M. D. ex rel. Stukenberg v. Abbott (907 F.3d 237, 2018)

EDITH BROWN CLEMENT, Circuit Judge:

Plaintiffs, a certified class of minor children in the Permanent Managing Conservatorship of the Department of Family Protective Services ("DFPS") in Texas, filed suit under 42 U.S.C. § 1983 seeking injunctive relief against the Governor of Texas, the Executive Commissioner of the Texas Health and Human Services Commission, and the Commissioner of DFPS (collectively "the State"). They allege that the State's maintenance of its foster care system exposes them to a serious risk of abuse, neglect, and harm to their physical and psychological well-being. The district court held that the State's policies and practices violated plaintiffs' constitutional right to be free from an unreasonable risk of harm, and granted plaintiffs a permanent injunction requiring sweeping changes to Texas's foster care system. The State appeals both the liability determination and the injunctive order. For the reasons stated below, we AFFIRM in part, REVERSE in part, VACATE, and REMAND for modification of the injunction.

I. Facts and Proceedings

The Texas Department of Family and Protective Services is responsible for roughly 29,000 children. When DFPS's Child Protective Services ("CPS") division determines that it is not safe for a child to remain with his legal guardian as a result of abuse and/or neglect, CPS petitions the court to remove the child to the Temporary Management Conservatorship ("TMC"). TMC is intended to be a nonpermanent custody arrangement. CPS places the TMC child with a relative or a certified caregiver while CPS attempts to reunify the child with his legal guardian, permanently place him with a relative, or arrange for him to be adopted. There are approximately 17,000 children in TMC, which lasts for one year unless the court extends it by six months. If CPS cannot achieve permanency for the child at the end of the TMC period, the child enters the Permanent Managing Conservatorship ("PMC"). There are approximately 12,000 children in PMC.

Many things change when a child is moved to PMC. As a general matter, PMC children get less attention from their caseworkers and other advocates than do TMC children. For example, according to requirements set by the state legislature, PMC children are entitled to fewer permanency review hearings, planning meetings, and status hearings per year. TMC children receive four service plan reviews in their first year, but PMC children receive only two reviews per year. Unlike TMC children, PMC children are not entitled to an attorney ad litem , and they are far less likely to have Court Appointed Special Advocate ("CASA") volunteers. As one state court-commissioned report put it, "[t]hough the State's responsibility for the child's life and well-being does not change-and arguably increases-the attention paid to the child's cases diminishes drastically." There is a sense among CPS staff that when a child transitions into PMC, "the clock stops ticking."

Children receive one of four "service level" designations upon entering state custody-Basic, Moderate, Specialized, or Intense-based on their physical and psychological needs. Placements must be licensed to care for children at specific service levels. DFPS has access to a variety of placement settings, though it directly manages only about 10% of them. The remaining 90% are managed by private child-placing agencies ("CPAs") contracting with the State. Relevant placement setting options, listed from least to most restrictive, include: 1) foster family homes that contain 1 to 6 children; 2) foster group homes that contain 7 to 12 children ("FGHs"); 3) general residential operations that contain 13 or more children ("GROs"); and 4) residential treatment centers ("RTCs"), which provide therapeutic treatment for children with more severe emotional or mental-health issues.

It is DFPS policy to find the most appropriate placement for foster children and to try to keep children in their home counties. Policy also specifies that children should be placed with their siblings whenever possible and in family-like settings as opposed to group homes if it is feasible and in the child's best interest. Because of practical limitations on placement availability, roughly 40% of children are placed "out of region." Approximately 64.7% of sibling groups are placed together. Just under 14% of PMC children under 12 are placed in FGHs, GROs, or RCLs. DFPS does not have a policy against mixing children of different ages, sexes, and service levels in FGHs, though girls and boys may not share the same bedroom.

Primary conservatorship caseworkers ("CVS caseworkers" or "caseworkers") are a foster child's most important point of contact within DFPS, and they are critical to the provision of safety for foster children. CVS caseworkers are responsible for, among other things, assessing children's placement needs, finding appropriate placement, monitoring the children to make sure they are safe, ensuring that they receive needed services, developing and implementing permanency plans, attending court hearings and plan meetings, updating the children's medical records, and conducting monthly face-to-face visits with the children and their foster families. Given caseworkers' sweeping responsibilities, the Child Welfare League of America ("CWLA") recommends that they carry a caseload that includes no more than 12 to 15 children. DFPS does not place any limits on the number of cases CVS caseworkers can carry. As of June 2014, nearly half of CVS caseworkers carried caseloads of 21 children or more, 22% carried caseloads of 26 or more, and nearly 10% carried caseloads of 31 children or more. Caseworkers report that they are overworked, and DFPS experiences a high rate of caseworker turnover.

Because placement availability is limited, it is often impracticable for caseworkers to make their monthly face-to-face visits with their children. DFPS often uses secondary workers to fill this gap. In some cases, when caseworkers are too busy or too far away, secondary "I See You" ("ISY") workers will take on the home visits. Caseworkers then rely on ISY workers' notes in case planning. ISY workers typically carry a large caseload, and their responsibilities are significantly more limited than are those of primary caseworkers. They are not required to follow up on a child's needs, and they are not involved in any aspect of a child's permanency plan outside of providing relevant information to the child's primary caseworker. ISY workers' primary responsibility is to see the child and confirm that the child "is still there." Testimony at trial strongly suggests that ISY visits are perfunctory and that the information they generate from the foster child is often superficial and unhelpful. Children do not feel comfortable sharing their problems with their revolving roster of ISY workers, who often fail to meet with them in private as required by DFPS policy.

With respect to recordkeeping, DFPS's methods are shockingly haphazard and inefficient. A significant portion of children's records are kept in DFPS's electronic IMPACT casework system. Data on abuse and neglect investigations are maintained by the Residential Child Care Licensing ("RCCL") division in its CLASS database. Caseworkers have access to CLASS, but the data is not merged with IMPACT files. RCCL allows CPAs to keep their own records. Medical records and related information is accessible via the STAR Health Passport, which is not synced with IMPACT, though IMPACT is supposed to include children's comprehensive medical information. Neither IMPACT nor STAR can "store" many requisite documents electronically, so documents such as medical assessments and birth certificates are maintained in paper files. Some children's files are maintained entirely in paper form, and casefiles are often inordinately long.

