2014-09001. Statewide Data Indicators and National Standards for Child and Family Services Reviews

AGENCY:

Children's Bureau (CB), Administration on Children, Youth and Families (ACYF), Administration for Children and Families (ACF), Department of Health and Human Services (HHS).

ACTION:

Notice of Statewide Data Indicators and National Standards for Child and Family Services Reviews.

SUMMARY:

The Children's Bureau reviews a state's substantial conformity with titles IV-B and IV-E of the Social Security Act through the Child and Family Services Reviews (CFSRs). Statewide data indicators are used to inform the Children's Bureau's determination of a state's substantial conformity relative to certain safety and permanency outcomes. This document advises the public of the Children's Bureau's plan to replace the statewide data indicators and the methods for calculating associated national standards on those indicators. We invite the public to comment on these indicators and methods before their use in CFSRs scheduled for Federal Fiscal Years (FFY) 2015 through FY 2018.

DATES:

Written comments must be submitted to the office listed in the ADDRESSES section below on or before May 23, 2014.

ADDRESSES:

Interested persons may submit written comments by any of the following methods:

• Federal eRulemaking Portal: http://www.regulations.gov. Follow the instructions for submitting comments.
• Mail or Courier Delivery: Miranda Lynch Thomas, Children's Bureau, Administration on Children, Youth and Families, Administration for Children and Families, 1250 Maryland Avenue SW., 8th Floor, Washington, DC 20024.

Instructions: If you choose to use an express, overnight, or other special delivery method, ensure that delivery may be made at the address listed under the ADDRESSES section. We urge interested parties to submit comments electronically to ensure that they are received in a timely manner. All comments received will be posted without change to http://www.regulations.gov. This will include any personal information provided.

FOR FURTHER INFORMATION CONTACT:

Miranda Lynch Thomas, Children's Bureau, 1250 Maryland Ave. SW., 8th Floor, Washington, DC 20024, [phone].

SUPPLEMENTARY INFORMATION:

Background

The Children's Bureau implemented the CFSRs in 2001 in response to a mandate in the Social Security Amendments of 1994 for the Department of Health and Human Services to issue regulations for the review of state child and family services programs under titles IV-B and IV-E of the Social Security Act (see section 1123A of the Social Security Act). The reviews are required for the Children's Bureau to determine whether such programs are in substantial conformity with title IV-B and IV-E plan requirements, implementing regulations, and relevant title IV-B and IV-E plans. The review process, as regulated at 45 CFR 1355.31-37, grew out of extensive consultation with interested groups, individuals, and experts in the field of child welfare and related areas.

The Children's Bureau conducted the first round of CFSRs from 2001 through 2004 and the second round from 2007 through 2010. The third round is scheduled to begin in FFY 2015. Information about the initiation of this latest round can be found in CFSR Technical Bulletin #7 issued in March 2014 (see http://www.acf.hhs.gov/​programs/​cb/​resource/​cfsr-technical-bulletin-7).

The CFSRs enable the Children's Bureau to: (1) Ensure conformity with federal child welfare requirements; (2) determine what is actually happening to children and families as they are engaged in child welfare services; and (3) assist states to enhance their capacity to help children and families achieve positive outcomes. The Children's Bureau conducts the reviews in partnership with state child welfare agency staff and other stakeholders involved in the provision of child welfare services. We have structured the reviews to help states identify strengths as well as areas needing improvement within their agencies and programs.

The CFSR assesses state performance on seven outcomes and seven systemic factors. The seven outcomes focus on key items measuring safety, permanency, and well-being. The seven systemic factors focus on key state plan requirements of titles IV-B and IV-E that provide a foundation for child outcomes. States that the Children's Bureau determines have not achieved substantial conformity in all the areas assessed in the review are required to develop and implement a program improvement plan within two years addressing the areas of nonconformity. The Children's Bureau supports the states with technical assistance and monitors implementation of their program improvement plans. States that are unable to complete their plans successfully have some of their federal child welfare funds withheld.

Most relevant to this document is the element of the reviews that provides for the Children's Bureau to determine whether the state is in substantial conformity with certain child outcomes based on national standards we set for state performance on statewide data indicators. The regulations at 45 CFR 1355.34(b)(4) and (5) authorize us to add, amend, or suspend any of the statewide data indicators when appropriate, and to adjust the national standards when appropriate. Statewide data indicators are aggregate measures and we calculate them using administrative data available from a state's submissions to the Adoption and Foster Care Analysis and Reporting System (AFCARS),^[1] the National Child Abuse and Neglect Data System (NCANDS),^[2] or a Children's Bureau-approved alternate source for safety-related data. If a state is determined not to be in substantial conformity with a related outcome due to its performance compared to the national standard for an indicator, the state will include that indicator in a program improvement plan. The Children's Bureau establishes performance goals for each indicator included in a program improvement plan, based on the state's prior performance, that the state is expected to reach by the end of the program improvement plan period. The goal to be achieved is relative to the state's baseline performance at the beginning of the program improvement plan period.

The Children's Bureau views the CFSR as a dynamic process and has made ongoing improvements in the process to best meet state and federal needs. Most recently, we solicited feedback from the public (see 76 FR 18677, published April 5, 2011) about how they would envision a federal review process that meets the statutory requirements in section 1123A of the Social Security Act and holds child welfare agencies accountable for achieving positive outcomes for children and families and continuously improving the quality of their systems for doing so. In addition, we hired a consultant that specializes in child welfare measurement to work with Children's Bureau data specialists. In 2013 we also tasked a contractor to the Children's Bureau to convene a panel of child welfare administrators and data measurement experts to develop recommendations and feedback about specific aspects of the review process, including the statewide data indicators, national standards, and program improvement. The information from these experts along with public comments has shaped our plan for replacing the statewide data indicators that will be used in the CFSRs.

Existing Statewide Data Indicators and Composite Measures and Planned Improvements To Address Feedback

For CFSR Round 2, the Children's Bureau developed six statewide data indicators and measures: two indicators related to safety and four composite measures related to permanency. The two safety-related statewide data indicators focused on recurrence of maltreatment and maltreatment of children in foster care and were used to inform an assessment of the state's substantial conformity with the safety outcome that children are, first and foremost, protected from abuse and neglect. The four permanency-related data composites were used to inform the assessment of a state's substantial conformity with the permanency outcome that children have permanency and stability in their living situations. The four permanency composites used during CFSR Round 2 were related to measures of timeliness and permanency of reunification; timeliness of adoptions, permanency for children in foster care for extended time periods, and placement stability.^[3] The Children's Bureau established national standards for each of the six data indicators and composites. The Children's Bureau thoroughly considered the alternatives available to us at the time and had a clear rationale for supporting the statewide data indicators, composites and methods chosen. However, we believe there is reason to make further improvements now given the additional concerns raised by stakeholders and the Children's Bureau that we summarize below and throughout this document as we describe the proposed new indicators.

