When Home is Still Unsafe: From Family Reunification to Foster Care Reentry

Abstract

Reuniting children with their families is the preferred outcome of foster care, yet many children reunited with their families reenter foster care. This study examined how parental substance abuse and mental health problems, and the time allotted for reunification, are associated with reentry risk. We used a complete cohort of children who entered the Texas foster care system in fiscal years 2008-2009 to identify the risk of foster care reentry within five years of reunification using selection-adjusted multi-level survival analysis. Approximately 16% of reunified children reentered care within 5 years. Substance abuse and mental health problems predicted higher rates of reentry. Reunification after 12 months was associated with increased reentry risk overall, but not among children commonly exempted from federal permanency timelines. Permanency guidelines that restrict the length of time to achieve reunification may have the unintended consequence of pushing reunification before maltreatment risks have been resolved.

Children enter foster care when they are unsafe in their homes and it is determined that no services or intervention short of removal from the home would ensure their immediate safety. Foster care is intended to provide a temporary living arrangement for children while their parents work to resolve the circumstances that culminated in the children’s removal. The majority of children exiting foster care are reunified with their families after a case services plan is completed (U.S. Department of Health and Human Services [USDHHS], 2017), yet, their safety is far from assured. Studies have estimated that 1 in 5 children reenter foster care within five years of reunification (Brook & McDonald, 2009; Wulczyn, 2004). Understanding why, and under what conditions, children reenter foster care is the key goal of this study.

Background

Although states have long been required to make reasonable efforts to reunify children in foster care (Adoption Assistance and Child Welfare Act of 1980), caseworkers struggle to identify and address families’ multiple and often complex needs (Dolan, Casanueva, Smith, & Ringeisen, 2012). Given difficulties in addressing families’ needs, foster care reentries may reflect unresolved, rather than newly emerging, risks (Kimberlin, Anthony, & Austin, 2009). Indeed, a large proportion of reentries occur within a year of reunification (Wulczyn, 2004). However, the extent to which children reenter care for the same reasons as the previous removal is less clear. Prior work in this area has largely focused on describing overall reentry rates and identifying family- or child-specific risk factors. Though these studies involved different geographic locations and sampling parameters, most estimated reentry rates at 20% to 40% within 1-5 years (Brook & McDonald, 2009; Lee, Jonson-Reid, & Drake, 2012; Wulczyn, 2004). A review of the reentry literature (Kimberlin et al., 2009) identified several risk factors at the child level (age; race; mental, physical, or behavioral problems), family level (poverty; parental substance abuse; lack of support; maltreatment type), and service level (number of placements; prior child protective services [CPS] involvement).

Parents’ substance abuse and mental health problems (SAMH) have steadily increased as a reason for removal: in 2016, over a third of child removals (foster care entries) involved parental substance abuse (USDHHS, 2017). Yet, SAMH alone does not constitute maltreatment; it typically must result in or create imminent risk of abuse or, more commonly, neglect of a child to culminate in removal and subsequent foster care placement (Child Welfare Information Gateway, 2012). Thus, SAMH is important to consider in the context of neglect, rather than as an independent risk factor. SAMH can be a barrier to reunification because agencies often face limited and expensive treatment options (Stensland, Watson, & Grazier, 2012), making it difficult for parents with SAMH to receive timely and appropriate treatment (Dolan et al., 2012). Even if intensive treatment is provided in the short term, services are not sustained for long (if at all) after reunification, which conflicts with long-term treatment and support needed for SAMH (Dawson, Goldstein, & Grant, 2007). There is thus a high potential for relapse, which may lead to higher risk for reentry when children are returned home after a short period of intervention.

Family risk factors are only part of the story of reentry; logistical constraints, such as efforts to limit children’s time in foster care, may inadvertently increase reentry risk. A decision to reunify must consider the policy requirements and the potential for future harm if a child were to reunify. Reunification has always been the preferred outcome of foster care and there are also pressures to reunify quickly: approximately two-thirds of reunifications occur in less than 12 months (USDHHS, 2013). The Adoption and Safe Families Act (ASFA) (Golden & Macomber, 2009) limited the time parents have to resolve the problems that lead to removal: once a child has been in care for 15 of the last 22 months, states must file for the termination of parental rights (legal severance of the parent-child relationship). States can exempt cases from this requirement at their choosing (U.S. General Accounting Office, 2002). Nevertheless, the median length of stay in foster care declined from 20 months in 1998 to 13.9 months in 2016 (USDHHS, 2006; 2017). In addition, federal Family and Child Services Reviews evaluate states on the percent of children exiting to permanency in less than 12 months, and states that underperform or fail to improve on this metric can be penalized by withholding federal funds (USDHHS, 2011; 2014).

