Reasonable efforts to preserve families? An examination of service utilization and child removal

Abstract

Background:

Recent federal policy has solidified the importance of preserving families, yet over 400,000 children enter foster care each year. Although a few studies have found that certain types of services, like intensive family preservation services, may reduce child removals, more research is needed.

Objective:

This study examined the relationship between family preservation, family support, and basic need service utilization and child removal among families with substantiated cases of maltreatment.

Methods:

We conducted a survival analysis using data from the National Child Abuse and Neglect Data System (NCANDS). We took a cohort of families with an investigation and substantiation of maltreatment in FY 2018 and followed them through FY 2019 to identify any that experienced a child removal during the study period. This included a total of n = 558,915 children.

Results:

Approximately 15.33% of children experienced a removal during the study period. Case management was the most frequently reported service, followed by transportation services and family preservation services. In the multivariable analysis, family preservation services (HR = 0.95, p < .01), home-based services (HR = 0.98, p < .001), and housing services (HR = 0.87, p < .001) decreased the hazards of child removal. Family support services (HR = 1.36, p < .001), transportation services (HR = 2.30, p < .001), education (HR = 1.13, p < .01), case management (HR = 1.83, p < .001), or daycare (HR = 1.26, p < .001) increased the hazards of child removal. Discussion: Findings from this study suggest that utilization of various services is associated with future child removals. While preservation and home-based services decreased the likelihood of removal, several basic needs services increased the hazards of child removal. This may reflect too little too late with services that may be better applied as primary or secondary preventive efforts. Implications for policy and future rollout of the Families First Prevention Services Act are explored.

1. Introduction

For the over 3 million referrals to Child Protective Services (CPS) each year, the priority is to maintain the family unit, while also ensuring child safety (Geiger & Schelbe, 2021). Family preservation and prevention of child removals have been solidified in recent federal policy through the Families First Prevention Services Act of 2018 (FFPSA). Indeed, expanding eligibility for funds is to cover preventive services that could enable families at-risk of removal to stay together is one of the three principles of FFPSA (Children’s Defense Fund, 2020). Despite the policy focus on preserving families, in 2019 in the U.S., 423,997 children across the nation were in the foster care system (USDHHS, 2020a). Thus, while the majority of cases referred to child protective services do not end in a child’s removal, there are almost a half million children for whom family preservation has not been achieved.

When a case is referred to child protective services for maltreatment and screened in, an investigation determines whether there was substantiated maltreatment and assesses the level of risk to a child’s safety. This may include consideration of parental risk factors, type of abuse substantiated, child age, and support systems in place (Hanna et al., 2021). Although recent federal policy has reiterated the importance of family preservation services, some experts have noted a disconnect, particularly in neoliberal child welfare systems such as that in the United States, between the services ordered by the court and available family supports (Broadhurst & Mason, 2017). Indeed, post-response services fall under the category of tertiary prevention as they are provided in response to a referral of child maltreatment (FRIENDS, 2021); this often results in families only receiving services after alleged child maltreatment has already occurred. Once tertiary services are provided, they may be limited in that they may be too short-term or crisis-oriented instead of ongoing holistic supports (Kemp et al., 2014). Furthermore, post-response services may target issues that could have been addressed through usual care of primary prevention efforts (Klevens & Whitaker, 2007). While post-response services in the U.S. often are only available after CPS has received a referral for alleged maltreatment, some child welfare systems abroad offer more universal, family-oriented support services. For example, family preservation and support has been an integral pillar of other child welfare systems, such as the family-oriented child welfare systems in Sweden (Heimer et al., 2018) or Norway (Križ & Skivenes, 2014).

