Monday, October 24, 2016: 10:00 AM - 11:00 AM
Bayshore Ballroom Salon F, Lobby Level (Westin Bayshore Vancouver)

Ahmed Bayoumi, MD, MSc
University of Toronto
Health Policy, Management and Evaluation

10:00 AM

Jeremy D. Goldhaber-Fiebert, PhD1, David Stauffer, MS1, Lea Prince, PhD1, Jason Andrews, MD, MS2, Sanjay Basu, MD, PhD1, Marcella Alsan, MD, PhD1, Eran Bendavid, MD, MS3, Margaret L. Brandeau, PhD4 and Douglas K. Owens, MD, MS5, (1)Stanford University, Stanford, CA, (2)Stanford University School of Medicine, Stanford, CA, (3)Division of General Medical Disciplines, Department of Medicine, Stanford University, Stanford, CA, (4)Department of Management Science and Engineering, Stanford University, Stanford, CA, (5)VA Palo Alto Health Care System, Palo Alto, CA

Purpose: Despite India's rapid economic development, it still has the world's largest number of active tuberculosis (TB) cases and hundreds of millions of people living in poverty. Assessments of TB control programs have primarily focused on health gains; yet, returns to such programs may be underestimated without also considering the relationship between TB and impoverishment. We developed, calibrated, and validated a model to consider the health and economic effects of TB in India.

Methods: We extended our prior dynamic transmission microsimulation of TB in India, stratifying it by socioeconomic status (SES), represented as real per-capita consumption expenditure. We parameterized SES in the model using the Indian National Sample Survey (1994-2012; n=5,548,989). The model captures relevant dynamics including: 1) secular economic growth (average improvement in real per-capita consumption expenditure over time) and shifts in the SES distribution; 2) TB transmission dynamics within and between members of SES groups; and 3) SES-specific risks of latent TB activation. We calibrated the model to overall demographic trends; WHO-reported TB trends (1996-2013); and age-, urban/rural- and SES-specific TB prevalence estimates from India's National Family and Health Survey-3 (NFHS-3). We validate the longer term impact of TB on SES using longitudinal data from the Indian Human Development Survey (IHDS) (2004/2005, 2011/2012; n=215,754).

Results: In addition to overall demographic trends, the model calibrates well to TB prevalence trends, and SES-specific TB prevalence in multiple subpopulations (Figure 1, Panels A-C). Figure 1 shows that while overall TB prevalence has declined, among the poorest group, it remains 2.5 times higher than in the wealthiest. While TB prevalence is generally higher in older ages, among the poorest of any age group, it is >3 times higher than similarly aged wealthy individuals. The model validates against longitudinal impacts of TB on SES. For example, both the model and IHDS data show that for individuals in the wealthiest SES groups, having TB in 2004 nearly doubles the chance of being in the poor SES group in 2011.

Conclusions: Our dynamic microsimulation captures relationships between TB and SES in India and strongly suggests that TB control could deliver substantial economic value beyond its direct health effects, an important consideration for future model-based economic evaluations of such programs.

10:15 AM

Jarrod E. Dalton, PhD, Cleveland Clinic and Case Western Reserve University, Cleveland, OH, Adam T. Perzynski, PhD, Case Western Reserve University at MetroHealth, Cleveland, OH, David A. Zidar, M.D., Ph.D., University Hospitals Case Medical Center, Cleveland, OH, Michael B. Rothberg, MD, MPH, Cleveland Clinic, Beachwood, OH, Claudia J. Coulton, Ph.D., Case Western Reserve University, Cleveland, OH, Alex T. Milinovich, BA, Cleveland Clinic, Cleveland, OH and Neal V. Dawson, MD, Case Western Reserve University at MetroHealth Medical Center, Cleveland, OH


   Inequality in health outcomes in relation to Americans' socioeconomic status (SES) is rising, despite recent evidence that the life expectancy gap between black and white Americans may be decreasing1,2. Cardiovascular disease, still the leading cause of death for Americans, merits study with respect to the socioeconomic spectrum. The objectives of our study were: i) to evaluate the relationship between neighborhood-level SES and major atherosclerotic cardiovascular disease (ASCVD)-related events (myocardial infarction, stroke, and cardiovascular death); and ii) to evaluate the relative extent to which neighborhood SES and physiological risk explain neighborhood-level variation in ASCVD event rates.


   We analyzed EHR data from 78,488 Cleveland Clinic patients living in Northeast Ohio who had an outpatient lipid panel drawn between 2007 and 2010, the date of which serving as study baseline. The follow-up time for major ASCVD events was 5 years. We applied Bayesian spatial analytic techniques3 (specifically, Weibull spatial autoregression with a Besag-York-Mollie4  covariance structure) to model ASCVD event rates across Northeast Ohio census tracts. Exposures of interest were census-tract-level socioeconomic status, which was defined as the first principal component of eight U.S. census measures (percent on Medicaid, percent uninsured, median income, percent with supplemental income, etc.); and the American College of Cardiology/American Heart Association Pooled Cohort Equations Risk Model (PCERM) estimated 5-year probability of major ASCVD events.


   We found substantial geographic variation in PCERM-adjusted ASCVD event risk that mirrored variation in neighborhood SES (see Figures). Neighborhood SES alone accounted for 29% of unexplained census-tract-level variation in ASCVD event rates, compared to 6.6% explained by the PCERM alone. Incrementally speaking, the PCERM explained 4.7% of this variation after adjusting for effects of neighborhood SES.


