Medical marijuana laws are associated with increases in substance use treatment admissions by pregnant women

Abstract

Background and Aims:

Between 2002 and 2014, past month marijuana use among pregnant women in the U.S. increased 62%, nearly twice the growth of the general population. This growth coincides with the proliferation of state medical marijuana laws (MMLs) authorizing physicians to recommend marijuana for approved conditions. We estimated the association between MMLs and substance use treatment utilization among pregnant and nonpregnant women of reproductive age. We also examined whether the association varied across MML provisions, age groups, and treatment referral sources to clarify potential pathways.

Design:

Nationwide administrative data from the 2002 to 2014 Treatment Episodes Data Set Admissions, and a difference-in-differences design that exploited the staggered implementation of MMLs to compare changes in outcomes before and after implementation between MML and non-MML states.

Setting:

21 MML and 27 non-MML U.S. states.

Participants:

Pregnant and nonpregnant women ages 12 to 49 admitted to publicly funded specialty substance use treatment facilities.

Measurements:

The primary outcome variable was the number of treatment admissions per 100,000 women ages 12 to 49, aggregated at the state-year level (N=606). Admissions for marijuana, alcohol, cocaine, and opioids were considered. The primary independent variable was an indicator of MML implementation in a state.

Findings:

Among pregnant women, the rate of marijuana treatment admissions increased by 4.69 [95%CI=1.32, 8.06] in MML states relative to non-MML states. This growth was accompanied by increases in treatment admissions involving alcohol [β=3.19; 95%CI=0.97, 5.41] and cocaine [β=2.56; 95%CI=0.34, 4.79], was specific to adults [β=5.50; 95%CI=1.52, 9.47], and was largest in states granting legal protection for marijuana dispensaries [β=6.37; 95%CI=−0.97, 13.70]. There was no statistically significant association between MMLs and treatment admissions by nonpregnant women.

Conclusions:

Medical marijuana law implementation in US states has been associated with greater substance use treatment utilization by pregnant adult women, especially in states with legally protected dispensaries.

1. INTRODUCTION

Marijuana is the most used psychoactive drug worldwide, with an estimated 183 million annual users [1]. Although marijuana is prohibited in most places, an increasing number of countries have legalized it for medical purposes. In the United States, 33 states and the District of Columbia have passed medical marijuana laws (MMLs) authorizing physicians to recommend the use of marijuana for approved conditions [2]. While clinical evidence supports the efficacy of marijuana for treatment of select conditions [3,4], there is considerable debate regarding its safety and concern that MMLs may increase the use of marijuana and other substances in non-patient populations, particularly the youth and pregnant women [5,6,7,8]. Previous work focusing on the youth has generally found no association between MMLs and marijuana use [9,10,11,12,13,14,15,16,17]. Work focusing on the general population or adults is somewhat inconclusive, with a subset of studies documenting increases in marijuana use, marijuana use disorder and related behavioral healthcare including treatment utilization for marijuana [9,12,13,17,18,19,20,21,22].

The MML literature has yet to consider how these laws may affect pregnant women, a vulnerable population with rapid growth in marijuana use during recent years. Between 2002–2014, past month marijuana use among pregnant women increased 62%, nearly twice the growth of the general population [23,24]. Pregnant women might be inclined to use marijuana not only recreationally for its psychoactive properties but also medically for its antiemetic properties, which may help alleviate the nausea and vomiting that commonly accompany pregnancy [25,26,27,28]. While evidence of the effect of marijuana use during pregnancy on fetal health is limited and causality is difficult to establish, observational studies suggest that marijuana is associated with lower birthweight and impaired brain development [29,30,31]. The American College of Obstetricians and Gynecologists discourages physicians from prescribing marijuana during preconception, pregnancy, or lactation and recommends that pregnant women and nonpregnant women contemplating pregnancy discontinue using marijuana and other substances [32]. Screening and substance use treatment are also recommended for optimizing the behavioral healthcare of women with substance use disorders (SUD), including marijuana [33]. Despite these recommendations, state MMLs are mute regarding medical marijuana use by pregnant women. Although no state MML lists pregnancy as an approved condition for medical marijuana, most do list nausea and do not prohibit, nor warn women of potential risks [1,27].