Traditionally, services offered to families in lieu of child removal may fall under family preservation or family maintenance (Lee & Ayon, 2007; Lloyd, 2020), also referred to as safety management services. Family preservation services vary by state and include case management, basic needs, family violence safety planning, and referrals to supplemental resources (Chaffin & Bard, 2011). Family preservation services tend to be home-based, culturally sensitive, and may last between six and twelve months, whereas family maintenance services focus on reducing harm or threats to child safety and tend to include more case management through child welfare agencies (Lee & Ayon, 2007). Services to prevent child removals may include education or skills-based services (e.g., home visiting, Triple P), mental health services, substance use services, or basic needs services (e.g., housing services, daycare) (Child Welfare Information Gateway, 2021; Strydom, 2012).

Based on 2019 reports, among states that reported post-response service utilization, the proportion of families referred to CPS who received services ranged from 20.6% (DC) to 100% (Tennessee) (USDHHS, 2020b). Among children referred to CPS and who were removed as a result of the investigation, between 5.3% (Kentucky) and 46.8% (Hawaii) of children received post-response services (USDHHS, 2020b). While there are few national studies to date of referrals to services and completion, some individual states have published data. For example, in Texas, the largest state-administered child welfare system in the United States, 19,769 families were referred to family preservation services in 2019. Of those, 18,957 completed services (Texas Department of Family and Protective Services, n.d.). However, to date, there are few studies that have looked at service utilization rates across the United States and subsequent removals.

1.1. Service utilization and family preservation

Evidence from prior studies has been inconclusive about the impact of post-response service utilization on subsequent child removal. Fuller and Nieto (2014) found that families who utilized in-home post-investigation services had a higher likelihood of recurring maltreatment than those who did not receive any. Tambling and Johnson (2021) found that the number of hours spent with families and the number of home visits were positively correlated with intensive family preservation service completion. However, they did not examine removals or future referrals for maltreatment.

One prior study (Loman & Siegel, 2015) examined the relationship between utilization of diverse services and future referrals or removals among families that received differential response instead of a traditional CPS investigation. In their study, Loman and Siegel (2015) found that children who received differential response instead of traditional CPS investigations received more services such as transportation, financial help, counseling, and housing; however, in this study, utilization of services was not linked to future removals or re-referrals.

For basic needs services, findings from prior studies have had had inconclusive findings. For example, in a study of 828 children in two states, Pergamit et al. (2017) found that utilization of housing services after initial referral did not reduce the odds of child removal. However, in a follow-up study, Pergamit et al. (2019) found that prioritizing child welfare-involved families for housing assistance decreased involvement with child welfare, although results varied significantly by site. In parallel, Gubits et al. (2015) found that offering housing vouchers to families experiencing homelessness significantly reduced foster care placement within the following year-and-a-half. Indeed, Wade (2018) argued that the lack of housing services under “reasonable efforts” legislation may lead to more child removals. This aligns with findings from a literature review Canfield et al. (2017) conducted, in which they concluded that educational training among mothers with substance use disorders could decrease future child welfare involvement. Similarly, based on qualitative interviews with parents involved in child welfare, Kokaliari et al. (2019) found that parents perceived more educational and employment training as a way to improve opportunities for family preservation and maintenance.

1.2. The current study

This study used a national, administrative dataset to explore the relationship between service utilization among families referred to Child Protective Services (CPS) and subsequent foster care entry. The following research question guided the study: What is the relationship between post-response service utilization and subsequent child removal among families referred to CPS? Based on prior literature, we hypothesized that service utilization would be negatively associated with foster care entry.

