Substance use treatment facility access and utilization among mothers in child welfare.

Substance use treatment facility access and utilization among mothers in child welfare - PMC Skip to main content An official website of the United States government Here's how you know Here's how you know Search Log in Primary site navigation Search Log in Search PMC Full-Text Archive Search in PMC Journal List User Guide As a library, NLM provides access to scientific literature. Inclusion in an NLM database does not imply endorsement of, or agreement with, the contents by NLM or the National Institutes of Health. Learn more: PMC Disclaimer | PMC Copyright Notice J Subst Use Addict Treat . Author manuscript; available in PMC: 2026 Jun 1. Published in final edited form as: J Subst Use Addict Treat. 2026 Feb 2;185:209890. doi: 10.1016/j.josat.2026.209890 Substance use treatment facility access and utilization among mothers in child welfare ☆ Dylan Jones a, * , Sarah Font b , Marci Cross a , Ezra Goldstein c Author information Article notes Copyright and License information a College of Health and Human Development, The Pennsylvania State University, United States of America b College of Criminology and Sociology, The Pennsylvania State University, United States of America c Georgia Institute of Technology, United States of America * Corresponding author. dxj5356@psu.edu (D. Jones). Issue date 2026 Jun. PMC Copyright notice PMCID: PMC13222658  NIHMSID: NIHMS2179085  PMID: 41638433 The publisher's version of this article is available at J Subst Use Addict Treat Abstract Introduction: Despite the severe consequences of substance use disorder (SUD), most Americans with SUD do not seek treatment. Uptake of treatment is especially important for mothers involved in the child welfare system (CWS), where SUD is common and can result in child maltreatment and loss of child custody. This study examines how local treatment accessibility influences utilization among CWS-involved mothers. Methods: Our sample includes 98,343 CWS referrals in Pennsylvania (2015–2019) in which the referred biological mother was receiving Medicaid. Using geocoded referral addresses and SUD treatment facilities statewide, we estimate how the drive time to and supply of SUD treatment is associated with maternal treatment uptake in the quarter following the referral. Given lack of consensus regarding the distance individuals will travel for SUD treatment (service areas), we consider multiple versions of our measures and allow operationalization of service areas to vary by county urbanicity and SUD treatment type. We use hierarchical logistic regression models with conditional autoregressive random effects to account for spatial autocorrelation to estimate treatment utilization. Results: Drive time is associated with lower outpatient utilization, particularly in large metropolitan areas with broader service area definitions. Greater treatment supply is associated with increased outpatient and inpatient utilization, with the magnitudes being higher for inpatient care. For both measures, levels of significance depend on urbanicity. Conclusions: The associations between the accessibility measures and treatment uptake are statistically significant but relatively small in magnitude. Greater attention to access barriers beyond drive time and supply of SUD treatment is needed to increase uptake among child welfare-involved mothers. Keywords: substance use treatment accessibility, substance use treatment supply, substance use treatment proximity 1. Introduction Although drug overdose deaths have recently begun to decline, almost 80,000 Americans continue to die each year from overdoses ( Ahmad et al., 2025 ), and millions more are affected by substance use disorder (SUD) ( Substance Abuse and Mental Health Services Administration, 2024 ). Although many individuals with SUD neither seek nor want treatment ( Substance Abuse and Mental Health Services Administration, 2021 ), others encounter barriers to participation. Limited accessibility, particularly for individuals who depend on Medicaid coverage, is widely believed to diminish SUD treatment utilization and prevent or delay recovery. Yet, there is mixed evidence on the extent to which reducing specific barriers to access, such as Medicaid non-eligibility or long drive times to treatment, improves uptake. For example, studies of Medicaid expansion find null or conditional effects on uptake of treatment for individuals with SUD ( Haight et al., 2022 ; Olfson et al., 2018 ), whereas other studies find that increases in the local supply of treatment facilities may reduce overdoses and related adverse events ( Corredor-Waldron & Currie, 2022 ). Uptake of SUD treatment has profound implications for families involved in the child welfare system (CWS). Parents with substance use problems are often brought to the attention of CWS because the acquisition and use of substances depletes resources, hinders supervision and attentiveness to children’s needs, and sometimes induces erratic or aggressive behavior ( Smith et al., 2016 ). It is well-established that substance use afflicts a large, perhaps even majority, share of caregivers involved with the CWS ( Goldstein & Font, 2025 ; Mark et al., 2024 ; Patton et al., 2020 ; Seay, 2015 ) and is a primary driver of CWS caseloads and foster care entries ( Palmer et al., 2024 ; Smith et al., 2016 ). When substance use necessitates CWS intervention, parents may be subject to ongoing drug testing, treatment, or other monitoring and required to desist in their substance use as a condition of retaining (or, if their child was removed from their care, regaining) custody. Under federal law, CWS is responsible for making “reasonable efforts” to preserve or reunify families, typically by providing or facilitating access to relevant services ( Congressional Research Service, 2014 ). Yet, studies suggest that most mothers with SUD who are involved with CWS are not in treatment ( Goldstein & Font, 2025 ; Mark et al., 2024 ). CWS relies on treatment providers in the local community. In the absence of sufficient treatment supply or other barriers to access, parents may be delayed or denied services that would help them safely care for their children. At the same time, mothers involved with CWS face myriad barriers to SUD treatment engagement, including lack of transportation and child care, and mental health comorbidities ( Bunger et al., 2024 ). It is not known whether accessibility plays a key role in their treatment utilization. Prior studies examining local availability of treatment have not considered CWS populations. This study examines associations between SUD treatment accessibility and utilization among a cohort of mothers involved with the Pennsylvania CWS in 2015–2019. Leveraging linked CWS and Medicaid claims data for 98,343 CWS referrals involving 63,878 mothers, we examine whether driving time and supply of SUD treatment are associated with outpatient and inpatient treatment utilization in the three months following a CWS referral. We consider multiple versions of our treatment accessibility measures, given the lack of consensus regarding service areas and how they vary among urban and rural areas ( Higgs, 2009 ). 