#  Radhika Tampi 

HEOR Associate, Analysis Group

 

 

 



*Dissertation Title*: "Simulation of Policy Levers Across the Tuberculosis Care Cascade"

Tuberculosis (TB) continues to impose substantial mortality and morbidity, particularly among children, for whom diagnosis remains challenging, and among adults with diabetes, whose comorbidity complicates treatment and worsens outcomes. In this dissertation, I develop simulation modeling and causal inference approaches to assess how failures across the TB care cascade shape outcomes in these populations.

In Chapter 1, I quantify the mortality and morbidity consequences of delays in diagnosis and treatment among children younger than 10 years with pulmonary TB. Using a microsimulation model, I evaluate how diagnostic sensitivity and delays in treatment initiation jointly affect outcomes. Delayed diagnosis increases mortality and morbidity across all age groups, with the largest absolute harms concentrated among children younger than two years and those living with HIV. Although improvements in diagnostic sensitivity yield meaningful gains in survival, particularly among the youngest children, these gains are substantially reduced by delays in treatment initiation. Similarly, immediate treatment initiation only partially mitigates the harms of imperfect sensitivity.

In Chapter 2, I evaluate the effects of trade-offs between health system access and diagnostic accuracy on health outcomes among children with presumptive TB. Using a microsimulation model, I assess four policy levers: access to care, access to chest x-ray, access to molecular WHO-recommended rapid diagnostics (mWRDs), and the sensitivity and specificity of treatment decision algorithms. These algorithms are highly sensitive standardized tools that combine clinical features and available diagnostic information to guide TB treatment initiation in primary care settings. I find that improving access to chest x-ray, mWRDs, or the sensitivity of treatment decision algorithms has limited population-level benefit if children do not access care.

Taken together, Chapters 1 and 2 show that the benefits of improving diagnostic sensitivity in children are limited unless barriers elsewhere in the care cascade are also addressed, particularly access to evaluation upstream and rapid treatment initiation downstream.

In Chapter 3, I examine whether the association between diabetes and unfavorable multidrug-resistant/rifampicin-resistant TB treatment outcomes (i.e., death, treatment failure, or loss to follow-up) differs by baseline glycemic control. I then use marginal structural models to examine how much of this risk may operate through intervenable pathways, namely disengagement from TB care and hyperglycemia. Using data from the endTB observational study, I find that, relative to no diabetes, uncontrolled diabetes at baseline is associated with a 43% higher risk of unfavorable outcomes, whereas controlled diabetes is not. When simulating hypothetical interventions in which all individuals remain engaged in treatment, the risk of unfavorable treatment outcomes is attenuated among those with uncontrolled diabetes at baseline, suggesting that disengagement from treatment may explain part, but not all, of the excess risk. Examining the impact of achieving glycemic control during treatment is difficult because, in this cohort, unfavorable outcomes occurred early in treatment.



 

 

 





 

 

- ## Dissertation Committee Member
    
     [Ankur Pandya](/dissertation-committee-member/ankur-pandya) [Molly Franke](/dissertation-committee-member/molly-franke) [Nick Menzies](/dissertation-committee-member/nick-menzies)
- ## Concentration
    
     [Decision Sciences](/conclabel/decision-sciences)
- ## Graduation Year
    
     [2026](/graduation-year/2026)
- ## Role
    
     [Alumni](/people/alumni)