The task of inspecting, investigating, and licensing placements is managed by RCCL. The Performance Management Unit ("PMU") is responsible for internal quality control for all of DFPS, including RCCL. RCCL investigates any reports of neglect and abuse. When RCCL investigates an allegation, it ascribes one of four outcomes upon completion: 1) Reason to Believe ("RTB"); 2) Ruled Out ("RO"); 3) Unable to Determine ("UTD") ; or 4) Administrative Closure. Two PMU studies of a random sample of UTD dispositions revealed a high rate of disposition errors.

RCCL investigates incidents of child-on-child abuse, but does not formally track or aggregate those statistics; rather, it labels child-on-child incident investigations "negligent supervision" cases. The only place RCCL records a child's history of abusing other children is in the perpetrating child's individual casefile. This means that this information is not easily accessible to caseworkers when they are evaluating whether a placement is appropriate for one of their children. It also means that incidents of child-on-child abuse are not included in the abuse rate/data DFPS provides to the federal government or the data that was provided to the district court.

Children "age out" of foster care when they reach the age of 18. Roughly 1,300-1,400 foster children "age-out" of the foster system annually. Of these, approximately 25-30% go back into extended foster care. Another small percentage with intellectual and developmental disabilities go into the guardianship of a separate program not maintained by DFPS. The rest, presumably, though it's not clear from the record, find a permanent living arrangement, make use of shelters and other non-profit programs for youths aging-out of foster care, or end up homeless. DFPS offers independent living classes to foster children over the age of 16, though DFPS apparently does not know what percentage of children actually utilize the program.

Plaintiffs, minor children in the PMC, filed suit through next friends in March 2011, alleging that the State violated their substantive rights under the Due Process Clause of the Fourteenth Amendment. They sought injunctive relief against the Governor of Texas, the Executive Commissioner of the Texas Health and Human Services Commission, and the Commissioner of DFPS. The district court granted their motion for class certification under Federal Rule of Civil Procedure 23. Following the Supreme Court's opinion in Walmart v. Dukes , 564 U.S. 338, 131 S.Ct. 2541, 180 L.Ed.2d 374 (2011), this court vacated and remanded the certification. M.D. ex rel. Stukenberg v. Perry (M.D. I ) , 675 F.3d 832 (5th Cir. 2012). After a three-day hearing, the district court concluded that the requirements of Rule 23(a) had been met, and certified a general class-all children now, or in the future, in the PMC in Texas-and three subclasses: 1) Licensed Foster Care Subclass ("LFC") ; 2) FGH subclass; and 3) Basic care GRO subclass . The State's interlocutory appeal of the certification order was dismissed as untimely. M.D. ex rel. Stukenberg v. Perry , 547 F. App'x 543 (5th Cir. 2013).

Following a two-week bench trial, the conscientious district court issued its liability opinion in December 2015. The district court ultimately found that DFPS's policies and practices with respect to caseloads, monitoring and oversight, placement array, and foster group homes violated plaintiffs' substantive due process rights. It ordered the State to "establish and implement policies and procedures to ensure ... PMC foster children are free from an unreasonable risk of harm" and enjoined DFPS from placing children in FGHs that lack 24-hour awake-night supervision.

Over the course of the trial, the court heard from twenty-eight fact witnesses: 1) six next friends and attorneys ad litem , who have roughly 80 years of experience in the child welfare system between them and who have dealt extensively with PMC children and CVS caseworkers; 2) five former PMC foster children; 3) a non-profit leader who runs a shelter and transitional living program for youths aging-out of foster care in Texas and who has personally fostered 65 children over the last 16 years; 4) two former CVS caseworkers; and 5) 14 current DFPS officers . The district court also heard testimony from twelve expert witnesses proffered by the plaintiffs and the State. It gave various weight to the experts' testimony according to the district court's credibility determinations. Ultimately, the district court disregarded the testimony of two proffered experts in their entirety as unreliable.

The court considered several reports detailing both independent and internal reviews of the Texas foster care system, including multiple reports that were commissioned by DFPS itself. The reports date back to 2004, and several of them reference earlier agency reviews and internal audits, most notably a report authored by the Governor's Committee to Promote Adoption ("GCPA") in 1996. Additionally, the district court considered national child welfare standards provided by the CWLA and the Council on Accreditation, and Child and Family Service Reviews ("CFSR") performed by the United States Department of Health and Human Services.

This court denied the State a stay pending appeal. The district court appointed Special Masters to address specific constitutional shortcomings at DFPS, and this court denied defendants' petition for writ of mandamus requesting the court vacate the appointment. The Special Masters studied DFPS and its policies for nearly two years and submitted a final list of findings and recommendations to the district court. The district court entered a final order granting plaintiffs a permanent injunction in January 2018 and appointed a Special Monitor.

The State appealed, and this court granted an administrative stay of the injunction, which was converted to a stay pending appeal by our panel on March 21. The State raises three primary objections to the district court's liability determination and the injunctive order: 1) the district court erred in concluding that DFPS policies affecting the PMC class violate plaintiffs' substantive due process right and, as such, plaintiffs are not entitled to injunctive relief; 2) the district court abused its discretion in certifying the general class and various subclasses; and 3) the scope of the district court's injunction is improper.

II. Standard of Review

Following a bench trial, this court reviews the district court's conclusions of law de novo and its factual findings for clear error. Cerda v. 2004-EQR1 L.L.C. , 612 F.3d 781, 786 (5th Cir. 2010). "The predicate findings of a substantial risk of serious harm and officials' deliberate indifference to the risk are factual findings reviewed for clear error." Ball v. LeBlanc , 792 F.3d 584, 592 (5th Cir. 2015) ; Gates v. Cook , 376 F.3d 323, 333 (5th Cir. 2004) ; cf. Farmer v. Brennan , 511 U.S. 825, 842, 114 S.Ct. 1970, 128 L.Ed.2d 811 (1994). Such findings are erroneous only if "[they are] without substantial evidence to support [them], the court misinterpreted the effect of the evidence, or this court is convinced that the findings are against the preponderance of credible testimony." Ball , 792 F.3d at 592 (quoting Petrohawk Props., L.P. v. Chesapeake La., L.P. , 689 F.3d 380, 388 (5th Cir. 2012) ). Whether the facts as found establish a violation of the Due Process Clause is a "legal conclusion based on factual inferences" subject to de novo review. See Dalheim v. KDFW-TV , 918 F.2d 1220, 1226 (5th Cir. 1990).

