Wednesday, October 23, 2013: 10:00 AM - 11:30 AM
Key Ballroom 3-4 (Hilton Baltimore)
Category Reference for Presentations
AHEApplied Health Economics DECDecision Psychology and Shared Decision Making
HSPHealth Services, and Policy Research METQuantitative Methods and Theoretical Developments

* Candidate for the Lee B. Lusted Student Prize Competition

Session Chairs:
Jesse D. Ortendahl, MS and Karen M. Kuntz, ScD
10:00 AM
Lauren E. Cipriano, MS1, Shan Liu, S.M.1, Thomas A. Weber, PhD2 and Jeremy D. Goldhaber-Fiebert, PhD1, (1)Stanford University, Stanford, CA, (2)Ecole polytechnique federale de Lausanne, Lausanne, Switzerland
Purpose: CDC guidelines recommend hepatitis C virus (HCV) screening for the 1945-1965 birth cohorts.  Since HCV prevalence is decreasing with birth-year, age-specific screening is less cost-effective in later cohorts.  To inform the optimal time to discontinue screening, collecting additional information may be valuable, though when this information should be collected is unclear.  Standard practice for value of information analyses does not include the option to delay information collection. However, delaying collection may be optimal when at least one model parameter is changing across cohorts as in this case.

Methods: We apply a Markov decision process framework to evaluate how long to continue HCV screening in US men. We identify the optimal information collection policy for two parameters assumed constant across cohorts - reductions in quality-of-life from awareness of HCV-positive status and the fibrosis-stage distribution at screen - detected diagnosis at age 50 - alone and in combination with information collection about HCV prevalence which is decreasing across cohorts.  We estimate lifetime costs and benefits using a previously-developed HCV screening model and HCV prevalence dynamics derived from NHANES.  The assumed willingness-to-pay threshold is $75,000 per QALY. 

Results: The presence of a parameter which varies across cohorts influences the per-person value-of-information about both time-varying and static parameters. In these cases, we show analytically that it may be optimal to delay information collection.  Given our prior beliefs, the optimal strategy is to collect sample information about the reduction in quality-of-life from awareness of HCV-positive status immediately and then, depending on the result of that study, collect information on HCV prevalence 3 to 20 years in the future.  This strategy, less the cost of information collection, increases the expected incremental net monetary benefit (INMB) by $2.3 million compared to a strategy of collecting information about both immediately.  The optimal time to collect information about the fibrosis-stage distribution is in 12 years, increasing the expected INMB by $1.7 million compared to a strategy of collecting information about both immediately. 

Conclusions: We demonstrate that when parameters vary across cohorts, the optimal information collection policy, for both time-varying and static parameters, may be to delay information collection until it is more likely to influence the decision. Our dynamic programming framework enables the consideration of delayed information collection in numerous contexts.

10:15 AM
Karl Claxton, PhD1, Steve Martin, PhD1, Marta Soares, Msc1, Nigel Rice, PhD2, D. Eldon Spackman, PhD1, Sebastian Hinde, MSc1, Nancy Devlin, PhD3, Peter C. Smith, PhD4 and Mark Sculpher, PhD1, (1)University of York, York, United Kingdom, (2)Centre for Health Economics, York, United Kingdom, (3)Office of Health Economics, London, United Kingdom, (4)Imperial College, London, United Kingdom
Purpose: To develop a conceptual framework for estimating cost effectiveness thresholds in the context of health care systems with budget constraints.  To use routinely available data in the English National Health Service to estimate a cost effectiveness threshold (in terms of cost per quality-adjusted life-year (QALY)) relevant to decisions made by the National Institute for Health and Care Excellence (NICE).

Method: Earlier econometric analysis estimated the relationship between differences in spending by local health care purchasers primary care trust (PCT) spending, across programme budgeting categories (PBCs), and associated disease-specific mortality.  This research has been extended in several ways including estimating the impact of marginal increases or decreases in overall NHS expenditure on spending in each of the 23 PBCs.  Further stages of work, using data sources including MEPS, link the econometrics to broader health effects in terms of QALYs. 

Result: The most relevant 'central' threshold was estimated at £18,317 per QALY (2008 expenditure, 2008-10 mortality).   Uncertainty analysis indicates that the probabilities that the threshold is less than £20,000 and £30,000 per QALY, respectively, are 0.64 and 0.92.  Additional 'structural' uncertainty suggests, on balance, that the central or best estimate is, if anything, likely to be an overestimate.  The health effects of changes in expenditure are greater when local purchasers are under more financial pressure and are more likely to be disinvesting than investing.  This indicates that the central estimate of the threshold is likely to be an overestimate for all technologies which impose net costs on the NHS and the appropriate threshold to apply should be lower for technologies which have a greater impact on NHS costs. 