Quick achievement of permanency is a clear priority, but understanding the role of time to reunification in the risk of reentry is difficult. We expect time to reunification to be positively associated with reentry, as families with the most serious risk factors would reasonably be expected to take longer to reunify and also be more likely to reenter due to recurring problems. Moreover, because the state has no explicit burden to prove that safety risks were resolved prior to reunification, whereas termination requires (among other things) clear and convincing evidence that a parent is unfit (Santosky v. Kramer, 1982), it is possible for there to be both a lack of grounds for termination of parental rights and ambiguity about whether reunification is appropriate. Thus, because permanency timelines require a decision to be made, it is plausible that reunification sometimes occurs before families’ presenting risks are resolved. By separately modeling groups that are commonly exempted from permanency timelines—children in kinship care and older children (U.S. General Accounting Office, 2002)—we gain insight into the likely impact of the timelines themselves. That is, when children are exempted from permanency timelines, they can be reunified if and when it is safe to do so, without regard to how long it takes parents to achieve safety; in contrast, children under the permanency timelines must be reunified within the allotted timeframe, which may result in higher-risk reunifications. If so, we would expect that time to reunification would not be associated with reentry when permanency timelines do not apply and a positive association when permanency timelines apply.

In this study, we addressed three research questions. Over and above known predictors of reentry: (1) To what extent are children reentering foster care for the same reasons as they previously entered care?; (2) How is parental SAMH associated with reentry risk as compared to neglect not involving SAMH and to no neglect?; and (3) How does time to reunification affect the risk of reentry, particularly for cases involving parental SAMH? We address these questions using a complete entry cohort of Texas foster children, followed from entry into foster care in 2008 or 2009 until 2016. Because children who are reunified with their families (“selected into” reunification) may have had fewer initial risks than children who were not reunified and remained in care, we used a selection model to create a predicted probability of reunification and then controlled for this probability in all subsequent models.

Method

Data

This study used data on all children who entered foster care in the state of Texas during the 2008 and 2009 fiscal years. The data include all foster care placements from the point of entry until May 2016. De-identified and restricted data were provided by the Texas Department of Family and Protective Services from their administrative databases. The data used for this study were a cohort of 25,777 children. We restricted the sample to children who exited foster care to reunification within 3 years of entry and were 16 years of age or younger at the time of exit (N=9,571). The latter exclusion was made to ensure that we observed children for a sufficient period prior to reaching the age of majority.

Measures

Reentry was operationalized as an entry into foster care that followed a legal discharge to reunification with a parent. The occurrence and date of a reentry was recorded by the child’s caseworker in the state’s administrative data system. Returns from trial reunifications (where a child returns home with the goal of reunification but remains under state supervision and custody) were not counted. We observed reentries that occurred before May of 2016.

Our primary explanatory variables of interest were time to reunification and reason for initial removal. Time to reunification for the initial foster care episode was calculated as the number of days between the date of removal from the home and the date of exit from the last foster placement. It was then segmented into 3 categories: less than 12 months, 12-18 months, and more than 18 months. These time segments align with Texas’s implementation of the policy guidance for permanency timelines under ASFA, as well as important federal performance indicators. ASFA stipulates that cases must have a permanency review hearing within 12 months and federal child and family services reviews evaluate states on the percentage of reunifications that occur within 12 months of entry. In Texas, a court order placing the child in foster care under state temporary managing conservatorship automatically expires after 12 months if a trial has not commenced to determine the merits of the petition that led to placement and no extension has been granted. Eighteen months is also an important decision point in Texas. Under ASFA, states are supposed to file for termination of parental rights (which negates reunification as an option) for children who have been in care for 15 of the last 22 months (with some exceptions). Under Texas Family Code, it is common practice that, after 12 months, a case is either dismissed (and the child returns home), or a 6-month extension is granted; after the 6-month extension, a case is dismissed or else proceeds to termination hearing.

