Prenatal drug use criminalization and health system avoidance: Evidence from births in Alabama, South Carolina, and Tennessee, 1989–2019

Abstract

Background

States are increasingly adopting legislation that explicitly criminalizes drug use during pregnancy. Evidence from Tennessee suggests such policies may lead to increases in health system avoidance, yet prior research has not explored this relationship in other contexts. We examined whether explicit criminalization policies in Alabama, South Carolina, and Tennessee were associated changes in prenatal care and facility-based delivery at the population-level.

Methods

Outcome data were derived from birth certificate records aggregated to the county-year-level between 1989 and 2019. Primary outcomes included the prevalence of any prenatal care and facility-based delivery, while secondary outcomes included captured dimensions of prenatal care quality, specifically timeliness (i.e., first trimester initiation) and adequacy (i.e., completed number of recommended visits). We used two-way fixed effects models to test associations, also controlling for co-occurring policy changes and other time-varying confounders.

Results

State adoption of an explicit prenatal drug criminalization policy was associated with 4396.29 fewer births per 100,000 receiving any prenatal care (95% Confidence Interval [CI]: −6176.07, −2616.51). Additionally, criminalization was associated with 1847.99 fewer facility-based deliveries (CI: −3688.29, −7.69), though some estimates were imprecise across several sensitivity analyses. We observed no clear changes in the prevalence of timely (−1328.43; [CI: −3108.21, 451.36]) or adequate (−792.21; [CI: −2571.99, 987.58]) prenatal care following criminalization, though decreases in timeliness were suggested.

Discussion

We found that policies explicitly criminalizing prenatal drug use were associated with some reductions in prenatal care and facility-based delivery, suggesting that such punitive approaches may be associated with health system avoidance.

Introduction

Health services utilization throughout the perinatal period is an important determinant of pregnancy outcomes. Comprehensive perinatal care reduces the risk of complications including certain adverse birth outcomes, and provides opportunities to engage pregnant patients in other preventive health services expected to improve outcomes (Office of the Surgeon General, 2020). While these benefits extend to all pregnancies, those among people who use drugs likely accrue some of the greatest benefits associated with early and regular care engagement (Jones et al., 2014). Despite greater need however, consistent evidence indicates that pregnant people who use drugs are significantly less likely than the general population to receive appropriate services (Schempf & Strobino, 2009).

Although multiple factors contribute to this disparity, one significant barrier is potential legal consequences associated with prenatal drug use. Starting in the early 2000s, there has been a significant upswing in the number and type of pregnancy-related substance use laws adopted both within the US and internationally (Gilmour et al., 2024; Kenny et al., 2023; Thomas et al., 2018). On one end of the spectrum, some policy strategies focus on improving access to pregnancy-specific drug treatment and harm reduction services—an approach consistent with the current scientific understanding of the etiology of, and most effective treatments for, substance use disorder as a health condition (Volkow, 2023). On the other end, a growing number of states are adopting policies that rely on punishment to deter prenatal substance use, especially use of nonlegal drugs. These punitive approaches, which are more consistent with criminal-legal system priorities (Volkow et al., 2017), include laws that enforce civil penalties (e.g., fines, custody loss, and court mandated drug treatment). Another set of punitive policies takes this approach a step further to directly designate prenatal drug use as a felony or misdemeanor crime.

The recent expansion of pregnancy-related drug policies has generated a growing body of research investigating potential unintended health consequences associated with punitive legal approaches (Bruzelius et al., 2024). This literature indicates that among people who use drugs, fear of incarceration and custody-loss can inhibit service utilization across the perinatal care spectrum. For example, qualitative studies illustrate how concerns about drug use identification in healthcare settings often lead pregnant people who use drugs to delay, underutilize or completely forgo needed services (Goodman et al., 2020; Leiner et al., 2021; O’Rourke-Suchoff et al., 2020; Syvertsen et al., 2021). Moreover, these fears may be magnified by policies and institutional practices that heighten prenatal drug use surveillance within health systems, such as mandatory reporting policies, which require health and social service providers to report prenatal drug use to relevant authorities (Jarlenski et al., 2017; Stone, 2015).