Cohorts Used for Statewide Data Indicators: The Children's Bureau is planning to calculate several statewide data indicators using entry cohorts to address concerns about the measurement techniques that relied on multiple exit cohorts in prior CFSR rounds. Some stakeholders noted that relying on exit cohorts and cross-sectional (also referred to as point-in-time) cohorts, rather than prospective entry cohorts, can statistically distort system performance outcomes because they represent biased sub-samples of all children who are served.^[4] For example, looking only at children who exit foster care in a given year ignores those children who did not leave; it is therefore not representative of all children in foster care who would be affected by state improvement efforts. Looking only at those children in foster care at the end of the year biases the sample to include more children with long lengths of stay in foster care. Entry cohorts, which include all children entering foster care in a given year, avoid these problems and provide a more complete assessment of overall system performance and recent practice trends.

While entry cohorts have methodological advantages, they have limitations in terms of assessing state performance with regard to children who have been in foster care for a long time because of the length of time we measure for state improvements. For example, with an entry cohort approach, children who had already been in foster care for two or more years could only start being tracked in a third year. To address this limitation, the Children's Bureau will still use other cohorts in some of its indicators.

Composites: The Children's Bureau plans to use individual indicators rather than composite measures for the purpose of establishing national standards in this round of review. We implemented the composite approach in CFSR Round 2 after consultation with states and national experts to support a more holistic approach to measuring state performance. The expansion to composites from the one-dimensional measures used in CFSR Round 1 allowed state performance on a particular domain to reflect broader performance, accounting for both strengths and weaknesses within the domain. Criticism of the composite measures used during CFSR Round 2 included the complexity of the composite scores for interpretation purposes.

To address these concerns and clarify expectations with regard to national performance, the Children's Bureau proposes to measure state performance with simplified statewide data indicators. We propose to maintain some of the advantages found with the composite approach by implementing companion measures during the program improvement plan period to provide an expanded and more effective measurement of a domain.

New Proposed Statewide Data Indicators and Methods

The Children's Bureau plans for the new statewide data indicators to measure maltreatment in foster care and re-report of maltreatment as a component in evaluating Safety Outcome 1: Children are, first and foremost, protected from abuse and neglect. We also plan for statewide data indicators to measure the achievement of permanency in 12 months for children entering foster care, permanency in 12 months for children in foster care for 2 years or more, re-entry to foster care, and placement stability. These four permanency indicators will be used as a component in evaluating Permanency Outcome 1: Children have permanency and stability in their living situations. Below is a description of each of the six proposed indicators including their definition, justification for inclusion, calculation and a discussion of relevant issues. Following the description of the indicators is information on the methods the Children's Bureau plans to use for calculating the national standards and our approach to measuring a state's program improvement on the indicators should we find that the state is not able to meet the national standard. We also provide additional information on how the Children's Bureau will share data information with states through profiles and data quality issues that impact these indicators and methods.

Attachment A provides a summary of each planned indicator including numerators, denominators, exclusions, and adjustments. Attachment B provides a comparison of the data measures used during CFSR Round 2 with the statewide data indicators we propose to use during Round 3. Attachment C provides information on the AFCARS and NCANDS data elements that are used to calculate the proposed indicators and national standards.

Statewide Data Indicators for CFSR Safety Outcome 1: Children Are, First and Foremost, Protected From Abuse and Neglect

Proposed Safety Performance Area 1: Maltreatment in Foster Care

Indicator Definition: Of all children in foster care during a 12-month period, what is the rate of victimization per day of foster care? ^[5] The indicator includes all cases of substantiated or indicated maltreatment while in foster care and all days for all children in foster care at any point during a 12-month period. The denominator is all child days in foster care over a 12-month period, and the numerator is the number of instances of substantiated or indicated maltreatment among children in foster care over that same period. The definition of “children” for this indicator (and all indicators) includes those under the age of 18. This indicator includes all maltreatment types by any perpetrator, which may include foster parents, facility staff members, parents, or others. In addition, this indicator includes all days for all children in foster care at any point during a 12-month period.

Some states provide incident dates in their NCANDS data submissions. If a state provides incident dates, records with an incident date occurring before the date of removal will be excluded. Children in foster care for less than 8 days and any report that occurs within the first 7 days of removal are excluded from this indicator. This indicator is calculated using data that match children across AFCARS and NCANDS using the AFCARS identifier.

Justification for Inclusion: This indicator provides a measure of whether the state child welfare agency is able to ensure that children do not experience abuse or neglect while in the state's foster care system. The indicator holds states accountable for keeping children safe from harm while under the responsibility of the state, no matter who perpetrates the maltreatment in foster care.

Relevant Issues and Discussion: During CFSR Round 2, the Children's Bureau had a safety data indicator related to maltreatment in foster care. For that indicator, the counts of children not maltreated in foster care were derived by subtracting the NCANDS count of children maltreated by foster care providers from the total count of all children in foster care on the last day of the year, as reported in AFCARS. Because of improved reporting by states, we can now link AFCARS and NCANDS data using the child identifier and determine whether maltreatment occurred during a foster care episode, improving accuracy on this indicator by using entry cohorts instead of the retrospective method used in CFSR Round 2.^[6] This technique also allows us to expand the indicator to include maltreatment regardless of the perpetrator type. This measure uses the report date reported in NCANDS to determine if a child is victimized while in foster care, discounting the first week to allow for a potential lag time between an incidence of maltreatment and report of maltreatment. For those states that provide incident dates, an adjustment will be made if the data indicates that the incident occurred prior to the foster care episode. We encourage states to report incident dates in NCANDS, which will improve the accuracy of this indicator. The Children's Bureau made this change in response to a suggestion from stakeholders with regard to the indicators used for the last round of reviews that we are now able to address.

Proposed Safety Performance Area 2: Re-Report of Maltreatment

Indicator Definition: Of all children who received a screened-in report of maltreatment during a 12-month period (regardless of disposition type), what percent were reported again within 12 months from the date of initial report? The denominator is the number of children with at least one screened-in report of alleged maltreatment in a 12-month period and the numerator is the number of children in the denominator that had another screened-in report with a disposition within 12 months of their initial report. Screened-in reports that have a disposition reported are included, regardless of whether the disposition is that the child is a victim or a non-victim. This indicator is calculated using data from NCANDS.

Justification for Inclusion: This indicator is included to provide an assessment of whether the agency took the necessary actions to prevent a future report of maltreatment for children previously the subject of a screened-in report to the agency.