The district court's decision to certify a class is reviewed for abuse of discretion. Yates v. Collier , 868 F.3d 354, 359 (5th Cir. 2017). This court recognizes "the essentially factual basis of the certification inquiry and [ ] the district court's inherent power to manage and control pending litigation." M.D. I , 675 F.3d at 836. "Nonetheless, this broad discretion must operate 'within the framework of Rule 23,' and we 'review de novo whether the district court applied the correct legal standards.' " Yates , 868 F.3d at 360 (quoting M.D. I , 675 F.3d at 836 ).

This court reviews the district court's grant of a permanent injunction for abuse of discretion. Ball , 792 F.3d at 598. The district court abuses its discretion if it "(1) relies on clearly erroneous factual findings when deciding to grant or deny the permanent injunction[,] (2) relies on erroneous conclusions of law when deciding to grant or deny the permanent injunction, or (3) misapplies the factual or legal conclusions when fashioning its injunctive relief." Symetra Life Ins. Co. v. Rapid Settlements, Ltd. , 775 F.3d 242, 254 (5th Cir. 2014) (internal quotation omitted).

III. Governing Law

We begin with an overview of the legal framework for evaluating plaintiffs' claims. In order to state a claim for a substantive due process violation under § 1983, the plaintiffs must demonstrate: 1) they were deprived of a cognizable constitutional right, see Rios v. City of Del Rio , 444 F.3d 417, 425 (5th Cir. 2006) ; 2) the State acted with "deliberate indifference" to the protected right, see Hernandez v. Tex. Dep't of Protective & Regulatory Servs. , 380 F.3d 872, 880 (5th Cir. 2004) ; and 3) the policies or practices complained of were the direct cause of the constitutional deprivation, see Piotrowski v. City of Houston , 237 F.3d 567, 578 (5th Cir. 2001).

1. The Substantive Due Process Right

As a general matter, the State is under no affirmative obligation to protect its citizens from private harm.

DeShaney v. Winnebago Cty. Dep't of Soc. Servs. , 489 U.S. 189, 197, 109 S.Ct. 998, 103 L.Ed.2d 249 (1989). The Supreme Court has recognized, however, that when the State assumes certain custodial roles with respect to an individual, it creates a "special relationship" that imparts to the State a limited duty to provide for that person's safety and general well-being. See id . As the Court explained:

The rationale for [these protections] is simple enough: when the state by the affirmative exercise of its power so restrains an individual's liberty that it renders him unable to care for himself, and at the same time fails to provide for his basic human needs-e.g. , food, clothing, shelter, medical care, and reasonable safety-it transgresses the substantive limits on state action set by the Eighth Amendment and the Due Process Clause.

Id . at 200, 109 S.Ct. 998 (citing Estelle v. Gamble , 429 U.S. 97, 103-04, 97 S.Ct. 285, 50 L.Ed.2d 251 (1976) ; Youngberg v. Romeo , 457 U.S. 307, 315-16, 102 S.Ct. 2452, 73 L.Ed.2d 28 (1982) ). The Court has found that a special relationship exists between the State and prisoners, Gamble , 429 U.S. at 103-04, 97 S.Ct. 285, involuntarily-committed mental patients, Youngberg , 457 U.S. at 315-16, 102 S.Ct. 2452, and suspected criminals injured in the course of being apprehended by police, City of Revere v. Mass. Gen. Hosp. , 463 U.S. 239, 244, 103 S.Ct. 2979, 77 L.Ed.2d 605 (1983).

In DeShaney, the Supreme Court found that no "special relationship" existed between the State and a child who had been placed in the temporary custody of a local hospital by an emergency court order. See DeShaney , 489 U.S. at 192-203, 109 S.Ct. 998. But, in holding that the State had no affirmative duty to intervene on the child's behalf, the Court highlighted that "the harms [the child] suffered occurred not while he was in the State's custody, but while he was in the custody of his natural father, who was in no sense a state actor." Id. at 201, 109 S.Ct. 998. It qualified that, "[h]ad the State by the affirmative exercise of its power removed [the child] from free society and placed him in a foster home operated by its agents, we might have a situation sufficiently analogous to incarceration or institutionalization to give rise to an affirmative duty to protect." Id . at 201 n.9, 109 S.Ct. 998.

Following DeShaney , the Fifth Circuit held that a "special relationship" exists between the State and children when the State "remove[s] them from their natural home and place[s] them under state supervision." Griffith v. Johnston , 899 F.2d 1427, 1439 (5th Cir. 1990). Having taken custody of a child, the State "assume[s] the responsibility to provide for constitutionally adequate care." Id . ; see also Doe ex rel. Magee v. Covington Cty. Sch. Dist. ex rel. Keys , 675 F.3d 849, 856 (5th Cir. 2012) ; Hernandez , 380 F.3d at 880.

This court has defined the substantive due process right enjoyed by children in the custody of the State's foster care system as a right to "personal security and reasonably safe living conditions." Hernandez , 380 F.3d at 880. Though the precise contours of "personal security" and "reasonably safe living conditions" have yet to be fleshed-out at length, it is clear that foster children are, at minimum, entitled to protection from physical abuse and violations of bodily integrity. See, e.g. , id. at 880-81. This court has not, however, required the State to guarantee the individual's betterment or unconditional stability.

But there is a significant amount of daylight between physical abuse and maximum personal psychological development, optimal treatment, or the most appropriate care. The district court held that the substantive right encompasses a right to protection from psychological abuse. We agree that plaintiffs' substantive right to "personal security and reasonably safe living conditions" includes the very limited right to be free from severe psychological abuse and emotional trauma-both of which are often inextricably related to some form of physical mistreatment or deprivation. See DeShaney , 489 U.S. at 200, 109 S.Ct. 998 ; Youngberg , 457 U.S. at 317, 102 S.Ct. 2452. Indeed, DeShaney notes that, in a custodial setting, the State assumes at least "some responsibility" for both an individual's "safety" and his "general well-being." Deshaney , 489 U.S. at 199-200, 109 S.Ct. 998. Other courts have recognized that the State's responsibility to protect foster children's "general well-being" requires it "to take steps to prevent children in state institutions from deteriorating physically or psychologically." See, e.g. , K.H. ex rel. Murphy v. Morgan , 914 F.2d 846, 851 (7th Cir. 1990).

