one-out bootstrap estimation was employed to correct for over-optimism in estimating missed HCV co-infections.
Main Results:
- Logistic regression missed the fewest HCV co-infections (8%) but required referral for 98% of patients.
- SKJ and CART models missed 12% and 29% of co-infections, respectively, while referring only about 30% of patients for testing.
- SKJ demonstrated the highest area under the ROC curve, suggesting superior discriminatory power.
Conclusions:
- Logistic regression yielded the highest log-likelihood and lowest missed co-infections, but SKJ offered a better balance for targeted screening.
- The likelihood ratios from SKJ may be more clinically interpretable than odds ratios from logistic regression or CART decision trees.
- CART provides flexibility without requiring pre-specified interactions or variable relationships, offering a robust alternative.
Background:
Hepatitis C virus (HCV) infection is a major global health problem. WHO guidelines recommend screening all people living with HIV for hepatitis C. Considering the limited resources for health in low and middle income countries, targeted HCV screening is potentially a more feasible screening strategy for many HIV cohorts. Hence there is an interest in developing clinician-friendly tools for selecting subgroups of HIV patients for whom HCV testing should be prioritized. Several statistical methods have been developed to predict a binary outcome. Multiple studies have compared the performance of different predictive models, but results were inconsistent.
Methods:
A cross-sectional HCV diagnostic study was conducted in the HIV cohort of Sihanouk Hospital Center of Hope in Phnom Penh, Cambodia. We compared the performance of logistic regression, Spiegelhalter-Knill-Jones and CART to predict Hepatitis C co-infection in this cohort. We estimated the number of HCV co-infections that would be missed. To correct for over-optimism, the leave-one-out bootstrap estimator was used for estimating this quantity.
Results:
Logistic regression misses the fewest HCV co-infections (8%), but would still refer 98% of HIV patients for HCV testing. Spiegelhalter-Knill-Jones (SKJ) and CART respectively miss 12% and 29% of HCV co-infections but would only refer about 30% for HCV testing.
Conclusions:
In our dataset, logistic regression has the highest log-likelihood and smallest proportions of HCV co-infections missed but Spiegelhalter-Knill-Jones has the highest area under the ROC curve. The likelihood ratios estimated by Spiegelhalter-Knill-Jones might be easier to interpret for clinicians than odds ratios estimated by logistic regression or the decision tree from CART. CART is the most flexible method, and no model has to be specified regarding presence of interactions and form of the relationship between outcome and predictor variables.