Relevant Issues and Discussion: During CFSR Round 2, the Children's Bureau had a safety performance area related to repeat maltreatment. That measure was derived from calculating what percent of all children who were victims in a substantiated or indicated maltreatment allegation during the first 6 months of the reporting period were not victims in another substantiated or indicated maltreatment allegation within a 6-month period. We are expanding this measure to count all children with screened-in reports of alleged maltreatment. The Children's Bureau believes that multiple reports regardless of whether maltreatment is substantiated or indicated is a viable measure of the agency's attempts to prevent maltreatment based on research indicating that families with screened-in but unsubstantiated reports are at a high risk of re-report, in some cases as high as substantiated cases.^[7] Because reports are included regardless of disposition, this indicator includes both victims and non-victims. In addition, this indicator expands the time period examined to 12 months to include more children. The indicator also tracks such children for 12 months, as opposed to 6 months as in the prior indicator.

The Children's Bureau made this change in response to a suggestion from stakeholders with regard to the indicators used for the last round of reviews that we are now able to address with the improved quality of data reported in NCANDS. In addition, the contractor's recommendations based on the expert panel convened in 2013 expressed support for a measure of screened-in reports to capture repeat maltreatment.

Previous CFSR data measures focused on substantiated and indicated reports of maltreatment. The growing implementation of differential response in the states (sometimes referred to as alternative response programs) where a substantial percentage of cases may bypass formal investigation altogether, however, makes a comparison of differential-response and non-differential-response states difficult. In addition, states that initiate or expand differential response during an improvement period could show improvement on a substantiation-based measure of repeat maltreatment merely as an artifact of adopting differential response. An indicator based only on screened-in reports is not affected by differential response which contributed to our selecting this indicator.

CFSR Permanency Outcome 1: Children Have Permanency and Stability in Their Living Situations

The permanency-related statewide data indicators exclude children entering foster care at age 18 and older or who are already 18 and older on the first day of the period under review. Although the amendments to title IV-E of the Social Security Act made by the Fostering to Success and Increasing Adoptions Act of 2008 (Pub. L. 110-351) permit states to provide foster care to youth who are age 18 and older, all states have not exercised such an option. Some states provide foster care to youth age 18 and older, however, there is no consistent inclusion of this population of youth across states and no consistent construct at this time for what achieving permanency means for such older youth. Therefore, the Children's Bureau believes that it is appropriate to limit the permanency statewide data indicators to children under age 18 in this way to maintain consistency as we have in prior rounds.

Proposed Permanency Performance Area 1: Permanency in 12 Months for Children Entering Foster Care

Indicator Definition: Of all children who enter foster care in a 12-month period, what percent discharged to permanency within 12 months of entering foster care? The denominator is the number of children who enter foster care in a 12-month period and the numerator is the number of children in the denominator who discharged to permanency within 12 months of entering foster care. For the purposes of this indicator, permanency includes a child's discharge from foster care to parent(s), living with relatives, guardianship, or adoption. Children who are in foster care for less than 8 days are excluded from this indicator. For children with multiple episodes during the same 12 month period, this measure will only evaluate the first episode within the period. This indicator is calculated using data from AFCARS.

Justification for Inclusion: This indicator provides a focus for the child welfare agency's responsibility to reunify or place children in safe and permanent homes as soon as possible after removal.

Relevant Issues and Discussion: During CFSR Round 2, the Children's Bureau included a similar measure as part of one of the permanency composites. That measure assessed the percent of all children entering foster care for the first time in a 6-month period that left foster care to reunification (or living with a relative) within 12 months of entering foster care. For CFSR Round 3, the indicator will stand alone and not be combined into a composite measure. The indicator is also more expansive than the prior to include: (1) All permanency outcomes, including guardianship or adoption; (2) all children entering foster care rather than first removals only; and (3) children entering foster care over the course of 12 months rather than 6 months.

The Children's Bureau made this change in response to suggestions from stakeholders with regard to the indicators used for the last round of reviews that we are now able to address. The indicator's expanded set of permanency outcomes recognizes that all forms of permanency represent equally successful outcomes for children. Although all permanency outcomes are included within this one indicator, states will still be able to analyze their data to determine which types of permanency they are achieving for children. The indicator's expanded population recognizes the Children's Bureau's desire to measure performance for all children entering foster care rather than first removals only. The expansion to 12 months, as opposed to 6 months in the prior indicator will yield more stable estimates of performance. A 12 month period is important for this indicator as this cohort will also serve as the basis for the denominator in the re-entry into foster care indicator (discussed further below). Re-entry into foster care after a discharge from foster care is a rarer event that is better captured over a longer period to accommodate variability. In addition, including a full 12 month period lessens the effect of potential seasonal differences between 6 month periods.

Please see the section on program improvement plans for more information on how this indicator may be used in program improvement.

Proposed Permanency Performance Area 2: Permanency in 12 Months for Children in Foster Care for 2 Years or More

Indicator Definition: Of all children in foster care the first day of the year who had been in foster care (in that episode) for 2 years or longer, what percent discharged to permanency within the next 12 months? The denominator is the number of children in foster care on the first day of a 12-month period who had been in foster care (in that episode) for 2 or more years, and the numerator is the number of children in the denominator who discharged to permanency within 12 months of the first day. Permanency includes discharge to parent(s), living with relatives, guardianship, or adoption. Children in foster care for less than 8 days are excluded from this indicator. This indicator is calculated using data from AFCARS.

Justification for Inclusion: The Children's Bureau is committed to maintaining focus on the key outcome of achieving permanency for all children in foster care and shortening the time to permanency. With a second indicator of permanency in 12 months specifically focused on the cohort of children who have been in foster care for 2 or more years, the Children's Bureau can monitor the effectiveness of the state child welfare agency in ensuring that states continue to emphasize permanency for children who have been in foster care for longer periods of time, to ensure they achieve permanency during the period under review.

Relevant Issues and Discussion: During CFSR Round 2, this same individual measure was evaluated as a part of a composite. There are substantial numbers of children that cannot achieve permanency in 12 months, and those children may present different challenges. Such children may have a higher likelihood of achieving permanency goals such as adoption and guardianship than those achieving permanency in the first year. The Children's Bureau continues this measure because of a commitment to monitor performance for children and youth who were already in foster care, and hold states accountable for attaining permanency for them, as well.

