Using Mathematical Models to Monitor and Evaluate the Impact of Public Health Interventions on Epidemics: The Case of the TB/HIV Co-pandemic in Africa.

Curtailing the public health crises generated by the TB/HIV co-pandemic poses many challenges, including monitoring difficulties that complicate our ability to accurately predict the efficacy of public health interventions. Unfortunately, the dire and urgent need to inform policy makers on the individual and population benefits to be obtained from different disease control measures may result in our prioritizing predicting per se, over conducting in-depth system studies that can maximize our ability to properly address the question(s) posed. In this chapter we illustrate these difficulties and exemplify how analytical studies can be used not only as a powerful predictive tool that can help guide public health policy, but also as a means of gaining important insight into the system's functioning that, in turn, can help the monitoring of pandemics. We present in detail the mathematical, statistical and computational approach and methodology behind our recent research focused on evaluating the potential public health benefits to be derived from reducing TB treatment duration from the standard 6- to 8-month treatments to a 2-month treatment in areas of high HIV prevalence. Because of the considerable uncertainties and spatial and temporal heterogeneity in parameter values we conducted a calibration to historical TB/HIV trends that increased the validity of our predictive results for high HIV prevalence areas. Unexpectedly, our calibration unveiled incongruent epidemiological patterns between the two interacting diseases in Kenya, which led us to evaluate reported TB/HIV co-dynamics for the whole of Africa with a new statistic that quantifies the relative change in TB numbers as compared to HIV numbers. Our initial analysis of pan-African patterns placed Kenya as an outlier. However, Kenyan TB and HIV estimates have since been revised, and now the relative change in the co-dynamics of the two diseases place Kenya in a more intermediate position. We additionally conducted a sensitivity analysis to evaluate which parameters had a greater influence on the output variables of choice. In this chapter we briefly explain the results of our model predictions and of our sensitivity analysis focusing on the general approach and methods because many of the lessons learned related to the reasoning and steps we followed, and not exclusively to the results of our investigation. Thus, here our final conclusions and recommendations are geared towards the analytical process more than the public health implications of our studies. Most importantly, this study taught us the power of contrasting trends of interacting diseases in global disease monitoring and control. This is particularly so for such a complex co-pandemic as TB/HIV, where our analysis of the joint co-dynamics provided important insights into the individual epidemiology of each disease.