Novel Implementation of State Reporting Policy for Substance-Exposed Infants

Data for this study were abstracted from the Connecticut CAPTA notification system. During the study timeframe (March 15, 2019 to July 21, 2021), 5526 notifications were made. Of these, 763 (13.8%) were dropped because of missing data. Each of these dropped notifications included the date of notification and only 3 variables: (1) the reporting hospital, (2) indication that the infant was tested for substances, and (3) indication that the infant did not have uterine exposure. Connecticut CAPTA policy does not require a notification for every infant who is tested for substances. Therefore, these dropped notifications should not have been submitted in the first place. No other information (ie, demographics) was available for these dropped cases. The remaining analytic sample size was 4763. Sample statistics are presented in Table 1.

To calculate percent and rate of total births, we used Department of Public Health birth data from March 15, 2019 to July 21, 2021. During this timeframe, 59 273 births were recorded in Connecticut. Notifications were made for ∼ 80 infants per 1000 live births.

The notification system directs the individual making the deidentified notification to report the infant for maltreatment using a separate child maltreatment reporting system if a safety concern exists. Safety concerns included: (1) infant tested positive as a result of maternal substance misuse (1 = no, 0 = yes), (2) there is a concern that mother’s substance use will impact parental functioning (1 = no, 0 = yes), and (3) family presents with suspicions of abuse or neglect (1 = no, 0 = yes). Additionally, if a dyad does not have a POSC by the time of notification, the system prompts the person making the notification to make a report. POSC can be developed during pregnancy, typically with a substance use treatment provider, or at the time of birth with a hospitalist or other hospital-based professional (ie, social worker). Using the safety concern and POSC variables, we calculated a categorical dependent variable to capture the 3 possible outcomes: 1= diversion with POSC (no safety concern and yes POSC), 2= maltreatment report with POSC (yes safety concern and yes POSC), and 3= maltreatment report without POSC (yes safety concerns and no POSC) in our analysis. These 3 outcomes were considered a nominal, multicategory dependent variable in our analysis.

Infant substance exposure type was recorded in the notification. Personnel could indicate any number of 11 substance categories (alcohol, buprenorphine, cocaine, methadone, prescription opioids, phencyclidine (PCP), prescription benzodiazepine, nonprescription opioids, other illegal or nonprescribed medication, the misuse of prescription drug or over-the-counter (OTC), and marijuana). Over 140 different combinations appeared in our data set. These data were recoded into 12 mutually exclusive dummy categories reflecting single substance types and polysubstance.

To further explore substance exposure type, we created additional nonmutually exclusive categories that captured whether a notification included single or polysubstance exposure to alcohol, marijuana, medication for opioid use disorder (MOUD; buprenorphine or methadone), other opioid (prescription opioid or nonprescription opioid), and other illegal drug (cocaine, phencyclidine PCP, other illegal or nonprescribed medication).

For simplicity, our regression model further grouped substances into 6 mutually exclusive categories on the basis of the exposure type: marijuana (reference group), alcohol, MOUD, prescription drug (prescription benzodiazepine, prescription opioid, and misuse of prescription drug or OTC), other illegal drug (cocaine, PCP, nonprescription opiates, and other illegal or nonprescribed medication), and polysubstance (any combination of 2 or more substances).

Mother’s age in years at birth was recorded as a whole number. Maternal race and ethnicity was recoded into 5 dummy variables for each of the race and ethnicity categories: non-Hispanic White, Non-Hispanic African American, Non-Hispanic multirace or other race, Hispanic (any race), unknown or not disclosed, with the largest, Non-Hispanic White, as the reference group.

We first performed a bivariate analysis between the 3 outcomes and independent variables described above. Next, we developed a multinomial logistic regression to identify the factors differentially associated with the 3 outcomes. We included all independent and control variables mentioned above, and findings were reported with robust standard errors at the hospital level. Finally, to better interpret our results, we reported relative risk ratio in the regression table. All research activities were approved by the University of Connecticut institutional review board.

Table 1 presents sample characteristics. In total, close to 8% of total births received notifications (n = 4763). The average age of mothers was about 28 years old. Non-Hispanic White mothers were most common (44%), followed by non-Hispanic Black (22%) and Hispanic mothers (20%). Twelve percent of notifications did not identify the mother’s race and ethnicity. Compared to the state population of births, mothers with notifications were younger on average. Non-Hispanic White and Hispanic mothers were proportionally underrepresented among notifications compared to state population data, and non-Hispanic Black mothers were overrepresented.

Across the 3 possible outcomes, racial and ethnic differences were observed. Non-Hispanic White mothers were more likely to have a report with POSC, non-Hispanic Black mothers more likely to have a report without POSC, and Hispanic mothers more likely to be diverted. The type of infant toxicology was also different across outcomes. Diverted cases were more likely to be tested with meconium and urine and less likely to be tested with urine only compared to reported cases.