2. Method

Data for this study come from the National Child Abuse and Neglect Data System (NCANDS) Child File, FY 2018, version 3, and FY 2019, version 2. The NCANDS Child File dataset contains information about investigated reports of maltreatment to state CPS agencies. Variables included child-level measures such as child age, sex, and race/ethnicity, parental risk factors, perpetrator relationship to child, and services utilized, among others. Each child has a unique child identifier in the dataset which can be linked to other datasets, such as other years of NCANDS as well as the Adoption and Foster Care Analysis and Reporting System (AFCARS), to track children who enter foster care or are re-referred to CPS after foster care discharge. For a complete description of NCANDS data collection and design, please see U.S. Department of Health and Human Services (2019). For the analytic sample, we included children who were referred to CPS in 2018 and had a substantiated case of child maltreatment (n = 731,773). Although some families have multiple children referred to CPS, there is no family indicator in the data. Therefore, we used child unique identifier to identify unique cases; this is noted in the Limitations section. We deleted duplicate cases, in which children experienced multiple referrals in one year (n = 116,859) and used the last referral in cases where there were multiple child identifiers in the data. To determine whether the child had entered foster care, we examined an indicator in the FY 2018 and FY 2019 NCANDS data that measured whether the child was provided with foster care services; in addition, we matched potential cases by the “StFCID” variable that gave a unique state and child identifier to children in NCANDS and AFCARS. We dropped cases of children whose families only received services the day of or after child removal (n = 55,999). Thus, we included children who had no date of service utilization (e.g., no services had been provided) and children who had a date of service utilization with either no removal date or a removal date that was at least one day after the service utilization date. This resulted in a final analytic sample of n = 558,915 children.

2.1. Measures

2.1.1. Dependent variable

Child removal into foster care was the main dependent variable. This was a dichotomous measure to compare children who were removed within 12 months of the investigation (1) to those who were not removed (0). Approximately 15.33% of children experienced removal within the study timeframe.

2.1.2. Time variable

This study utilized Cox survival analyses to test the relationship between service utilization and foster care entry. This method requires identifying a time variable that measures the likelihood of an outcome over a given period of time while also including a censored variable to control for cases where the event occurred (e.g., foster care entry) and cases in which the event did not occur. For this study, time was measured using the number of days between the referral report and the removal. For children who were not removed during the study period, time was calculated as the number of days between the last day of the fiscal year (September 30, 2019) and the referral report date. Cases were observed for 1 to 637 days.

2.1.3. Service utilization

Two types of services were included in the analyses: family preservation/support services, and basic needs services. All services were based on caseworker report regarding whether the family received the service (yes/no). Included in our analyses were three family support/preservation services: 1) family support services; 2) family preservation services; and 3) home based services. Also included were five basic needs services: 1) daycare services; 2) education services; 3) housing services; 4) transportation services; and 5) case management services. Each of these service types was included in the model as a dummy-coded variable indicating whether the family received the service (1 = yes) or not (0 = no).

2.1.4. Demographics and covariates

Child race/ethnicity, child sex, child age, prior victim status, parental risk factors, and type of maltreatment were all entered as covariates. Child race/ethnicity was a categorical, mutually-exclusive variable that included non-Hispanic White children (0), non-Hispanic Black children (1), non-Hispanic American Indian children (2), Hispanic children of any race (3), non-Hispanic Asian children (4), non-Hispanic Native Hawaiian/Pacific Islander children (5), and non-Hispanic Multiracial children (6). Child sex was a dichotomous variable that included female children (1) and male children (0). Child age was measured in years at the time of report to CPS. Prior victim status was a dichotomous variable in which children were identified as having had a previous referral (1) and those who had not had a prior referral (0). Parental risk factors were dichotomous measures reported by the caseworker and included parental emotional disturbance, domestic violence, housing instability, and parental drug use (dummy-coded, 1 = yes, 0 = no) Type of maltreatment was also a dichotomous measure reported by the caseworker and included physical abuse, neglect, and sexual abuse (dummy-coded 1 = yes, 0 = no).

2.2. Missing data

Using the “mi set,” “mi misstable summarize,” and “mi misstable pattern” commands in Stata, version 16, we ran descriptive statistics on the analytic sample to identify missingness and patterns in missingness. Model variables ranged from zero to 60% missing data, with the largest proportion of missing data for parental risk factors and data were not missing completely at random. Table 1 displays the percent of missing data on each variable prior to imputation. Congruent with prior research (Azur et al., 2011), we conducted multiple imputation chained equations (MICE) to impute missing values. MICE is the preferred approach to multiple imputation when data are missing for both categorical and continuous variables (White & Royston, 2009).