Proposed Permanency Performance Area 3: Re-Entry to Foster Care

Indicator Definition: Of all children who entered foster care in a 12-month period who were discharged within 12 months of that entry to reunification, living with a relative, or guardianship, what percent re-entered foster care within 12 months of their discharge? The denominator is the number of children who entered foster care in a 12-month period who discharged within 12 months to reunification, living with a relative, or guardianship, and the numerator is the number of children in the denominator who re-entered foster care within 12 months of their discharge from foster care. Children in foster care for less than 8 days are excluded from this indicator. If a child re-enters foster care multiple times within 12 months of when they left, only the first re-entry into foster care is selected. This indicator is calculated using data from AFCARS.

Justification for Inclusion: Although the Children's Bureau believes that it is important to reunify children with their families as quickly as possible, we also believe that children should be reunified when safe and appropriate and with sufficient supports in place to prevent a subsequent removal. This indicator enables the Children's Bureau to monitor the effectiveness of programs and practice that support reunification and other permanency goals.

Relevant Issues and Discussion: During CFSR Round 2, this performance area was evaluated using a similar measure as a part of a composite. That measure was derived by calculating what percent of all children discharged from foster care to reunification or living with a relative in a 12-month period re-entered foster care in less than 12 months from the date of discharge. This indicator differs from the measure used for CFSR Round 2 by limiting the children eligible for re-entry to the entry cohort. The CFSR Round 2 measure counted all children who left foster care to reunify or live with a relative, regardless of when they entered foster care. The purpose of this focus on current practice is in keeping with the rationale that new interventions may best be monitored in an entry cohort. We also expanded the denominator to include children who leave foster care for guardianship in an effort to reflect a more comprehensive definition of permanency.

The Children's Bureau made this change in response to suggestions from stakeholders with regard to the indicators used for the last round of reviews that we are now able to address. The indicator attempts to capture the rate of “permanency” for children who leave foster care by measuring whether children re-enter foster care. For this indicator, adoption is not included as a permanency outcome because it is not always possible to identify children who re-enter foster care following adoption.

Please see the section on program improvement plans for more information on how this indicator may be used in program improvement.

Proposed Permanency Performance Area 4: Placement Stability

Indicator Definition: Of all children who enter foster care in a 12-month period, what is the rate of placement moves per day of foster care? ^[8] The denominator is, among children who enter foster care in a 12-month period, the total number of days these children were in foster care as of the end of the 12-month period. The numerator is, among children in the denominator, the total number of placement moves during the 12-month period. The initial placement in foster care is not counted, but all subsequent moves occurring within the 12-month period are included in the calculation. Children in foster care for less than 8 days and other settings a child may be placed in, that are not considered as placement settings for AFCARS purposes such as trial home visits, are not included in this indicator. This indicator is calculated using data from AFCARS.

Justification for Inclusion: Placement stability is a critical component of the permanency and well-being of children in foster care. States are responsible for ensuring that children who are removed from their homes by the state experience stability while they are in foster care.

Relevant Issues and Discussion: During CFSR Round 2, this performance area was evaluated using a similar measure as a part of a composite. That measure was derived from calculating what percent of all children served in foster care during the 12-month period had two or fewer placement settings. The new indicator controls for the length of time children spend in foster care so that we can examine moves per day of foster care, rather than children as the unit of analysis, as was employed during CFSR Round 2. The measure used for CFSR Round 2 was unable to differentiate between children who moved twice from children who moved more. The new indicator counts each move to accurately capture the rate of placement moves, rather than the number of children affected.

The Children's Bureau believes that placement stability is important to the permanency and well-being of children in foster care regardless of how long they have been in foster care. Even so, our analysis of AFCARS data indicates that most placement moves occur within a child's first 12 months of foster care, which is why we plan to focus this indicator on that time period. With this refined focus, the Children's Bureau and states can monitor the period during which placement moves are most likely to occur and the state's most recent performance. In the CFSR Round 2 measure, placement moves were monitored over the life of the case which meant that placement instability for a child in the early years of foster care placement would affected the assessment of the state's CFSR performance in a more recent period under review. We also believe that by confining the indicator to this period of time, we are better able to measure a state's improvement in a subsequent 12-month period. The Children's Bureau made this change in response to suggestions from stakeholders with regard to the indicators used for the last round of reviews that we are now able to address.

National Standards

National standards will be established for all indicators. By measuring state performance against national standards on statewide data indicators, the Children's Bureau can assist states in continuously monitoring their performance on child outcomes and help practitioners and administrators better understand the entirety of their child welfare systems.

We propose that the national standard for each indicator be set at the national observed performance for that particular indicator. The national standards will remain constant over the entire round of review, as has been the case in prior rounds. In CFSR round 2, national standards were based on the 75th percentile (approximately) of all states' performance, with an adjustment for sampling error. For this round, we believe that the national observed performance, which will be similar to the average performance across all states, is a more reasonable benchmark and would appropriately challenge states to improve their performance.

The national standard set at the national performance level for each indicator is a benchmark that is easily communicated to and understood by stakeholders, and a reasonable goal given the reality that states still need to improve practice in multiple areas. Setting the national standard at the national performance for each indicator is rooted in strategies central to an effective performance management system focused on continuous quality improvement.

Methodology: We propose that state performance on each statewide data indicator be assessed using a multi-level (i.e., hierarchical) model that risk-adjusts for select child- and state-level characteristics. Multi-level modeling is a widely accepted statistical method that enables fair evaluation of relative performance among states with different case mixes. The technique calculates how much variance in performance is due to (1) children's individual risk factors; (2) random measurement errors (due to modest sample sizes); and (3) the state's long-run ability to achieve the desired outcome.

Risk Adjustment: The model we propose to use will incorporate some risk adjustment. By incorporating risk adjustment, the multi-level model takes into account and controls for factors that differ across the states and that can affect outcomes regardless of the quality of services the state provides. The goal of risk adjustment is to minimize differences in outcomes that are due to factors over which states have little control, such as the age of children coming into foster care. For example, for the statewide data indicator of permanency in 12 months for children entering foster care, a state may discharge 40% of its children to permanency by 12 months. Forty percent is the state's observed performance, and is simply the number of children discharged to permanency by 12 months divided by the number of children eligible for such an exit. But this state's risk-adjusted performance might be 45%. That the state's risk-adjusted performance is higher than its observed performance means permanency was achieved for more children than expected, given the state's case mix and how other states, on average, performed with a similar case mix.

The Children's Bureau will finalize risk adjustment variables after receiving public comments on this document. The contractor's recommendations to us based on feedback from the expert panel convened in 2013 support the use of risk adjustment. The Children's Bureau's consideration of particular risk-adjustment variables will be based initially on the research literature, recommendations based on feedback from the expert panel and expert consultants, and the availability of data. The Children's Bureau will test proposed variables and retain only those variables that have a statistically significant relationship to the outcome for each statewide data indicator. For example, the Children's Bureau has tested for possible inclusion child's age, sex, number of prior removals, and interactions among these with the proposed statewide data indicators. The Children's Bureau will consider risk-adjustment variables at both the level of the child (e.g., age at entry) and the state (e.g., foster care entry rate).