Table 2 summarizes exposure types documented in notifications including the percentage of total births and the rate per 1000 births based on Department of Public Health birth data. Table 2 also includes P values reflecting results from unreported chi-square tests comparing percents of exposure types across the 3 outcomes. The polysubstance subcategories and combined single or polysubstance categories in Table 2 are not mutually exclusive. Among single substance notifications (84% of notifications), the most identified substances were marijuana only (69% of notifications; 55.6 per 1000 births) and MOUD only (buprenorphine and methadone; combined 6.5%).

We observed extensive variability in exposure type across the 3 outcomes. The diverted cases recorded 48% single substance exposures, reported with POSC cases recorded 15.5% single substance exposures, and reported without POSC cases recorded 33.3% single substance exposures. Cocaine (0.4%), other illegal or nonprescription drug (0.3%), PCP (0.1%), and misuse of prescription drug or OTC (0.08%) were less frequently associated with the diverted group. Alcohol (0.3%), prescription benzodiazepine (0%), and misuse of prescription drug or OTC (0.3%) were less frequently associated with the POSC with report group.

Table 2 also summarizes the 15% of notifications that included polysubstance exposure (12.6 per 1000 births). Polysubstance exposure was most common at 29% of the reported with POSC group, followed by 23% of the reported without POSC group and 7% of the diverted group.

Finally, Table 2 summarizes notifications that included either single or polysubstance use of four substance types. Notifications documenting any alcohol exposure (single and polysubstance) totaled 3% of notifications and a rate of 2.2 per 1000 births. Notifications documenting any marijuana exposure were most prevalent at 79% of notifications and a rate of 63.2 per 1000 births. Notifications documenting any opioid exposure, including legal and illegal opioids in single or polysubstance exposures, were second in prevalence at 21% of notifications and a rate of 16.5 per 1000 births. A total of 14% of notifications involved MOUD a rate of 11.4 per 1000 births).

Table 3 reports the multinomial logistic regression to understand factors associated with 3 outcomes. The model used the diverted cases as the reference group. The first half of Table 3 presents the risk factors associated with being reported with a POSC. As shown in the table, infants exposed to illegal drugs (including cocaine, nonprescription opiates, PCP, and other illegal drugs) had a relative risk ratio of 3.9 times more likely to be reported with POSC than diverted. Prescription drug exposures were 0.6 times more likely to be reported with POSC, whereas polysubstance exposure were 5.2 times more likely to be reported with POSC than diverted.

The second half of the table presented the risk factors associated with being reported without POSC. Mother’s age and infants’ substance exposure types were significantly associated with being reported without POSC. Infants with polysubstance exposures had the highest relative risk ratio (3.8) for being reported without POSC. In contrast, those exposed to prescription drug had the lowest relative risk ratio (0.4) for being reported without POSC.

Across both models, exposure to any illegal drug and polysubstance exposures had the largest relative risks for any maltreatment report with and without POSC. The lowest relative risk for maltreatment report was observed for prescription drugs.

Although the intent of CAPTA is to address opioid misuse, its actual implementation has revealed significant marijuana usage in pregnant people. Connecticut CAPTA identified 4 marijuana exposures for every 1 opioid exposure. In states in which marijuana is legal, exposures are likely much higher.⁸  Marijuana use in pregnancy is controversial because of its prevalence, shifting legal status, racialized implications, and the lack of clarity of its impact on the fetus.^9,10  Studies suggest that prenatal exposure, even exposure in very early pregnancy, may adversely impact physical and neurocognitive outcomes.^10–13  Marijuana is also illegal in most states and subject to CPS surveillance.¹⁴  Because of this, continued use of marijuana throughout pregnancy despite known risks to the pregnant person and her neonate suggests at least a mild substance use disorder according to DSM-5 criteria.¹⁵  CAPTA, then, represents 1 pathway to identify people with marijuana use disorder.

Although Connecticut CAPTA may aid in identification of marijuana, it is likely underidentifying alcohol exposure. Connecticut policy goes further than federal CAPTA language and requires notifications for fetal alcohol spectrum disorder (FASD) and alcohol exposure. However, our rate of alcohol detection (0.2% of births) was over 47 times lower than the national average rate of self-reported alcohol use in pregnancy.¹⁶  Like marijuana, there exists considerable debate regarding alcohol use in pregnancy as well as varied opinions regarding risk to infants.^17,18  Depending on quantity and timing, alcohol usage in pregnancy causes significant and intractable teratological and cognitive impairments in children.¹⁹  FASD are often underdiagnosed, and children with FASD are overrepresented in foster care and criminal justice systems.^20,21  As implemented in Connecticut, CAPTA represents a missed opportunity to identify people with alcohol use disorder.