The idea that punitive drug policies may lead to health system avoidance is supported by a complementary literature documenting negative health outcomes associated with other types of punitive policies. For example, studies highlight how similar system avoidance strategies are commonly deployed to limit contact with authorities following immigration policy changes—often in ways that can often increase risk of negative outcomes through reduced utilization of services and resources (Brayne, 2014; Carbonaro, 2022; Fong, 2019; Haskins & Jacobsen, 2017). Another key insight from this related literature is that fear of detection can itself be a severe stressor with health implications. Chronic vigilance associated with concealment of immigration status has been shown to increase risk of adverse pregnancy and birth outcomes, and may similarly apply in other concealment settings, such as with prenatal drug use (Novak et al., 2017; Ro et al., 2020; Torche & Sirois, 2019).

Much of what we know about the specific health impacts of punitive prenatal drug policies, including explicit criminalization, comes from evaluations of Tennessee’s Fetal Assault Law, first enacted in 2014 before expiring due to a sunset provision in 2016. Although Tennessee is not the only state to have adopted such a policy—South Carolina and Alabama adopted explicit criminalization via court decisions in 1997 and 2013—Tennessee’s law received significant attention when it was enacted as an ostensible public health solution to rising incidence of prenatal opioid use in the state (Ko et al., 2020; Winkelman et al., 2018). The Fetal Assault law authorized prosecutors to charge pregnant people with assault if they gave birth to a child deemed by the courts to have been harmed by illegal use of a narcotic drug during pregnancy (Boone & McMichael, 2021; Shearer et al., 2019). The law was initially enacted with the stated intent of motivating pregnant people with drug use disorders to seek treatment, as it contained a provision excluding those enrolled in treatment from prosecution. However, emerging evidence suggests that the law may have counteracted its intended goals, as it coincided with a decrease, rather than an increase, in drug treatment admissions following implementation (Bach, 2022).

Beyond reductions in drug treatment utilization, studies also suggest that Fetal Assault may have been associated with negative repercussions for pregnancy and birth-related outcomes. Boone and McMichael (Boone & McMichael, 2021) reported several notable though somewhat imprecise associations between adoption of the Fetal Assault law, and increases in infant mortality and stillbirth, as well as decreases in Apgar score, gestational age and prenatal care. In another recent analysis, Choi et al (S. W. Choi et al., 2023), found that the odds of out-of-state births increased 13–17% among Tennessee residents following Fetal Assault implementation—supporting provider concerns of law-related obstetrical care avoidance (Amnesty International, 2017). Relatedly, some Tennessee providers anecdotally reported higher incidence of pregnant people giving birth outside of healthcare settings to circumvent drug testing at delivery, a common practice in some hospitals (Howard, 2017; SisterReach, 2020). Increases in out-of-hospital births have also been observed in the context of punitive immigration policies (Amuedo-Dorantes et al., 2022), suggesting that system avoidance strategies affecting health service utilization generally could also apply to pregnant people’s decisions about birth setting.

These initial studies, together with work documenting the adverse effects of punitive civil prenatal drug laws (Bruzelius et al., 2024), provide important insights into the ways in which a punitive policy approach may discourage pregnant people who use drugs from seeking pregnancy-related care. However, given that the only studies directly examining prenatal drug use criminalization were conducted in a single state, and concerned a law in effect for only a short period, potential consequences remain underexplored. We therefore build on this work by testing whether prenatal drug use criminalization was associated with reductions in the prevalence of any prenatal care receipt in Alabama, South Carolina, and Tennessee over a 30-year timeframe. We further probed whether criminalization affected two specific dimensions of recommended prenatal care: timeliness—i.e., care initiated in the first trimester, and adequacy—i.e., an appropriate number of visits occurred. Lastly, given concerns that punitive approaches could push pregnant people who use drugs to avoid hospital-based births (Bach, 2022; Howard, 2017; SisterReach, 2020), we examined the impacts of criminalization on a previously unexplored facet of pregnancy-related healthcare utilization—facility delivery, meaning whether the birth occurred in a hospital, birth center, or clinic as opposed to a residence or other setting.

Methods

These secondary data analyses were approved by the Columbia University institutional review board (AAAT0797). Outcome data were from U.S. National Center for Health Statistics (NCHS) National Vital Statistics System (NVSS) Natality files, 1989–2019. We selected the 1989 start date to coincide with the introduction of the 1989 birth certificate revision (Supplemental Table 1), and to capture the period prior to the first instance of explicit criminalization (South Carolina, 1997). The 2019 end date was selected to precede the onset of the COVID-19 pandemic and related healthcare disruptions affecting pregnancy care recommendations (American College of Obstetricians and Gynecologists, 2020a, 2020b). Our primary interest was in understanding how prenatal drug use criminalization may affect care-seeking behavior among pregnant people who use drugs—the population most likely to be impacted by such laws. However, because drug use is not recorded in birth certificate data, our study population necessarily included all births.