Conclusion: The methods go some way to providing an empirical estimate of the scale of opportunity costs the NHS faces when considering whether the health benefits associated with new technologies are expected to offset the health that is likely to be lost elsewhere in the NHS.  The study also starts to make the other NHS patients, who ultimately bear the opportunity costs of such decisions, less abstract and more ‘known’ in social decisions.  This work has implications for the Government's proposals to move to a system of value-based pricing for new prescription pharmaceuticals.

10:30 AM
Jingshu Wang, PhD, Ruifeng Xu, PhD, Keaven Anderson, PhD and James M. Pellissier, PhD, Merck Research Laboratories, North Wales, PA
Purpose: Frequently in oncology drug development, model-based projections of treatment outcomes are sought that must be based on limited data from early-phase clinical trials.  The assessment of treatment benefit may be improved by adjusting for baseline risk factors.  Small sample sizes and inclusion of covariates implies small degrees of freedom, which makes probabilistic analyses very important.  If probabilistic analyses are conducted by independently drawing parameters of the survival functions for progression-free survival (PFS) and overall survival (OS), the relationship between PFS and OS may not be realistic. This work presents an illustrative case study using bootstrapping in the context of fitted statistical models to assess treatment benefit and its uncertainty with respect to PFS and OS estimates for decision-analytic models.

Methods: A randomized phase II trial (PRECEDENT) evaluated the efficacy of vintafolide plus pegylated liposomal doxorubicin (V+PLD) vs. PLD alone in platinum-resistant ovarian cancer treatment. PRECEDENT showed the efficacy of V+PLD was present in patients who tested to have 100% folate receptor positive tumors (FR [100%]).  For FR[100%] patients (n=37), median PFS were 24.0 vs. 6.6 weeks for V+PLD vs. PLD patients (HR 0.381; p=0.018).  Our study sought to extend the estimation of the PFS and OS for use in modeling.  Parametric Weibull survival models were estimated including treatment indicator and pre-specified baseline factors.  Survival probabilities were estimated for all patients given a treatment and baseline covariates.  The mean of all patients’ calculated survival probabilities at each time point yielded estimated population survival curves, with area under the curve providing the mean survival time.  Variability was assessed by repeatedly drawing samples with replacement from the trial data (bootstrapping). 

Results: Modeled OS and PFS by group fit Kaplan-Meier curves well. Predicted means with confidence intervals are shown for 2-year results:

Mean time (months)


95% CI


95% CI



(1.4, 5.7)


(5.3, 15.4)



(3.9, 10.9)


(11.0, 17.6)



(-0.8, 8.3)


(-2.9, 10.1)

Conclusions: This case study illustrates an approach to statistically fit and capitalize on bootstrapping to estimate and capture the uncertainty in modeling PFS and OS for a small, randomized Phase II trial.  This methodology can also be used for other outcomes important to decision-analytic models.   The techniques described will help modelers working with limited clinical data.

10:45 AM
Sun-Young Kim, PhD1, Louise B. Russell, PhD2, Jeehyun Park, PhD2, Jennifer R. Verani, MD3, Shabir A. Madhi, MD, PhD4, Clare L. Cutland, MBBCh4, Stephanie J. Scharg, DPhil3 and Anushua Sinha, MD, MPH5, (1)University of Texas School of Public Health, San Antonio, TX, (2)Rutgers University, New Brunswick, NJ, (3)National Center for Immunization and Respiratory Diseases, Centers for Disease Control and Prevention, Atlanta, GA, (4)Medical Research Council: Respiratory and Meningeal Pathogens Research Unit and University of the Witwatersrand, Johannesburg, South Africa, (5)New Jersey Medical School, Rutgers University, Newark, NJ

In low-and middle-income countries, cost-effectiveness analyses of new vaccine introduction have typically compared vaccine against doing nothing. We illustrate the impact of competing new vaccines against other realistic prevention alternatives, using maternal GBS vaccination, which is currently in trials including in South Africa, as an example. 


We developed a probabilistic decision-analytic model for an annual cohort of pregnant women and their babies that simulates maternal GBS colonization status and the natural history of early onset (EOGBS) and late onset GBS disease (LOGBS) in infants. We compared four strategies: doing nothing; risk factor-based intrapartum antibiotic prophylaxis (RFB-IAP) which is used in some South African hospitals; the potential new maternal GBS vaccine; and vaccination plus RFB-IAP.  


Compared to doing nothing, RFB-IAP would prevent 10% of EOGBS/LOGBS cases in South African infants, vaccination 42%, and vaccination plus RFB-IAP 48%. Incremental comparisons show that RFB-IAP would cost $240 per DALY averted compared with doing nothing (2010 US$); at a vaccine price of $20/dose, RFB-IAP alone has the highest probability of being cost-effective when willingness-to-pay falls between $280 and $1,800. Vaccination alone would cost $1,998/DALY compared with RFB-IAP alone. Vaccination plus RFB-IAP would cost $596/DALY compared with vaccination alone. The weak domination of vaccination alone does not, in this case, point to a realistic policy alternative; vaccination is delivered months before delivery and RFB-IAP given at delivery, based on risk factors present at that time. 