Assessing State Performance

To assess state performance, the Children's Bureau proposes to estimate each state's risk-adjusted performance and the corresponding 95% interval estimate. The Children's Bureau can be 95% confident that a state's true performance lies somewhere between the lower and upper limit of this interval. This interval also provides a way to judge whether a state's performance is above or below the national average in a statistically meaningful way.

The Children's Bureau plans to compare each state's interval estimate to the national observed performance, and assign each state to one of three groups:

• “No different than national performance” if the 95% interval estimate surrounding the state's risk-adjusted performance includes the national observed performance.
• “Higher than national performance” if the entire 95% interval estimate surrounding the state's risk-adjusted performance is higher than the national observed performance.
• “Lower than national performance” if the entire 95% interval estimate surrounding the state's risk-adjusted performance is lower than the national observed performance.

Whether it is desirable for a state to be higher or lower than the national performance depends on the indicator. For the two permanency measures, a higher value is more desirable; for the remaining measures, a lower value is desirable.

The methodology described above is similar to that used by the Centers for Medicare & Medicaid Services to measure hospital performance as part of its Hospital Inpatient Quality Reporting program.^[9] The methodology is also consistent with the use of such models in education and health care to distinguish statistically high- and low-performing schools and hospitals.^[10]

Data

Data Profiles: The Children's Bureau will provide data profiles of state performance to each state before the state's CFSR on the statewide data indicators and other contextual data available from AFCARS and NCANDS. This data profile will assist the state to develop its statewide assessment. In addition, the Children's Bureau will provide data profiles semi-annually to assist states in measuring progress toward the goals identified in the program improvement plan.

Data Quality: Setting national standards and measuring state performance on statewide data indicators for CFSR purposes relies upon the states submitting high-quality data to AFCARS and NCANDS. The Children's Bureau has provided states with consultation and technical assistance before and throughout both rounds of reviews to address data quality issues. Additionally, the Children's Bureau has provided states with tools for AFCARS and NCANDS that allow the agency to examine its data for accuracy and encourages states to incorporate these in their ongoing quality assurance process to review data.

During the first two rounds of the CFSR, there have been occasions in which the Children's Bureau cannot use a state's data in aggregate calculations of the national standard. In isolated circumstances, these data quality issues have been significant enough to prevent us from relying on the state's data as an accurate assessment of its performance on a statewide data indicator. The Children's Bureau would like to be clear about the level of data quality issues that prevent state data from being used for CFSR purposes as described below.

Data Quality: Excluding States From National Standards or State Performance

We analyzed every data element from AFCARS and NCANDS that is relevant to each statewide data indicator (as listed in Attachment C) and performed data quality checks across files both over time as well as between files.^[11] Examples of these checks included looking for the presence of the same child identifier in the AFCARS and NCANDS file and reviewing for consistent reporting of a child in AFCARS from the time the child entered foster through discharge and with an associated reason for discharge. This analysis revealed the scope of data quality issues present in current AFCARS and NCANDS submissions.

Based on this analysis, we developed thresholds to identify data quality concerns and either accept or exclude the files when calculating national standards and state performance. For those data quality issues that are contained to one data file submission, we will consider a threshold of 5%. In other words, any state that has more than 5% of data missing or invalid ^[12] will be excluded from the model used to calculate the national standard (i.e., the national observed performance) and estimate states' risk-adjusted performance. For cross-file checks, we are setting a higher threshold for exclusion. For the maltreatment in foster care measure, a state will be excluded from the national standards calculation and performance estimate if more than 10% of NCANDS victims are missing an AFCARS identifier. For the permanency indicators, a state will be excluded if it has more than 10% of dropped cases across two six month file submissions.

Data Quality: Case-Level Exclusions

For those states that do not exceed the data quality thresholds but still have identified data quality problems, we will include the state in national standards calculations and measure state performance but we will exclude child-level records with missing or invalid data on elements needed to determine the child's outcome and perform the risk-adjustment. For example, if the risk-adjustment for an indicator includes age at entry, a child whose age at entry cannot be determined (due to a missing date of birth) will not be include in the analysis. We believe this exclusion will result in more accurate estimates of performance for those states with minor data quality issues. For each indicator, the Children's Bureau will provide each state with a list of records that were excluded from the analyses.

Program Improvement Plans

States that fall below the national standard on any given indicator will be required to include that indicator in a program improvement plan. Regardless of which indicators a state is required to include in its program improvement plan, the Children's Bureau will provide each state with a data profile that includes information on the state's performance on all of the statewide data indicators.

Companion Measures: For two of the statewide data indicators, permanency in 12 months for children entering foster care and re-entry to foster care, the Children's Bureau proposes to consider performance for program improvement purposes in concert with the other indicator. This means that if a state has a program improvement plan that includes improving on the indicator permanency in 12 months for children entering foster care, the Children's Bureau's determination of whether the state has improved successfully will take into consideration its performance on the re-entry to foster care indicator as a companion measure. Specifically, the state must stay above a threshold for the companion re-entry to foster care indicator as well as achieve its goal on the permanency in 12 months for children entering foster care indicator, to successfully complete the program improvement plan. The reverse is also true. For details about threshold calculations, please see the section below. If a state must improve on the re-entry to foster care indicator in its program improvement plan, it must also include the permanency in 12 months for children entering foster care indicator as a companion measure. Although the Children's Bureau believes that it is important to reunify children with their families as quickly as possible, we also believe that children should be reunified when safe and appropriate and sufficient supports are in place to prevent a subsequent removal. The recommendations to us based on the expert panel convened in 2013 also support the use of companion measures in program improvement.

Methods for Setting State Performance Baselines, Goals, and Thresholds: The key components for setting performance goals and monitoring progress over the course of a program improvement plan involve calculating baselines, goals, and thresholds. The Children's Bureau methodology for specifying state improvement goals is statistically grounded, producing goals that are programmatically challenging, yet attainable. In addition, goals and thresholds should reflect each state's own performance history and demonstrated capacity for change. In CFSR Round 2, the percent of improvement required for each state was the same for all states, and was applied to state-specific baselines. While this standardization was easy to communicate how it applied to each state, it also meant that high-performing states were asked to make larger absolute improvements than lower-performing states. This approach did not take into account the variability in performance that states have shown over time. For CFSR Round 3, the Children's Bureau plans to set improvement goals relative to each state's past performance.

