Preventing Osteoporotic Fractures in Men Living with HIV: Model Calibration and Economic Evaluation Studies

I aimed to provide Ontario’s public healthcare payers with an economic evaluation of fracture prevention strategies for HIV-positive men who take antiretroviral therapy (ART). When developing microsimulations models for this purpose, calibration is critical and computationally expensive. Studies that explore efficient calibration methods for microsimulation models are limited. As such, I sought to determine whether simulated annealing or Nelder-Mead is the best-performing parameter search algorithm to calibrate a microsimulation model in project one; determine whether a discrete-event simulation model of fractures leads to more efficient calibrations than a patient-level, discrete-time simulation model in project two; and, estimate the cost-effectiveness of osteoporosis screening and treatment strategies in ART-treated men over a lifetime horizon in project three. For projects one and two, I constructed a Markov microsimulation model that tracked bone loss and incident fractures. All calibrations were based on 4 calibration targets, 12 calibration inputs, a binomial log-likelihood goodness-of fit measure, and first-order searches in a simulated sample of 1000. I assessed calibration performance according to differences in the goodness-of-fit and the time to identifying a good parameter set. In project one, both algorithms produced good sets that had similar fit, but simulated annealing identified the sets two times faster than Nelder-Mead. In project two, both model structures generated similarly accurate good sets, while the discrete-event simulation generated these three times as quickly as the discrete-time simulation. In project three, I evaluated 13 strategies for initiating osteoporosis treatment based on Fracture Risk Assessment Tool (FRAX) scores, with and without bone mineral density (BMD). The base-case included 50-year-old, HIV-positive men who took ART for ≥3 years and did not have a history of fractures or osteoporosis treatment. I incorporated the calibrated results from projects 1 and 2 in project 3’s probabilistic analysis. I discounted outcomes annually by 1.5%. At a willingness-to-pay threshold of $50,000/QALY, it is most cost-effective to assess FRAX without BMD, offer treatment if FRAX is ≥10%, and re-screen annually. Collectively, these studies suggest using simulated annealing and discrete-event simulation as efficient calibration methods; as well as that the payer should not pay for BMD screening in older HIV-positive men.

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.