We stress, however, that there are significant limits on the scope of the right to be free from certain forms of psychological harm. The Fourteenth Amendment does not entitle plaintiffs to receive optimal treatment and services, nor does it afford them the right to be free from any and all psychological harm at the hands of the State. See, e.g. , Griffith , 899 F.2d at 1439 ; Drummond , 563 F.2d at 1208-09 ; Feagley , 868 F.2d at 1441. Many inherent features of the foster care system, such as the ambulatory nature of children's placements, have negative psychological consequences. Such negative consequences are regrettable, but they are not the type of significant, abuse-related psychological damage the Constitution prohibits. In sum, egregious intrusions on a child's emotional well-being-such as, for example, persistent threats of bodily harm or aggressive verbal bullying-are constitutionally cognizable. Incidental psychological injury that is the natural, if unfortunate, consequence of being a ward of the state does not rise to the level of a substantive due process violation.

2. Deliberate Indifference

In order to hold the State liable for violating plaintiffs' substantive due process rights, plaintiffs are required to demonstrate that the State's conduct "shocks the conscience." County of Sacramento v. Lewis , 523 U.S. 833, 846, 118 S.Ct. 1708, 140 L.Ed.2d 1043 (1998). The Supreme Court has explained that the Due Process Clause protects individual citizens from "arbitrary action of government," and that under the "shocks the conscience" standard, "only the most egregious official conduct can be said to be arbitrary in the constitutional sense." Id . at 845-46, 118 S.Ct. 1708 (internal quotations omitted). "While the measure of what is conscience shocking is no calibrated yard stick," id . at 847, 118 S.Ct. 1708, it is not enough that the conduct "offend[s] some fastidious squeamishness or private sentimentalism." Rochin v. California , 342 U.S. 165, 172, 72 S.Ct. 205, 96 L.Ed. 183 (1952). Most recently, the Court reiterated that the " 'shocks the conscience' standard is satisfied where the conduct was 'intended to injure in some way unjustifiable by any government interest,' or in some circumstances if it resulted from deliberate indifference." Rosales-Mireles v. United States , --- U.S. ----, 138 S.Ct. 1897, 1906, 201 L.Ed.2d 376 (2018) (quoting Lewis , 523 U.S. at 849-50, 118 S.Ct. 1708 ). Furthermore, "liability for negligently inflicted harm is categorically beneath the threshold of constitutional due process." Lewis , 523 U.S. at 849, 118 S.Ct. 1708.

"Consistent with [these] principles," this court has required plaintiffs to show that the State "at a minimum acted with deliberate indifference toward the plaintiff." Hernandez , 380 F.3d at 880 (quoting McClendon v. City of Columbia , 305 F.3d 314, 326 (5th Cir. 2002) ). Demonstrating that the State acted with deliberate indifference is "a significantly high burden for plaintiffs to overcome." Id . at 882 (citing Doe v. Dall. Indep. Sch. Dist. , 153 F.3d 211, 218 (5th Cir. 1998) ). "To act with deliberate indifference, a state actor must consciously disregard a known and excessive risk to the victim's health and safety." Id . at 880 (citing Farmer , 511 U.S. at 837, 114 S.Ct. 1970 ). Stated differently, "the [State] must be both aware of facts from which the inference could be drawn that a substantial risk of serious harm exists, and [it] must also draw that inference." Id . at 881 (quoting Farmer , 511 U.S. at 837, 114 S.Ct. 1970 ). This is "a degree of culpability beyond mere negligence or even gross negligence; it 'must amount to an intentional choice, not merely an unintentionally negligent oversight.' " James v. Harris Cty ., 577 F.3d 612, 617-18 (5th Cir. 2009) (quoting Rhyne v. Henderson Cty. , 973 F.2d 386, 392 (5th Cir. 1992) ). Moreover, the State is not deliberately indifferent to a substantial risk of serious harm if, aware of the risk, it "respond[s] reasonably ... even if the harm ultimately was not averted." Farmer , 511 U.S. at 844, 114 S.Ct. 1970.

Though deliberate indifference is a subjective standard of recklessness focusing on what the State actually knew, rather than what it should have known, McClendon , 305 F.3d at 326, "this court has never required state officials to be warned of a specific danger." Hernandez , 380 F.3d at 881 (emphasis added). "[R]ather, it is enough that the [State] acted or failed to act despite [its] knowledge of a substantial risk of serious harm"-the plaintiffs need not show that the State anticipated the exact form the harm would take. Id . This court in Hernandez explained: "as a state official may not escape deliberate indifference liability by arguing that the risk of harm arises from a source not contemplated, a defendant also cannot avoid such liability by contending that the particular method of harm, i.e. how the abuse was carried out, was not envisioned." Id . at 882. Accordingly, to overcome the culpability standard, the plaintiffs were required to demonstrate only that the State "knew of the underlying facts indicating a sufficiently substantial danger and that [it] did not believe that the risks to which the facts gave rise [were] insubstantial or nonexistent." Id . (citing Rosa H v. San Elizario Indep. Sch. Dist. , 106 F.3d 648, 659 (5th Cir. 1997) ).

The Supreme Court has also explained that the deliberately indifferent state of mind can be inferred "from the fact that the risk of harm is obvious." Hope v. Pelzer , 536 U.S. 730, 737, 122 S.Ct. 2508, 153 L.Ed.2d 666 (2002) ; see also Farmer , 511 U.S. at 842, 114 S.Ct. 1970. In other words, if the risk of severe abuse is sufficiently apparent, a court is entitled to find that the State was deliberately indifferent. See Hernandez , 380 F.3d at 881. Furthermore, plaintiffs may be protected against future harm not yet realized through a prospective injunction. See Helling v. McKinney , 509 U.S. 25, 33, 113 S.Ct. 2475, 125 L.Ed.2d 22 (1993) (stating that "[i]t would be odd to deny an injunction to inmates who plainly proved an unsafe, life-threatening condition in their prison on the ground that nothing yet had happened to them[,]" and "a remedy for unsafe conditions need not await a tragic event"); see also Hoptowit v. Spellman , 753 F.2d 779, 783-84 (9th Cir. 1985) ( Hoptowit II ); Gates v. Collier , 501 F.2d 1291, 1304 (5th Cir. 1974).