Table 1.

Imputed sample characteristics (N = 558,915).

	Percent of imputed sample	95% CI		% Missing (prior to imputation)
		LL	UL
Child race
Non-Hispanic White	46.24	46.10	46.37	5.53
Non-Hispanic Black	21.70	21.59	21.81	5.53
Hispanic	24.38	24.26	24.50	5.53
Native American	1.37	1.34	1.40	5.53
Asian	1.02	0.99	1.05	5.53
Native Hawaiian	0.19	0.18	0.20	5.53
More than 1 race	5.06	5.01	5.12	5.53
Child gender				0.40
Male	51.72	51.59	51.85
Female	48.27	48.14	48.40
Parental risk factor
Parental drug abuse	32.18	31.86	32.49	46.11
Parent emotionally disturbed	10.76	10.38	11.14	60.42
Interpersonal violence	29.10	28.93	29.27	30.94
Parental housing instability	8.09	7.98	8.20	44.38
Type of maltreatment
Physical abuse	23.12	23.01	23.23	0
Sexual abuse	11.23	11.15	11.31	0
Neglect	76.18	76.07	76.29	0
Prior victim of maltreatment	70.39	70.27	70.51	1.05
Removed	15.33	15.24	15.43	0
	M	95% CI		% Missing (prior to imputation)
		LL	UL
Child age at removal	7.08	7.07	7.10	0.12
Length of stay	578.18	577.66	578.70	0

2.3. Data analysis

We conducted descriptive statistics to examine sample characteristics, bivariate statistics to examine service utilization by race/ethnicity, and a series of multivariable Cox regression (survival analyses) to explore the relationship between service utilization and subsequent child removal while controlling for other variables. Survival analysis, specifically the Cox regression approach, was selected as the best analytic tool to answer our research questions because it enabled us to consider not only likelihood of the outcome variable occurring, but the time it took for the outcome to occur (Crowder, 2012). Child welfare involvement, including receipt of prevention interventions, is a longitudinal process where the outcome of interest (i.e., child removal) could occur at any time within a long window. Survival analysis permits sampling using an entry cohort method, meaning that we could make statistical inferences on an entire cohort of cases from a starting point to an ending point (the last date within a data set), even though many of the cases within the cohort had not achieved the outcome by the last date of observation (i.e., censored cases; Crowder, 2012). Furthermore, the Cox regression approach to survival analysis includes a standardized effect size coefficient, the Hazard ratio, which provides an estimate of the likelihood that a case characterized by the independent variable (i.e., using a given service) will experience the outcome (i.e., removal) on a given day during the study. We followed a bottom-up model building approach; the first model was an unadjusted model, where each variable was introduced in a separate model to explore the relationship between each independent variable and removal without adjusting for other factors. In Model 2, we entered all independent variables and controls in the same model to explore the adjusted Hazard ratios. Given the use of a national dataset, we utilized robust standard errors to account for clustering within states.

3. Results

Table 1 displays the sample characteristics. The largest percentage of children in the sample was non-Hispanic White (46.24%), followed by Hispanic (24.38%), non-Hispanic Black (21.70%), non-Hispanic Multi-Racial (5.06%), non-Hispanic American Indian (1.37%), Asian (1.02%), or non-Hispanic Pacific Islander (0.19%). There were slightly more male (51.72%) than female (48.27%) children. The most frequently reported parental risk factor was parental drug abuse (32.18%) followed by interpersonal violence (29.10%). Neglect was the most frequently reported type of maltreatment (76.18%). The majority of children had experienced a prior history of maltreatment victimization (70.39%). Approximately 15.33% of the sample entered foster care before the study end date as a result of the CPS referral made in 2018.