Our primary outcomes included county-level prevalence of any prenatal care and facility-based delivery. A prenatal visit was defined as a healthcare encounter in which the provider directly examined or counseled the pregnant person, and births were classified as having received any prenatal care if one or more visits occurred. Facility-based delivery was defined as a birth occurring in a health care facility (hospital, freestanding birth center, or clinic), as opposed to a home or other non-facility settings. These data were derived from birth certificate records, based on information abstracted from medical records rather than parental self-report, reducing the potential for measurement error. However, prior studies have noted variability in data quality across states and specific data elements (Martin et al., 2013; Osterman & Martin, 2018).

Defining our main outcome as receipt of any prenatal care ignores potentially important dimensions of care quality. As prenatal care is ideally expected to begin within the first trimester, and to increase in frequency as the pregnancy approaches term, we therefore additionally examined prevalence of timely and adequate prenatal care as secondary outcomes. We classified births where prenatal care began within the first trimester as having received timely prenatal care, whereas births receiving ≥80% of observed-to recommended prenatal care visits—based on the adequacy of prenatal care utilization index (APNCU)—were defined as adequate. For all four measures, we aggregated individual birth outcomes to the county-year-level using total county-year births as denominators to generate prevalences per 100,000. Because information on prenatal care timeliness and adequacy require month of initiation to calculate, which was not consistently captured across the study period due to the variation in state-level adoption of the 2003 birth certificate revision, several state-years were missing from the analyses for these outcomes (Supplemental Table 2). To maintain comparability between treatment and comparison groups and to minimize bias from systematic differences in reporting, timeliness and adequacy were therefore designated as secondary outcomes, with results interpreted in light of their more limited geographic and temporal representation.

Exposures and Data Source

Our primary exposure was state-level explicit criminalization of prenatal drug use. States were considered to have adopted criminalization if they enacted legislation or had a supreme court decision that explicitly allowed prosecutors to charge a pregnant person with a felony or misdemeanor crime following alleged prenatal drug use. This definition, consistent with existing legal and public health literature (Bach, 2022; Boone & McMichael, 2021; Bridges, 2020; Bruzelius et al., 2024; Howard, 2017), includes states where case law rather than statute explicitly permitted such prosecutions. Under most state and federal laws, it is drug possession—not drug use itself—that is typically criminalized. However, these pregnancy-specific policies uniquely create criminal liability for drug use during pregnancy, meaning that the same conduct would not be subject to criminal prosecution outside the context of pregnancy. We focused on criminal penalties, as distinct from civil penalties or mandatory reporting laws because criminalization represents a qualitatively different and arguably more severe legal exposure. While state appellate court decisions may also shape the legal environment for prenatal drug use prosecutions, they were beyond the scope of our data collection and are not reflected in our exposure classification. Based on this definition, we classified counties in South Carolina after 1997, Alabama after 2013, and Tennessee between 2014 and 2016 as exposed to explicit criminalization policies.

Confounders and Data Sources

Potential confounders were time varying policies and characteristics expected to be associated with criminalization and pregnancy care (Supplemental Figure 1, Supplemental Table 3). First, given that research documents potential impacts of other types of prenatal substance use-related policies (Bruzelius et al., 2024), we controlled for state adoption of legislation that considers prenatal substance use grounds for child abuse or neglect substantiation, parental rights termination, or civil commitment. We further controlled for state mandatory reporting policies that require health and social services providers to report prenatal drug use to authorities for investigative purposes, as well as policies aimed at supporting treatment among pregnant people, specifically limitations on criminal prosecution, priority treatment access and state Medicaid expansion status, which increased access to drug treatment services and pregnancy care (S. Choi et al., 2021; Faherty et al., 2022). Models additionally controlled for factors expected to be associated with time varying shifts in criminal-legal system and substance use treatment priorities funding priorities, including police officers and substance use treatment facilities per capita, and state political party. Additionally, we included urbanicity as a proxy for factors associated with changing access to hospital obstetric services (Kozhimannil et al., 2025; Office of the Surgeon General, 2020). Lastly, birth certificate version was included to address potential changes in data collection procedures.