3. Causation

In addition to establishing that they were deprived of a constitutional right and that the State acted with the requisite level of culpability, plaintiffs must show that the State is the " 'moving force' behind the deprivation." Kentucky v. Graham , 473 U.S. 159, 166, 105 S.Ct. 3099, 87 L.Ed.2d 114 (1985) (quoting Monell v. N.Y.C. Dep't of Soc. Servs. , 436 U.S. 658, 694, 98 S.Ct. 2018, 56 L.Ed.2d 611 (1978) ); see also Piotrowski , 237 F.3d at 578. "[T]hus[ ] ... the entity's 'policy or custom' must have played a part in the violation of federal law." Graham , 473 U.S. at 166, 105 S.Ct. 3099. This court has cautioned that culpability and causation requirements "must not be diluted, for '[w]here a court fails to adhere to rigorous requirements of culpability and causation, [state entity] liability collapses into respondeat superior liability.' " Snyder v. Trepagnier , 142 F.3d 791, 796 (5th Cir. 1998) (quoting Bd. of Cty. Comm'rs of Bryan Cty. v. Brown , 520 U.S. 397, 415, 117 S.Ct. 1382, 137 L.Ed.2d 626 (1997) ). Causation is, however, an "intensely" fact-bound inquiry. Morris v. Dearborne , 181 F.3d 657, 673 (5th Cir. 1999). "Because the district court is better positioned [...] to decide the issue, our review of the ... cause determination is deferential." Brown v. Plata , 563 U.S. 493, 517, 131 S.Ct. 1910, 179 L.Ed.2d 969 (2011) (internal quotations omitted).

Establishing a "direct causal link between the [State] policy and the constitutional deprivation" is a "high threshold of proof." Piotrowski , 237 F.3d at 580. "This connection must be more than a mere 'but for' coupling between cause and effect." Fraire v. City of Arlington , 957 F.2d 1268, 1281 (5th Cir. 1992) (citing City of Canton v. Harris , 489 U.S. 378, 388-89, 109 S.Ct. 1197, 103 L.Ed.2d 412 (1989) ). "It follows that each and any policy [or practice] which allegedly caused constitutional violations must be specifically identified by a plaintiff, and it must be determined whether each one is facially constitutional or unconstitutional." Piotrowski , 237 F.3d at 579-80. Concerned with the distinction between an offending policy and isolated instances of violative conduct by individual bad actors, the court in Piotrowski emphasized the need to "disaggregate[ ]" the policies or customs causing the alleged constitutional deprivation. See id . at 581. Neither Piotrowski nor this court's related precedent regarding the § 1983 causation requirement, however, suggests a plaintiff is required to demonstrate that a challenged policy or practice is the exclusive cause of the constitutional deprivation. See id . ; Fraire , 957 F.2d at 1281 ; see also Graham , 473 U.S. at 166, 105 S.Ct. 3099 (noting that the entity's policy or practice "must have played a part in the violation of federal law" (emphasis added) ).

The district court stated that, with respect to causation, it "[understood] Plaintiffs' argument as saying that each policy and practice does not, on its own, have to result in a constitutional violation." It pointed to this court's opinion in Alberti v. Klevenhagen for the proposition that, "[i]n determining the constitutional question, we need not separately weigh each of the challenged institutional practices and conditions, for we instead look to 'the totality of conditions.' " 790 F.2d 1220, 1224 (5th Cir. 1986) (quoting Ruiz v. Estelle, 679 F.2d 1115, 1139 (5th Cir.) ( Ruiz VII ), modified on other grounds, 688 F.2d 266 (5th Cir.1982) ; see also Rhodes v. Chapman , 452 U.S. 337, 347, 101 S.Ct. 2392, 69 L.Ed.2d 59 (1981) ("Conditions ... alone or in combination, may deprive inmates of the minimal civilized measure of life's necessities."); Hutto v. Finney , 437 U.S. 678, 686-87, 98 S.Ct. 2565, 57 L.Ed.2d 522 (1978) (stating that whether punitive isolation is unconstitutional will depend on the conditions of confinement). Alberti , Ruiz , and the related Supreme Court precedent stand for the logical proposition that, under the Eighth Amendment, the question of whether a particular policy or practice causes a constitutional violation necessarily depends on context-i.e. , how that policy or practice is interacting with other prison conditions. A certain condition could amount to a constitutional violation in the Prison A environment, but not in the Prison B environment. See, e.g. , Finney , 437 U.S. at 686-87, 98 S.Ct. 2565.

Requiring plaintiffs to identify with specificity the policies they allege directly cause constitutional harm is not irreconcilable with the idea that, in assessing the harm caused by a particular policy, the court may consider how other policies or practices exacerbate or ameliorate its effect. In an Eighth Amendment case from the Ninth Circuit, the court explained that, in applying the "totality of conditions" test, courts are not entitled to simply add together a number of conditions, "each of which satisfy Eighth amendment requirements," and then "rely on a vague conclusion that the 'totality of conditions' violates the Eighth Amendment." Hoptowit v. Ray , 682 F.2d 1237, 1247 (9th Cir. 1982) ( Hoptowit I ), overruled on other grounds by Sandin v. Conner , 515 U.S. 472, 115 S.Ct. 2293, 132 L.Ed.2d 418 (1995). Instead, courts must consider whether each specific condition amounted to cruel and unusual punishment. See Hoptowit II , 753 F.2d at 783-84. The court went on, however, to explain that "[e]ach condition of confinement does not exist in isolation; the court must consider the effect of each condition in the context of the prison environment, especially when the ill-effects of particular conditions are exacerbated by other related conditions." Hoptowit I , 682 F.2d at 1247 (internal quotations omitted). This does not absolve the reviewing court from having to identify individual, deficient conditions. See id . It is merely "a recognition that a particular violation may be the result of several contributing factors." Id .

In sum, the § 1983 causation component requires that the plaintiffs identify, with particularity, the policies or practices they allege cause the constitutional violation, and demonstrate a "direct causal link." See Piotrowski , 237 F.3d at 580. We do not, however, read our precedent to require the court to consider each policy or practice in a vacuum. The court may properly consider how individual policies or practices interact with one another within the larger system. Though the district court apparently accepted the "totality of conditions" approach, it did address each of the State's specific policies and practices on an individual basis. It also considered how the harmful effects of some policies are exacerbated by others. For example, the district court explained that DFPS's refusal to track child-on-child abuse-a policy related to monitoring and oversight-compounded the problem with excessive caseworker workloads.

