Disclaimer
The data used in this publication, Dataset 277, AFCARS Foster Care File, 6-month periods (FY2016A – 2022B), were obtained from the National Data Archive on Child Abuse and Neglect and have been used in accordance with its Terms of Use Agreement licence. The Administration on Children, Youth and Families, the Children’s Bureau, the original dataset collection personnel or funding source, NDACAN, Cornell University and their agents or employees bear no responsibility for the analyses or interpretations presented here.
The information provided in this report is published “as is” and without any warranties, whether express or implied, including but not limited to warranties of merchantability and fitness for a particular purpose. The author(s) do not warrant or assume any legal liability or responsibility for the accuracy, completeness, or usefulness of any information disclosed.
1. Summary
This study investigates the factors influencing the likelihood of aging out of foster care in a jurisdiction in the western US using the Adoption and Foster Care Analysis and Reporting System (AFCARS) dataset. The analysis focuses on a cohort of children aged 12-14 at the start of the 2016 Fiscal Year, tracking their outcomes through the 2022 Fiscal Year. The study employs descriptive statistics, correspondence analysis, and logistic regression to identify key predictors and quantify their effects.
Key findings include:
- Children with emotional disturbances and those removed due to caretaker inability to cope or relinquishment have higher probabilities of aging out.
- Title IV-E foster care payments are associated with a reduced likelihood of aging out.
- Placement settings significantly impact outcomes, with children in group homes or runaway situations facing higher risks of aging out.
2. Introduction
Youth who leave foster care without achieving permanency by age 18 face significant challenges during their transition to adulthood (Dworsky, Napolitano, and Courtney 2013; Fowler et al. 2017). Identifying the factors that influence the likelihood of aging out of foster care is crucial for developing targeted interventions and support systems to mitigate these difficulties. This study uses the Adoption and Foster Care Analysis and Reporting System (AFCARS) dataset to investigate these factors in a jurisdiction in the western US, aiming to inform efforts to enhance support systems and improve outcomes for these individuals.
The AFCARS dataset provides comprehensive information on children in foster care and those adopted through state public child welfare agencies. It includes demographic details, foster care placements, service goals, adoption readiness, entry and exit dates, funding sources, and information on both biological and foster parents. This rich dataset allows for a detailed analysis of the factors influencing aging out, helping to identify areas where interventions can be most effective.
In this study, we focus on a cohort of children from foster care system in the jurisdiction in the western US who were aged 12-14 at the start of the 2016 Fiscal Year and were reported in March 2016. We track whether these children aged out of foster care in any subsequent 6-month period up to the 2022 fiscal year. By analyzing this cohort, we aim to uncover the key predictors of aging out and provide insights into how these factors can be addressed to support youth in foster care.
Our methodology involves a combination of descriptive statistics, correspondence analysis, and logistic regression modeling. We start by examining the descriptive statistics to compare characteristics of children who aged out versus those who did not. Correspondence analysis is then used to visualize the relationships between categorical variables and identify key associations. Finally, we employ logistic regression to quantify the effects of various predictors on the likelihood of aging out.
3. Data
The Adoption and Foster Care Analysis and Reporting System (AFCARS) is a federally required data collection system that provides detailed information on children in foster care and those adopted through state public child welfare agencies (Children’s Bureau Administration on Children, Youth and Families 2023). This dataset contains child demographic details, data on previous foster care placements, service goals, adoption readiness, entry and exit dates, funding sources, and information on both biological and foster parents. Data is gathered semi-annually and disseminated by the National Data Archive on Child Abuse and Neglect (NDACAN), covering all U.S. states, the District of Columbia, and Puerto Rico.
3.1 Data Preprocessing
The AFCARS dataset, originally in SPSS format, was loaded into R for preprocessing. The SPSS-stored dates, represented as the number of seconds since October 14, 1582, were converted into R Date objects. Character variables were cleaned by removing leading and trailing whitespace, and records with empty record number (RecNumbr) fields were excluded.