Statistical Analysis

To test associations, we compared pre- and post- criminalization pregnancy care utilization prevalence ratios for each outcome in criminalization counties compared to non-criminalization counties using separate regression models, characterized by equation 1 below.

$$\mathit{\text{Utilizatio}}{n}_{\mathit{\text{tcs}}}={\beta }_{\mathit{0}}+{\beta }_{\mathit{2}}{\text{T}}_t+{\beta }_{\mathit{3}}{\text{C}}_c+{\beta }_{\mathit{4}}{Z}_{tcs}+{\beta }_{\mathit{1}}\mathit{\text{Polic}}{y}_{tcs}+{\epsilon }_{\mathit{\text{tcs}}}$$ (E1)

Where, using the example of any prenatal care, prevalence of prenatal care (Utilization_tcs) per 100,000 births in year t and county c in state s, are regressed on county- and year-fixed effects (C and T, respectively), a vector of controls (Z), and a policy indicator (Policy_tcs), set to 1 for explicit criminalization counties beginning in the year that the policy was adopted, and 0 otherwise. The error term (ε_tcs) reflects heteroskedasticity-robust standard errors clustered at the state-level to account for potential heterogeneity in treatment effects across states (Mansournia et al., 2021). Models were specified using linear regression as linear models have been shown to produce more reliable estimates in the context of high-dimensional fixed effects. Secondary analyses tested alternative specifications using count models (Greene, 2004).

Assumption Testing

The overall difference-in-differences approach assumes that the outcome trajectories of counties in non-criminalization states provide a reasonable counterfactual for trajectories in counties in criminalization states, absent criminalization. This assumption is more plausible if there is an indication that trends in exposed and unexposed groups did not differ significantly in the years prior to policy adoption (Roth et al., 2023). To explore the plausibility of this assumption, we used regression approaches to test for significant differences in outcome trends in pre-policy years relative to policy adoption, using policy leads. Evidence of significant differences in these pre-trend tests suggests that the parallel trends assumption may not hold.

Sensitivity Analyses

We conducted a series of additional analyses to consider potential sensitivity to specific modeling choices. First, in the primary analyses we used a strict definition of criminalization for counties in Tennessee that limited exposure to the period when the law was in effect in the state (i.e., 2014–2016). However, there are some data that potential inhibiting effects of criminalization on birth outcomes persisted even after the formal law expired in 2016 (Boone & McMichael, 2021). Therefore, in a refinement of the exposure criteria, we conducted analyses considering counties in Tennessee exposed from 2016 onwards. These models retained the original exposure timing definitions for counties in other criminalization states (i.e., Alabama post 2012, South Carolina post 1997). Second, we included an additional set of county-level control variables (age and racial and ethnic composition, median household income) to explore whether changes in county composition over the period may have been partially driving results (Stuart et al., 2014). Third, although we elected to use linear models in primary analyses, we also conducted supplementary analyses fitting Poisson models to analyze the relationship between criminalization policies and each outcome, including the total number of births in a county-year as offset term.

Results

Descriptive Statistics and Assumption Testing

Table 1 presents the characteristics for the 83,300 county-years in the sample, representing over 124 million births. Overall, counties in states that never went on to adopt explicit criminalization tended to have higher levels of any prenatal care, timely prenatal care, and facility-based delivery, compared to those in states that eventually adopted criminalization. In contrast, prenatal care adequacy was generally lower among counties in never-criminalization states in the beginning of the observation period, but the trend reversed towards the end. Counties in ever and never criminalization states were also generally similar but tended to differ in terms of adoption of other, non-criminalization pregnancy-related policies.

Table 1.

Policy, and demographic characteristics of counties by prenatal drug use criminalization status: NVSS Natality, 1989–2019