Vaccination would be very cost-effective in South Africa by World Health Organization’s gross domestic product-based guidelines. Interpretation of this finding is influenced by inclusion of an alternative, RFB-IAP. Although it prevents only 10% of cases, RFB-IAP is the most cost-effective alternative to doing nothing. The combined strategy of vaccination plus RFB-IAP prevents more disease and costs more than vaccination alone, and is consistently very cost-effective. Realistic comparators in addition to doing nothing should be included in cost-effectiveness analyses of vaccines whenever possible, to provide low- and middle-income countries’ decision makers with more complete information about policy alternatives.

11:00 AM
Amilcar Azamar-Alonso, MSc1, Sergio Bautista-Arredondo, MSc1, Gilberto Sanchez-Gonzalez, MSc2 and Juan Sierra-Madero, MD3, (1)National Institute of Public Health, Cuernavaca, Mexico, (2)national Institute of Public Health, Cuernavaca, Mexico, (3)National Institute of Medical Sciences and Nutrition, Mexico, Mexico
Purpose: The purpose of this study was to analyze two patient-level reminder interventions aimed to increase adherence levels to HAART IN Mexico. Clinical evidence shows that adherence levels ≥90% are required to maximize HAART effectiveness, lower levels increase the disease progression and therefore the probability of death on HIV patients. In Mexico, universal access to HAART; however, average adherence level is 79.8% (95% CI: 77.8-81.8).

Method: The study design was a cost-effectiveness analysis from the governmental perspective. All the costs were expressed in 2010 constant USD. A natural history of disease dynamic model for HIV was used to estimate the following parameters: CD4 and CD8 cell replication and cell mortality rates, as well as infectivity rates of individuals simulated. Also, we analyzed data from a national representative survey of HIV patients on HAART (N=2289) and presenting at 50 governmental hospital/clinics to obtain adherence levels. With these parameters we used a Markov model to estimate life expectancy, total patients’ care costs, and therefore incremental cost-effectiveness ratios. Patients were classified as adherent (90%) and non-adherent (<90%). We evaluated two patient-level reminder interventions to increase adherence to HAART: (1) three reminder text messages (SMS) sent daily to the patient’s cell phone, and (2) a pill bottle with alarm (pill reminder). Both were modeled throughout the patients’ lives. We performed probabilistic sensitivity analysis for both adherence levels and costs.

Result: Of the 2289 patients, 26% were adherent (≥90%) (mean adherence level: 79.8%). We did not find statistically significant differences between adherents and non-adherents in sociodemographic characteristics. Seventy percent reported that HAART daily intake omission is the main reason for non-adherence. Interventions increase life expectancy by 2.6 years (SMS) and 3.1 years (pill reminder) with an incremental cost of $4050 and $5552, respectively. Incremental cost-effectiveness ratios are $207 and $637 per year life gained (3% annual discount rate).

Conclusion: Both interventions are below one GDP per capita in Mexico; therefore, they are cost-effective and could be considered for implementation in our country.

11:15 AM
Hawre Jalal, MD, MSc and Karen M. Kuntz, ScD, University of Minnesota, Minneapolis, MN
   Purpose: Expected value of sample information (EVSI) is a key concept in Bayesian decision theory and illustrates the advantage of the decision-analytic framework over traditional statistical power calculations. We propose a simple and practical framework to compute measures of EVSI.       Methods:   Our approach entails three steps: (1) conduct a probabilistic sensitivity analysis (PSA), (2) regress the model parameters on the incremental net benefit (INB) of the new intervention compared to the standard of care using linear regression metamodeling (LRM), and (3) compute EVSI using the unit normal loss integral (UNLI) method.  The key concept in our approach is that EVSI relies on the correct estimation of the variance of the INB posterior to collecting new data from a new sample (n).  We achieved this goal using LRM, which assumes a linear relationship between the INB and the uncertain model parameters.  Then we adopted the UNLI as a parametric approach to compute EVSI from the fraction of the INB variance explained by the parameters of interest.  We illustrate our approach using a previously published decision model, which compared a new treatment to the standard of care for treating a serious condition.  The new treatment is effective, but associated with additional risk of a critical event.  The uncertain model parameters are (1) the probability of the critical event with the standard care (pC), the probability of side-effects following the new treatment (pSE), the number of quality-adjusted years after a critical event (QE), and the odds ratio of the efficacy of the new treatment compared to standard care (OR).       Results:   The Figure shows the EVSI calculated using our approach and those from the published model for various sample sizes (n).  The EVSI is shown for individual model parameters and for all the parameters combined.       Conclusion:   Our results closely predicted the results from the published model.  While PSA, LRM and UNLI are rooted in simulation studies, they have never been combined in this capacity to express EVSI.  In addition, our approach avoids complex mathematical notations, requires one PSA, and allows for correlation among model parameters.