Setting Baselines: The Children's Bureau plans to set the baseline for each statewide data indicator included in a program improvement plan at the state's observed performance on that indicator for the most recent year of data available before the beginning of program improvement plan implementation. We also considered using the state's three-year average in this calculation, but that approach occasionally produced goals that fell below the state's performance in the most recent year. Consistent with CFSR Round 2, we will use the most recent year as the baseline, because it represents a reasonable estimate of the state's current performance.

Setting Goals and Thresholds: We plan to set performance goals and thresholds will be based on the variability in the state's observed performance in the three most recent years of data. There will be improvement factors, as in CFSR Round 2, but these will be driven by the variability in performance that the particular state has shown in the last three years. We will apply the improvement factors to the baseline to produce the concrete performance goal. The state's amount of improvement required for a program improvement plan will be more than what is likely, in a statistical sense, to occur by chance alone. Conversely, we plan to set thresholds as the inverse of goals, which will identify a point by which a state is demonstrating a performance decline for companion measures that is more statistically than what might be expected by chance.

We will use a technique called bootstrapping to develop goals and thresholds. The method calls for the Children's Bureau to repeatedly sample a state's past three years of performance estimates to construct a larger sample, and from that the calculation of a grand mean and standard deviation. The grand mean reflects that state's “average” performance and the standard deviation reflects how much normal fluctuation in performance the Children's Bureau might expect for that state, given its past performance. Then the standard deviation is used to calculate an estimate that would represent a level of change above and beyond the typical fluctuation that would otherwise be expected. The Children's Bureau will set the magnitude at four standard deviations from the grand mean. At that level we can say with confidence that—if we were to randomly estimate a state's performance on the indicator (using their past performance), and did so 100 times—we would expect to see performance at this level less than 6% of the time (or fewer than 6 times out of 100). Six times out of 100 is rather rare, which is why we can treat it as representing a statistically meaningful change has occurred in the program.^[13]

To determine exactly how much a state will need to improve, we must first calculate an improvement factor, which is the percentage difference between the grand mean and four standard deviations above the grand mean. We then apply that to the baseline, which is the observed performance in the most recent year. To demonstrate a sample calculation:

A state may have a grand mean of 50%, a grand mean plus four standard deviations = 52%, and a year 3 value of 51. This will give us an improvement factor of 52/50 = 1.04. If that is applied to the baseline of 51%, the program improvement plan goal will be 51% × 1.04 = 53.04%.

We will use a comparable technique to set thresholds for companion measures, subtracting rather than adding four standard deviations to the grand mean (when higher performance on an indicator is better), which can be used to identify a state's decline in performance. To provide an example, if a goal was calculated to be three percentage points higher than the baseline percent, the threshold would be three percentage points below it. Thresholds are only relevant to companion measures.

By design, states with less variation in performance from year to year have more modest goals, while those showing greater variation have more aggressive goals. Overall, we believe that the goals are reflective of each state's own prior experience and performance levels, with goals that are achievable and substantively meaningful. We acknowledge that a few states with the lowest variation in performance for each indicator are assigned very modest goals, while a few states with the highest variation in performance are assigned very aggressive goals.

To address these problems at the extreme ends, we propose to establish minimum and maximum improvement factors. Specifically, a floor will be set at the top of the bottom fifth, and the bottom of the top fifth, ordered by the size of the improvement factor. While the impact of this rule varies somewhat from indicator to indicator, overall we believe it provides a consistent basis for producing goals that are achievable and substantively meaningful. The inverse would be done for the thresholds. The contractor's recommendations to us based on the feedback from the expert panel convened in 2013 support the setting of maximum and minimum thresholds for improvement goals at the level of performance of top and bottom quintiles.

Process for Final Indicators

We are interested in comments on all aspects of the statewide data indicators proposed and the methods to calculate national standards and program improvement. After considering the feedback to this docuemnt, we plan to publish a final list of indicators and methods that will be used in the CFSRs along with the actual national standards.

(Authority: 42 U.S.C. 1320a-1a; 45 CFR 1355.31-37.)

Mark Greenberg,

Acting Commissioner, Administration on Children, Youth and Families.

Attachment A: Proposed Statewide Data Indicators

Category	Measure title	Measure description	Denominator	Numerator	Exclusions	Notes
Safety	Maltreatment in foster care	Of all children in foster care during a 12-month period, what is the rate of victimization per day of foster care?	(From AFCARS) Among children in foster care during a 12-month period, total number of days these children were in foster care as of the end of the 12-month period a	(From NCANDS) Among children in the denominator, total number of substantiated or indicated reports of maltreatment (by any perpetrator) during the 12-month period b	—If a state provides incident dates, records with an incident date occurring before the date of removal will be excluded —Children in foster care <8 days —Any report that occurs within the first 7 days of removal	Cases are matched across AFCARS and NCANDS using AFCARS ID.
Safety	Re-report of maltreatment	Of all children with a screened-in report of alleged maltreatment in a 12-month period (regardless of disposition), what percent had another screened-in report within 12 months of their initial report?	(From NCANDS) Number of children with at least one screened-in report of alleged maltreatment in a 12-month period	(From NCANDS) Number of children in the denominator that had another screened-in report within 12 months of their initial report	None	Reports are included regardless of the type of disposition, so this indicator includes both victims and non-victims.
Permanency	Permanency in 12 months for children entering foster care	Of all children who enter foster care in a 12-month period, what percent discharged to permanency within 12 months of entering foster care? c	(From AFCARS) Number of children who enter foster care in a 12-month period	(From AFCARS) Number of children in the denominator who discharged to permanency within 12 months of entering foster care or by the time they reached 18	—Children in foster care <8 days —Children who enter foster care at age 18 or more	Youth who turn 18 while in foster care who were included in the denominator will not be counted as having achieved permanency, regardless of discharge reason.
Permanency	Permanency in 12 months for children in foster care 2 years or more	Of all children in foster care on the first day of a 12-month period, who had been in foster care (in that episode) for 2 or more years, what percent discharged to permanency within 12 months of the first day?	(From AFCARS) Number of children in foster care on the first day of a 12-month period who had been in foster care (in that episode) for 2 or more years	(From AFCARS) Number of children in the denominator who discharged to permanency within 12 months of the 1st day or by the time they reached 18	—Children age 18 or more on the first day of the year	Youth who turn 18 while in foster care who were included in the denominator will not be counted as having achieved permanency, regardless of discharge reason.
Permanency	Re-entry to foster care	Of all children who enter foster care in a 12-month period who discharged within 12 months to reunification, live with relative, or guardianship, what percent re-entered foster care within 12 months of their discharge? a	(From AFCARS) Number of children who enter foster care in a 12-month period who discharged within 12 months to reunification, live with relative, or guardianship	(From AFCARS) Number of children in the denominator who re-enter foster care within 12 months of their discharge	—Children in foster care < 8 days —Children who enter or exit foster care at age 18 or more	If a child has multiple re-entries to foster care within 12 months of their discharge, only his first re-entry is selected.
Permanency	Placement Stability	Of all children who enter foster care in a 12-month period, what is the rate of placement moves per day of foster care?	(From AFCARS) Among children who enter foster care in a 12-month period, total number of days these children were in foster care as of the end of the 12-month period d	(From AFCARS) Among children in the denominator, total number of placement moves during the 12-month period e	—Children in foster care < 8 days —Children who enter foster care at age 18 or more	The initial removal from home (and into foster care) is not counted as a placement move.
a For example, if during the 12-month period there were two children in foster care, one child for 10 days (1st episode), the same child for 40 days (2nd episode), and the other child for 100 days (his only episode), the denominator would = 150 days (10+40+100).
b For example, if during the 12-month period there were two children in foster care, and one child with 3 substantiated or indicated reports and the other with 1 such report, the numerator would = 4 reports (3+1).
c If a child has multiple entries during the year, only his or her first entry is selected.
d For example, if during the 12-month period two children entered foster care, one child for 10 days and the other child for 100 days, the denominator would be 110 days (10+100).
e For example, if during the 12-month period two children entered foster care, and one child had 3 moves and the other had 1 move, the numerator would = 4 moves (3+1).