Characteristic	No explicit criminalization (n=81,533 county-years)	Explicit criminalization (n=1,767 county-years)
Punitive civil law	27,511 (33.77)	0.00 (0.00)
Supportive law	28,960 (35.54)	0.00 (0.00)
Mandatory reporting law	29,806 (36.61)	497 (28.12)
Medicaid expansion	5,945 (7.30)	0.00 (0.00)
Governor’s political party, Democrat	46,015 (56.43)	1,638 (92.7)
Police force size	9.02 (4.51)	10.76 (1.22)
Substance use treatment facilities per 1,000 population	4.71 (3.06)	2.66 (0.46)
Metropolitan county
Population 1 million or more	11,773 (13.69)	152 (8.61)
Population 250,000–1 million	9,533 (11.72)	444 (25.11)
Population < 250,000	9,131 (11.25)	258 (14.67)
Nonmetropolitan county
Population 20,000 or more, metro adjacent	6,315 (7.71)	199 (11.38)
Population 20,000 or more, metro nonadjacent	3,268 (4.08)	19 (1.12)
Population 2,500–19,000, metro adjacent	16,591 (20.40)	523 (29.64)
Population 2,500–19,000, metro nonadjacent	13,318 (16.31)	77 (4.44)
Population <2,500/completely rural, adjacent	4,406 (5.41)	65 (3.74)
Population <2,500/completely rural, nonadjacent	7,159 (4.08)	29 (1.66)

Parallel pre-trends testing suggested no significant differences in pre-policy trends for the prenatal care outcomes, however, there was an indication of some potential pre-existing differences between counties in criminalization and non-criminalization states for the facility-based delivery outcome (Supplemental Table 4). Specifically, significant pre-trend differences were noted 8 and 10 years prior to criminalization policy adoption for this measure.

Model Results

Table 2 shows the primary results from the primary difference-in-differences models. After controlling for covariates, adoption of an explicit criminalization policy was associated with a decrease in the prevalence of births receiving any prenatal care (−4396.29/100,000 births; [95% Confidence Interval [CI]: −6176.07, −2616.51]). Similarly, criminalization adoption was associated with a decrease in the prevalence of facility-based deliveries (−1847.99 [CI: −3688.29, −7.69]), although a wide confidence interval was observed.

Table 2.

Association between prenatal drug use criminalization and prevalence of any prenatal care and facility-based births, NVSS Natality, 1989–2019

	Unadjusted Models			Adjusted Models
Primary Outcomes	Estimate	95% CI Lower	95% CI Upper	Estimate	95% CI Lower	95% CI Upper
Any PNC	−4554.45	(−6471.71,	−2,637.18)	−4396.29	(−6176.07,	−2616.51)
Facility Delivery	−1229.12	(−3074.14,	615.89)	−1847.99	(−3688.29,	−7.69)
Secondary Outcomes
Timely PNC	−792.21	(−2571.99,	987.58)	−1328.43	(−3026.82,	369.96)
Adequate PNC	−1328.43	(−3026.82,	369.96)	−792.208	(−2570.00,	987.58)

PNC: Prenatal Care; CI: Confidence Interval. Models include county and year fixed effects with standard errors clustered at the state-level adjusted for co-occurring policy adoption (punitive civil laws, supportive laws, Medicaid expansion), police officers per capita, SUD treatment facilities per capita, political affiliation and urbanicity.

For the secondary outcomes, models indicated that adoption of explicit criminalization was associated with decreases in both the prevalence of timely (−1328.43 [CI: −3108.21, 451.36]) and adequate (−792.21 [CI: −2571.99, 987.58]) prenatal care (Table 2). However, for both measures, the magnitudes of the associations were attenuated relative to any prenatal care measure, and associated confidence intervals included the null.

Sensitivity Analysis Results

Table 3 shows the results of the sensitivity analyses. First, the inverse association between criminalization adoption and prevalence of any prenatal care was highly robust to analytic choices. For example, models altering the exposure criteria (−4374.56 [CI: −5944.86, −2804.26], and controlling for additional county characteristics (−4320.35 [CI: −6101.65, −2539.06]), each produced estimates and intervals similar to those observed in primary analyses (−4396.29 [CI: −6176.07, −2616.51]). The Poisson models also produced the same conclusion as the primary analysis (Prevalence Ratio [PR]: 95.50 [CI: 95.48, 95.51]). In contrast, estimates of association for facility-based delivery were more sensitive to the exposure definition (−211.37 [CI −1627.27, 1204.53]) and the inclusion of additional county covariates (−1301.43 [CI: −3134.36, 531.50])—although the direction of associations remained consistently negative. Using a Poisson specification, produced the same conclusion as in the primary analysis (PR: 97.94 [CI: 97.44, 98.44]).

Table 3.