Reasons for removal (both initial and, when applicable, at reentry) were recorded by the child’s caseworker at or near the date of removal from the home. The available options for removal reasons are uniform across the state. They were not mutually exclusive (caseworkers could select multiple reasons) and overlap between neglect-related reasons was high. Caregiver SAMH was rarely the only reason and usually co-occurred with physical or supervision neglect. Thus, we created a neglect variable for which 0 = no neglect, 1 = physical or supervision neglect without SAMH, and 2 = SAMH. In addition to neglect type, we included in all models non-mutually-exclusive dichotomous (0,1) indicators for physical abuse, sexual abuse, emotional abuse, and parent unavailable due to abandonment, incarceration or other reason.

General risk factors (derived from Kimberlin et al., 2009) included whether a child had a behavior problem, a mental health problem, a cognitive or learning disability, or a physical disability, along with their child protective services history (number of prior investigations, any prior removal). Demographic variables were child age at reunification, child race or ethnicity (Black, White, Hispanic, other), child sex, and number of children on the case. These variables were measured at the initial observed removal and were based on information recorded by the child’s current or previous caseworker. Lastly, we included an indicator of whether the child was living with a relative (kinship care) prior to reunification.

Analytic Approach

Our main analyses were completed in two stages. First, we constructed a selection model of the foster care exit process in order to adjust for the conditionality of reentry on reunification. Reentry is only observed for those who were reunified, and reunification is a selective process. Numerous studies indicate that reunified children are less likely to be Black, to be under 1 year of age, to have a disability or mental health problem, or to have been placed in a kinship care setting (Connell, Katz, Saunders, & Tebes, 2006; Putnam-Hornstein & Shaw, 2011). We would expect that, if processes for estimating risk of future maltreatment were entirely accurate, then children who did not reunify would be those who, if reunified, would have been most likely to reenter. Thus, if unaccounted for, the characteristics associated with selection into reunification would be mistaken as directly influencing reentry risk. The general approach for addressing this concern in analyses is a Heckman selection model. If the factors associated with selection into a specific type of foster care exit are known, the selection process can be modeled as a sort of latent construct. Thus, we built two multinomial selection models, one for children who were initially placed with kin, and one for all other children, to estimate the probability of exiting in each of four possible ways: reunification, adoption or custody to a relative, adoption or custody to a non-relative, and other. Reunification included both returns to the parents or caregivers from whom the child was initially removed as well as transfers of custody to non-custodial parents. “Other” referred to non-permanency exits (emancipation or running away). Information on exit type was drawn from a combination of three variables: discharge reason as recorded in the legal custody data, caseworker-reported reason for placement end for children’s last observed placement, and the exit type as recorded in the foster care records. Three sources of information (prioritizing exit type) were used due to missing data or non-informative responses on one or more of these variables. From the multinomial logistic regressions, we calculated the predicted probability of reunification and included it as an explanatory variable in our main models.

We modeled the reentry process using multilevel parametric survival models. Survival analysis was appropriate because our data were censored, meaning that our length of post-reunification observation varied across children and many remained at risk of reentry past our observation period. This approach allowed us to identify the risk of reentry within a given time interval (we used one month) among only those at risk during that period. Exposure to risk ended for a child once they reentered care, turned 18 years of age, or reached the end of the observation period. We defined the observation period to be within five years of reunification. Our analyses were multilevel in that we included a random intercept for the county from which the child was removed (n=204 counties). This adjustment was made because CPS is generally organized at the county level and counties have considerable variation in their practices and types of cases. We estimated the factors associated with reentry within 5 years for the overall sample, and by age at reunification and last placement setting prior to reunification.

Results

Characteristics of Children Who Reenter Foster Care

Table 1 shows the characteristics of reunified children by reentry status. Nearly 16% of children reentered care. Those who reentered were a year younger on average, had fewer siblings on their cases with them, and were more likely to have behavioral problems or cognitive disabilities than those who did not reenter. Children who reentered had more prior (before the initial removal) substantiated maltreatment reports and were more likely to have had a removal prior to the observed foster care episode. They also were less likely to have been in kinship foster care. There were no differences in reentry by gender or race. Those who reentered were more likely to have been previously removed for caregiver substance abuse or mental health problems (SAMH). Those previously removed for non-neglect reasons, particularly sexual or emotional abuse, were underrepresented in the reentry group. Of those with an initial neglect type of SAMH, 18.8% reentered, compared with 14.2% of those initially removed for neglect without SAMH and 8.7% for those initially removed for a reason other than neglect (not shown).