3.2 Study Cohort
The study cohort comprises children from foster care system in the jurisdiction in the western US who were aged 12-14 at the start of the 2016 Fiscal Year and were reported in March 2016 (N = 7,653). By the end of the study period in 2022, all individuals in this cohort are 18 years or older.
3.3 Target
Following Ahn, Gil, and Putnam-Hornstein (2021), we frame the task of predicting aging out of foster care as a binary classification problem. The objective is to determine whether a child will age out (AgedOut = 'Yes') or not (AgedOut = 'No') during the study period. To construct our binary target variable, we track whether the children in the study cohort aged out of foster care in any subsequent 6-month period up to the 2022 fiscal year.
3.4 Covariates
The covariates for the analysis were carefully selected to ensure relevance and accuracy. Age at first removal (AgeAtFirstRem) was derived from the difference between the first removal date (Rem1Dt) and the birth date (DOB). Variables with a substantial number of missing values were excluded. Missing values for the number of placements (NumPlep) were replaced with the most common value (1). Redundant race indicators, uninformative metadata, highly correlated variables, and variables with fewer than 20 observations were removed. Additionally, all variables that could leak the target variable were eliminated. For example, the variable Exited represent whether a child exited the foster care system, which directly indicates the target variable AgedOut. By removing such variables, we ensured that the model was trained on a clean dataset with no data leakage.
3.5 Descriptive Statistics
Table 1 presents a comparison of chosen covariates, categorized by whether they aged out of the system or not. The “Aged Out” column is divided into “No” (N = 5,939) and “Yes” (N = 1,714) indicating the number and percentage of children who did not age out versus those who did.
Table 1. Descriptive Statistics
| Charateristic | Description | Aged Out N = 5,939 | Did not Age Out N = 1,714 |
|---|---|---|---|
| Sex | Child Sex | ||
| Male | 3,127 (53%) | 855 (50%) | |
| Female | 2,812 (47%) | 859 (50%) | |
| VisHear | Visually Or Hearing Impaired | 737 (12%) | 737 (12%) |
| PhyDis | Physically Disabled | 18 (0.3%) | 10 (0.6%) |
| EmotDist | Emotionally Disturbed | 900 (15%) | 399 (23%) |
| OtherMed | Other Diagnosed Condition | 1,087 (18%) | 389 (23%) |
| TotalRem | Total Number Of Removals | ||
| 1 | 4,027 (68%) | 1,057 (62%) | |
| 2 | 1,397 (24%) | 455 (27%) | |
| 3 | 392 (6.6%) | 150 (8.8%) | |
| >= 4 | 123 (2.1%) | 52 (3.0%) | |
| NumPlep | Placement Settings in Current FC Episode | ||
| 1 | 2,262 (38%) | 418 (24%) | |
| 2-5 | 3,022 (51%) | 937 (55%) | |
| 6-10 | 503 (8.5%) | 263 (15%) | |
| >= 11 | 152 (2.6%) | 96 (5.6%) | |
| ManRem | Removal Manner | ||
| Voluntary | 106 (1.8%) | 34 (2.0%) | |
| Court ordered | 5,499 (93%) | 1,603 (94%) | |
| Not yet determined | 334 (5.6%) | 77 (4.5%) | |
| PhyAbuse | Removal Reason-Physical Abuse | 730 (12%) | 250 (15%) |
| SexAbuse | Removal Reason-Sexual Abuse | 294 (5.0%) | 87 (5.1%) |
| Neglect | Removal Reason-Neglect | 4,327 (73%) | 1,173 (68%) |
| AAParent | Removal Reason-Alcohol Abuse Parent | 161 (2.7%) | 31 (1.8%) |
| DAParent | Removal Reason-Drug Abuse Parent | 600 (10%) | 117 (6.8%) |
| DAChild | Removal Reason-Drug Abuse Child | 14 (0.2%) | 8 (0.5%) |
| ChilDis | Removal Reason-Child Disability | 13 (0.2%) | 14 (0.8%) |