Sensitivity analyses of the association between prenatal drug use criminalization and prevalence of any prenatal care and facility-based births, NVSS Natality, 1989–2019

	Exposure RefinementA			Additional Covariate AdjustmentB			Poisson SpecificationC
Primary Outcomes	Estimate	95% CI Lower	95% CI Upper	Estimate	95% CI Lower	95% CI Upper	OR	95% CI Lower	95% CI Upper
Any PNC	−4374.56	(−5583.08,	−3166.03)	−4320.36	(−6101.64,	−2539.1)	95.50	(95.48,	95.51)
Facility Delivery	−211.37	(−1914.26,	1491.53)	−1301.4	(−3134.36,	531.5)	97.94	(97.44,	98.44)
Secondary Outcomes
Timely PNC	−1686.48	(−3245.85,	−127.12)	−1708.61	(−3413.09,	−4.14)	96.94	(96.85,	96.96)
Adequate PNC	−1193.27	(−2819.32,	432.78)	−1120.32	(−2906.34,	665.70)	98.90	(98.87,	98.94)

PNC: Prenatal Care; CI: Confidence Interval; OR: Odds Ratio.

^AReflects estimates after reassigning counties in Tennessee (TN) to exposed for 2014 onward (i.e., ignoring the 2016 expiration of Fetal Assault).

^BIncludes secondary county-level composition covariates, specifically age, race and ethnicity, and median household income.

^CSpecified using Poisson as opposed to linear regression with county-year births offsets. All models include county and year fixed effects with standard errors clustered at the state-level.

Additional analyses exploring secondary outcomes also showed some sensitivity to analytic choices. Confidence intervals for inadequate prenatal care crossed the null across all specifications except in the Poisson sensitivity analysis (exposure reassignment: −1193.27 [CI: −2819.32, 432.78]; additional covariate adjustment: −1120.32 [CI: −2906.34, 665.70]; Poisson specification PR: 98.90 [CI: 98.87, 98.94]). For prenatal care timeliness, we noted evidence suggestive of prevalence decreases following criminalization adoption in some sensitivity analyses. For example, refining the exposure criteria increased the magnitude of the association between criminalization and timeliness. Using this approach, adoption of prenatal drug use criminalization was associated with 1686.48 fewer births receiving first trimester prenatal care (CI: −3245.86, −127.12). Similarly, inclusion of additional covariates pushed the bounds of the confidence interval to exclude the null (−1708.61 [CI: −3413.09, −4.14]), indicative of an association between explicit criminalization and reduced timeliness. Lastly, in the Poisson-specified models we observed a prevalence ratio of 96.94 (CI: 96.85, 96.96), suggesting that adoption of an explicit criminalization policy was associated with a small relative reduction in the prevalence of timely prenatal care in counties with criminalization compared to counties without criminalization.

Discussion

Extending existing state-specific examinations of prenatal drug use criminalization, we identified an association between state adoption of an explicit prenatal drug use criminalization policy and reductions in the prevalence of any prenatal care. Moreover, while we found few indications that explicit criminalization was associated with overt reductions in prenatal care adequacy, criminalization policies may have contributed to decreases in prenatal care initiation within the first trimester. Lastly, we find suggestive evidence that criminalization may negatively impact the prevalence of facility-based delivery. Anecdotal reports have suggested that some pregnant people with drug use disorders may elect to give birth outside of healthcare settings to avoid hospital drug testing (Howard, 2017). HHowever, this is the first study to our knowledge to test the association empirically.

This work supports several prior studies documenting the adverse health impacts of punitive prenatal drug policies, including explicit criminalization. Our findings suggest three potential insights that can help enhance understanding of how punitive responses to prenatal drug use impact health. First, our finding that explicit criminalization is associated with a decrease the prevalence of any prenatal care supports a potential link between punitive policies and adverse neonatal outcomes. Multiple studies have found that the adoption of some punitive civil policies is associated with increases in neonatal drug withdrawal syndrome (NDWS)(Atkins & Durrance, 2020; Faherty et al., 2019, 2022; Meinhofer et al., 2022)—a term describing the withdrawal signs that can occur in neonates following chronic intrauterine exposure to opioids and other drugs (Devlin et al., 2022). In addition, prior work evaluating the health impacts of Fetal Assault adoption in Tennessee (Boone & McMichael, 2021), and other punitive civil laws generally, suggests that criminalization may also increase low birthweight, preterm birth, stillbirth, and infant mortality (Austin et al., 2022; Meinhofer et al., 2022). Our results are therefore consistent with the idea that punitive laws could be increasing some of these adverse birth outcomes by minimizing opportunities for the identification and management of pregnancy complications that would have occurred in the context of appropriate prenatal care, absent criminalization. Similarly, as prenatal care avoidance could further undermine opportunities to connect pregnant people with drug use disorder to effective drug treatment services, associations between punitive laws and increases in NDWS are potentially mediated through the reduced prenatal care pathway. It is important to note, however, that in the context of NDWS, medications for opioid use disorder (MOUD) can themselves produce withdrawal signs (Goodman et al., 2019). Additional work is therefore needed to clarify these mechanisms, and to identify effective state policies to address NDWS through health-supporting channels. As well, additional work examining other potential pathways, especially stress-related pathways, between criminalization and adverse birth outcomes is warranted. Significant research links exposure to policies and practices that promulgate stigma with adverse birth outcomes across numerous contexts (Boen et al., 2023; Chegwin et al., 2023; Everett et al., 2022; Samari et al., 2020; Torche & Sirois, 2019).

