David Beheshti

The onset of the opioid crisis coincided with the beginning of nearly 15 years of declining labor force participation in the US. Furthermore, the areas most affected by the crisis have generally experienced the worst deteriorations in labor market conditions. Despite these time series and cross-sectional correlations, there is little agreement on the causal effect of opioids on labor market outcomes. I provide new evidence on this question by leveraging a natural experiment which sharply decreased the supply of hydrocodone, one of the most commonly prescribed opioids in the US. I identify the causal impact of this decrease by exploiting pre-existing variation in the extent to which different types of opioids were prescribed across geographies to compare areas more and less exposed to the treatment over time. I find that areas with larger reductions in opioid prescribing experienced relative improvements in employment-to-population ratios, driven primarily by an increase in labor force participation. The regression estimates indicate that a 10 percent decrease in hydrocodone prescriptions increased the employment-to-population ratio by about 0.7 percent. These findings suggest that policies which reduce opioid misuse may also have positive spillovers on the labor market.

The US is currently in the midst of the worst drug overdose epidemic in its history, with nearly 64,000 drug overdose deaths in 2016. In response, pharmaceutical companies have begun introducing abuse-deterrent painkillers, pills with properties that make the drug more difficult to misuse. The first such painkiller, a reformulated version of OxyContin, was released in 2010. Previous research has found no net effect on opioid mortality, with users substituting away from OxyContin toward heroin. This paper explores health effects of the reformulation beyond mortality.  Exploiting variation across states in OxyContin misuse prior to the reformulation, I find large relative increases in the spread of hepatitis B and C in states most likely to be affected by the reformulation. In aggregate, the estimates suggest that absent the reformulation we would have observed between 66-75% fewer cases of hepatitis C and 46-60% fewer cases of hepatitis B. I document further evidence that points to the likely cause of these effects: the reformulation led individuals to substitute from OxyContin to heroin, which is substantially more likely to be injected, increasing exposure to blood-borne diseases.

In this paper, we examine the importance of individual physicians in explaining the significant variation in prescription drug spending in Medicare Part D. By tracking prescribing behavior before and after physician relocations, we find that movers' prescribing converges toward the average of their new location. However, this convergence is far from complete, highlighting the importance of idiosyncratic physician-specific factors. Overall, these physician-specific factors explain about 60 to 70 percent of the cross-sectional variation in prescription drug spending, suggesting that physicians are one of the most important supply-side determinants of this variation. We investigate several potential mechanisms behind this partial convergence.

A growing literature has examined how mandatory access prescription drug monitoring programs (MA PDMPs), laws that require providers to consider a patient's prescription history before prescribing controlled substances, affect opioid-related outcomes. However, little is known about their impact on non-opioid-related outcomes. In this paper, we examine the effect of MA PDMPs on prescribing patterns of stimulants and benzodiazepines. Using a difference-in-differences event study design, we show that MA PDMPs led to decreases in stimulant prescribing. In contrast, we find suggestive evidence that benzodiazepine prescriptions increase following the implementation of a MA PDMP. Our findings highlight that MA PDMPs do have effects on non-opioid drug prescribing, but these effects differ substantially across drug types. 

This paper studies how the prevalence of opioids affects joint physician-patient decisions over medical procedures. Following Alpert et al. (2019), we utilize variation in OxyContin exposure due to state policies that affected OxyContin's marketing and market entry. Preliminary results suggest that higher availability of opioids led to a substantial increase in the number of discretionary surgical discharges (e.g., knee replacements, hip replacements, and back surgeries), but did not meaningfully affect discharges for non-discretionary surgical procedures or medical discharges. 

Over the last two decades the federal and state governments have implemented a wide variety of policies intended to reduce unnecessary opioid prescribing, diversion, and abuse. An important policy action has been statewide mandatory access prescription drug monitoring programs (PDMPs) that help prescribers identify patients at ‘high risk’ for suspected misuse, diversion, and doctor shopping, and have been found to significantly reduce opioid prescribing, but have been also linked to increased use of illicit substances like heroin and related overdose mortality. Patient advocacy groups have expressed concerns that restrictions, like mandatory PDMPs, which reduce access to necessary pain medications may explain some of the unintended substitution towards illicit substances by individuals suffering from severe pain, and who are no longer able to receive adequate opioid analgesics. This study examines whether supply side restrictions on opioid prescribing, as a result of mandatory access state PDMPs, have led to increased pain in settings of outpatient, hospital inpatient and long-term nursing home residents. Using multiple self-reported and proxy measures of pain (use of over-the-counter (OTC) pain medication), across several proprietary data sets capturing nationally representative samples of commercially insured, Medicare and difference-in-differences framework, we find no evidence that these programs have led to increased pain in these populations. These results suggest that targeted policies like state PDMPs may be successful in reducing inappropriate prescribing, without restricting access to necessary pain medication.

This project studies the real-world impact of pre-exposure prophylaxis (PrEP) on the incidence of human immunodeficiency virus (HIV). Although PrEP has exhibited high efficacy in clinical trials, the decline in HIV infections in the U.S. plateaued following its approval in 2012. To estimate the real-world effectiveness of PrEP, we leverage the fact that the drug is almost exclusively taken by men who have sex with men. We use a synthetic-control estimator comparing counties with similar trends in HIV infections but different male same-sex partnership rates. While our causal estimates are consistent with evidence from clinical trials, they raise the question of why PrEP did not result in a reduction in aggregate HIV infections. In a battery of exercises, we show that the gap between clinical efficacy and real-world effectiveness is driven by differential take-up between Black and White men.

Syphilis, a sexually transmitted infection that can lead to serious health complications, was almost eliminated in the United States by 2000. But since then, its incidence began to increase, recently reaching a 60-year peak. We suggest that the introduction of the highly active antiretroviral therapy (HAART) drug regimen, which transformed HIV into a manageable chronic disease, is partly responsible, as HIV+ and HIV- individuals altered their sexual behavior after the introduction of HAART. To test this empirically, we exploit variation in HAART takeup based on spatial variation in pre-HAART AIDS prevalence, sex, and time in a triple differences framework. We find that a one standard deviation increase in the pre-HAART AIDS prevalence rate led to a 21 percent increase in the syphilis incidence rate, and that in the absence of HAART, there would have been 78 percent fewer syphilis cases between 1996 and 2006. These results highlight the need to consider unintended consequences that could stem from behavioral changes following the introduction of life-saving medical innovations.