Table 1.

Sample Characteristics by Reentry (means or proportions)

	No reentry 8,068	Reentry 1,513	Significance of group differences
Age at reunification	6.29	5.16	***
Number of children on case	2.93	2.82	*
Male	.51	.52
Black	.16	.14
White	.35	.37
Hispanic	.42	.42
Other race	.06	.07
Behavior problem/delinquency	.11	.13	*
Cognitive or learning disability	.08	.17	***
Mental health problem	.05	.05
Physical disability	.03	.03
Number of prior CPS investigations	1.11	1.41	***
Prior removal	.05	.09	***
Last placement in kinship	.37	.30	***
Time to reunification
<12 months	.52	.46	***
12-18 months	.32	.35	*
>18 months	.16	.19	**
Initial removal neglect type
None	.19	.10	***
Neglect without SAMH	.27	.24	**
SAMH	.53	.66	***
Other initial removal reasons
Unavailable/unwilling caregiver	.29	.32
Sexual abuse	.13	.06	***
Physical abuse	.30	.29
Emotional abuse	.03	.01	**

Note: CPS = child protective services; SAMH = parent substance abuse or mental health problem.

^*p < .05,

^**p < .01,

^***p < .001.

Reasons for Reentry

The first step in our analyses was to examine reasons for reentry as a function of the reason the child previously entered foster care (Figure 1). Most reentries cited multiple reasons, with SAMH and supervision neglect co-occurring in most SAMH cases. Of those who were initially removed due to SAMH and reentered, SAMH was listed as a reason for reentry in 68% of cases, versus 36% of reentries in the neglect without SAMH group, and 29% in the no neglect group. Supervision neglect was cited in over 60% of reentries in the neglect without SAMH and SAMH groups, and in 44% of reentries in the no neglect group. An unavailable or unwilling caregiver was the most commonly cited reentry reason for the no neglect group. Physical neglect contributed to 30% of reentries in the neglect without SAMH group, a higher percentage than for the other groups. Reentries due to sexual or emotional abuse were rare across groups.

Figure 1.

Reason for Reentry by Neglect Type at Initial Removal

Note: Reasons for reentry were not mutually exclusive.

Main Analyses

Our main sets of analyses were survival models predicting the hazard of reentry. In Table 2, we show the results of a basic survival model for reentry within 5 years that included time to reunification, reason for initial removal, and a host of covariates (including the probability of reunification from the selection model). Among those with the highest probabilities of reunification (as predicted by the selection model), the risk of reentry was about 46% lower than for those with the lowest probability. We found that time to reunification was positively associated with reentry risk, such that the hazard of reentry among those reunified in 12-18 months or after 18 months was 15% and 30% higher, respectively, than among children reunified in less than 12 months. Compared with children removed for a reason other than neglect, those removed for SAMH had a 57% higher risk of reentry and those removed for neglect without SAMH had no significant difference in reentry risk.

Table 2.

Risk of Reentry Within 5 Years (N=9,571)

	HR	(SE)
Probability of reunification (from selection model)	0.560	(0.122)**
Time to reunification (reference <12 months)
Reunified in 12-18 months	1.149	(0.076)*
Reunified after 18 months	1.295	(0.107)**
Neglect type (reference: no neglect)
Neglect without SAMH	1.228	(0.139)
SAMH	1.567	(0.172)***
Other reasons for removal (not mutually exclusive)
Unavailable / unwilling caregiver	1.137	(0.070)*
Sexual abuse	0.765	(0.097)*
Physical abuse	0.978	(0.063)
Emotional abuse	0.611	(0.141)*
General risk factors
Number prior CPS investigations	1.067	(0.019)***
Any prior removal	1.446	(0.151)***
Behavior problem/delinquency	1.346	(0.130)**
Cognitive/learning disability	1.987	(0.156)***
Mental health problem	1.091	(0.155)
Physical disability	0.667	(0.114)*
Demographics
Male	0.957	(0.053)
Anglo	1.184	(0.113)
Hispanic	0.935	(0.089)
Other race	1.226	(0.156)
Age at Reunification	0.947	(0.009)***
Number children on case	1.001	(0.019)
Was reunified from kinship care	0.669	(0.045)***
ln(p) [shape parameter]	-.373	(.026)***
County variance component	.382	(.094)