A second related issue is that, in addition to finding that criminalization decreases receipt of any prenatal care, we also identified inconsistent evidence that criminalization decreases prenatal care timeliness. This finding provides empirical support for provider-raised concerns that in the context of criminalization, pregnant people with drug use disorder may be postponing prenatal care entry to initiate detoxification (Amnesty International, 2017). If these cessation efforts are not medically supervised, such an approach is less likely to be effective, given that behavioral interventions along with MOUD when appropriate, lead to more favorable outcomes (American College of Obstetricians and Gynecologists, 2021; Substance Abuse and Mental Health Services Administration, 2018). Moreover, drug detoxification is not recommended during pregnancy due to insufficient safety information. Detoxification can further increase risks of relapse and non-fatal and fatal overdose (Terplan et al., 2018)—a growing concern during pregnancy (Bruzelius & Martins, 2022).

A final insight from this study is that prenatal drug use criminalization may reduce the number of pregnant people who give birth in a healthcare facility. Given the sensitivity of this measure to model specifications, however, additional work is needed to replicate the finding. Although decreases in facility deliveries are not necessarily indicative of adverse outcomes directly, reductions signal greater pregnancy care avoidance with potentially wider implications. Home-based births are not currently recommended for most high-risk pregnancies including those among people with drug use disorders. Delivering outside of a facility setting may also decrease access to immediate neonatal intensive care unit, and specialty care, which may be an important consideration for certain types of births, such as those affected by opioid use, which can require higher acuity services under certain conditions (Jones et al., 2014). In addition to further research replicating this finding, work is also needed to explore whether other elements of the labor and delivery and also postpartum experiences of people who use drugs could be negatively affected by punitive laws.

Limitations

Several limitations should be considered while interpreting our results. First, we lacked data on individuals’ drug use or their awareness of criminalization policies, which limits our ability to directly assess associations among the subgroups most likely to be impacted. As a result, our analyses include all births, even though the behavioral response to criminalization is likely concentrated among pregnant people who use drugs. This mismatch between the study population and the target population of interest may have attenuated effect sizes and contributed to wide confidence intervals for some outcomes, as estimates reflect averages across a broader population that includes individuals unlikely to be affected. It is plausible that, within a sample restricted to people with drug use disorders, the estimated associations would be larger and more precise. While our findings are directionally consistent with our hypotheses and prior related work (Austin et al., 2022; Boone & McMichael, 2021; S. W. Choi et al., 2023; Meinhofer et al., 2022), several estimates remain imprecise, with confidence intervals that include or approach the null. These findings should therefore be interpreted with caution and warrant replication. A second related data concern is that, as previously noted, due to birth certificate version changes, we observed significant missing outcome data, especially for prenatal care timeliness and adequacy. Both measures require information on the month of prenatal care initiation to appropriately calculate. Given that data for this variable were not consistently provided across states over the full observation period, further work is needed to replicate and interrogate associations.

Second, although we tried to minimize bias throughout the study, two concerns persist. First, residual confounding remains a possibility, especially given the 30-year timeframe. While we controlled for a broad set of closely related policies, changes in related policy domains—such as broader child welfare, reproductive health, substance use, or healthcare access—may also have influenced outcomes. In line with other drug policy evaluations (Schuler et al., 2020), and consistent with other research on the health effects of punitive prenatal drug policies specifically (Austin et al., 2022; Meinhofer et al., 2022; Tabatabaeepour et al., 2022), we controlled for multiple co-occurring policies expected to be associated with policy adoption and pregnancy care, including punitive civil policies, treatment supportive policies and Medicaid expansion. Within the policy evaluation literature this approach is typically considered necessary for valid inferences (Matthay, Gottlieb, et al., 2021). However, controlling for multiple co-occurring policies can also reduce precision. Future work in this area might consider alternative dimension reduction methods for addressing potential policy co-linearity (Matthay, Hagan, et al., 2021; Schuler et al., 2021).