The State's overarching causation argument essentially boils down to the contention that, since most children are already "damaged" upon entering foster care, it would be nearly impossible to prove that it was the State's policies or practices, rather than their experiences prior to State custody, that inflicted the damage. It seems to argue that there is just no way to quantify how much harm came before custody and how much harm was inflicted while the children were in State care. But the State provides no support for the proposition that in order to prove causation plaintiffs are required to measure with absolute precision how much more damage was done before rather than after children enter the foster system.

First, Plaintiffs do not dispute that many, if not most, children enter the foster care system having already experienced significant physical or emotional trauma that may have a lasting impact on their psychological well-being. It cannot be the case, however, that because a child has experienced some form of abuse before the State intervenes, he is not capable of being further harmed by additional abuse or neglect while in foster care. It is illogical to argue that because a child comes in already "damaged" the State cannot be held liable for inflicting further harm that compounds that damage-even if it cannot be measured with mathematical certainty.

Furthermore, there was a wealth of evidence at trial establishing that many children experience some degree of concrete harm after entering the State's care. For example, almost all of the named plaintiffs entered the system at a "Basic" level of care. By DFPS's own standards, a "Basic" child is the least "damaged" an intake can be. Most saw their level of care increase markedly over the course of their time in PMC as a result of abuse and continued lack of permanency. Their experiences map the accounts of the former foster children who were presented as fact witnesses at trial and are consistent with testimony from attorneys ad litem , former DFPS caseworkers, and experts. Moreover, as noted in Section III(2), supra , plaintiffs need not show that every member of the class has actually been harmed while in State custody; they need only demonstrate that they face a risk of serious harm as a result of the State's policies and that the State was deliberately indifferent to the risk. See Helling , 509 U.S. at 33, 113 S.Ct. 2475.

Before examining the merits of the plaintiffs' claims with respect to the individually challenged policies, we note globally that the State relies heavily, as it did in the district court, on its performance in a preliminary phase of the Child and Family Services Review ("CFSR") conducted by the federal government in 2014 as evidence that foster children do not face significant safety risks in the State's care. It notes that it outperformed the national standard on 6 of the 7 statewide safety and permanency indicators. But 2 of the 6 indicators in which Texas exceeded the standard do not incorporate any data for PMC children at all-they pertain only to children in TMC. The other 4 contain both TMC and PMC data. The only indicator that uses exclusively PMC data is the one that the State failed.

Moreover, the CFSR the State cites was preliminary and based entirely on data the State provided to the federal regulators. In preparing its final report, federal regulators conduct an independent, on-site review. The results of that review were significantly less flattering. Texas failed all 7 outcome measures pertaining to child safety, permanency, and well-being, and it failed 4 out of the 7 statewide indicators. The report also noted that the independent review "raised numerous concerns regarding the quality of the state's self-assessment of its case practices and the accuracy of case ratings."

IV. DFPS Policies and Practices

Plaintiffs' claims regarding DFPS's policies and practices fall into four overarching categories. Specifically, plaintiffs contend that DFPS's policies and practices with respect to 1) caseload management, 2) monitoring and oversight, 3) placement array, and 4) foster group homes, violate their right to be free from an unreasonable risk of harm while in State custody. We will examine each challenged policy area in turn.

1. Caseloads: General Class

DFPS concedes that caseworkers are critical to ensuring children's safety and that "almost every day these caseworkers can make life and death decisions about the children in their care." It also admits that "if [caseworkers] really are too busy" to do their job, it would create a safety risk. Notwithstanding that admission, DFPS does not impose any limit on caseloads, and it has not conducted a workload study to determine how many cases a caseworker can safely manage. Moreover, given the lack of reliable, up-to date-statistics, it is not even clear from the record how many children, on average, caseworkers are responsible for. As the district court lamented, "caseworker caseloads are still something of an open question despite years of litigation and weeks of trial."

Nonetheless, even by DFPS's charitable estimates, most caseloads exceed the maximum recommended by professional standards and experts. The CWLA recommends a caseload range of 12 to 15 children, while the Council on Accreditation recommends a range of 8 to 15 children. The most recent, comprehensive count estimates that nearly half of CVS caseworkers carry caseloads of 21 children or more, 22% carry caseloads of 26 children or more, and nearly 10% carry caseloads of 31 children or more. And numbers supplied by DFPS undersell the scope of the problem.

The data is problematic for a host of reasons. To begin with, DFPS calculates caseloads in terms of "stages," each representing a segment of a child's care plan, rather than by the number of individual children for whom each caseworker is responsible. This makes it difficult to assess how many children each caseworker actually has. DFPS claims that, by its calculation, caseworkers are responsible for between 17 and 19 children. In calculating caseload distribution, however, DFPS counted secondary workers-who are not primary CVS caseworkers and some of whom never interact with the child face-to-face-as well as part-time caseworkers and non-human workers "created out of overtime." Accordingly, the 17 to 19 estimate is exceedingly generous. Indeed, it is internally inconsistent : DFPS represented to the Texas Senate Committee on Finance in 2017 that additional caseworkers and salary increases were necessary to bring the caseload down to 26 children per caseworker. The Assistant Commissioner of CPS, Lisa Black, stated at trial that she believes caseworkers carry an average caseload of 28.1 children.

Caseload figures reported by neutral outside auditors demonstrate that the numbers presented by the State at trial are artificially low. A report commissioned by the Texas Supreme Court states that "CPS caseworkers are routinely handling around 30 cases," and "[i]n the larger urban jurisdictions, caseworkers are commonly assigned 40-plus cases at a time, and the 'cases' often involve multiple children from the same family." The Adoption Review Committee estimated that caseloads were between 30 and 35 cases per caseworker, "often twice what is deemed best practice." Former DFPS caseworker Beth Miller testified that she routinely carried 40 to 60 cases, "sometimes higher." Her former colleague, Katrina Voelkel, similarly testified that she remembers regularly carrying between 30 and 50 cases. Significantly, caseload averages "do not reflect spikes in caseloads that can occur when a co-worker goes on extended leave or quits." In short, the record amply supports the district court's finding that CVS caseloads are extremely high.