Note: HR = hazard ratio; CPS = child protective services; SAMH = parent substance abuse or mental health problem. Reference group for race is black. Exponentiated coefficients (hazard ratios). Model included random intercept for county.

^*p < .05,

^**p < .01,

^***p < .001.

Those removed due to an unavailable or unwilling caregiver were also at heightened risk of reentry, whereas sexual and emotional abuse were associated with decreased reentry risk. Number of prior maltreatment investigations and any prior removal were positively associated with reentry. Risk of reentry was heightened for children with behavior problems or delinquency (HR=1.35, p<.01) and those with cognitive or learning disabilities (HR=1.99, p<.001), but not for children with mental health problems (HR=1.09, ns) or physical disabilities (HR=0.67, p<.05). Age at reunification and final placement with relatives were associated with decreased reentry risk. Gender, race, and number of children on the case were not associated with reentry.

In Table 3, we show an interaction model in which we tested whether time to reunification moderated the association between type of neglect and reentry. We show only the main coefficients of interest, but the model included all covariates shown in Table 2. Cell sizes for the groups created by the interaction terms range from 170 to 2,389 (median=973). To ease interpretation of the interaction terms, we graphed the predicted hazards and 95% confidence intervals for each subgroup created by type of neglect and time to reunification in Figure 2. We did find evidence of moderation, such that the association between SAMH and reentry was weaker for children who remained in care between 12 and 18 months as compared with children in care for less than 12 months or more than 18 months. Among children who reunified within 12 months, the hazard of reentry was twice as high among those removed for SAMH compared to those removed for no neglect, and 43% higher compared to those removed for neglect without SAMH. Yet for those reunified within 12-18 months, the predicted hazard of reentry among SAMH cases was only 39% as compared with no neglect, and was no different from neglect without SAMH. Among those who remained in care beyond 18 months, hazard of reentry was 52% and 68% higher for SAMH removals than non-neglect or neglect without SAMH removals, respectively. Put another way, among children removed for reasons other than SAMH, the risk of reentry was significantly higher for those reunified in 12-18 months as compared with less than 12 months, but there was no difference in reentry risk for those removed for SAMH. Due to a higher initial risk of reentry, SAMH cases were more likely to reenter overall, but the difference between SAMH and other cases was largest for those reunified in less than 12 months.

Table 3.

Moderation of the Association between Neglect Type and Risk of Reentry within 5 years (N=9,571)

	Reentry within 5 years
	HR	(SE)
Probability of reunification (from selection model)	0.564	(0.122)**
Time to reunification (reference <12 months)
Reunified in 12-18 months	1.808	(0.369)**
Reunified after 18 months	1.638	(0.442)
Neglect type (reference: no neglect)
Physical or supervision neglect without SAMH	1.387	(0.227)*
SAMH	2.007	(0.309)***
Interaction terms
Reunified in 12-18months # Physical or supervision neglect without SAMH	0.786	(0.189)
Reunified in 12-18 months # SAMH	0.555	(0.121)**
Reunified after 18 months # Physical or supervision neglect without SAMH	0.789	(0.242)
Reunified after 18 months # SAMH	0.763	(0.216)
ln(p) [shape parameter]	-.373	(.026)***
County variance component	.378	(.094)

Note: SAMH = substance abuse or mental health. Reference group for race is black. Exponentiated coefficients (hazard ratios). Model included random intercept for county. Model included all covariates shown in Table 2.

^*p < .05,

^**p < .01,

^***p < .001.

Figure 2.

