Click here for list of publications on google scholar.

Forthcoming — Accepted Manuscripts

Provider Practice Competition and Adoption of Medicare's Oncology Care Model

Journal: Medical Care

Background: There is concern that the Oncology Care Model (OCM), a voluntary bundled payment program, may incentivize mergers and acquisitions among physician practices leading to reduced competition and price increases. These concerns are heightened if OCM is preferentially adopted in competitive healthcare markets, but little is known about the characteristics of markets where OCM is adopted. 

Objective: We test the hypothesis that adoption of OCM was greater in more competitive medical oncologist markets.

Research Design: The Herfindahl-Hirschman Index (HHI), a measure of competition, was calculated for Hospital Referral Regions (HRRs) using secondary data from the Centers for Medicare and Medicaid Services. The relationship between HHI and OCM adoption was assessed using a two-part regression model adjusting for market-level number of practices, physician density, average practice size, sociodemographic characteristics, and medical resources. 

Subjects: 10,788 providers who billed Medicare for oncology services in 2015. 

Results: OCM was adopted in 114 (37%) of the 306 HRRs. Practices in competitive healthcare markets were more likely to adopt OCM than in non-competitive markets with a 15.7 percent difference in the predicted probability of adoption comparing the interquartile range of HHI. Average size and number of practices in an HRR were positively associated with adoption, while rate of full-time equivalent hospital employees to 1,000 residents was negatively associated with adoption.

Papers Under Review (Primary Author)

Minimum Wage and Infant Mortality

Journal: Journal of Human Resources

Interest in the non-economic impact of minimum wage laws have led to a growing literature examining how population health measures respond to state minimum wage increases in the US. These studies rely on the canonical two-way fixed effect estimator, which is an unreliable method to infer causal effects with spatially heterogeneous populations. In this paper, I study the impact of minimum wage differences across states and within-state ordinances on infant mortality rates. I isolate the causal effect of the minimum wage by employing a spatial regression discontinuity design using contiguous counties in different states that share a common state border as local treatment and control populations. I find that a 10 percent increase in local cost-of-living adjusted minimum wage reduces infant mortality rates by 3.2 percent among lower-educated mothers--a group more likely affected by minimum wage changes. I find that reductions in infant mortality is concentrated in the post-neonatal period. I also demonstrate that the traditional empirical approach employed in this literature will produce lower estimates of the health impact of the minimum wage. My findings provide robust evidence that minimum wage increases are causally linked with improved infant health.

Cost-Effectiveness of Nusinersen in a Universal Screening Setting for Spinal Muscular Atrophy

Journal: Pediatrics

Background:Nusinersen has been demonstrated to be effective in treating spinal muscular atrophy (SMA). Despite efficacy, initial studies have concluded that nusinersen is not cost-effective given the high dosage price ($125,000). However, little is known about the role of genetic screening on the economic value of nusinersen. The objective of this study was to evaluate the cost-effectiveness of nusinersen for infantile-onset SMA, the most severe case, in the setting of universal newborn screening.

Methods:Markov simulation model using data from published and preliminary results of clinical trials with U.S. epidemiologic and cost data was developed. Primary interventions studied were nusinersen treatment in a screening setting, nusinersen treatment in a non-screening setting, and standard care. Analysis was conducted from a societal perspective. 

Results:Nusinersen treatment was weakly dominated by nusinersen treatment and screening in the base-case analysis. The incremental cost-effectiveness ratio (ICER) for nusinersen with screening was $339,793 per event-free LYs saved. We find that at a per-dose price of $23,361, nusinersen and screening is a cost-effective strategy given a willingness-to-pay (WTP) threshold of $50,000 per event-free LYs saved, $41,813 at a WTP of $100,000, and $60,266 at a WTP of $150,000. Preliminary data from the NURTURE trial indicated an 85.7% improvement in expected LYs saved compared to our base results. In probabilistic sensitivity analysis, nusinersen and screening was a preferred strategy at a $500,000 WTP threshold proposed for rare diseases.

Conclusions: Expanded newborn screening for SMA provides improved economic value for payers and patients if nusinersen is available.

Working Papers (Primary Author)

Subjective Life Expectancy as a Source of Health Disparities?

A large volume of literature links social and economic factors with mortality and life expectancy. This essay examines to what extent these socio-economic factors, specifically income, impact subjective life expectancy (SLE) or an individual’s belief in surviving to a certain age. I first develop a model to describe how SLE may play an important role in linking wage increases to favorable health behavior by making future time costs of adverse health events binding. I then estimate a heterogeneous choice proportional odds model using previously unexplored data from the Federal Reserve's 2013 Survey of Household Economics and Decision-making (SHED) to study this relationship. I find that Income and higher education is strongly associated with improved confidence in survival to age 75. I also find that periods of unemployment, disability, divorce, negative life event associated with the great recession were negatively associated with SLE. My study contributes to the health disparities literature by providing evidence for a new channel connecting socio-economic factors and health outcomes. 

Cost of Primary Single Level Lumbar Discectomy


Improving value in surgical care requires a detailed understanding of the true costs of surgical and operating room resources.