This study is the first to consider how medical marijuana laws may affect the behavioral healthcare of reproductive age women, including pregnant and nonpregnant women ages 12 to 49, by focusing on treatment utilization for marijuana and other substances. Substance use treatment can help improve the health and social functioning of individuals with SUD and is particularly relevant for pregnant and nonpregnant women contemplating pregnancy as it can also improve fetal wellbeing. Since treatment utilization is a function of the number of substance users and the probability of entering treatment among substance users, MMLs may affect treatment utilization for marijuana by inducing changes in either of these components. Changes in the number of users may occur through greater access to marijuana, or reductions in perceived risk, fear or stigma of marijuana use. Changes in the probability of entering treatment may also occur through these or other pathways, even in the absence of changes in users. One example is increases in the rate of treatment referrals, which may occur if MMLs lead to greater marijuana use surveillance by healthcare providers and the criminal justice system. MMLs may affect treatment utilization for other substances as well, depending on whether a given substance is a complement or a substitute to marijuana or on whether marijuana serves as a “gateway” toward the use of another substance.

To determine whether MMLs may affect substance use treatment utilization among reproductive age women, we analyzed administrative data capturing admissions to publicly funded specialty substance use treatment facilities for the period 2002–2014 in 48 states. We aimed to (1) estimate the association between MMLs and treatment utilization for marijuana, alcohol, cocaine, prescription opioids, and heroin among pregnant and nonpregnant women of reproductive age and (2) assess whether the strength of the association varied across MML provisions (i.e. legally protected and operational dispensaries), age groups (i.e. youth/adults), and treatment referral sources (i.e. criminal justice, healthcare providers) as to clarify potential pathways. Our analysis provides policymakers with important and timely information regarding the implications of MMLs on a vulnerable yet understudied subpopulation for whom the need for substance use treatment might be especially crucial.

2. METHODS

2.1. Design

Our analytic sample comprised a panel of 13 years (2002–2014) and 48 states, 21 MML states (8 implementing between 1996–2001 and 13 implementing between 2002–2014) and 27 non-MML states. We removed 3 states (DC,NH,NM) due to missing data issues. We estimated difference-in-differences models at the state-year level that exploited the staggered implementation of MMLs to compare average changes in treatment admissions before and after MML implementation between MML (treatment group) and non-MML (comparison group) states.

2.2. Data and Measures

2.2.1. Substance Use Treatment Utilization

Admissions to substance use treatment facilities were drawn from the Substance Use and Mental Health Services Administration’s (SAMHSA) 2002–2014 Treatment Episode Data Set Admissions (TEDS-A). TEDS-A captures nearly all admissions to publicly funded specialty substance use treatment facilities in the U.S. and collects information on approximately two million admissions each year. Information includes the primary, secondary, and tertiary substances reported at the time of admission, along with client demographics, referral source, pregnancy status, and geographic identifiers.

We identified treatment admissions involving marijuana, alcohol, prescription opioids, heroin, or cocaine among pregnant and nonpregnant women ages 12–49. In secondary analyses, we categorized treatment admissions by referral source and age group. Referral sources included healthcare providers (i.e. substance use and other healthcare providers), criminal justice, and other sources (i.e. self, school, employer, community). Age groups included the youth (ages 12–17) and adults (ages 18–49).

Treatment admissions by pregnant and nonpregnant women ages 12–49 were aggregated into state-year counts, divided by state population estimates of women ages 12–49, and multiplied by 100,000. When analyzing data by age group, state-year counts of women ages 12–17 were divided by population estimates of women ages 12–17, and those of women ages 18–49 by population estimates of women ages 18–49.

2.2.2. State Medical Marijuana Laws

The effective dates of state MMLs were drawn from ProCon (Appendix Table A1) [1]. Using these dates, we constructed an MML indicator that was equal to one if a state had an MML in a given year and zero otherwise. In secondary analyses we considered the time when dispensaries became legally protected in a state. Since the effective date states grant legal protection for dispensaries does not always coincide with the date dispensaries become operational, we followed Powell, Pacula and Jacobson, (2018) and constructed a dispensary indicator based on the time when dispensaries became both legally protected and operational [34].