Predicted Hazards of Reentry by Neglect Type and Time to Reunification

In Table 4, we show separate models by last placement prior to reunification (kin versus non-kin) and age group (0-5, 6-10, 11-16). Cell size ranges for the subgroups created by the interaction terms were: kin (40-1,067, median=275); non-kin (130-1,322, median=677); age 0-5 (60-1,400, median=465); age 6-10 (45-618, median=295); age 11-16 (65-371, median=185). In all cases, the smallest cells were for reunifications after 18 months among children removed for reasons other than neglect. Predicted hazards of reentry among those last placed in kin and non-kin foster care by neglect type and time to reunification are shown in Figure 3.

Table 4.

Models for reentry within 5 years by last placement setting and age group

	Non-Kinship Care		Kinship Care		Reunified at Age 0-5		Reunified at Age 6-10		Reunified at Age 11-16
	(n=6,102; 946 reentries)		(n=3,469; 405 reentries)		(n=5,026; 813 reentries)		(n=2,633; 331 reentries)		(n=1,912; 207 reentries)
	HR	(SE)	HR	(SE)	HR	(SE)	HR	(SE)	HR	(SE)
Time to reunification (reference <12 months)
Reunified in 12-18 months	2.029	(0.460)**	1.178	(0.581)	2.348	(0.731)**	4.182	(1.901)**	0.920	(0.323)
Reunified after 18 months	1.570	(0.456)	1.537	(1.195)	2.212	(0.980)	3.498	(1.937)*	0.814	(0.361)
Neglect type (reference: no neglect)
Physical or supervision neglect without SAMH	1.135	(0.215)	2.179	(0.763)*	1.992	(0.480)**	2.290	(0.928)*	1.044	(0.301)
SAMH	1.855	(0.328)***	2.101	(0.714)*	2.840	(0.649)***	4.148	(1.603)***	1.054	(0.305)
Interactions
Reunified in 12-18 months # Physical or supervision neglect without SAMH	0.847	(0.230)	0.678	(0.380)	0.556	(0.198)	0.411	(0.214)	1.494	(0.688)
Reunified in 12-18 months # SAMH	0.434	(0.107)***	1.109	(0.564)	0.487	(0.158)*	0.197	(0.095)***	0.896	(0.383)
Reunified after 18 months # Physical or supervision neglect without SAMH	0.873	(0.293)	0.820	(0.699)	0.659	(0.320)	0.456	(0.280)	0.636	(0.382)
Reunified after 18 months # SAMH	0.759	(0.233)	0.883	(0.703)	0.671	(0.305)	0.314	(0.182)*	1.104	(0.569)
ln(p) [shape parameter]	-.381	(.031)***	-.329	(.048)***	-.332	(.034)***	-.391	(.054)***	-.434	(.067)***
County variance component	.310	(.096)	.687	(.206)	.243	(.081)	.110	(.091)	.402	(.203)

Note: SAMH = substance abuse or mental health. Exponentiated coefficients (hazard ratios). Model included random intercept for county. Model included all covariates shown in Table 2.

^*p < .05,

^**p < .01,

^***p < .001.

Figure 3.

Predicted Hazards of Reentry by Last Placement Setting, Neglect Type, and Time to Reunification

We found that among children not placed with kin who were removed for non-neglect or neglect without SAMH reasons, reunification in 12-18 months was associated with increased risk of reentry as compared with reunification in less than 12 months. This was, however, especially true for the no neglect and neglect without SAMH groups. Children removed for SAMH were more likely to reenter if reunified after 18 months (versus either 12-18 or less than 12 months), whereas there the opposite trend was observed for the no neglect and neglect with SAMH children. In contrast, among children last placed with kin, there were no consistent associations between time to reunification and reentry. The results of the age group models were generally similar to our main models, with one exception. Among children removed for SAMH who left care between ages 6 and 10, reentry risk was lower for those reunified in 12-18 months versus less than 12 months. For children who were reunified between 11 and 16 years old, there were no direct or moderated associations of time to reunification or neglect type with reentry.

Conclusion

This study examined the associations of parent SAMH and time to reunification with foster care reentry among a statewide foster care entry cohort from Texas. We found that over 15% of all Texas children who were reunified before age 16 reentered foster care within 5 years. Though concerning, this rate is lower than the 20-30% reentry rates reported in other large studies across multiple states (Brook & McDonald, 2009; Wulczyn, 2004).