Attachment B: Comparison of Data Measures—CFSR Round 2 and Round 3

Category	Measure title	Proposed CFSR round 3 indicator	Comparable CFSR round 2 measure	How and why it's changed
Safety	Maltreatment in foster care	Of all children in foster care during a 12-month period, what is the rate of victimization per day a of foster care?	Of all children in foster care during the reporting period, what percent were not victims of substantiated or indicated maltreatment by a foster parent or facility staff member?	In the CFSR 2 measure, counts of children not maltreated in foster care are derived by subtracting the NCANDS count of children maltreated by foster care providers from the total count of all children placed in foster care, as reported in AFCARS. Because of improved reporting by states, we now link AFCARS and NCANDS data using the child ID and determine if maltreatment occurred during a foster care episode, improving accuracy on the indicator. This also allows us to expand the measure to include all types of perpetrators (including, for example, parents) under the assumption that states should be held accountable for keeping children safe from harm while in the care of the state, no matter who the perpetrator is.
Safety	Re-report of maltreatment	Of all children with a screened-in report of alleged maltreatment in a 12-month period, what percent had another screened-in report within 12 months of their initial report?	Of all children who were victims of substantiated or indicated maltreatment allegation during the first 6 months of the reporting period, what percent were not victims of another substantiated or indicated maltreatment allegation within a 6-month period?	We are expanding the measure to count all children with screened-in reports of alleged maltreatment, because research suggests children with prior reports are at greater risk. In addition, by limiting only to victims, we could face measurement challenges as states implement Differential Response during a monitoring cycle, which could have an impact on numbers of substantiations. We also propose using a full 12-month period rather than only 6 months to capture the denominator, to create more stable estimates.
Permanency	Permanency in 12 months for children entering foster care	Of all children who enter foster care in a 12-month period, what percent discharged to permanency within 12 months of entering foster care?	Composite 1.3: Of all children entering foster care for the first time in a 6-month period, what percent discharged to reunification (or live with relative) within 12 months of entering foster care or by the time they reached 18?	We now count all types of permanency (reunification, live with relative, adoption or guardianship) as having `met' the indicator. We also expanded the measure to include all children who entered foster care that year; not just those on their first removal episode. We also expanded the window of time for the entry cohort to a full year instead of 6 months; this will yield more stable estimates.
Permanency	Permanency in 12 months for children in foster care for 2 years or more	Of all children in foster care on the first day of a 12-month period who had been in foster care (in that episode) for 2 or more years, what percent discharged to permanency within 12 months of the first day?	Composite 3.1: Of all children in foster care on the first day of a 12-month period who had been in foster care (in that episode) for 2 or more years, what percent discharged to permanency within 12 months of the first day or by the time they reached 18?	Same measure; no change. The difference is that it is now evaluated on its own, rather than as just one part of a composite measure. By including the Entry Cohort Permanency indicator (listed above) as well as the Legacy Cohort Permanency indicator, we hold states accountable not only for children in their first year of foster care, but also those children/youth who have been in foster care for long periods of time.
Permanency	Re-entry in 12 months	Of all children who enter foster care in a 12-month period and discharged within 12 months to reunification, live with relative, or guardianship, what percent re-entered foster care within 12 months of their date of discharge?	Composite 1.4: Of all children discharged from foster care to reunification or live with a relative in a 12-month period, what percent re-entered foster care in less than 12 months from the date of discharge?	The new indicator is limited to those children who entered foster care during the year, whereas the CFSR Round 2 measure counted all children who discharged to reunification or live with relative, regardless of when they entered foster care. The purpose of this focus is in keeping with the rationale that new interventions may best be monitored in an entry cohort. This indicator will also be used as a companion measure with permanency in 12 months, to ensure that states working to improve permanency rates in their entry cohort do not see worsening performance on rates of re-entry to foster care. We also expanded the denominator to allow discharges to guardianship, in an effort to capture more discharges to permanency. Exits to adoption are not included because they cannot be tracked reliably, as some states issue new child identifiers if a child who was previously adopted enters foster care.
Permanency	Placement stability	Of all children who enter foster care in a 12-month period, what is the rate of placement moves per day b of foster care?	Composite 4.1: Of all children served in foster care during the 12-month period, what percent had two or fewer placement settings?	The proposed indicator controls for length of time in foster care, so we are looking at moves per day of foster care, rather than children as the unit of analysis. The rationale for using an entry cohort rather than all children served is that our analysis shows children entering foster care tend to move much more than those children/youth in foster care for longer periods of time, whose placements may have stabilized. In CFSR Round 2 measure, moves that took place prior to the monitoring period were counted. Now we only count those moves that occur during the monitoring period. The initial placement is not counted. The CFSR Round 2 measure treated children who moved 2 times in an episode the same as children who moved 15 times; both were a failure to meet the measure. The new indicator counts each move, so it continues to hold states accountable for those children/youth who have already moved several times.
a The rate may be expressed per 100,000 days because it is such a rare event. Using this metric gives us numbers greater than zero, which are easier to communicate.
b The rate is expressed per 1,000 days to convert the rate to a metric that gives us numbers greater than zero.

Attachment C: Data Elements Required for Statewide Data Indicators

For instruction regarding AFCARS data elements, refer to http://www.acf.hhs.gov/​programs/​cb/​resource/​afcars-tb1.

