|Category Reference for Presentations|
|AHE||Applied Health Economics||DEC||Decision Psychology and Shared Decision Making|
|HSP||Health Services, and Policy Research||MET||Quantitative Methods and Theoretical Developments|
* Candidate for the Lee B. Lusted Student Prize Competition
Purpose: We developed and compared the performance of two novel dynamic control algorithms for determining personalized time between testing for patients diagnosed with open angle glaucoma (OAG) against fixed interval monitoring schedules.
Methods: We developed a Kalman filter which combines population disease dynamics with the individual patient's health to predict the patient's future health state. Logistic regression was then used to map the Kalman filter's Gaussian confidence region for the forecasted health states to a probability that a patient is experiencing OAG progression. Two control algorithms (for scheduling the times at which tests will be taken) were created: (1) test when the worst-case point of the confidence region exceeds a threshold on the probability of progression; or (2) test when a proportion of the confidence region exceeds a progression threshold. These algorithms were compared against fixed interval scheduling using longitudinal data from 571 patients who were enrolled in the Advanced Glaucoma Intervention Study (AGIS) and Collaborative Initial Glaucoma Treatment Study (CIGTS) randomized clinical trials.
Results: Our control algorithms achieved Pareto dominance over the fixed yearly monitoring schedule with respect to efficiency in progression detection and detection delay. Using the same average number of scheduled tests as the fixed yearly schedule for each patient, both algorithms increased efficiency in identifying OAG progression by 29% (p<0.0001) and detected progression 57% earlier (p=0.02). Furthermore, the algorithm's performance using the conservative single point of progression was near identical to the algorithm's performance using the more robust proportion of the region method.
Conclusions: We proposed and validated two new and robust algorithms for designing personalized monitoring schedules based on the patient's disease dynamics (which are updated as each new test is performed). Our two control algorithms reallocate monitoring visits to schedule testing closer to the time when progression occurs and to avoid testing when no significant change in the patient's health state is predicted. We demonstrate that equivalent solutions may be obtained by considering two optimization based approaches: a single point of progression threshold or a region crossing the progression threshold. More broadly, the novel modeling framework we developed can be applied to determine personalized monitoring schedules for a variety of chronic diseases which require longitudinal monitoring.
Method: We requested research groups with expertise in modeling HIV epidemiology to simulate the potential impact of standardized policy alternatives in four settings: India, Vietnam, South Africa, and Zambia. Competing policies involved expanding ART eligibility to individuals with higher CD4 cell counts, prioritizing high-risk groups, increasing treatment coverage. A standardized costing framework was developed. Policy options were compared in terms of epidemic impact, health system costs, and cost-effectiveness over 20 years. Research groups met in November 2012 to finalize the analytic approach. Synthesized results from 12 mathematical models were reported to the WHO guideline committee in early 2013.
Result: All models predicted that broad expansions in ART eligibility and coverage would substantially reduce HIV transmission, reduce HIV-related mortality, and increase total costs. In generalized epidemics (South Africa, Zambia), most models estimated that expanding eligibility to CD4 ≤500 cells/µL—a key policy under consideration— would cost US$500-$1,500 per disability-adjusted life year averted, versus current guidelines. Other expansions in eligibility and coverage also produced favourable cost-effectiveness ratios. In concentrated epidemics (India, Vietnam), expanding ART eligibility and increasing coverage in high-risk populations appeared highly cost-effective compared to conventional benchmarks, but efforts to raise coverage in the general population did not. While common themes emerged, translating findings into policy messages was challenging: there was substantial variation on several modelled outcomes, models differed in the set of interventions they could simulate, costing approaches and reported outcomes had to be simplified to allow consistent application across models, and the project timeframe was constrained by predefined WHO policy-setting timelines, limiting the depth of analysis possible. Challenges were particularly acute when attempting to calculate the incremental cost-effectiveness of multiple competing interventions, and standard approaches for synthesising modelled evidence proved unsatisfying for summarizing results across multiple models.
Conclusion: Modeled analyses provided useful input to guideline revisions, identifying policies that would likely be cost-effective in most settings. Synthesizing the results of multiple models is challenging, and must balance the transparency of reporting disparate outcomes with the need to identify summary messages.
There is increasing demand for transcatheter aortic valve replacement (TAVR) as the primary treatment option for patients with severe aortic stenosis (AS) who are high risk surgical candidates or inoperable. TAVR is typically limited to centers of excellence with restricted capacity, thereby causing prolonged wait-times. Our objective was to use mathematical simulation models to estimate the hypothetical effectiveness of TAVR with increasing wait-times, when compared to either conservative medical therapy in inoperable candidates or conventional surgical aortic valve surgery in high risk candidates.
We developed a fully probabilistic discrete event model, using input data from the randomized controlled Placement of Aortic Transcatheter Valves (PARTNER) trials. We evaluated two populations separately: a) in the high risk surgical cohort, we compared TAVR to conventional aortic valve surgery; b) in the inoperable cohort, we compared TAVR to conservative medical therapy. We evaluated 7 scenarios with hypothetical TAVR wait-times ranging from 10 days to 180 days. The main outcome was 1-year mortality and wait-time deaths.