2. Method 2.1. Data and sample This study leverages linked CWS and Medicaid claims records in Pennsylvania. The sample comprises confirmed CWS cases in Pennsylvania from 2015 through June of 2019, where the biological mother was identified as receiving Medicaid in the three months following the referral date (98,382 referrals for 63,879 mothers). In Pennsylvania, CWS cases are expunged if they are not substantiated as child abuse or otherwise validated as an alternative response case. Thus, our sample includes only referrals with one or more confirmed allegations. Each referral is geocoded to the census block group level based on the incident or primary address associated with the referral. Additionally, mothers in block groups without neighboring block groups were removed. Specifically, these were mothers whose neighboring block groups did not have any CWS referrals and were thus not represented in the data. This step was taken to ensure valid analytic inference when implementing the hierarchical logistic regression models with conditional autoregressive random effects ( Bivand, n.d. ). The final sample is 98,343 referrals for 63,876 mothers. 2.2. Measures 2.2.1. Specialized SUD treatment utilization Each biological mother was linked to her Medicaid claims in the 12 months preceding and 3 months following the CWS referral date. Data linkage was probabilistic, based on Personally Identifiable Information in each database. Within the subset of claims billed by a specialized SUD treatment provider and with a date of service between 0 and 90 days after the CWS referral, we derived two outcome variables: inpatient and outpatient substance use treatment. Outpatient SUD treatment is equal to 1 if the mother received any service from a drug or alcohol treatment provider not in an emergency or inpatient setting, as defined by provider specialty and place of service, and 0 otherwise. Inpatient SUD treatment is equal to 1 if the mother received any inpatient or residential care from a drug or alcohol treatment provider, and 0 otherwise. We focus on SUD treatment facilities specifically, excluding general health care providers who may provide medication for opioid use disorder (MOUD), for several reasons. First, the CWS population has high rates of polysubstance use ( Goldstein & Font, 2025 ) and comorbid mental illness ( Mark et al., 2024 ) that would implicate a need for more comprehensive and specialized care ( Jarpe-Ratner et al., 2015 ). Standalone treatment with MOUD reduces fatal overdose risk ( National Academies of Sciences, Engineering, and Medicine, 2019 ) but does not address the functional impairments (and thus risks to child safety) associated with ongoing substance misuse, particularly in the context of polysubstance use. Second, we found that 83% of all mothers who had a Medicaid claim with a primary diagnosis of SUD during the outcome observation period received care from a specialty provider, suggesting that mothers involved with CWS are unlikely to receive MOUD from a general practitioner in the absence of other interventions. Lastly, as a practical matter, it is challenging to determine which non-specialist providers would prescribe MOUD and under what circumstances, and it would be inaccurate to assume anyone with access to a general practitioner has access to SUD treatment generally or MOUD specifically. 2.2.2. SUD treatment facilities accessibility Treatment facilities in Pennsylvania were identified from the Substance Abuse and Mental Health Services Administration (SAMHSA) 2016–2020 National Directory files. The directory includes facility addresses and information about the type of treatment provided and populations served. In the five-year period, a total of 1200 unique facilities were identified as being open for at least one year. Of the 2541 unique facility-years in Pennsylvania, we exclude facilities that only serve men ( n = 171) or do not accept Medicaid ( n = 348). Of these facility-years, 2326 provided outpatient treatment, and 1385 provided inpatient/residential treatment. To establish measures of accessibility, we first needed to establish a criterion for designating facilities as within the area in which a given mother (within a block group) would most likely access services and conversely designating block groups as within the service providing area of a given treatment facility. For each block group and treatment facility, we assign a service area defined by its size, which we measure using drive time. The service area size represents the maximum drive time that a mother is able or willing to spend seeking treatment. Drive times between block group centroids and treatment facilities were calculated using the osrm package in R ( Giraud, 2022 ). Because the distribution of service area sizes is unknown, we explore three different scales to approximate narrow , mid-range , and wide service area sizes. The specific service area sizes at each scaling vary according to two dimensions: treatment type (inpatient or outpatient) and urbanicity. We anticipate that individuals will travel further for inpatient treatment because they would need to make the trip fewer times (only at entry and exit to treatment). Prior literature has shown that the distribution of travel times to SUD treatment facilities varies substantially by urbanicity ( Kiang et al., 2021 ), potentially more so in our low-income sample where mothers may rely on public transportation or have inconsistent access to a vehicle. Urbanicity is based on county and is categorized into three values using the 2013 Urban-Rural Classification Scheme for Counties (URCSC) ( Ingram & Franco, 2014 ): large metro (codes 1 and 2), medium and small (medium/small) metro (codes 3 and 4), and nonmetro (codes 5 and 6). A facility is considered in the service area of a block group if drive time from the block group centroid to the facility is less than the service area’s size, conditional on scale, treatment type and urbanicity. Block groups are considered in the service area of facilities using the same formulation. Values for service area sizes are given for both outpatient and inpatient treatment in Table 1 (outpatient and inpatient, respectively). Literature on how much time individuals are willing to spend driving to seek treatment is limited, particularly among disadvantaged populations and with consideration of treatment type and urbanicity. However, Kiang et al. (2021) provide a starting point for defining the variation by county urbanicity. They report median drive times to the closest opioid treatment facilities for each 2013 URCSC code. Although these values do not themselves represent the maximum amount of time a person is able or willing to travel to seek treatment, they provide a lower bound with how long they would have to spend to obtain treatment. Compared to URCSC codes 1 and 2 (large metro), the median drive time is approximately 50% greater for URCSC codes 3 and 4 (medium/small metro) and approximately three times greater for codes 5 and 6 (nonmetro). We use these ratios to define service area sizes for outpatient treatment. For inpatient treatment, we assume less variation, given the lower frequency of trips needed for treatment. Table 1. Measures of outpatient treatment drive time and supply by county urbanicity and service area. Scale. All counties Large metro Medium/small metro Nonmetro # counties 67 13 24 30 # block groups 8974 4239 3438 1297 # referrals 98,373 38,790 40,611 18,972 Outpatient facilities Narrow service area Size 20, 30, or 60 20 30 60 Drive time, M (SD) 15.3 (10) 10.9 (3.6) 13.6 (7.3) 33 (10.6) Supply, M (SD) 42.7 (19.5) 40.4 (20.4) 41.6 (17.8) 52.8 (17.9) Mid-range service area Size 30, 45, or 60 30 45 90 Drive time, M (SD) 22.2 (15.2) 15.2 (5.4) 19 (9.4) 53.2 (10.7) Supply, M (SD) 39.9 (14.2) 36.6 (13.4) 39.8 (12.9) 50.4 (14.9) Wide service area Size 40, 60, or 120 40 60 120 Drive time, M (SD) 29.5 (19.4) 19.3 (7.5) 26.7 (9.9) 70.4 (11.2) Supply, M (SD) 38.3 (10.8) 34.7 (8.6) 39.4 (10.2) 47.3 (12.6) Inpatient facilities Narrow service area Size 60, 60, or 90 60 60 90 Drive time, M (SD) 30.8 (15.7) 25.9 (11.8) 27.8 (11.4) 54.9 (14.7) Supply, M (SD) 22 (7.4) 23.6 (6.3) 19.1 (7) 24.7 (9.2) Mid-range service area Size 90, 90, or 105 90 95 105 Drive time, M (SD) 41.3 (19.8) 30.6 (16.3) 45.7 (15.9) 64.9 (14.2) Supply, M (SD) 22.1 (7.2) 22.5 (6.3) 20.1 (6.7) 26 (9) Wide service area Size 120 120 120 120 Drive time, M (SD) 49.2 (24) 32.2 (17.5) 60.9 (18) 73.6 (14.6) Supply, M (SD) 21.6 (6.8) 21.7 (6.3) 20.2 (6.4) 24.9 (8.1) Open in a new tab Note: Service area size refers to the driving minutes to the furthest away facility that is included in the service area. Driving time is defined as the 25% percentile drive time. Supply is per 100,000 adults on Medicaid. We measure two components of treatment accessibility. First, drive time to treatment facilities is defined as the 25th percentile of drive time across all (Medicaid-accepting) treatment facilities within the service area of that block group. We opted for the 25th percentile over the median, mean, or other measures under the expectation that mothers would prefer closer facilities to facilities to those that are further away. Second, we measure supply to capture the relative treatment capacity available to mothers in each neighborhood, accounting for both proximity and service population. Unlike a simple count of nearby facilities, our proposed measure incorporates how many people are likely to rely on each facility and adjusts for shared access across block groups. As such, it provides a proxy for whether a given mother is likely to find available treatment slots in her area—reflecting not just presence of services, but practical accessibility. Specifically, supply is equal to the number of treatment facilities f in the service area of a given block group b divided by the sum of the “service population” on each of those facilities. For the denominator, we sought to estimate the number of individuals who – were they to seek SUD treatment –would depend on each of the facilities in the service area of b . We first define population of b as the count of non-elderly adults (19–65) on Medicaid ( US Census Bureau, 2020 ). We then estimate the “service population” on each facility across all block groups that may use that facility. The measure takes into account that individuals often have multiple facilities in their service areas and facilities provide services to multiple block groups. The calculation of supply is implemented in two steps, given in Eqs. (1) and (2) . In Eq. (1) , the service population sp for each treatment facility f is calculated by: s p f = ∑ b ϵ B f population b F b (1) where B f denotes the set of block groups in the service area of treatment facility f , population b denotes the number of individuals aged 19–65 enrolled in Medicaid in block group b , and F b denotes the set of treatment facilities in the service area of block group b . Each block group allots to each facility in its service area the size of its population divided by the number of its treatment facilities. With this formulation, a block group with many other treatment facilities in its service area will allot a relatively small proportion of its population to treatment center f . Supply is then calculated as: Supply b = F b ∑ b ϵ F b s p f (2) where the supply for block group b is the number of treatment facilities in the service area divided by the sum of the service population on each of those treatment facilities. Both drive time and supply are calculated separately for inpatient and outpatient treatment, with only facilities offering that treatment modality considered. 