Fig. 1 displays the percentage of children/families that utilized each service during the study period. The most frequently utilized service was case management, followed by transportation, daycare services, family preservation, housing, family support services, education, and home-based services.

Fig. 1.

Service utilization among children referred to CPS in FY 2018.

3.1. Service utilization and removal

Table 2 presents the results of Model 1 (unadjusted) and Model 2 (adjusted) that explored service utilization and removal across the entire sample.

Table 2.

Service utilization and removals from the 2018–2019 National Child Abuse and Neglect Datasystem (NCANDS), imputed sample, N = 558,915.

Characteristics	Model 1 unadjusted			Model 2 adjusted
	HR	95% CI		HR	95% CI
		LL	UL		LL	UL
Family support	2 41***	2.35	2.46	1.36***	1.23	1.33
Family preservation	1.99***	1.96	2.03	0.95**	0.92	0.97
Home based services	2 70***	2.61	2.80	0.98***	0.93	1.03
Daycare	2 87***	2.81	2.94	1.26***	1.22	1.30
Education	3.49***	3.31	3.69	1.13**	1.06	1.20
Housing	3.40***	3.29	3.50	0 87***	0.82	0.92
Transportation	4.36***	4.25	4.47	2.30***	2.21	2.39
Case management	2.58***	2.55	2.62	1.83***	1.79	1.86
Child race (ref. white)
Non-Hispanic Black	0.91***	0.89	0.93	1.00*	0.97	1.02
Native American	1.51**	1.44	1.58	1.20***	1.13	1.28
Hispanic	0.96***	0.95	0.98	0.91***	0.89	0.92
Asian	0.57***	0.52	0.62	0.68***	0.62	0.74
Native Hawaiian	1.02	0.87	1.19	1.17	0.99	1.37
More than 1 race	1.33***	1.29	1.37	1 19***	1.15	1.23
Female child	0 92***	0.91	0.93	0.99	0.97	1.00
Child age	0 97***	0.96	0.97	0 97***	0.97	0.97
Child prior victim	0.60***	0.59	0.61	0.57***	0.56	0.58
Parental risk factor
Parental emotional disturbance	1 78***	1.73	1.84	1.11**	1.05	1.17
Domestic violence	1.03***	1.01	1.05	0 92***	0.90	0.93
Housing instability	214***	2.08	2.20	1.32***	1.28	1.36
Parental drug use	1 78***	1.72	1.84	1.28***	1.23	1.33
Type of maltreatment
Physical abuse	0.94***	0.93	0.96	1 10***	1.07	1.12
Neglect	1.85***	1.82	1.89	1.35***	1.31	1.38
Sexual abuse	0.49***	0.48	0.50	0.91***	0.88	0.94
Model fit
Average RVI					0.73
F					1312.59***

Notes. HR = hazard ratio; CI = confidence interval.

^*p < .05.

^**p < .01.

^***p < .001.

In the unadjusted model, utilization of family support services (HR = 2.41, p < .001), family preservation services (HR = 1.99, p < .001), and home based services (HR = 2.70, p < .001), increased the hazards of removal. Among the basic needs services, utilization of daycare services (HR = 2.87, p < .001), education services (HR = 3.49, p < .001), housing (HR = 3.40, p < .001), case management (HR = 2.58, p < .001), or transportation (HR = 4.63, p < .001) also increased the hazards of removal. Before adjusting for other factors, children who were Black, Hispanic, or Asian, had lower hazards of removal than their White peers, while those who were American Indian or Multiracial had higher hazards of removal than White children. Female children had lower hazards of foster care entry than their male counterparts, and child age was negatively associated with foster care entry. Before adjusting for other factors, children whose parents had an emotional disturbance, history of domestic violence or drug use, or who experienced housing instability had higher hazards of foster care entry than those without. Children who experienced physical abuse or sexual abuse had lower hazards of foster care entry than those with other types of maltreatment, while those who experienced neglect had higher hazards of foster care entry.