A related difference-in-differences policy evaluation concern is that recent methodological work points to challenges in interpreting associations when either pre-trends do not fully hold, as was the case with facility-delivery, or when potential exposure effects are heterogenous by group or timeframe (Goodman-Bacon, 2021). While several newer estimators can address some forms of heterogeneity bias (Callaway & Sant’Anna, 2021; Roth et al., 2023), their implementation in this setting was limited by the small number of treated states, uneven lengths of pre- and post-policy periods, and incomplete overlap between treated and control units in several periods. Combined with the sensitivity of the facility-based delivery estimates to analytic specifications and potential pre-trends violations, our results should therefore be interpreted cautiously. Nonetheless, across all outcomes and model specifications, results were consistently negative—suggesting no benefit associated with criminalization and indicating potential harms.

Third, compared to much of the previous research on punitive laws, we specifically restricted our focus to enforcement of criminal rather than civil penalties. As criminalization arguably represents the most extreme example of a punitive policy, we conceptualized this choice as a necessary tradeoff between increasing internal validity with a narrow exposure definition and capturing the full scope of punitive policy implementation—particularly given that prior studies using broad punitive definitions have yielded mixed results (Bruzelius et al., 2024; Tabatabaeepour et al., 2022). Additionally, this prioritization of construct validity likely comes at the expense of statistical power, as only three states have adopted policies explicitly authorizing criminal prosecution. In contrast to our criminalization-focused approach, other studies have found that civil laws with custody implications can also prompt healthcare system disengagement (Austin et al., 2022). Moreover, custody-loss, arrests and prosecutions are known to have occurred in virtually every state, even in contexts with no legal foundation for such actions (Paltrow & Flavin, 2013). This fact suggests that pregnant people with drug use disorders may avoid pregnancy care independent of the actual legal status of prenatal drug use in their state. The consistency of our findings, in conjunction with other work in this area, suggests that formal laws do make a difference. Future work should explore how best to identify the types of laws, and ‘active’ components of laws, most likely to deter pregnant people from needed services. As well, studies are needed to examine potential synergies between punitive and supportive approaches, their interplay with mandatory reporting requirements, and their links to broader reproductive health and substance use policy contexts.

Fourth, a limitation of this study is that we did not assess how prenatal drug use criminalization could differentially affect pregnancy care within subgroups of pregnant people who use drugs. Research shows that the groups most likely to experience prenatal drug use prosecution are those who already experience the lowest levels of pregnancy care access and highest levels of pregnancy complications—namely, Black and Indigenous pregnant people, and economically marginalized pregnant people (Bach, 2022; Flavin & Paltrow, 2010; SisterReach, 2020). It is therefore reasonable to expect that any negative effects of criminalization on pregnancy care could amplify existing disparities. Given limited evidence of a relationship between criminalization and pregnancy care avoidance prior to this study, and potential statistical challenges associated with estimating stable county-level, group-specific outcome trends, we limited our focus to estimating overall associations. However, a critical next step for future work will be to assess differential policy effects, and to identify the types of nonpunitive policies with the greatest potential for increasing care utilization among all pregnant and postpartum people.

Finally, our analysis only covers the period up to 2020, but the legal landscape surrounding pregnancy has continued to evolve, particularly in the post-Dobbs era. These shifts are particularly relevant for pregnant people with substance use disorders, who may now face growing legal obligations to continue a pregnancy while simultaneously being subject to criminalization and other punitive responses. As states continue to adopt or reinterpret laws targeting pregnancy-related behaviors, the effects of punitive drug policies on prenatal care and birth outcomes warrant ongoing monitoring and further research.

Conclusion

Prenatal drug use criminalization is often adopted under the guise of public health, but evidence to date does not support the effectiveness of this type of approach. Our findings, which suggested reductions in prenatal care and facility-based birth prevalence following the adoption of explicit criminalization policies, contribute to a growing body of evidence finding that punitive policy approaches to prenatal drug use can generate unintended negative consequences.