2.2.3. Covariates We include controls for characteristics of the CWS referral, mother, and household. For CWS referrals, we include the following non-mutually exclusive allegation types: physical abuse, parent substance use, and child behavior problems. We caution however, that allegations may not fully represent family circumstances; prior research using these linked data shows that approximately half of cases that do not have an allegation related to parent substance use involve a mother with SUD ( Goldstein & Font, 2025 ). Household characteristics are the number of children listed on the case and the age of the youngest child. Maternal age and race/ethnicity are also included. Race/ethnicity was coded as non-Hispanic Black, non-Hispanic White, Hispanic, and other/unknown. Mothers who identified as American Indian/Alaskan Native ( n = 17), Asian ( n = 53) or Native Hawaiian or Other Pacific Islander ( n = 11) were coded into the other/unknown category as their observed counts in the data were too small for model convergence. We do not make any attempt to interpret the other/unknown parameter. Lastly, county and year indicator variables were included. 2.3. Analytic approach We model the probability of treatment utilization using hierarchical logistic regression models with conditional autoregressive random effects. Random effects were specified at the block group level, and the correlation structure was based on block group adjacency. Models are fit using the spaMM package in R ( Rousset & Ferdy, 2014 ). For each treatment modality (outpatient and inpatient), models were fit for (a) the full sample, (b) large metro counties, (c) medium/small metro counties, and (d) nonmetro counties. Within each of these samples, we apply the different scales for service area (narrow, mid-range, and wide, respectively). The values of the primary independent variables (drive time and supply) are standardized. 3. Results 3.1. Sample description A detailed sample description is found in Table A1 in Appendix A . Of the 98,343 total referrals, approximately equal numbers occurred in large metro and medium/small metro counties, 38,790 and 40,611, respectively. Twenty percent occurred in nonmetro counties. The majority (64.5%) of the referrals identified non-Hispanic White mothers. Twenty-two percent of the referrals identified non-Hispanic Black mothers, and 10.5% identified Hispanic mothers. For 2.8% of referrals, the mothers’ race/ethnicity was identified as other or was unknown. In the full sample, 66.5% of referrals had either an allegation of caregiver substance use (39.4%) or the mother had been diagnosed with an SUD prior to the observation period end (58.1%). In the full sample, mother’s accessed outpatient treatment following 20.6% of referrals and accessed inpatient treatment following 3.8% of referrals. Treatment utilization was higher in large metro counties (outpatient: 24.7%; inpatient: 5.2%), compared with medium/small metro counties (outpatient: 17.7%; inpatient: 3%) and nonmetro counties (outpatient: 18.4%; inpatient: 2.8%). 3.2. Description of accessibility measures Table 1 presents the descriptive statistics for the two measures of accessibility: drive time and supply. For outpatient facilities in the full sample, the mean drive time varied from 15.3 (narrow service area) to 29.5 (wide service area) minutes. Drive time was shorter for large metro counties (10.9–19.3 min) than for medium/small metro (13.6–26.7) and nonmetropolitan (33.0–70.4) counties, which was by design. Outpatient supply in the full sample ranged from 38.3 (wide service area) to 42.7 facilities (narrow service area) per 100,000 non-elderly adults on Medicaid. Supply was highest in nonmetro counties. In the full sample of inpatient care, the average drive time was 30.8 min (narrow service area) to 49.2 min (wide service area). Although the service areas are all larger for inpatient treatment, the supply was lower than for outpatient facilities. In the full sample, supply of inpatient facilities ranged from 21.6 (wide service area) to 22.1 (mid-range service area) facilities per 100,000 non-elderly adults on Medicaid. Nonmetropolitan counties had the highest levels of inpatient supply, and medium/small metro counties had the lowest. 3.3. Regression results Fig. 1 presents the odds ratio (ORs) and confidence intervals for drive time and supply from the outpatient treatment utilization models. The drive time and supply values are in standard deviation (SD) units and all covariates previously described are included in the models, but only the main coefficients are presented. Full results can be found in Table A2 in Appendix A . Fig. 1. Open in a new tab Predicting outpatient treatment utilization: Odds ratios for drive time and supply for outpatient models. Note: Drive time and supply are in standard deviation units. Biological mothers accessed outpatient treatment following 20,266 (20.6%) referrals. In the full sample, drive time is negatively associated with outpatient treatment utilization, with modestly larger coefficients in the models with larger service area scales. For a wide service area scale, a 1 SD increase in drive time is associated with an 11% decrease in odds of outpatient treatment utilization. A 1 SD increase in supply is associated with a 5% increase in odds of outpatient treatment for narrow and mid-range service areas. The association lessens in wide service areas. In the models subgrouped by county urbanicity, we find that that associations between drive time and outpatient treatment utilization are concentrated within large and medium/small metro counties. Specifically, for the mid-range service area, a 1 SD increase in drive time is associated with a 6% decrease in the odds of outpatient treatment in large metro counties and 5% decrease in medium/small metro counties. Coefficients are greater in magnitude in the wide service area specification, but smaller and statistically nonsignificant in the narrow service area specification. In addition, supply is associated with increased treatment uptake in large metro counties but not medium/small metro counties. Within nonmetro counties, patterns of association for drive time are inconsistent and generally statistically non-significant. The association for supply is significant for mid-range (OR = 1.25) and wide (OR = 1.23) service areas. Fig. 2 presents the same information as above but for inpatient treatment utilization. Note that the x -axes cover a wider range than in Fig. 1 . In the full sample and across county types, drive time is often not significantly associated with inpatient treatment utilization, although the coefficients are similar in magnitude to those in the outpatient models. In the full sample, supply is significantly positively associated with inpatient treatment for all three service area specifications. For a wide service area scale, a 1 SD increase in supply is associated with a 35% increase in odds of treatment utilization. The subgroup models suggest that the association between supply and inpatient treatment uptake remains consistent across county types, except for large metro counties with a wide service area specification and nonmetro counties with a narrow and mid-range specification. Fig. 2. Open in a new tab Predicting inpatient treatment utilization: Odds Ratios for driving time and supply for inpatient models. Note: Drive time and supply are in standard deviation units. Biological mothers accessed inpatient treatment following 3777 (3.8%) referrals. 