2.3. Analysis

2.3.1. MMLs and Substance Use Treatment Utilization

We estimated difference-in-differences models with ordinary least squares regressions and clustered standard errors at the state level [35]. All analyses were conducted using Stata 15. The main independent variable was the MML indicator. Each model controlled for state fixed effects to account for time invariant differences between states and year fixed effects to account for nationwide trends in admissions. Despite these controls, estimates may be subject to bias if we fail to account for other variables affecting outcomes and changing differentially over time in MML states. To minimize any such bias, we controlled for state differences in unemployment rates, beer excise tax rates, Medicaid income eligibility thresholds for pregnant women, the Affordable Care Act Medicaid expansions, pain clinic laws, prescription drug monitoring program operations and mandates, and recreational marijuana laws. Data sources of these control variables were summarized in the Appendix. To further alleviate concerns that state-year differences or other issues could be biasing estimates, we tested the robustness of main findings to changes in sample period and additional controls (Appendix Tables A3–A7). Additional controls included the rate of substance use treatment programs with specialized services for pregnant women, supportive state policies that fund substance use treatment programs for pregnant women, and punitive state policies considering substance use during pregnancy to be child abuse, requiring perinatal substance use testing, or requiring provider reporting of suspected perinatal substance use. We also classified admissions by setting (i.e. residential) to assess whether SAMHSA’s Services Grant Program for Residential Treatment for Pregnant and Postpartum Women could be biasing estimates (Appendix Table A8).

Difference-in-differences models are only valid under the assumption of parallel trends, implying that in the absence of the intervention, the difference in outcomes between treatment and comparison groups would continue to be constant over time. In an effort to assess this assumption and whether the strength of the association between MMLs and outcomes changed over time, we presented graphical evidence based on an event study approach that controlled for t leads and lags of the intervention. We did this by eliminating the MML indicator and instead including indicators that an admission occurred t years before (-t,…,−2,−1) or after (1,2,…,t) MML implementation. The reference group is t=0, the year right before an MML became effective in a given state. A test result of non-significance for the indicators of the pre-MML implementation period is suggestive of the plausibility of the parallel trends assumption.

2.3.2. Potential Pathways

MMLs may affect treatment utilization by inducing changes in either the number of substance users or in the probability of entering treatment among substance users.

Changes in the number of substance users may occur through easier access to medical marijuana, reductions in perceived risk, fear or stigma of marijuana use, or through illegal drug market responses such as lower cost/higher potency street drugs. Greater substance use may lead to a higher number of individuals developing SUD that require treatment or testing positive for substance use during perinatal drug screenings and being referred to treatment. Greater substance use might also induce risky sexual behaviors that increase the number of women who become pregnant [21]. To assess if changes in outcomes were consistent with greater substance use, we explored whether the strength of the association varied by ease of access to medical marijuana. Previous work has shown that MMLs granting legal protection for dispensaries increase access and are more strongly associated with changes in marijuana use [13,22]. Ease of access to medical marijuana may also vary by age group. For instance, minors must have permission of a parent or legal caregiver in order to use medical marijuana, which restricts access to patients under the age of 18 [14]. We considered the time when dispensaries became legally protected and operational in a state by including a dispensary indicator along with the MML indicator and classified pregnant women into age group subsamples (i.e. youth/adults).

Changes in the probability of entering treatment may also be affected through these or other pathways, and may occur even in the absence of changes in the number of substance users. One such pathway is increases in treatment referral rates. This might occur if increases in visits to healthcare providers among pregnant women who already use substances and are seeking medical marijuana lead to increases in treatment referrals rather than in marijuana use. This might also occur if MMLs reduce stigma and fear of disclosing marijuana use to healthcare providers, family, and friends, which could lead to increases in treatment referrals. Alternatively, MMLs may generate responses from the criminal justice system such as greater law enforcement of punitive policies for reducing prenatal substance use, which could also affect treatment referrals rather than substance use. To assess if changes in outcomes were consistent with increases in the probability of entering treatment, we classified admissions into referral sources (i.e. healthcare providers, criminal justice, other). We wanted to examine whether changes were primarily driven by referrals from a given source. Note that one would still expect changes across all referral sources if MMLs increase substance use in our population of interest.