In Model 2, all variables were entered simultaneously to adjust for the effects of parental risk factor, type of maltreatment, child gender, age, race/ethnicity, and service utilization. In this model, utilization of family preservation services (HR = 0.95, p < .01), home-based services (HR = 0.98, p < .001), or housing services (HR = 0.87, p < .001), reduced the hazards of foster care entry. In contrast, utilization of family support services (HR = 1.36, p < .001), daycare services (HR = 1.26, p < .001), education (HR = 1.13, p < .01), case management services (HR = 1.83, p < .001), or transportation (HR = 2.30, p < .001) increased the hazards of removal. After adjusting for other factors, children who were American Indian or Multiracial had higher hazards of entry than their White peers, while those who were Hispanic or Asian had lower hazards. As child age increased, the hazards of removal decreased. Children whose parents had an emotional disturbance, history of drug use, or housing instability had higher hazards of foster care entry than those without, while those who had been exposed to domestic violence had lower hazards of foster care entry. After adjusting for other factors, children who experienced physical abuse or neglect had higher hazards of foster care entry, while those who experienced sexual abuse had lower hazards of foster care entry.

4. Discussion

This study examined the relationship between service utilization after a report to child protective services and subsequent child removal. This study built upon some prior intervention studies to explore trends with a larger sample from diverse states. Our first hypothesis was partially supported by our findings. Notably, some family preservation services, such as home based services and general preservation services, decreased the hazards of foster care entry during the study period. Therefore, it is possible that family preservation services may prevent foster care entries in a short/intermediate term. However, it is also worth noting that family support services were associated with a higher likelihood of foster care entry among children in our sample. Our findings contradicts some previous research (e.g., Chaffin et al., 2001; Lindsey et al., 2002) but aligns with more recent research that has found specific types of services such as intensive family preservation to decrease the likelihood of child removal (Bezeczky et al., 2020; Kirk & Griffith, 2004). Our findings also respond to recent calls to conduct larger-scale research on the impact of family preservation and support services (O’Reilly et al., 2010). Notably, several cases were dropped from our analytic sample because they only received services at or after removal. Therefore, although family preservation services, housing, and in-home or home-based services were associated with a lower likelihood of foster care entry among families in our sample after adjusting for other factors, more research is needed to understand which families are utilizing services and at what point of their case.

In contrast to our hypothesis, basic needs services such as transportation, education, and daycare increased the hazards of foster care entry. In fact, prior to adjusting for other factors, utilization of any service increased the hazards of foster care entry. After adjusting for parental risk factors, types of maltreatment, and other factors such as prior victimization and child age, housing was the only basic service that reduced the likelihood of foster care entry. It is possible that our inclusion of only substantiated cases of child maltreatment increased the likelihood of overall child removal. It is also plausible that services were provided as the investigation wrapped up and were not completed prior to the decision to remove the child. With recent changes to the Title IV-E program under the Families First Prevention and Services Act of 2018, it is also possible that caseworkers make more efforts to secure services for families in cases where a removal looks likely. FFPSA requires social service agencies to make reasonable efforts to preserve families prior to removing a child (Child Welfare Information Gateway, 2019). This includes services like family preservation, family support, and home-based services. It is worth noting that we removed several cases from the analytic sample because they were only provided services at or after removal. In contrast to family-oriented child welfare systems, the child protection and safety focus of the U.S. may still only provide support services to cases deemed high-risk, where foster care entry is imminent. Therefore, it is also possible that children whose families received services mere days before removal had already been identified as high-risk for foster care entry. Future research could compare the relationship between service utilization and child removals across different types of child welfare systems, such as child safety-focused compared to family service-oriented approaches more common in some of Western Europe (Burns et al., 2017; Pösö et al., 2021).