For instruction with regard to NCANDS data elements, refer to http://www.ndacan.cornell.edu/​datasets/​pdfs_​user_​guides/​178-NCANDS-child2012v1-User-Guide-and-Codebook.pdf.

Primary data elements required for calculation	Permanency by 12 months	Re-entry by 12 months	Placement stability	Re-report of maltreatment	Maltreatment in foster care
AFCARS FC Element #1: a Title IV-E Agency	✓	✓	✓	NA	✓
AFCARS FC Element #4: Record Number	✓	✓	✓	NA	✓
AFCARS FC Element #21: Date of Latest Removal	✓	✓	✓	NA	✓
AFCARS FC Element #23: Date of Placement in Current Foster Care Setting	NA	NA	✓	NA	NA
AFCARS FC Element #24: Number of Placement Settings during this Removal Episode	NA	NA	✓	NA	NA
AFCARS FC Element #56: Date of Discharge from FC	✓	✓	✓	NA	✓
AFCARS FC Element #58: Reason for Discharge	✓	✓	NA	NA	NA
NCANDS CF Element #4: Child ID	NA	NA	NA	✓	NA
NCANDS CF Element #6: Report Date	NA	NA	NA	✓	✓
NCANDS CF Element #27: Child Maltreatment 1—Disposition Level b	NA	NA	NA	NA	✓
NCANDS CF Element #29: Child Maltreatment 2—Disposition Level	NA	NA	NA	NA	✓
NCANDS CF Element #31: Child Maltreatment 3—Disposition Level	NA	NA	NA	NA	✓
NCANDS CF Element #33: Child Maltreatment 4—Disposition Level	NA	NA	NA	NA	✓
NCANDS CF Element #34: Maltreatment death	NA	NA	NA	NA	✓
NCANDS CF Element #145: AFCARS ID	NA	NA	NA	NA	✓
a The elements are numbered by their position in the flat ASCII files submitted by states to these reporting systems. These numbering schema are specific to the files utilized by ACYF. Files obtained through the National Data Archive on Child Abuse and Neglect (NDACAN) may have a slightly different order.
b Definition of `victim' includes all children with a disposition level (for any of up to four maltreatments per child) of: a) Substantiated, or b) Indicated. These do not propose including differential response victims. Victims also include children who died as a result of maltreatment.

Additional data elements required for risk-adjusted Analysis c	Permanency by 12 months	Re-entry by 12 months	Placement stability	Re-report of maltreatment	Maltreatment in foster care
AFCARS FC Element #6: Child's Date of Birth	✓	✓	✓	NA	✓
AFCARS FC Element #7: Child Sex	✓	✓	✓	NA	✓
AFCARS FC Element #19: Total # of Removals	✓	✓	✓	NA	✓
NCANDS CF Element #14: Child Age	NA	NA	NA	✓	NA
NCANDS CF Element #17: Child Sex	NA	NA	NA	✓	NA
US Census Bureau: Child Population, by State	✓	✓	NA	✓	✓
c In addition to those data elements used for risk adjustment, a few more are used to make necessary adjustments to outcomes; for example, we use the child's current placement setting to determine if he or she is in Trial Home Visit before Reunification and, if so, time in foster care is adjusted consistent with adjustments for trial home visits used in CFSR Round 2. If a state provides NCANDS CF Element #146, Incident Date, an adjustment will be made to the maltreatment in foster care indicator to improve accuracy.

Footnotes

1.  AFCARS collects case-level information from state and Tribal title IV-E agencies on all children in foster care and those who have been adopted with title IV-E agency involvement. Title IV-E agencies must submit AFCARS data to the Children's Bureau twice a year.

2.  NCANDS collects child-level information on every child who receives a response from a child protective services agency due to an allegation of abuse or neglect. States report this data to the Children's Bureau voluntarily. In FFY 2013, all 50 states, the District of Columbia, and Puerto Rico submitted NCANDS data.

3.  More detailed information on the CFSR Round 2 measures can be found on the Children's Bureau Web site at http://www.acf.hhs.gov/​sites/​default/​files/​cb/​data_​indicators_​for_​the_​second_​round_​of_​cfsrs.pdf.

4.  See Testa, M., Poertner, J. Koh, E. (2008). Can AFCARS be rescued? Fixing the statistical yardstick that measures state child welfare performance. Urbana, IL: CFRC; and Testa, M. & Poertner, J. (Eds.). (2010). Fostering accountability: Using evidence to guide and improve child welfare policy. Oxford: Oxford University Press.

5.  Rates are calculated per day of foster care. However, we will multiply the rate by 100,000 to produce larger and more meaningful numbers.

6.  Almost all states report AFCARS identifiers in the NCANDS data. For those states that do not, a Children's Bureau-approved alternate source will be required to assess performance on this indicator.

7.  See Drake, Jonson-Reid, Way, & Chung (2003). Substantiation and Recidivism. Child Maltreatment. Vol. 8, No. 4, 248-260.

8.  Rates are calculated per day of foster care. However, we will multiply the rate by 1,000 to produce larger and more meaningful numbers.

9.  Yale New Haven Health Services Corporation, & Center for Outcomes Research & Evaluation. (2013). 2013 Measures Updates and Specifications Report: Hospital-Level 30-Day Risk-Standardized Readmission Measures for Acute Myocardial Infarction, Heart Failure, and Pneumonia (Version 6.0). Retrieved from http://www.qualitynet.org/​dcs/​ContentServer?​c=​Page&​pagename=​QnetPublic/​Page/​QnetTier4&​cid=​1219069855841.

10.  See Goldstein & Spiegelhalter (2007). League Tables and Their Limitations. 159(3), 385-443; Normand & Shahian (2007). Statistical and Clinical Aspects of Hospital Outcomes Profiling. Statistical Science, 22(2), 206-226.

11.  This data quality analysis is specific to indicators proposed for CFSR Round 3. It is a separate analysis from the standard data quality checks or validation that are performed when states submit their AFCARS or NCANDS data.

12.  For example, the date of birth is greater than the date of latest removal.

13.  These guidelines are based on a theorem known as Chebyshev's inequality. When all you have is an overall mean and standard deviation (which is what we have for each state's performance), the theorem guarantees that a certain percent of values will be k standard deviations away from the mean. Specifically, at least 75% of the values will be within two standard deviations of the mean, at least 89% within three standard deviations, at least 94% within four standard deviations, at least 96% within five standard deviations, and at least 97% within six standard deviations (Chong et al. (2012). Improving Generalization Performance in Co-Evolutionary Learning. IEEE, vol. 16, no. 1, 70-85; Sheppard (2011). Environmental Study-Soil Sample Analysis for the Department of Ecology at Hanford).