In this study, we sought to identify predictors of health care costs and operating room resource use for patients undergoing initial single-level discectomies using a unique institutional cost dataset from the value driven outcomes (VDO) program developed by the University of Utah.


A retrospective cost analysis of primary single-level lumbar discectomies.


A total of 622 patients undergoing initial single-level lumbar discectomy between January 2014 to April 2016 at the University of Utah Orthopaedic Center.


Health care costs, clinical length of stay, and operating room time.


Univariate and multivariate generalized linear models (GLM) performed to identify predictors of health care costs, clinical length of stay (LOS, days) and operating room time (OR, minutes) using a unique costing tool. Modified park test procedure was implemented to determine GLM error distribution specifications and standard errors were clustered by provider. Cost outcomes were normalized using mean costs for a patient with normal body mass index (BMI) and a healthy American Society of Anesthesiologists (ASA) classification system. Average marginal effects in multivariate analysis were reported as percentage of normalized costs.


Advanced age, male gender, Hispanic, Black, unemployment, and being retired were significant positive predictors of costs in univariate analysis. Asian, Native American, outpatient procedures, and being a student were significant negative predictors of costs. Univariate analysis also showed that obesity, higher ASA class and insurance status were also positive and significant predictors of costs. In multivariate analysis, we found that obesity led to higher average marginal total direct (9%), total facility (15%), and facility OR costs (22%), as well as an average marginal effect of 24 more OR minutes per surgery. Conversely, being underweight was associated with lower average marginal total direct (-23%), total non-facility (-54%), total facility (-8%), facility OR costs (-15%), as well as lower clinical LOS (-.42 days), and 18 less OR minutes. While being overweight was not significantly associated with greater total direct costs, it was associated with higher total facility (8%), and facility OR costs (12%), with an average marginal effect of 11 more OR minutes per surgery. Age was associated with a longer LOS but not with OR costs in multivariate analysis. As expected, outpatient surgical costs, LOS and OR time were significantly lower than inpatient procedures. Patients with incapacitating disease ASA Class had the highest predicted OR time in GLM estimates with an average marginal effect of 28 minutes, while severe systematic disease was associated with greater total and nonfacility costs. In addition, Medicare patients had higher facility costs (14%) compared to privately insured patients.


Although many of our observed variables were significant predictors of costs in univariate analysis, multivariate GLM estimates showed that the main variables determining cost and operating room time and resource use were obesity (BMI), general health (ASA class), Medicare status and advanced age, and whether the surgery was an outpatient or inpatient procedure.

Cost effectiveness of current and potential serum based colorectal screening strategies: can a serum based test do better?

Background: Despite efficacy of colorectal cancer (CRC) screening, recent trends in screening rates have not improved. Furthermore, national society recommendations have not prioritized a single screening modality. Both flexible sigmoidoscopy (Flex-Sig) and guaiac fecal occult blood test (gFOBT)) have been recommended. Studies suggest that noninvasive tests may improve screening rates, however, compliance remains modest for such tests. A recent study found that health system transition from gFOBT to a fecal immunochemical test (FIT) moderately improved test compliance rates. However, FIT is an expensive substitute for gFOBT and sensitivity of fecal based tests for precursor lesions in the colon remain inferior to structural examinations. Serum-based blood tests (SBT) may provide higher compliance rates, but necessary accuracy for such a test to be a dominant or cost-effective screening strategy over current recommendations have not been determined. This study analyzed the cost-effectiveness of multiple CRC screening strategies and estimated minimum test characteristics of a hypothetical SBT to be a preferred screening strategy over current modalities.

Methods: A Markov microsimulation model was developed to analyze the costs and effects, (quality-adjusted life-years, QALYs) of CRC screening following a hypothetical cohort of 10,000 individuals at age 50 until death. Our model considered three strategies: 1) initial colonoscopy followed by Flex-Sig every 5 years, 2) initial colonoscopy with yearly gFOBT, and 3) initial colonoscopy with yearly FIT. An initial colonoscopy with a hypothetical yearly SBT was also examined at a range of test sensitivities. A long-term payer perspective was assumed for professional and facility costs and lifetime costs of cancer treatment. Model probabilities and utility values were obtained from the literature or calculated from National Vital Statistics Reports. 

Results: Annual gFOBT was the least costly strategy ($3,242) while annual FIT was eliminated through extended dominance. Flex-Sig was the costliest strategy at $4,667 but had the highest expected QALYs at 19.68. The incremental cost-effectiveness ratio (ICER) of Flex-Sig relative to gFOBT was $35,615. Our secondary analysis demonstrated that a minimum joint sensitivity of 70% and 60% is required for an SBT at 80% and 100% compliance rates, respectively, to achieve extended dominance of both Flex-Sig and FIT by SBT and gFOBT. ICER’s of SBT ranged from $12,984 to $4,779 in these sensitivity ranges compared to gFOBT with expected QALYs as high as 19.96. Total individual costs in the model varied from $365 to $111,535.

Conclusion: With improved sensitivity and high compliance rates, annual screening for CRC via a non-invasive SBT is a cost-effective approach compared to structural examinations and currently recommended gFOBT and FIT.