3.4. Supplemental analyses In supplemental models, we divide the sample by whether the CWS referral pertained to caregiver substance use ( Fig. A1 Appendix ). This supplemental analysis was only conducted for outpatient models, as the inpatient models were underpowered for the sample subset to mothers without a substance use allegation. For both outpatient and inpatient models, we divided the sample by whether the biological mother was 24 years old or younger ( Fig. A2 for outpatient and A4 for inpatient), and whether the CWS referral included one child ( Fig. A3 for outpatient and A5 for inpatient). Results of these supplemental models are generally consistent with the main analysis, with some exceptions. For example, associations between drive time and outpatient treatment utilization appear to diminish for younger mothers. Further research with a larger sample is needed to confirm these supplemental analyses. 3.5. Study limitations Several limitations in our methodology should be noted. First, there is substantial uncertainty in the definitions of service areas. Many mothers in our sample may rely on public transportation, primarily in metro counties: if so, their driving time understates their travel time. Second, important aspects of accessibility cannot be measured in this study. For example, recent evidence suggests that the majority of the time spent accessing opioid treatment programs for individuals with a high need for treatment consist of out-of-vehicle travel components, such as walking or waiting in line ( Kim et al., 2024 ), time not captured in the drive time metric used in this study. Mothers may also vary in access to childcare, including informal care support from relatives, which is likely particularly valuable for mothers seeking inpatient treatment. Third, parents in child welfare system often face stigma related to their substance use, which may impact their willingness to engage in services ( Dir et al., 2025 ). However, treatment participation may also be a condition of retaining or regaining custody of children. Fourth, our measure of service population (non-elderly adults on Medicaid) includes individuals who would never have a need for SUD treatment. If the share of non-elderly Medicaid participants who need SUD treatment is relatively stable across census block groups, this would have no effect on our analysis. However, substantial variation across block group in the prevalence of SUD treatment need among non-elderly adults enrolled in Medicaid could distort out measure of treatment supply and ultimately bias our findings. Also, Pennsylvania expanded Medicaid under the Affordable Care Act, reducing the share of low-income parents without health insurance. Barriers to treatment access may be more pronounced in other states where a larger share of parents are uninsured. Lastly, our measure of SUD treatment is limited in several respects. First, we consider only specialty providers: mothers may make use of an array of other services – such as MOUD prescribed by a general practitioner or community-based twelve-step meetings – that are not captured by our measures. Facilities in the SAMHSA files are recorded yearly and select facilities may open or close during the study window. We also cannot measure the capacity of each facility (i.e., number of beds) as SAMHSA does not include this information in their provider directories. 4. Discussion This study examines how SUD treatment accessibility is associated with utilization among mothers involved with the CWS in Pennsylvania. The majority (66.5%) of mothers in our sample have either a CWS case related to substance use or a diagnosed SUD; yet a small share received outpatient (20.6%) or inpatient SUD treatment (3.8%) within three months of their CWS referral. Thus, constraints on access could play an important role in treatment participation, which in turn may impact mothers’ ability to retain or regain custody of their child(ren). In examining treatment accessibility, we consider both different service area sizes and measures of accessibility (drive time and supply). We highlight two key findings. First, drive time to nearby SUD treatment facilities was negatively associated with treatment utilization, particularly outpatient treatment. The higher importance of driving time for outpatient treatment facilities relative to inpatient treatment facilities is logical given the number of trips to and from the facility that a client would need to make. Mothers’ outpatient treatment uptake was more sensitive to drive time in large metro counties compared with medium and small metro counties or nonmetro counties. This finding may be related to an increased reliance on public transit in large metro areas, which may take longer than driving and impose logistical challenges (i.e., aligning appointment times with transit schedules). Overall, findings regarding drive time align with prior literature linking distance and drive time to treatment utilization ( Alibrahim et al., 2022 ; Amiri et al., 2018 ; Cantor et al., 2025 ; Rosenblum et al., 2011 ). Given various efforts to increase utilization of medications for opioid (and, to a lesser extent, alcohol) use disorders, broadening the set of providers with the ability (and expertise) to prescribe these medications may increase treatment utilization. Efforts to distribute MOUD via pharmacies for at-home use have grown in popularity, as pharmacies are more readily accessible than opioid treatment facilities ( Kleinman, 2020 ). Yet, it is important to remember that such medications can pose a danger in the home: thousands of children have been poisoned by medications for opioid use disorder ( Land et al., 2020 ) – presumably medications improperly stored by the parent to whom it was prescribed, or due to illicit diversion of such medications for recreational use ( Chilcoat et al., 2019 ; Smith et al., 2020 ). This population also exhibits high rates of polysubstance use (see Table A1 ), meaning that MOUD may address only one component of their substance use. Prior research has shown that this group has a distinct risk profile compared with other individuals seeking SUD treatment ( Grella et al., 2006 ). Evidence further suggests that SUD treatment delivered in conjunction with psychosocial interventions may yield the most favorable outcomes ( Amato et al., 2008 ). Thus, broader efforts to mitigate distance as a barrier to treatment are needed, particularly for the highest-risk patients for whom at-home medications are insufficient as a standalone treatment. Second, the population-weighted supply of treatment at a wide service area scale was associated with a 35% increase in the odds of uptake of inpatient treatment, suggesting that lack of facilities remains an important barrier to treatment in this high-need population. In Pennsylvania, the supply of inpatient (Medicaid-accepting) treatment is much lower than for outpatient treatment. Even with the extended service area scales, the supply of inpatient treatment is roughly half that of outpatient treatment. Even if the need for inpatient services is in general less than for outpatient treatment at a county or state level, it may be that mothers in need of inpatient treatment are less likely to find openings than mothers in need of outpatient treatment. Inpatient treatment is often used for more severe or long-term SUDs, sometimes after unsuccessful outpatient treatment attempts. Delays to accessing more intensive treatment may not only place mothers at risk for adverse events, such as overdose, but also compromise the safety of their children and increase the probability that mothers lose custody. Mothers involved with CWS are more likely to rely on Medicaid than any other form of insurance; although CWS agencies can sometimes purchase services separately from Medicaid, they have limited resources with which to do so. Expanding the supply of providers available to mothers in the CWS may require improving the Medicaid reimbursement rate for such services, which is far lower than Medicare and private insurance ( Clemans-Cope et al., 2022 ). Appendix A Table A1. Descriptives statistics. All counties Large metro M/S metro Nonmetro Data at the biological mother - referral level Biological mother-referrals (#) 98,373 38,790 40,611 18,972 Outpatient treatment within 90 days a (%) 20.6 24.7 17.7 18.4 Inpatient treatment within 90 days (%) 3.8 5.2 3 2.8 Substance use allegation (%) 39.4 41.8 38 37.4 Physical abuse allegation (%) 5.1 5.2 5.6 3.9 Child behavioral problem (%) 22.6 21.4 22.8 24.9 Number of children (M,SD) 2.1 (1.3) 2.1 (1.3) 2.1 (1.3) 2 (1.2) Youngest child age (M,SD) 5.1 (5) 4.8 (5) 5.2 (4.9) 5.5 (5.1) Biological mother age (M,SD) 31.9 (7.4) 32.1 (7.4) 31.7 (7.4) 31.7 (7.4) SUD diagnoses prior to referral (%) At least one SUD 58.1 65.4 53.6 52.9 More than one SUD 22.8 28 19.3 19.5 Alcohol 23.4 26.9 22.3 18.9 Opioid 31.6 35.7 27.5 32.2 Stimulant 5.9 4.9 5.5 8.9 Cocaine 13.8 19.8 10.9 7.3 Hallucinogen 2 3.5 1.2 0.6 Sedative 7.4 10.8 5.4 4.6 SUD diagnosis or substance use allegation b (%) 66.5 71.8 63.4 62.2 Race/ethnicity (%) Non-Hispanic Black 22 39.4 14.4 2.9 Hispanic 10.5 7.3 16.6 4.2 Other/unknown 3.2 3 3.9 2.2 Non-Hispanic White 64.2 50.3 65.1 90.7 Open in a new tab a Specific to drug and alcohol treatment providers. The percent of mothers with an outpatient Medicaid claim with a primary diagnosis of SUD within 90 days of the referral, regardless of provider type, was 24.6. See Limitations section. b Measures the presence of either a substance use disorder diagnosis prior to referral or a substance use allegation identified on the referral. Table A2. Odds ratios with 95% confidence intervals for the full sample outpatient regression models. Narrow Mid-range Wide Drive time 0.99 (0.95, 1.02) 0.94 (0.9, 0.99) 0.89 (0.83, 0.94) Supply 1.04 (1.01, 1.08) 1.05 (1.01, 1.09) 1.02 (0.98, 1.09) Substance use allegation 4.19 (4.03, 4.36) 4.21 (4.05, 4.38) 4.21 (4.04, 4.37) Physical abuse allegation 1.22 (1.11, 1.33) 1.23 (1.12, 1.34) 1.23 (1.13, 1.34) Child behavioral problem 0.78 (0.74, 0.83) 0.78 (0.74, 0.83) 0.78 (0.74, 0.83) Number of children 0.83 (0.82, 0.85) 0.83 (0.82, 0.85) 0.83 (0.82, 0.85) Youngest age of child 0.78 (0.76, 0.8) 0.79 (0.77, 0.81) 0.79 (0.77, 0.81) Biological mother age 1.19 (1.16, 1.21) 1.18 (1.15, 1.21) 1.18 (1.15, 1.21) Race/ethnicity (ref = Non-Hispanic White) Non-Hispanic Black 0.45 (0.43, 0.48) 0.45 (0.43, 0.48) 0.45 (0.43, 0.48) Hispanic 0.72 (0.68, 0.77) 0.73 (0.68, 0.77) 0.72 (0.68, 0.77) Unknown 0.67 (0.6, 0.74) 0.68 (0.61, 0.75) 0.68 (0.61, 0.75) Open in a new tab Note: Year and county fixed effects included but not shown. Drive time and supply are in standard deviation units. Table A3. Marginal effects for drive time and supply coefficients for outpatient models (given as change in probability). All Large metro Drive time Narrow −0.004 −0.002 Mid-range −0.011 * −0.01 * Wide −0.019 * −0.014 * Supply Narrow 0.008 * 0.016 * Mid-range 0.007 * 0.013 * Wide 0.005 0.011 Med/sm metro Nonmetro Drive time Narrow −0.003 −0.002 Mid-range −0.008 * 0.011 * Wide −0.011 * 0.007 Supply Narrow 0.001 0.003 Mid-range −0.002 0.034 * Wide −0.007 0.031 * Open in a new tab Notes: Baseline probabilities specified as the proportion each sample that utilized outpatient treatment. Drive time and supply are in standard deviation units. * Significant coefficient at the p < 0.05 level. Table A4. Marginal effects for drive time and supply coefficients for inpatient models (given as change in probability). All Large metro Drive time Narrow −0.002 −0.003 Mid-range −0.003 −0.003 Wide −0.005 * −0.004 Supply Narrow 0.007 * 0.015 * Mid-range 0.006 * 0.013 Wide 0.01 * 0.005 Med/sm metro Nonmetro Drive time Narrow −0.001 −0.001 Mid-range −0.005 * 0.001 Wide −0.005 −0.002 Supply Narrow 0.006 * 0.002 Mid-range 0.007 * 0.004 Wide 0.011 * 0.011 * Open in a new tab Notes: Baseline probabilities specified as the proportion of each sample that utilized inpatient treatment. Drive time and supply are in standard deviation units. * Significant coefficient at the p < 0.05 level. Fig. A1. Open in a new tab Predicting