Joseph LeCates: Research

• Home
• Curriculum Vitae
• Research
• Personal
• Contact

Current Research

"Accounting for Inaccuracies in Retrospective Duration Data: A Monte Carlo Study of Smoking Cessation "

with Donald Kenkel and Feng Liu. November 2014.

• Abstract: Even when contemporaneous data do not exist, retrospectively collected information provides opportunities to analyze the determinants of past behaviors. However, such data are often plagued by "heaping, the tendency of survey responses to be concentrated at certain values due to recall bias. In the context of a discrete-time hazard model, heaping results in misclassification of a binary outcome, which may substantially bias estimated regression coefficients and marginal effects. We present a model of the heaping process in a discrete-time hazard setting with significant heaping in respondent recall: smoking cessation. The 2002 Current Population Survey Tobacco Use Supplement provides a basis for a Monte Carlo analysis to quantify the bias introduced from heaping and compare several methods proposed to account for such bias. Results suggest that the bias in estimated regression coefficients and marginal effects in a discrete-time hazard model of smoking cessation are modest: less than 7% difference for estimated coefficients and 5% difference for marginal effects. Methods intending to account for such bias typically performed worse than the naïve model that ignores heaping. However, additional simulations of a policy intervention suggest that under some circumstances the biases from heaping are substantial.

Papers in Publication

"Identification and Comparison of Patients with Hemophilia A Receiving Prophylactic versus On-Demand Factor VIII Treatment Regimens Using Health Care Claims Data" (Under Review, Blood Clotting and Fibrinolysis)

with Jennifer Pocoski, Abhijit Gadkari, Mei Sheng Duh, and Francis Vekeman. October 2014.

• Abstract: Health care claims databases have sufficient patient counts to provide real-world evidence comparing prophylactic versus on-demand factor VIII ("FVIII") treatment for hemophilia A ("HemA"), yet lack treatment regimen information available in specialty pharmacy databases necessary for those comparisons. We used age and FVIII units, data elements available in both data sources, from 3,645 specialty pharmacy patients with HemA, age 2 years or older, receiving only one regimen, and without bypassing agents, to calibrate a flexible algorithm for predicting treatment regimen in claims data. The algorithm maximizing the product of sensitivity and specificity identified the following age groups and annual FVIII IU thresholds: 2-7 [16,480]; 8-12 [32,770]; 13-16 [66,920]; 17-21 [122,757]; 22+ [204,552]. Annual FVIII use exceeding these thresholds predicted prophylactic treatment. We applied these results to a claims data sample of Medicaid patients with HemA, at least 2 filled FVIII prescriptions after age 2, continuous coverage for at least 6 months before and 12 months after the eligible FVIII dispensing, and without bypassing agent or desmopressin use. Four hundred forty-eight patients met the eligibility criteria, with 225 (50.2%) identified as receiving prophylactic FVIII treatment. Younger patients were more likely to receive prophylaxis (percent [age group]: 61% [2-7]; 70% [8-12]; 63% [13-16]; 43% [17-21]; 20% [22+], mean (SD) age, prophylaxis: 10.7 (9.8); on-demand: 18.7 (14.3) years old). With average follow-up greater than 5.5 years for both treatment groups, using the algorithm's results with Medicaid claims allows the opportunity for long-term, real-world comparisons of prophylactic versus on-demand FVIII treatment.

"Positive Subjective Measures in Abuse Liability Studies and Real-World Non-Medical Use: Potential Impact of Abuse Deterrent Opioids on Rates of Non-Medical Use and Associated Health Care Costs" (Accepted, Journal of Opioid Management)

with Alan G. White, Howard G. Birnbaum, Wendy Cheng, Carl L. Roland, and Jack Mardekian. September 2014.

• Objective: The objective of this study was to quantify the potential impact of reductions in positive subjective measures from human abuse liability studies on real-world rates of non-medical use of prescription drugs and associated health care resource utilization and costs.
• Design: Positive subjective endpoints "overall drug liking," in-the-moment "drug liking," and "drug high" E_maxs (peak effects) were recorded from published studies. Non-medical use data were obtained from the 2010 NSDUH and DAWN surveys. Multivariate regressions evaluated the association between the positive subjective endpoints and non-medical use rates, controlling for prescription volume, whether the drug is an opioid, and controlled substance schedule. A published budget-impact model was used to assess health care resource utilization and cost impacts of abuse deterrent opioid formulations.
• Results: A 5-point reduction in overall drug liking/drug liking/drug high E_max was associated with a 0.25/0.10/0.05 (standard errors: 0.11/0.12/0.07) percentage point decrease in the NSDUH lifetime non-medical use rates. Those decreases yielded a 11.3/4.2/2.1-% reduction compared to the samples' lifetime non-medical use rates of 2.21/2.38/2.36-%. Based on a number of assumptions, these reductions were associated with private payer cost reductions for a morphine and oxycodone abuse deterrent formulation in the ranges of $147.9 - 324.1 million and $230.7 - 958.7 million, respectively.
• Conclusions: Reductions in overall drug liking were significantly associated with reduced real-world non-medical use, healthcare utilization and costs. Associations using drug high and drug liking were directionally consistent with this finding though not statistically significant. A reduction in positive subjective measures associated with an abuse deterrent formulation has potential to reduce abuse and associated healthcare utilization and costs.