In addition to disparate identification of types of substance exposures, our study revealed significant racial disproportionality among infants identified with PSE. Racial disproportionality in toxicology testing at birth and maltreatment reporting among IPSE is documented in thirty years of research.^22,23  Consistent with 2 previous population-based studies of prenatal substance exposure diagnoses at birth,^24,25  compared to the Connecticut state population, Black mothers were disproportionally overrepresented among notifications and Hispanic mothers were disproportionally underrepresented. Unlike these earlier studies, both of which found that White mothers were disproportionally overrepresented, White mothers in Connecticut were underrepresented. Notably, assessing proportionality using population parameters as the denominator is, in some ways, misleading, because hospitalists typically determine which infants are tested. Notifications measure infants who are tested and who test positive. In an attempt to overcome provider testing bias in measuring actual rates and type of substance use in pregnancy, some self-report studies suggest different rates and types of substance use in pregnancy according to race group, although as is the case in other studies on racial disproportionality in maternal-IPSE outcomes, once socioeconomic and health risk factors are accounted for, racial differences in substance use in pregnancy are no longer significant.²⁶  In our study, once accounting for substance type, race group was not significantly associated with outcomes.

Of greater importance than simple identification of substance exposure is the consequence of that identification. CAPTA aims to ensure a consistent public health response to identified dyads using universal POSC. Unfortunately, we found that several patient characteristics were associated with not receiving a POSC. Our results are consistent with other studies documenting different responses, including among health care providers, depending on substance type.^24,27 

Two main factors work against CAPTA’s public health goals. First, CAPTA requires a series of informal hospitalist assessments: whether to verbally screen, order toxicology, interpret toxicology as an indication of substance misuse, and infer substance use and other factors as indications of child maltreatment risk. This leaves physicians to rely on their own opinions, training, or hospital policies. Most birthing hospitals, including those in Connecticut, selectively screen for substance exposure, resulting in likely underdetection, particularly for older, more resourced, White women.^28,29  Even hospitals that use criteria-based screening practices do not overcome racial disparities in drug testing.³⁰  Moreover, routine toxicology rarely includes alcohol. Second, and relatedly, CAPTA is child welfare policy but has considerable implications for hospital-based physicians. With no financial stream for implementation and limited federal guidance on implementation, ensuring that hospitalists adopt CAPTA practices depends on adequate training and belief that the new approach will result in better outcomes for their patients. Existing literature documents low levels of CAPTA knowledge among health care providers,³¹  uncertainty regarding the effects of certain exposures,³²  providers feeling ill-equipped to respond to patients who admit use, and believing use will result in inevitable child welfare involvement.³² 

Addressing the problems of racial disproportionality and inconsistent use of POSC begins with universal screening for substance use, including alcohol, and mental health disorders. The American Academy of Pediatrics recommends universal substance use screening in pregnancy.³³  From a child safety perspective, cooccurring disorders present much higher risk than either substance use or mental health disorder alone,³⁴  and mood disorders are common among people who use substances in pregnancy.³⁵  Early identification and intervention facilitated by the POSC can mitigate many health and safety risks.³⁶  Therefore, federal, state, and local support to implement all CAPTA elements including universal POSC and connection to needed services is needed.³⁷ 

Our study is limited in several ways. As such, results must be interpreted cautiously. First, these data were administratively collected and deidentified. Records could not be validated against hospital discharge data, so we do not know whether all IPSE received a notification. Second, and relatedly, these are cross-sectional data collected at the time of the infant’s birth. We have no way of knowing when or for how long the substance exposure occurred or how the dyad fared after the notification was made. Third, although we could control for maternal age and race and ethnicity on likelihood of outcomes, we could not account for salient risk factors such as prenatal care, insurance status, or previous child welfare system involvement. Fourth, and similarly, we could not account for salient protective factors. For instance, we have no way of measuring substance use treatment use in pregnancy. Finally, these findings reflect 1 state’s implementation that may not generalize to other settings.

Future directions for research are many. Longitudinal research is needed to help pediatric hospitalists understand whether diverted cases remain safe in the community. CAPTA is implemented differently across states, and more research is needed to understand the effect of specific policy approaches, especially on identification across substance types. Additionally, research is needed to establish an evidence base regarding POSC and impact on service receipt. States vary regarding who develops and who monitors these plans.³⁸  Finally, research is needed to understand the therapeutic versus punitive effects of a public health policy implemented under the purview of CPS. Whether an historically punitive system can truly support these families is unclear and, according to many, unlikely.

The complex needs and risks for IPSE warrant a nuanced, thoughtful, and multi-system policy strategy. Connecticut CAPTA is a novel policy approach that aims to divert IPSE without safety concerns away from CPS. Despite limitations including administratively collected deidentified data, our study found that over half of all identified infants were diverted to receive supportive services in the community. However, racial disproportionality in identification was apparent. The type of substance exposure was the most prominent factor associated with diversion. Alcohol exposure was significantly underidentified. Future research is needed to refine our understanding of CAPTA’s effects.