We wish to highlight two important findings of our study. First, our findings suggest that the timelines for permanency are an important part of the reentry story. As noted previously, we would expect that families with more serious or entrenched problems would both take longer to rectify those problems and be less likely to sustain their progress. The implicit assumption behind ASFA’s permanency timelines is that, if there has not been substantial progress toward reunification within a year, there is unlikely to be substantial progress in the near future. Hence, time to reunification is likely to generally correlate with higher rates of reentry, which our results do show. This does not itself prove anything about the effects of the timelines themselves. Yet, we also found that time to reunification was not consistently associated with reentry risk among the populations for whom the permanency timelines are often waived, namely children in kinship care and older children. That the risk of reentry is higher among those to whom the timelines are most often applied, and that time to reunification is associated with reentry primarily among those groups, suggests that forcing agencies and courts to either terminate parental rights or send children home may have the unintended result of some children being returned to unsafe homes. This is not to suggest that more time would improve reunification outcomes. Rather, based on the generally null associations of time to reunification and reentry for older children and children in kinship care, where timelines are less often applied, there appears to be no inherent harm or benefit to additional time, at least for reentry risk. However, our findings should not be interpreted to promote termination of parental rights in all cases that cannot be resolved within a year, however. With more than 100,000 children in foster care each year waiting for an adoptive family (USDHHS, 2016) and many prospective adoptive families reluctant to consider foster children (Zhang & Lee, 2011), terminating parental rights to children who then are unable achieve permanency through guardianship or adoption may be unproductive.

Second, our study provides new information of the role of SAMH in reentry. Although prior work (Kimberlin et al., 2009) has highlighted the high reentry rate among children who previously entered due to SAMH, we found that substance abuse and supervision neglect (which often co-occur) are the primary drivers of foster care reentry, even among children who initially entered foster care for other reasons. Given that caregiver SAMH accounts for 1 in 3 foster care entries and that nearly 3 of every 4 foster care entries are due to neglect (USDHHS, 2016), it is unsurprising that SAMH would be a strong driver of reentries overall. However, we show that this was attributable both to recurring and emerging SAMH problems. In addition, our multilevel survival models showed that, reentry rates were highest among SAMH cases where reunification occurred after 18 months. Studies of substance abuse relapse have shown that relapse rates remain high following treatment and the first two years are when most recovering addicts return to using (Walitzer & Dearing, 2006). Predicting who will relapse is difficult, and courts are likely reluctant to keep children in foster care when parents have demonstrated several months of sobriety. Thus, post-reunification services may bridge the gap between sending children home when there is a high risk of relapse and keeping children in care indefinitely. Post-reunification services may help agencies identify substance relapse early on and provide treatment services before the family reaches the point of reentry. In addition, the development and implementation of predictive risk assessments for reentry in this population could be helpful in addressing this problem.

A primary limitation of this study was that we did not have access to all variables with potentially importance to the reunification or reentry process. For example, we could not account for the types of services families received pre- or post-reunification or parental cooperation. Despite our use of a selection model to adjust for differential probability of reunification, it is possible that unmeasured attributes resulted in biased estimates. In addition, despite a relatively long follow-up period as compared with past studies, we were not able to examine risk of reentry throughout childhood. We found that the risk of reentry declined over time, but it is possible that reentry risk is nonlinear throughout childhood. For example, among children who were in care very young, reentry risk could increase again in late adolescence as new behavioral challenges emerge. Lastly, we note that, because these data were collected for administrative, rather than research purposes, some of the measures are reductive and it is not possible to affirm their reliability or validity. Despite these limitations, our study had several strengths, including a large and diverse statewide sample, multiple years of observation, and a rigorous modeling strategy.

In sum, foster care reentry is a common problem among reunified families, particularly those where substance abuse and mental health problems are present. Currently, each state’s CPS system has a strong policy-driven and well-intentioned emphasis on quickly returning children home from foster care; however, children are too often being reunified only to reenter foster care at a later date. We have found that children in foster care whose families have substance abuse problems are at high risk of reentry no matter how long they remain in foster care prior to reunification. CPS systems across the states must improve their ability to detect risk of reentry in this population. The advancement of administrative data and predictive analytics are key resources in making that progress. It is also crucial that parents affected by SAMH be offered treatment and counseling services as early and frequently as possible to increase the chance and success of reunification with their children.