Development of a clinical decision support system to assess the risk of lung cancer for prevention and earlier diagnosis in people with HIV

The advent of antiretroviral therapy (ART) changed the prognosis of HIV. People with HIV (PWH) live longer lives but are susceptible to the same age-related disease as people without HIV. Cancer is a leading cause of death for PWH, and lung cancer is the leading cause of cancer-related death. Smoking and increased age are the primary causes of lung cancer in the general population. Still, they do not explain the significant increase in lung cancer incidence in PWH. Risk factors specific to HIV, such as immunocompetence and respiratory disease, present more lung cancer risk in PWH. Existing guidance for lung cancer prevention exclude PWH and do not consider HIV specific risk factors. This deficiency has led to increased incidence and mortality from lung cancer and at younger ages and more advanced stages. This study addresses the urgent and unmet need to develop a lung cancer risk assessment tool for prevention and earlier diagnosis in PWH. Using an integrated set of public-use data of 7,607 HIV-positive men and women from Johns Hopkins University, knowledge-based risk factors specific to PWH were formed using logistic regression models predicting lung cancer. Twenty-one HIV-specific risk factors were identified from 51 candidates using bivariable logistic regression models conditioned on gender, race, and smoking status. Similarly, stratified multivariable logistic regression models determined 14 lung cancer predictors for PWH. A multiplicate risk score was created for each risk factor and then summed to create a single risk score. Risk stratifications for low, medium and high risk were identified using the quartiles of risk. The risk score in the knowledge-base had a sensitivity of 77%. It led to the developing of the inference engine and a web-based Clinical Decision Support System (CDSS) for PWH and their clinicians. A random sample of 20 patients reserved for validating the CDSS had a sensitivity of 80%, while criteria from existing guidelines for the general population missed between 97 and 100% of lung cancer cases. These results demonstrate that a risk assessment tool, such as this CDSS, predicts lung cancer in PWH better than existing guidance and can prevent or diagnose it earlier.Ph.D.Includes bibliographical reference