Alternatively, it is possible that the increased likelihood of child removal among families that used basic needs service reflects the “failure of past and current social policies to ensure provision of the institutional supports required to enable parents to meet their children’s development needs” identified by other researchers (McGowan, 1990, pp. 65–66). Families in our sample already had a referral to CPS that was substantiated for maltreatment; therefore, it is possible that this reflects the failure of first-line efforts to stabilize housing and improve job and educational opportunities among parents as a means of preventing initial maltreatment. It is also possible that families in our sample experienced a pile-up of strain due to factors such as financial instability, housing instability, or mental health issues, all of which may have contributed to decreased parenting skills. Prior research has found that increased strain can impact outcomes related to family adaptation, including parenting stress (Meleady et al., 2020). In our study, we posited that post-response services would increase parental resources for coping, reduce stressors, and help parents better understand the events that led to their child’s referral. However, it is possible that the basic needs services provided did not adequately integrate resources for coping and may have been “too little, too late” for families that were struggling (Beyer, 1996). It is also possible that basic needs services may provide some temporary relief for families, but that intermediate or long-term effects are limited due to the time-limited nature of prevention services under FFPSA. FFPSA limits the use of Title IV-E funds to cover only 12 months of services for children who are candidates for foster care (National Conference of State Legislatures [NCSL], 2020). Therefore, as our study period spanned more than 12 months after the initial referral, it is possible that any reduction in strain experienced by families who received basic needs services was negated once the service provision ended.

4.1. Limitations

Many of the limitations in this study involve the scope and depth of administrative data. First, although NCANDS provides a large sample, it is based on observational data; therefore, it is likely that families who received services differed from families who did not. As such, it is not possible to establish causality (Wang et al., 2015). In addition, the service utilization variables listed a service date, but this did not differentiate families that completed services from those that did not. We were not able to tell how long service were utilized for, or whether families accessed other services between the time of initial disposition date and the end of the study period. This may have impacted removals and/or other future referrals, as Tambling and Johnson (2021) found that service completion and number of sessions were both associated with reductions in child removals. Future research could use administrative data from a specific state to include dates of service utilization, completion, and to fully capture additional services that might be provided during the span of a case.

Furthermore, different types of services may have been grouped together, such as diverse types of home visiting/home-based programs. Therefore, there is a need for future research to determine which specific models of family preservation services may be evidence-based and most effective in preventing child removal (Lindell et al., 2020; Putnam-Hornstein et al., 2021). In addition, it is possible that there was variation in how states reported service utilization or parental risk factors. In parallel, some variables had large amounts of missing data. Although we used MICE to impute data, the patterns of missing data, particularly among parental risk factors and service utilization, are an important limitation to note. Finally, we were unable to identify sibling groups. Although one recent study used report identifier as a base to identify children referred as part of the same report, in that study authors used foster care data to triangulate entry dates (Drake et al., 2021). Since our study only used NCANDS data, we did not use report identifier, instead using unique child identifier, as consistent with several prior studies (Choi et al., 2021; Luken et al., 2021).

5. Conclusion

Family preservation and maintenance has been prioritized as a goal in recent policy. Findings from this study suggest that family preservation, housing services, and in-home services may reduce the likelihood of child removal, while utilization of basic needs services may increase the likelihood of child removal. As states continue to implement and align their practices with FFPSA, more research is warranted to explore specific models of preservation and maintenance services, and to continue to evaluate the effectiveness of these efforts. Future randomized studies could control for potential differences in referrals to services. Furthermore, it is important to consider primary prevention services that may reduce family strain and stress before initial child maltreatment occurs. For example, several families in our sample utilized housing, education, transportation, daycare, or employment services as part of post-response services; however, these may be more effective as part of primary prevention to increase parental resources and decrease strain, which in turn could help reduce initial child maltreatment rates. Future research is needed to determine whether timely services are truly provided to all families that need them, or just those already flagged as high-risk for foster care entry. As more focus is placed on family preservation and prevention of foster care entry, it is crucial to ensure that families have access to timely services and are able to receive support prior to child removal.