In the inoperable cohort, the mean 1-year mortality for the conservative medical therapy arm was approximately 50%. When the TAVR wait-time was 10 days, the mean TAVR wait-time mortality was 1.9% with a 1 year mortality of 31.5%. Mean TAVR wait-time deaths increased to 28.9% with a 180-day wait, with a corresponding mean 1-year mortality of 41.4%.
In the high risk cohort, the wait-time deaths and mean 1-year mortality for the surgical patients were 2.5% and 27% respectively in all scenarios. The TAVR wait-time deaths increased from 2.2% at a 10-day wait to 22.4% at a 180-day wait, with a corresponding increase in 1-year mortality from 24.5% to 32.6%. The 1-year mortality in the TAVR group exceeded that in the surgical group at wait-times greater than 60 days.
We found that modest increases in TAVR wait-times would have substantial impact on the effectiveness of TAVR in both inoperable patients and high risk surgical candidates. In the high-risk surgical candidates, at wait-times beyond 60 days, TAVR was less effective on average compared to conventional surgery. Our results highlight the importance of aggressive wait-time management for TAVR in severe AS.
Method: We calibrated a previously published dynamic model of HPV transmission to fit observed HPV prevalence and cervical cancer incidence in Norway. Under various scenarios of catch-up vaccination in females, we projected reductions in HPV incidence over multiple birth cohorts, including both direct and indirect benefits, and applied these reductions to a microsimulation model of cervical cancer and incidence-based models for non-cervical HPV-related diseases. We adopted a societal perspective and assumed that vaccination of females age >19 years would incur higher delivery costs (i.e., through their family physician). Scenarios reflecting 50% coverage of women up to age 20, 22, 24 or 26 were compared to a baseline strategy assuming that these cohorts were not vaccinated. Sensitivity analyses were conducted on vaccine cost (market vs. tender price) and differential uptake among targeted women.
Result: The marginal benefit of the vaccine decreased as the upper bound of the catch-up age increased. For example, at 50% coverage, the cohort of girls aged 18-years-old in 2014 gained an absolute 21% in cumulative reduction in HPV-16 incidence, compared to no catch-up campaign, while for the cohort of girls aged 26-years-old, this gain was only 10%. Cost-effectiveness followed a similar trend. At the current market price of the vaccine, catch-up can only be extended to age 22 while still remaining below Norway’s willingness-to-pay threshold (≈$83,000/QALY), compared with vaccinating 12-year-old girls only. However, the tender price of the vaccination (not publicly available) is believed to be less than 50% of the market price, in which case a catch-up program to age 26 falls below the threshold. Results remained stable for a catch-up campaign achieving only 30% coverage.
Conclusion: At current market price, a one-year catch-up program up to age 22 is likely to be cost effective; however, at the assumed tender price, HPV vaccination may be extended to age 26 while remaining cost-effective.
Purpose: To test the efficacy of an expanded version of the BRCA Gist Intelligent Tutoring System (ITS) designed to help women understand and make decisions about genetic testing for breast-cancer risk with a broader range of participants recruited on the web, in local communities, and at two universities.
Methods: This interactive tutorial guided by Fuzzy-Trace Theory is the first use of an ITS in patients' medical decision-making. Three female avatars of varying ethnicities present tutorial information orally, visually, in brief video clips, and in writing (a screen shot is shown below). Tutorial dialogues address questions such as, "how do genes affect breast cancer risk?" Expanded content addresses inherited genetic mutations, what should be considered before genetic testing, how breast cancer spreads, stages of breast cancer, and the Gail model. Using "expectations texts" and Latent Semantic Analysis, a conversational agent (avatar) "understands" and responds to participants' typed questions and comments using natural language. Information pertaining to breast cancer and genetic risk was taken from the National Cancer Institute (NCI) web site, and vetted by medical experts.
The efficacy of the BRCA Gist ITS was tested in two randomized, controlled experiments equating time on task. Participants in both a laboratory experiment (n=210) and a field experiment (n=180) were randomly assigned to one of three conditions: BRCA Gist; studying pages from the NCI web site covering comparable materials; or a control tutorial on nutrition. Participants were then given a test of content knowledge about breast cancer and genetic risk, twelve scenarios applying their knowledge about assessing breast cancer risk, a gist comprehension measure, and questions about the participant's interest in being tested.
Results: In both the laboratory and field experiments, the BRCA Gist group performed significantly better than the control group on content knowledge about breast cancer and genetic risk, gist comprehension, knowledge application, and risk assessment. Participants' interest in testing was significantly lower following the BRCA Gist tutorial. Effect sizes were generally large.
Conclusions: The BRCA Gist ITS may be fruitfully applied in assisting laypeople in preventive health and medical decision-making by effectively teaching content knowledge, enhancing gist-based comprehension, and increasing the ability to apply knowledge to assess risk and make testing decisions about genetics and breast-cancer.