3. RESULTS

Our analytic sample contained N=606 observations based on 282,955 treatment admissions from pregnant women and 6,273,520 treatment admissions from nonpregnant women ages 12–49. Summary statistics for outcome and control variables are in Appendix Table A2.

We found that MML implementation was associated with an increase of 4.69 [95%CI: 1.32, 8.06] marijuana treatment admissions by pregnant women per 100,000 women ages 12–49, a 33% increase over the baseline mean (Table 1). The event study plot suggests that the positive association between MMLs and marijuana treatment admissions strengthened over time, and that the parallel trends assumption held (Figure 1). Treatment admissions involving alcohol [β=3.19; 95%CI=0.97, 5.41] and cocaine [β=2.56; 95%CI=0.34, 4.79] also increased significantly. However, there was no statistically significant association between MML implementation and treatment admissions involving prescription opioids or heroin. We also found no statistically significant association between MML implementation and treatment admissions by nonpregnant women, regardless of the substance involved (Table 2).

Table 1.

Association between medical marijuana laws (MML) and substance use treatment admissions by pregnant women per 100,000 women

	Marijuana		Alcohol		Rx Opioids		Heroin		Cocaine
MML	4.69***		3.19***		1.93		0.76		2.56**
	[1.32, 8.06]		[0.97, 5.41]		[−1.34, 5.21]		[−2.38, 3.90]		[0.34, 4.79]
Mean	14.34		13.04		7.51		11.16		10.41
N	606		606		606		606		606

Source: Treatment Episode Data Set Admissions, 2002–2014.

Notes: Coefficients are based on a difference-in-differences approach that estimates changes in outcomes and accounts for controls. Controls include state unemployment rates, beer excise tax rates, Medicaid income eligibility thresholds for pregnant women, the Affordable Care Act Medicaid expansions, pain clinic laws, prescription drug monitoring program operations and mandates, and recreational marijuana laws. State-year treatment admission counts of pregnant women ages 12 to 49 are divided by state population estimates of women ages 12 to 49 and multiplied by 100,000. The mean captures average outcomes 2 years prior to MML implementation for MML states. Confidence intervals are in parentheses.

^***p<0.01,

^**p<0.05,

^*p<0.1

Figure 1: Association between medical marijuana laws (MML) and marijuana treatment admissions by pregnant women, by time since MML implementation.

Source: Treatment Episode Data Set Admissions, 2002–2014.

Notes: Coefficients and 95% confidence intervals are based on an event study approach that estimates leads and lags of the intervention and that accounts for controls. Controls include state unemployment rates, beer excise tax rates, Medicaid income eligibility thresholds for pregnant women, the Affordable Care Act Medicaid expansions, pain clinic laws, prescription drug monitoring program operations and mandates, and recreational marijuana laws. The reference year is t=0, the period right before MML implementation.

Table 2.

Association between medical marijuana laws (MML) and substance use treatment admissions by nonpregnant women per 100,000 women

	Marijuana		Alcohol		Rx Opioids		Heroin		Cocaine
MML	24.62		11.6		21.91		−9.04		−16.49
	[−10.33, 59.58]		[−42.17, 65.38]		[−16.09, 59.91]		[−65.04, 46.95]		[−55.16, 22.18]
Mean	308.03		498.28		170.66		269.11		288.5
N	606		606		606		606		606

Source: Treatment Episode Data Set Admissions, 2002–2014.

Notes: Coefficients are based on a difference-in-differences approach that estimates changes in outcomes and accounts for controls. Controls include state unemployment rates, beer excise tax rates, Medicaid income eligibility thresholds for pregnant women, the Affordable Care Act Medicaid expansions, pain clinic laws, prescription drug monitoring program operations and mandates, and recreational marijuana laws. State-year treatment admission counts of nonpregnant women ages 12 to 49 are divided by state population estimates of women ages 12 to 49 and multiplied by 100,000. The mean captures average outcomes 2 years prior to MML implementation for MML states. Confidence intervals are in parentheses.