Oversized workloads are also the primary cause of the exceedingly high rate of caseworker turnover. Indeed, the State's own expert on child welfare policy, Dr. Jane Burstain, authored an article in 2009 that stated there was "a fairly direct relationship ... between caseloads and voluntary turnover." This relationship has "remained consistent from year to year." Over 25% of the roughly 2,000 CVS caseworkers leave CPS annually. More than 25% of caseworkers leave within their first year, and 43% leave within their first two years. And Burstain has said that turnover rates are likely understated, as they do not account for caseworkers who leave their positions for others within the agency. To keep pace with the attrition rate, DFPS has to hire approximately 500 new caseworkers every year just to maintain a full-capacity workforce of 1,000. DFPS struggles to hire and train workers quickly enough to fill all of the available positions. As a result, CPS has a high volume of positions that remain vacant.

Turnover is also an enormous fiscal burden for DFPS. The Sunset Commission estimated in 2014 that the loss of caseworkers over the prior year resulted in a $72.7 million impact to the agency. Experts estimate that retaining even a portion of lost workers for an extra year could save DFPS roughly $25 million. Turnover is not only costly, "[i]t also creates a negative environment that reduces productivity as well as feeds more turnover." Internal DFPS reviews consistently reveal the general sentiment within CPS that management practices are "unfair, unsupportive, bullying, unreasonable, and fear-driven." Agency employees even expressed concern about retaliation for their cooperation with an external review commissioned by the state legislature. Caseworkers feel that agency supervisors are singularly focused on ensuring that caseworkers meet arbitrary metrics rather than assessing whether they are making meaningful progress with their cases and providing quality services. The added stress of keeping up with the requirements of a purely numbers-driven management approach contributes to caseworkers' feelings of hopelessness and frustration and reduces overall productivity. The vicious cycle is never-ending: unmanageable workloads and a caustic work environment lead to high rates of caseworker turnover; turnover further exacerbates caseworker burnout, low morale, and a negative agency culture, which feeds more turnover.

High turnover compounds the workload problem, as caseloads have to be redistributed as caseworkers leave. New caseworkers do not receive a full caseload for at least six months after joining CPS. Accordingly, the remaining CVS caseworkers are forced to add additional cases to their already bloated caseloads in the interim. Of the new hires, one out of every six leaves within the first few months, and the process of hiring and training starts anew as DFPS scrambles to fill those slots. This means that DFPS cannot relieve veteran caseworkers of their additional "temporary" workload in a timely manner. New caseworkers that remain after the training process is complete face a daunting learning curve. Again, it took the district court 462 hours-eleven uninterrupted workweeks-just to read the 358,102 pages of casefiles for 20 PMC children. Reading the information containing in the casefile is, of course, only the beginning. All in all, DFPS estimates it takes roughly two years for a caseworker to get up to speed on a new case. The high rate of turnover year after year means that this arduous process is duplicated many times over.

The combination of unmanageable caseloads and high caseworker turnover creates a "cycle of crisis" that allows children to "fall through the cracks." A comprehensive agency analysis commissioned by DFPS found that the workload level "is qualitatively reducing CPS caseworkers' ability to keep children safe." This conclusion is unsurprising. A logical result of inconsistent and perfunctory contact with caseworkers is that children don't have material access to an advocate when something goes wrong in their placement. Attorneys ad litem and former foster children testified that caseworkers were often too busy to answer or return phone calls-even after repeated attempts to reach them. Former foster child Darryl Jackson testified that "it was just hard to get in contact" with his caseworker, "to even have a conversation with her, you know, tell her I needed things." Colleen McCall, the Director of Field Operations at CPS, stated in an Action Memorandum that "[d]ue to the shortage of staff, required caseworker documentation, such as Child Service Plans and documentation of children's medicals and dentals, are not being completed timely, if at all." More than 55% of caseworkers report that "they do not have adequate time during the workday to successfully do their job."

Caseworkers are routinely unable to make regular, face-to-face contact with their children; even when they are able to make visits, the contact is often "cursory." As a result of high caseloads and administrative burdens, both of which are exacerbated by the abysmal state of DFPS's recordkeeping systems, CVS caseworkers spend only 26% of their work hours actually interacting with foster children and families. "[T]his number is clear evidence that the agency is doing more compliance than care." John Specia, the DFPS Commissioner at the time of trial, called this face-time figure "disturbing."

Several named plaintiffs and former foster children testified that they would often go months without seeing their primary caseworker. When caseworkers do manage to visit their foster children in person, the interaction is likely to be extremely brief. As a reference point, one foster parent of seven years reported that caseworkers visiting the home typically spent no more than five minutes with each child. Though caseworkers are required to conduct face-to-face meetings with their children in private, foster children report that they are frequently interviewed in the presence of their caregiver and other children. As a result, reporting issues or abuse-already a difficult and intimidating task for vulnerable children-becomes near-impossible. Under these conditions, it is entirely unsurprising that children struggle to establish meaningful, productive relationships with their caseworkers. Children do not trust their caseworkers to follow-up on problems or to keep them safe.

Moreover, because of turnover, children are cycled through multiple caseworkers. In some cases, children do not even know who their caseworker is. This further inhibits the development of a trusting relationship in which children feel safe communicating their needs or reporting abuse. Former foster child Patricia Virgil, who had a total of 10 caseworkers throughout her seven years in DFPS custody, explained that because her caseworkers changed so frequently, "whenever I had issues in some of the homes, I didn't know who to go to, I didn't know who to trust and so I just-most of the time I just kept my mouth shut." Though she attempted to report being sexually abused at one of her foster homes-apparently through some sort of central DFPS phone line rather than to her absent caseworker-no one from the agency ever followed up on the investigation, and her caseworker never once visited her at that placement.

In many instances, caseworkers lack the time to be thorough when evaluating the safety or appropriateness of a placement on the front-end. This means that important red flags may get overlooked. Even assuming that a "red flag" regarding a placement has been documented, a caseworker would have to navigate tens of thousands of pages of records that are scattered across multiple databases and paper files that are not consistently maintained chronologically in order to stumble upon it. And because records and case files are outdated and woefully incomplete, there is no guarantee the information caseworkers' need was ever recorded in the first place. Caseworkers do not have the time to perform fundamental aspects of their job; clearly, they do not have the bandwidth to replicate a needle-in-a-haystack search several times over for each individual child every time they have to move him. This limited ability to rigorously evaluate placement choices and permanency plans substantially increases the chance that a child will be exposed to serious safety risks. The risk is further compounded by the fact that DFPS does not centrally track child-on-child abuse and that RCCL investigations have an exceedingly high error rate.