Working Papers

"Perception Spillovers Between Bundled Goods: A Model of Physician Learning from Chemotherapy Choice"

February 2014.

• Abstract: When consumers evaluate a product that is part of a bundle of goods, they can learn about the value of other, similar bundles. Such perception spillovers affect the adoption rate of new products, particularly for entrants that are strong complements to available goods. I extend current models of consumer behavior by developing a Bayesian model of learning which allows across-alternative perception spillovers in the context of chemotherapy for colorectal cancer. Using cancer surveillance data with patient and physician characteristics, I estimate a model of treatment choice to test whether there are informational spillovers across chemotherapy regimens. Despite sharing their major component drug, regimens entering the market in 1991 were not initially considered by physicians to be significantly correlated (ρ = 0.02). In contrast, physicians' initial beliefs for those regimens with irinotecan, introduced in 1996, exhibited considerable correlation (ρ = 0.71); physicians updated their beliefs for both regimens from experiences with either. Simulated counterfactuals demonstrate that learning and spillovers mitigate the costs of uncertainty by more than 60%. These results provide a richer perspective for consumer learning; in particular, physicians use patient feedback to update their beliefs about regimen quality, even across non-prescribed regimens.

"Patient Preferences and Willingness to Pay for Enhancements to Subcutaneous Injection Devices Used to Treat Rheumatoid Arthritis"

with Alfred A. Cividino, Udayasankar Arulmani, Peggy Lin, Namita Tundia, Martha Skup, Jingdong Chao, Parvez Mulani, and Yanjun Bao. July 2013.

• Background/Purpose: Subcutaneous (SC) injection is used to deliver certain biologic therapies for moderate to severe rheumatoid arthritis (RA). Characteristics of injection devices may affect patient preferences for therapy administration. This study assessed patient preferences for enhancements to SC injectable devices.
• Methods: An online patient survey employing a discrete choice experimental design was conducted in patients with RA recruited from a representative research panel. Patients were at least 18 years of age and, based on self-report, had a medical diagnosis of RA and were currently using prescription oral or injected medication for RA. Patients were asked to choose between hypothetical treatments with varying SC injection device characteristics: needle size (27-gauge (G) or thinner 29-G), manufactured without natural rubber (yes/no), and storage temperature requirements (requires constant refrigeration or can be left at room temperature for up to 2 weeks). To assess the value of the improved SC injectable device, we estimated the willingness of patients to pay (WTP) for those improvements in terms of their out-of-pocket copayment. Multivariate logistic regression was used to estimate associations between treatment choices and device-characteristics, and WTP for preferred treatment choices was calculated. Data were analyzed for all patients and for the subgroups of injection-experienced and injection-naïve patients.
• Results: The majority of the 797 study participants were women in their mid- to late-50s with mean duration of RA about 14 years; 32% had previously used injected drugs as therapy for RA. A device manufactured without natural rubber that had a 29-G needle and could be stored without refrigeration for up to 2 weeks was preferred by 74% of all participants compared to a device manufactured with natural rubber that had a 27-G needle and required refrigeration. WTP for these enhancements was $29.03 for a without natural rubber device, $26.88 to reduce needle size from 27G to 29G, and $16.90 for the convenience of storing the device at room temperature for up to 2 weeks. Preference for enhancements was similar among injection-experienced and injection-naïve patients (see Figure). No statistically significant differences in WTP were seen for injection-experienced vs injection-naïve patients (manufactured without natural rubber, $34.01 vs. $26.75; thinner needle, $31.92 vs $24.69; ability to store the device at room temperature for 2 weeks, $16.05 vs $16.82). Overall, patients were willing to add $72.81 (184%) to their $39.62 average, monthly, RA-specific drug copay for the new device.
• Conclusions: Regardless of previous experience with injectable therapy, patients with RA highly valued a SC injection device that had a thinner needle, did not require continuous refrigeration, and was manufactured without natural rubber.