^***p<0.01,

^**p<0.05,

^*p<0.1

The association between MML implementation and marijuana treatment admissions by pregnant women was stronger in states with legally protected and operational dispensaries (Table 3). In these states, there was an additional increase of 6.37 [95%CI=−0.97, 13.70] admissions relative to MML states without dispensaries. Moreover, the association between MMLs and marijuana treatment admissions by pregnant women held for adults but not for youths. There was an increase of 5.50 [95%CI=1.52, 9.47] marijuana treatment admissions by pregnant women ages 18–49 per 100,000 women ages 18–49, a 35% increase over the baseline mean. Finally, increases in marijuana treatment admissions by pregnant women were not driven by increases in criminal justice referrals and only partially driven by healthcare provider referrals [β=1.12; 95%CI=−0.13, 2.37]. Other sources such as self, school, employment, or community referrals played a more important role [β=2.55; 95%CI=0.48, 4.63].

Table 3.

Association between medical marijuana laws (MML) and marijuana treatment admissions by pregnant women per 100,000 women, potential pathways

	Ease of Access				Tx Referral Source
	Legal Dispensaries Full sample	Youth Subsample	Adult Subsample			Healthcare Provider Subsample	Criminal Justice Subsample	Other Sources Subsample
MML	4.11***	0.32	5.50***			1.12*	0.6	2.55**
	[1.24, 6.99]	[−3.01, 3.65]	[1.52, 9.47]			[−0.13, 2.37]	[−0.51, 1.71]	[0.48, 4.63]
Dispensary	6.37*
	[−0.97, 13.70]
Mean	14.34	6.36	15.77			2.74	4.03	7.22
N	606	606	606			602	602	602

Source: Treatment Episode Data Set Admissions, 2002–2014.

Notes: Coefficients are based on a difference-in-differences approach that estimates changes in outcomes and accounts for controls. Controls include state unemployment rates, beer excise tax rates, Medicaid income eligibility thresholds for pregnant women, the Affordable Care Act Medicaid expansions, pain clinic laws, prescription drug monitoring program operations and mandates, and recreational marijuana laws. The dispensary indicator captures the differential effect of MMLs in states with operational and legally protected dispensaries. The youth subsample includes marijuana treatment admissions by women ages 12 to 17 and the adult subsample includes marijuana treatment admissions by women ages 18 to 49. The mean captures average outcomes 2 years prior to MML implementation for MML states. Confidence intervals are in parentheses.

^***p<0.01,

^**p<0.05,

^*p<0.1

4. DISCUSSION

This study provides the first estimates of the association between MMLs and substance use treatment utilization among reproductive age women. Substance use treatment is an essential behavioral healthcare outcome for individuals with SUD, especially those who are pregnant or contemplating pregnancy as it can help improve both maternal and fetal wellbeing. Using treatment admissions data and a difference-in-differences design, we documented several key findings.

First, marijuana treatment admissions by pregnant women increased 33% over the baseline mean after MML implementation. While proportionately large, the effect size is rather small, implying an annual increase of about 3,800 admissions if MMLs were enacted nationwide. These effects were accompanied by increases in admissions involving alcohol and cocaine, possibly due to a complementarity between these substances or a gateway mechanism. Admissions involving prescription opioids or heroin did not change significantly. Our findings for marijuana are in line with previous work documenting increases in problematic use of marijuana after MML implementation [12,18,19,21]. For instance, Wen et al. (2015) found that past month and daily marijuana use among adults 21+ increased 14% and 15%, respectively. Baggio et al. (2018) found even larger proportional growth, with a 33% increase in past month marijuana use among adults ages 21–30, the prime childbearing years. As for other substances, our findings for alcohol coincide with studies documenting increases in binge drinking [12], but differ from studies documenting decreases in opioids [34,36,38,39,40]. While speculative, differences between pregnant women and individuals with chronic pain, who may be more likely to substitute opioids with marijuana, might explain this divergence.