The evidence in the record establishes that the State is deliberately indifferent to the risks posed by its policies and practices toward caseload management. The State is well-aware that caseworkers have unmanageable workloads. It also knows that high caseloads-which are a direct cause of high turnover rates-have a negative impact on PMC children's welfare. Numerous reports, internal audits, and comprehensive studies of the system conducted over several years-including some that were commissioned by DFPS itself-have informed the agency that caseloads are too high and that, as a result, children are at a greater risk of harm. The findings and recommendations have been replicated repeatedly over the past two decades. Every single one of the reports in the record identifies unmanageably high caseloads as one of the most urgent problems DFPS faces and explicitly warns that high caseloads compromise caseworkers' ability to keep children safe.

DFPS has also been cautioned recurrently since 1996 that high turnover rates exacerbate the caseload problem and contribute to the agency's inability to provide quality services to the children and families in its care. The reports themselves acknowledge that they tend to merely reiterate the problems and potential solutions that have been proffered to the agency time and time again. As the Texas Adoption Review Committee lamented, "[s]adly ... many of the same problems identified in 1996 still exist in the current child welfare system in Texas," and, as a result, "[m]any of our recommendations are, sadly, ones that have been made in prior years."

DFPS has repeatedly acknowledged that high caseloads cause the quality of casework to suffer and put foster children at an increased risk of harm. Legislative appropriations requests regularly state that caseload increases result in "significant" child safety issues. Top DFPS officials, including former Commissioner Specia, admit that there is a causal connection between high caseloads and negative safety outcomes for children. DFPS is also aware that frequent turnover exacerbates the workload problem and further threatens child safety and well-being. The State does not contend that DFPS was unaware of the numerous reports in the record. Indeed, it cites two of them several times in its brief on appeal. Many reports were directly commissioned by the Texas legislature or the judiciary. Commissioner Specia himself was on the policy development team for the 2010 Texas Appleseed Report before he was appointed to lead DFPS.

The State contends that DFPS is actively managing caseloads and making improvements to its workforce. Reasonable steps to cure the problem, even if ultimately ineffective, would negate the district court's finding that the State was deliberately indifferent to the risk of harm. See Farmer , 511 U.S. at 844, 114 S.Ct. 1970. The State contends that DFPS has taken several steps to address the risks associated with high caseloads, including hiring more caseworkers, employing secondary workers and support staff, and initiating a new program to improve caseworker training. Under the circumstances, none of these steps constitute a "reasonable" response to the systemic issues.

The most fundamental problem DFPS faces with respect to addressing the caseload issue is the lack of informative workload data or internal agency caseload standards. McCall-who is the individual tasked with ensuring that CVS caseworkers have manageable caseloads-admitted at trial that she had "no idea what size of child caseload [ ] conservatorship workers should have in order to do their jobs properly." She conceded that she had never even attempted to count the number of children, on average, that each CVS caseworker is responsible for. Indeed, at the time of trial, DFPS had not performed a comprehensive workload study in over a decade . Despite being explicitly informed by the State Auditor's Office in 2009 that the 2004 study was outdated and should be redone, "DFPS did not implement this recommendation and continues to use the 2004 information."

In response to the district court's 2015 liability opinion, DFPS provided the Special Masters with a limited workload survey conducted from August 2015 to March 2016 which purported only to estimate how much time was actually spent on casework during that time period. It made no attempt to quantify how much time caseworkers should be spending on casework or how many cases a caseworker could safely manage. Despite being reprimanded by the district court in 2015 for its inclusion of ISY workers in its workload estimates, DFPS again included ISY workers in its 2016 study. The Special Masters extrapolated from the data DFPS provided that an average caseworker has adequate time to manage 14 PMC cases at one time, to the exclusion of TMC cases. When the Special Masters asked DFPS to determine how many additional caseworkers it would need to achieve workloads of 14 children per caseworker, DFPS declined to provide the information, responding that it was "not feasible" to do so.

The district court was prepared to consider caseload standards promulgated by the agency, but DFPS did not have any. The Special Masters were prepared to work with DFPS to develop standards, but DFPS declined to do so. DFPS's response to inquiries regarding appropriate workload levels is, essentially, that the volume of cases an individual caseworker can shoulder is dependent on a number of different variables, including the complexity of the cases and the caseworker's level of experience. This is undoubtedly true. But the fact that caseworkers' workload capacities will vary, on an individual basis, according to the types of cases a caseworker is assigned does not obviate the need for general guidelines that identify an appropriate caseload range.

The lack of agency standards colors the lens through which we view the steps DFPS claims it has taken toward addressing the caseload management problem. For example, the district court took judicial notice of the fact that the legislature approved DFPS's request for additional caseworkers and salary increases. Considered in a vacuum, asking for more money to hire more people seems logical. But DFPS has included a request for more money to hire caseworkers in every appropriations request it has submitted to the legislature in the past two decades-it is a standard, boilerplate request. And, of course, DFPS has no choice but to continually hire more caseworkers every year. The number of children in DFPS custody is steadily increasing. Moreover, because of turnover, DFPS has to replenish roughly a quarter of its caseworker workforce every year in addition to the staff necessary to accommodate the influx of more children.

Simply adding more employees has continually proven ineffectual. High-volume hiring is not a solution. Critically, because of the lack of internal standards, DFPS does not even know how many caseworkers it actually needs to reduce the caseloads to safe levels. Without a target number, the agency is hiring blind. Lastly, while DFPS focuses primarily on high-volume hiring to fill the gaps left by the mass exodus of caseworkers every year, it repeatedly fails to address the internal management issues that motivate many caseworkers to leave so quickly after joining CPS. Thus, not only is a portion of DFPS's yearly budget allocated by the state legislature to hire more workers and reduce caseloads effectively wasted, but the underlying problem remains unsolved.

The State points to DFPS's use of secondary workers as evidence that it has made a reasonable effort to alleviate the burden on caseworkers. The district court was entitled to find that the risks associated with overburdened caseworkers were not sufficiently mitigated by these secondary workers, particularly ISY workers.

ISY workers are not nearly as intimately involved with an individual child's case as is a primary caseworker. They do not participate in a child's long-term placement plan nor are they required to do any follow up on the child's needs. Many primary caseworkers have never met a single ISY wo