   Rates of major ASCVD events in low-SES communities were over three times that of high-SES communities, and neighborhood SES explained over four times the amount of neighborhood-level variation in event rates than the established ASCVD risk model. The incremental explained variation attributable to the physiological variable-based PCERM – beyond the explained variation from neighborhood SES – was small. SES needs to be incorporated into risk-based decision-making procedures for primary prevention of ASCVD.


[1] Bosworth (2016), Brookings Institution.
[2] Kochanek (2015), NCHS.
[3] Rue (2009), Journal of the Royal Statistical Society-B
[4] Besag (1991), Annals of the Institute of Statistical Mathematics.


10:30 AM

Ankur Pandya, PhD1, Djora Soeteman, PhD1, Ajay Gupta, MD2, Hooman Kamel, MD2 and Meredith Rosenthal, PhD1, (1)Harvard T.H. Chan School of Public Health, Boston, MA, (2)Weill Cornell Medicine, New York, NY

Purpose: Healthcare payers in the U.S. are increasingly tying provider payments to quality or value using pay-for-performance (P4P) policies. Cost-effectiveness analysis (CEA) quantifies value in healthcare but is not currently used to design P4P policies. We used a stroke simulation model to demonstrate how value-based acute ischemic stroke (AIS) payment adjustments (P4P incentives) could be determined using CEA.

Methods: We used a previously-published AIS simulation model to calculate the difference in population-level net monetary benefit (NMB, willingness-to-pay of $100,000/quality-adjusted life year [QALY]) accrued under current Medicare policy (stroke payment not adjusted for performance) compared to various hypothetical P4P scenarios. Performance measurement was based on time-to-thrombolytic treatment with tissue-type plasminogen activator (tPA). tPA within 0-3 hours of stroke onset leads to superior acute outcomes quantified by the modified Rankin Scale (mRS). In the model, time-to-tPA influenced acute mRS outcomes; discounted lifetime QALYs and stroke-related costs (payer perspective) were projected based on mRS outcomes. Time-to-tPA was modeled by adding time from: 1) stroke onset to hospital arrival; to 2) hospital arrival to tPA (“door-to-needle”). In P4P scenarios, we modeled door-to-needle time reductions of 0-30 minutes (range based on the national Get With The Guidelines initiative average 10 minute reduction) and tPA payment (unadjusted payment=$6,270) increases of 0-50% for timely (cost-effective) treatment.

Results: Without performance-based payment (i.e, current payment), tPA versus no tPA had incremental cost-effectiveness ratios of: $14,300/QALY in the 0-3 hour treatment window; $27,100/QALY in the 3-4.5 hour window; and was dominated in the 4.5-6 hour window. Compared to current payment, equivalent population-level NMB was achieved in P4P scenarios with 10 minute door-to-needle time reductions (6,392 more AIS cases/year in the 0-3 hour window) incentivized by increasing tPA payment by as much as: 10.5% for 0-4.5 hours; 12% for 0-3 hours and 6% for 3-4.5 hours; or 13.75% for 0-3 hours only. Population-level NMB differences between current and P4P payment scenarios depended on the sizes of performance improvements and payment incentives (Figure).    

Conclusions: Determining the optimal size of financial incentives used in P4P is an important challenge facing policy developers who seek to improve the value of healthcare delivered in the U.S. Value-based AIS payment adjustments can be set using CEA and a NMB framework that could be generalized to other quality measures across health conditions.

10:45 AM

Yao-Hsuan Chen, Ph.D.1, Paul G. Farnham, Ph.D.1, Christopher Goodrich, B.S.2, Benjamin Allaire, M.S.2 and Stephanie L. Sansom, PhD, MPP, MPH1, (1)Centers for Disease Control and Prevention, Atlanta, GA, (2)RTI International, Research Triangle Park, NC


We assessed the potential for HIV elimination in the United States, i.e., decreasing the HIV reproduction number to less than 1, and we evaluated the most effective improvements in the HIV care continuum to achieve this goal.


We used the HIV Optimization and Prevention Economics (HOPE) compartmental model to estimate the HIV reproduction number in the United States from 2006 to 2015. In this model, we stratified the population by transmission group, race/ethnicity, age, circumcision status, and HIV risk level and the HIV-infected subpopulations by disease progression and continuum of care stage. Transition rates among the approximately 4,000 subpopulations were based on the literature, expert opinion, and model calibration methods, ensuring that this model closely reflected both demographics critical to the spread of HIV and HIV infection in the United States between 2006 and 2010.

We estimated the HIV control reproduction number, Rc, using the next generation matrix method (NGM). Rc is an estimate of the average number of secondary infections from an infected person in a completely susceptible population calculated under the additional assumption that the susceptible population may be taking steps to prevent HIV acquisition.

We conducted a sensitivity analysis to find the degree of change from a base-case Rc in 2015 that would occur by altering model transition rates associated with the HIV care continuum such as being diagnosed with HIV, linked to care, and achieving viral load suppression (VLS). 


The estimated values of Rc were 1.46 and 1.50 for the periods 2006 - 2010 and 2011 – 2015, respectively. Holding other aspects of the care continuum constant, we found that Rc was insensitive to changes in rates related to diagnosis (% deviation from the base case Rc:  [-1%, 1%]) and linkage to care [-9%, 6%], but was significantly affected by rates related to VLS [-47%, 106%] (Figure 1). Substantially decreasing the VLS dropout rate was the only measure that reduced the value of Rc to less than one.


Current progress of HIV-infected persons along the care continuum is not sufficient to result in the potential elimination of HIV in the United States (Rc < 1). To achieve this goal, ensuring that HIV-infected persons achieve and maintain VLS will be an important focus.