top of page

Impact of Counselling to Reduce
Risk of HIV Contraction Among Couples

Executive Summary

The primary focus of this medical data analytics project (Section 3 of the document) is on assessing the impact of counseling interventions on reducing unprotected sexual acts using Poisson and Negative Binomial regression models. The project also covers the exploration of linear regression assumptions and logistic regression models in Sections 1 and 2, enhancing the analytical breadth. Key insights include identifying significant predictors of unprotected sexual acts and evaluating model performance under different statistical assumptions. Recommendations emphasize the importance of proper intervention methods in reducing risky behaviors.

Medical Problem

A clinic seeks to understand the impact of counseling interventions on reducing unprotected sexual acts to mitigate the risk of HIV transmission. The objective is to analyze whether counseling both partners, just the female, or no counseling at all has a significant impact on the number of unprotected sexual acts post-intervention.

Methodologies

  • Literature Review: Investigated existing research on the impact of counseling interventions on HIV prevention.

  • Data Collection: Utilized real-world data from a study involving 329 couples.

  • Exploratory Data Analysis: Conducted exploratory plots to visualize the data.

  • Model Building: Poisson Regression and Negative Binomial Regression

  • Model Evaluation: Compared models using residual analysis and dispersion metrics.

Skills

  • Programming Languages: R

  • Libraries and Tools: ggplot2, tidyverse, MASS

  • Statistical Techniques: Poisson Regression, Negative Binomial Regression, Logistic Regression, Linear Regression

  • Data Processing: Data cleaning, handling missing values, data visualization

  • Model Evaluation: Residual analysis, overdispersion assessment, confidence intervals

Results 

Results & Reccomendations

  1. Impact of Interventions: Counseling both partners significantly reduces the number of unprotected sexual acts post-intervention.

  2. Model Performance: The Negative Binomial model better accounts for overdispersion in the data compared to the Poisson model.

  3. Predictors: Baseline unprotected acts, type of intervention, and gender are significant predictors.

​

Model Performance

  • Poisson Regression: Indicated overdispersion with a dispersion parameter of 10.32, suggesting the need for a more flexible model.

  • Negative Binomial Regression: Provided a better fit with less sensitivity to outliers and improved residual diagnostics.

Recommendations and Next Steps

  • Focus on Joint Counseling: Emphasize counseling both partners to achieve significant reductions in unprotected sexual acts.

  • Further Research: Conduct additional studies to validate these findings across different demographics and settings.

  • Data Enrichment: Collect additional data on behavioral and psychological factors influencing unprotected sexual acts.

  • Advanced Modeling: Explore hierarchical models or generalized estimating equations to account for potential clustering in the data.

  • Policy Implementation: Design targeted intervention programs based on the findings to maximize impact on reducing HIV transmission.

bottom of page