Theses Doctoral

In cases of opioid overdose, do medical marijuana laws matter? A case-control study among National Health Interview Survey participants, 1986-2011

Kim, June H.

While the proportion of U.S. ambulatory, office-based visits with a primary symptom or diagnosis of pain has remained consistent from 2000 to 2010, the frequency of opioid prescribing among these visits nearly doubled from 11.3% to 19.6% [1]. Concurrently, the U.S. experienced an epidemic of opioid-related morbidity and mortality [2]. Medical marijuana, allowed in states with medical marijuana laws (MMLs), may serve as an alternative to opioids in the treatment of severe or chronic pain [3]. If marijuana is a substitute for opioids, MMLs that increase marijuana use may also be inadvertently decreasing opioid use. It has been hypothesized that this mechanism, marijuana for opioid substitution, may also be driving reductions in opioid-related morbidity and mortality [4, 5]. This dissertation has three aims. The first aim is to assess whether the hypothesis, that state MMLs can reduce opioid-related mortality, is plausible and whether it is consistent with the available literature. The second aim is to replicate prior state-level finding using individual-level data among participants surveyed in the National Health Interview Survey (NHIS), between 1986 and 2009. These individuals are followed up for mortality up to December 31st, 2011. The final aim is to assess whether MMLs have a heterogeneous impact across subpopulations defined by age, sex, and/or race/ethnicity.
In Chapter 1, we find evidence, from a limited number of quantitative studies, that show associations between more liberal marijuana policies and reductions in opioid prescribing, opioid positivity (i.e., use), opioid-related treatment admissions, and opioid-related overdose. From surveys, we found that a majority of medical marijuana patients use marijuana for indications where opioids are commonly prescribed and report reductions in prescription drug use, including opioids specifically. We found the overall quality of the quantitative studies to be moderate to strong. While results were farily consistent across studies, the reviewed studies all shared similar designs and assumptions. Further, regional heterogentiy in MMLs as well as opioid overdoses is never addressed.
In Chapter 2, among all NHIS adult participants eligible for mortality follow-up and surveyed between 1986 and 2009, we observed 791 cases who died of an opioid overdose. Compared to controls, cases were more likely to be male, middle-aged, non-Hispanic White, separated/divorced; less educated, and have a family income below the poverty threshold. After adjusting for matched calendar year, participant sex, age, race/ethnicity, marital status, educational attainment, and poverty level, we find no overall association between state MMLs and the rate of opioid overdose. Adjusting for region depreciated the association towards a protective effect. Upon stratifying by region, we find that state MMLs were associated with a reduced rate of opioid overdoses in the West between 2006-2011, but not in the Northeast.
In Chapter 3, we find no evidence that the association between state MMLs and opioid overdose is heterogeneous by race/ethnicity or sex. However, we do find evidence that age-dependent heterogentiy is present, and that this heterogeneity is magnified in the West. We find that Western MMLs are associated with a reduced overdose rate for individuals under the age of 60, but not for older adults. In the final chapter, we provide an overview of our findings in the context of the available literature, a discussion of the major strengths and weakness of our study findings, and a recommendation for the direction of future studies.
In conclusion, we find that hypothesis that MMLs can reduce opioid-related mortality is plausible, and that the likely mechanism is substitution. However, in our study, our results were not consistent with this hypothesis overall, and signficant reductions were only present after stratifying by region and by sampling frame. The discrepancy between our findings and prior studies should be explored, particularly in light of how regional variations may impact measures of association.


  • thumnail for Kim_columbia_0054D_14293.pdf Kim_columbia_0054D_14293.pdf application/pdf 522 KB Download File

More About This Work

Academic Units
Thesis Advisors
Martins, Silvia S.
Ph.D., Columbia University
Published Here
October 26, 2017