outpatient treatment utilization in all counties: Odds ratios for drive time and supply for outpatient logistic regression models subset by substance use being identified on referral. Note: Drive time and supply are in standard deviation units. Fig. A2. Open in a new tab Predicting outpatient treatment utilization in all counties: Odds ratios for drive time and supply for outpatient logistic regression models subset by biological mother’s age Note: Drive time and supply are in standard deviation units. Fig. A3. Open in a new tab Predicting outpatient treatment utilization in all counties: Odds ratios for drive time and supply for outpatient logistic regression models subset by number of children on referral. Note: Drive time and supply are in standard deviation units. Fig. A4. Open in a new tab Predicting inpatient treatment utilization in all counties: Odds ratios for drive time and supply for inpatient logistic regression models subset by biological mother’s age. Note: Drive time and supply are in standard deviation units. Fig. A5. Open in a new tab Predicting inpatient treatment utilization in all counties: Odds ratios for drive time and supply for inpatient logistic regression models subset by number of children on referral. Note: Drive time and supply are in standard deviation units. Footnotes ☆ Disclaimer: Data for this project was provided by the Pennsylvania Department of Human Services, Office of Children Youth and Families and Office of Medical Assistance Programs, but these entities do not certify the accuracy of results or endorse any positions taken by the authors. This research was funded with help from the National Institute of Child Health and Human Development R01 HD095946 and P50HD089922. CRediT authorship contribution statement Dylan Jones: Writing – review & editing, Writing – original draft, Methodology, Formal analysis, Conceptualization. Sarah Font: Writing – review & editing, Writing – original draft, Methodology, Formal analysis, Conceptualization. Marci Cross: Writing – review & editing, Writing – original draft, Methodology, Formal analysis, Conceptualization. Ezra Goldstein: Writing – review & editing, Writing – original draft, Methodology, Formal analysis, Conceptualization. Declaration of competing interest The authors have nothing to declare. References Ahmad F, Cisewski J, Rossen L, & Sutton P (2025). Provisional drug overdose death counts. National Center for Health Statistics. [ Google Scholar ] Alibrahim A, Marsh JC, Amaro H, Kong Y, Khachikian T, & Guerrero E (2022). Disparities in expected driving time to opioid treatment and treatment completion: Findings from an exploratory study. BMC Health Services Research, 22(1), 478. 10.1186/s12913-022-07886-7 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Amato L, Minozzi S, Davoli M, & Vecchi S (2008). Psychosocial and pharmacological treatments versus pharmacological treatments for opioid detoxification. The Cochrane Database of Systematic Reviews, 3. 10.1002/14651858.CD005031.pub2 [ DOI ] [ PubMed ] [ Google Scholar ] Amiri S, Lutz R, Socías ME, McDonell MG, Roll JM, & Amram O (2018). Increased distance was associated with lower daily attendance to an opioid treatment program in Spokane County Washington. Journal of Substance Abuse Treatment, 93, 26–30. 10.1016/j.jsat.2018.07.006 [ DOI ] [ PubMed ] [ Google Scholar ] Bivand R. No-neighbour observation and subgraph handling. Retrieved April 7, 2025, from https://cran.r-project.org/web/packages/spdep/vignettes/subgraphs.html . Bunger AC, Chuang E, Girth AM, Lancaster KE, Smith R, Phillips RJ, … Aarons GA (2024). Specifying cross-system collaboration strategies for implementation: A multi-site qualitative study with child welfare and behavioral health organizations. Implementation Science, 19(1), 13. 10.1186/s13012-024-01335-1 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Cantor J, Hernandez HG, Kofner A, Lai J, Agniel D, Simon KI, … Taylor EA (2025). Medicare beneficiary receipt of methadone by drive time to opioid treatment programs. JAMA Network Open, 8(4), Article e253099. 10.1001/jamanetworkopen.2025.3099 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Chilcoat HD, Amick HR, Sherwood MR, & Dunn KE (2019). Buprenorphine in the United States: Motives for abuse, misuse, and diversion. Journal of Substance Abuse Treatment, 104, 148–157. 10.1016/j.jsat.2019.07.005 [ DOI ] [ PubMed ] [ Google Scholar ] Clemans-Cope L, Lynch V, Payton M, & Aarons J (2022). Medicaid professional fees for treatment of opioid use disorder varied widely across states and were substantially below fees paid by medicare in 2021. Substance Abuse Treatment, Prevention, and Policy, 17(1), 49. 10.1186/s13011-022-00478-y [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Congressional Research Service. (2014). Child welfare: The adoption incentive program and its reauthorization. Corredor-Waldron A, & Currie J (2022). Tackling the substance use disorder crisis: The role of access to treatment facilities. Journal of Health Economics, 81, Article 102579. 10.1016/j.jhealeco.2021.102579 [ DOI ] [ PubMed ] [ Google Scholar ] Dir AL, Batch BL, & Aalsma MC (2025). Understanding parents’ experience navigating substance use treatment and the child welfare system. Journal of Substance Use and Addiction Treatment, 172, Article 209672. 10.1016/j.josat.2025.209672 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Giraud T (2022). Osrm: Interface between R and the OpenStreetMap-based routing service OSRM. Journal of Open Source Software, 7(78). 10.21105/joss.04574 [ DOI ] [ Google Scholar ] Goldstein EG, & Font SA (2025). Prevalence and treatment of maternal substance use disorder in child welfare. JAMA Health Forum, 6(3), Article e250054. 10.1001/jamahealthforum.2025.0054 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Grella CE, Hser Y-I, & Huang Y-C (2006). Mothers in substance abuse treatment: Differences in characteristics based on involvement with child welfare services. Child Abuse & Neglect, 30(1), 55–73. 10.1016/j.chiabu.2005.07.005 [ DOI ] [ PubMed ] [ Google Scholar ] Haight SC, Yoon J, Luck J, Harvey M, Shapiro-Mendoza C, Li R, & Ko JY (2022). Medicaid expansion in Oregon and postpartum healthcare among people with and without prenatal substance use disorder. Drug and Alcohol Dependence Reports, 5, Article 100096. 