Second, we found no evidence of increases in treatment admissions by nonpregnant women, which contrasts with findings for pregnant women. Several factors specific to pregnancy and affecting substance use or the probability of entering treatment might help explain this heterogeneity. One such factor is that pregnant women might be inclined to use marijuana for its antiemetic properties, which may help alleviate pregnancy related nausea and vomiting [25,26,27,28]. Indeed, observational studies document that pregnant women who experience these symptoms are significantly more likely to use marijuana than those who do not [26,28]. Additionally, MMLs might increase the number of women who use substances and become pregnant by inducing risky sexual behaviors along with greater substance use. A recent study found a 3% increase in birth rates after MML implementation, implying an annual increase of over 100,000 births if MMLs were enacted nationwide [21]. A third factor is that conditional on substance use, the probability of entering substance use treatment is higher for pregnant women. According to the National Survey of Substance Use and Health (NSDUH), 18% of pregnant women and 13% of nonpregnant women with marijuana use disorder received any treatment in the past year. One reason for these differences is that pregnant women might be more able to finance treatment since pregnancy is an eligibility criteria for Medicaid. About 50% of pregnant women in our sample had Medicaid, while only about 30% of nonpregnant women had Medicaid and the majority were uninsured. Another reason is that women are generally more exposed to the healthcare system during pregnancy, which creates opportunities for drug screening that could lead to treatment. This might be especially salient in states with punitive policies requiring prenatal drug testing or authorizing civil commitment of pregnant women who use drugs.

Another key finding was that the strength of the association between MMLs and marijuana treatment admissions was not driven by any particular referral source but did vary by ease of access to medical marijuana. Increases in marijuana treatment admissions by pregnant women were largest in states granting legal protection for dispensaries. This is not surprising in light of previous work showing that dispensaries are associated with greater marijuana use in the general population [13]. Another study shows that 69% of dispensaries in Colorado recommend marijuana products for treating nausea and vomiting in pregnancy [41]. We also found that increases in marijuana treatment admissions by pregnant women were specific to adults. Consistent with previous work finding no association between MMLs and marijuana use among the youth, marijuana treatment admissions by pregnant women ages 12–17 did not change. Previous work also shows that teenagers in need of treatment are less likely to receive specialty treatment than adults, which might also help explain a lack of effects for youths [24].

4.1. Limitations

This study has several limitations. First, while many studies have relied on TEDS-A to study MMLs [13,14,18,34] or pregnant women more generally [42,43,44,45,46], this data is not without limitations. On any given year, one to four states fail to report pregnancy status for a significant proportion of the population (the percentage of missing pregnancy status among females in our sample is about 6%). Since this issue affects only a minority of concentrated states, we removed three states especially subject to missing data. Moreover, TEDS-A does not capture substance use treatment at non-specialty settings such as primary care or self-help groups. Nonetheless, since according to NSDUH about 67% of pregnant women in substance use treatment receive it at specialty settings, TEDS-A should capture a large proportion of our population of interest. Additionally, TEDS-A measures admissions rather than individuals and there is a possibility of counting a same individual who had multiple admissions over the years. Another shortcoming is that we do not observe substance use but rather observe substance use treatment utilization. Finally, it is possible that unobserved factors might be biasing our estimates.

5. CONCLUSION

We found that MML implementation is associated with greater substance use treatment utilization by pregnant women, especially in states with legally protected and operational dispensaries and specifically among adults. Taken together, our findings appear to be consistent with greater use of marijuana and other substances through greater access to marijuana. It is plausible that increases in access to marijuana were accompanied by reductions in perceived risk, fear or stigma of marijuana use during pregnancy, which might also affect substance use. A study using nationally representative data found that the proportion of pregnant women reporting past month marijuana use and believing marijuana use poses no harm increased 153% over the last decade [47]. Another study found that 31% of pregnant women who ever used marijuana did not believe marijuana use during pregnancy was harmful for the baby [48]. Since we do not directly observe the use of marijuana or other substances, we cannot rule out the possibility that MMLs increased the rate of treatment referrals without increasing substance use among pregnant women or that MMLs induced other market responses affecting the probability of entering treatment. The specific pathways through which MMLs might increase substance use treatment utilization by pregnant women remain a topic for future research. Regardless of the pathway, at minimum our findings suggest that policymakers in MML states must ensure sufficient availability of comprehensive behavioral healthcare services to meet the treatment needs of pregnant women with SUD.

While self-reported prevalence of marijuana use among pregnant women is small, given nationwide growth in marijuana use by pregnant women, changes in public perceptions about the safety of marijuana use during pregnancy, and findings from our study linking MMLs to increases in substance use treatment utilization by pregnant women, more research is needed to understand the effects of MMLs and marijuana use on maternal and fetal health to generate a clear public health message that can guide MML legislation and educate healthcare providers and expectant mothers.