10.1016/j.dadr.2022.100096 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Higgs G (2009). The role of GIS for health utilization studies: Literature review. Health Services and Outcomes Research Methodology, 9, 84–99. [ Google Scholar ] Ingram D, & Franco S (2014). In National Center for Health Statistics (Ed.), 2013 NCHS urban–rural classification scheme for counties (Vital Health Stat 2(166)). [ PubMed ] [ Google Scholar ] Jarpe-Ratner E, Bellamy JL, Yang D-H, & Smithgall C (2015). Using child welfare assessments and latent class analysis to identify prevalence and comorbidity of parent service needs. Children and Youth Services Review, 57, 75–82. 10.1016/j.childyouth.2015.07.019 [ DOI ] [ Google Scholar ] Kiang MV, Barnett ML, Wakeman SE, Humphreys K, & Tsai AC (2021). Robustness of estimated access to opioid use disorder treatment providers in rural vs. urban areas of the United States. Drug and Alcohol Dependence, 228, Article 109081. 10.1016/j.drugalcdep.2021.109081 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Kim J, Lee J, Thornhill TA, Dennett J, Lu H, Howell B, … Gonsalves G (2024). Accessibility of opioid treatment programs based on conventional vs perceived travel time measures. JAMA Network Open, 7(2), Article e240209. 10.1001/jamanetworkopen.2024.0209 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Kleinman RA (2020). Comparison of driving times to opioid treatment programs and pharmacies in the US. JAMA Psychiatry, 77(11), 1163–1171. 10.1001/jamapsychiatry.2020.1624 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Land ME, Wetzel M, Geller RJ, Kamat PP, & Grunwell JR (2020). Analysis of 207,543 children with acute opioid poisonings from the United States National Poison Data System. Clinical Toxicology (Philadelphia, Pa.), 58(8), 829–836. 10.1080/15563650.2019.1691731 [ DOI ] [ PubMed ] [ Google Scholar ] Mark TL, Dolan M, Allaire B, Parish W, Strack C, Poehler D, … Butler V (2024). Untreated psychiatric and substance use disorders among caregivers with children reported to child protective services. JAMA Health Forum, 5(4), Article e240637. 10.1001/jamahealthforum.2024.0637 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] National Academies of Sciences, Engineering, and Medicine. (2019). The effectiveness of medication-based treatment for opioid use disorder. In Mancher M, & Leshner AI (Eds.), Medications for opioid use disorder save lives. National Academies Press. https://www.ncbi.nlm.nih.gov/books/NBK541393/ . [ PubMed ] [ Google Scholar ] Olfson M, Wall M, Barry CL, Mauro C, & Mojtabai R (2018). Impact of Medicaid expansion on coverage and treatment of low-income adults with substance use disorders. Health Affairs, 37(8), 1208–1215. 10.1377/hlthaff.2018.0124 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Palmer L, Font S, Eastman AL, Guo L, & Putnam-Hornstein E (2024). What does child protective services investigate as neglect? A population-based study. Child Maltreatment, 29(1), 96–105. 10.1177/10775595221114144 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Patton D, Liu Q, Kersten E, Lucenko B, & Felver BEM (2020). Substance use disorder treatment penetration among child welfare–involved caregivers. Washington State Department of Social and Health Services. Retrieved February 4, 2025, from https://www.dshs.wa.gov/sites/default/files/rda/reports/research-7-121.pdf . [ Google Scholar ] Rosenblum A, Cleland CM, Fong C, Kayman DJ, Tempalski B, & Parrino M (2011). Distance traveled and cross-state commuting to opioid treatment programs in the United States. Journal of Environmental and Public Health, 2011, Article 948789. 10.1155/2011/948789 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Rousset F, & Ferdy J-B (2014). Testing environmental and genetic effects in the presence of spatial autocorrelation. Ecography, 37(8), 781–790. 10.1111/ecog.00566 [ DOI ] [ Google Scholar ] Seay K (2015). How many families in child welfare services are affected by parental substance use disorders? A common question that remains unanswered. Child Welfare, 94(4), 19–51. [ PMC free article ] [ PubMed ] [ Google Scholar ] Smith KE, Tillson MD, Staton M, & Winston EM (2020). Characterization of diverted buprenorphine use among adults entering corrections-based drug treatment in Kentucky. Drug and Alcohol Dependence, 208, Article 107837. 10.1016/j.drugalcdep.2020.107837 [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Smith VC, Wilson CR, Committee on Substance Use, Prevention, Ryan SA, Gonzalez, … Walker LR (2016). Families affected by parental substance use. Pediatrics, 138(2), Article e20161575. 10.1542/peds.2016-1575 [ DOI ] [ PubMed ] [ Google Scholar ] Substance Abuse and Mental Health Services Administration. (2021). National Survey of Substance Abuse Treatment Services (N-SSATS): 2020. Data on substance abuse treatment facilities. Substance Abuse and Mental Health Services Administration. [ Google Scholar ] Substance Abuse and Mental Health Services Administration. (2024). Key substance use and mental health indicators in the United States: Results from the 2023 National Survey on Drug Use and Health (HHS publication no. PEP24-07-021, NSDUH Series H-59). Center for Behavioral Health Statistics and Quality, Substance Abuse and Mental Health Services Administration. [ Google Scholar ] United States Census Bureau Types of health insurance coverage by age (B27010). (2020). https://data.census.gov/table/ACSDT5Y2023.B27010?q=COOL+COVERS . ACTIONS View on publisher site PDF (1.8 MB) Cite Collections Permalink RESOURCES Similar articles Cited by other articles Links to NCBI Databases Follow NCBI NCBI on X (formerly known as Twitter) NCBI on Facebook NCBI on LinkedIn NCBI on GitHub NCBI RSS feed Connect with NLM NLM on X (formerly known as Twitter) NLM on Facebook NLM on YouTube National Library of Medicine 8600 Rockville Pike Bethesda, MD 20894 Web Policies FOIA HHS Vulnerability Disclosure Help Accessibility Careers NLM NIH HHS USA.gov Back to Top

Related documents

• Cross-system collaboration between the child welfare system and substance use treatment providers: a qualitative analysis of strategies, facilitators, and barriers.#645536
• Feasibility, Acceptability and Preliminary Treatment Effects of A-CRA for Youth in Compulsory Institutional Care#839803
• A Substance Use Adaptation of Fathering Through Change#871298
• Evaluation of Conditional Cash Transfers to Increase Retention in PMTCT Services in Akwa Ibom State, Nigeria#979203
• Treatment for Adolescent Marijuana Abuse#849223