| ChBehPrb | Removal Reason-Child Behavior Problem | 359 (6.0%) | 88 (5.1%) |
| PrtsDied | Removal Reason-Parent Death | 30 (0.5%) | 15 (0.9%) |
| PrtsJail | Removal Reason-Parent Incarceration | 143 (2.4%) | 32 (1.9%) |
| NoCope | Removal Reason-Caretaker Inability Cope | 1,964 (33%) | 659 (38%) |
| Abandmnt | Removal Reason-Abandonment | 54 (0.9%) | 15 (0.9%) |
| Relinqsh | Removal Reason-Relinquishment | 25 (0.4%) | 22 (1.3%) |
| Housing | Removal Reason-Inadequate Housing | 254 (4.3%) | 72 (4.2%) |
| CurPlSet | Current Placement Setting | ||
| Foster home, relative | 1,936 (33%) | 240 (14%) | |
| Pre-adoptive home | 259 (4.4%) | 0 (0%) | |
| Foster home, non-relative | 2,666 (45%) | 955 (56%) | |
| Group home | 521 (8.8%) | 292 (17%) | |
| Institution | 394 (6.6%) | 181 (11%) | |
| Supervised independent living | 2 (<0.1%) | 1 (<0.1%) | |
| Runaway | 50 (0.8%) | 33 (1.9%) | |
| Trial home visit | 111 (1.9%) | 12 (0.7%) | |
| PlaceOut | Out Of State Placement | 139 (2.3%) | 20 (1.2%) |
| IVEFC | Title IV-E Foster Care Payments | 2,537 (43%) | 575 (34%) |
| IVEAA | Title IV-E Adoption Assistance 157 (2.6%) | 157 (2.6%) | 5 (0.3%) |
| IVDCHSUP | Title IV-D Child Support Funds | 460 (7.7%) | 147 (8.6%) |
| XIXMEDCD | Title XIX Medicaid | 5,842 (98%) | 1,694 (99%) |
| SSIOther | SSI Or Social Security Act Benefits | 236 (4.0%) | 85 (5.0%) |
| IsTPR | Parents have relinquished parental rights | 637 (11%) | 87 (5.1%) |
| Race | Derived Race Variable | ||
| White | 1,853 (31%) | 487 (28%) | |
| Black or African American | 1,226 (21%) | 419 (24%) | |
| American Indian or Alaska Native | 102 (1.7%) | 32 (1.9%) | |
| Asian | 121 (2.0%) | 29 (1.7%) | |
| Hawaiian or Other Pacific Islander | 18 (0.3%) | 4 (0.2%) | |
| More Than One Race | 354 (6.0%) | 124 (7.2%) | |
| Race Unknown | 2,265 (38%) | 619 (36%) | |
| AgeAtFirstRem | Age at First Removal | ||
| 0 | 327 (5.5%) | 127 (7.4%) | |
| 1-5 | 998 (17%) | 372 (22%) | |
| 6-11 | 2,482 (42%) | 711 (41%) | |
| 12-14 | 2,132 (36%) | 504 (29%) |
4. Multiple Correspondence Analysis
Multiple Correspondence Analysis (MCA) is a multivariate statistical technique particularly useful for exploring relationships between categorical variables (Husson, Lê, and Pagès 2017; Lê, Josse, and Husson 2008). Given the categorical nature of many variables in the AFCARS dataset, MCA offers valuable insights into associations that might not be immediately apparent through other methods. The technique helps visualize complex interrelationships within the data, making it easier to identify patterns and clusters of related variables.
MCA is primarily an exploratory tool, and its application in this study is justified by the need to understand the underlying structure of categorical data. This initial exploration is crucial for identifying potential predictors and guiding subsequent, more formal analyses like logistic regression. By using MCA, the study can uncover significant associations and trends that warrant further investigation.
Limitations:
- Exploratory Nature: MCA does not provide definitive conclusions but rather suggests associations that need further validation through other methods.
- Dimensionality Reduction: The reduction of complex data into two or three dimensions may result in the loss of some information.
- Interpretation Challenges: Interpreting the dimensions and the placement of points in the biplot can be subjective and may require careful consideration of the context.
Assumptions:
- Categorical Data: MCA assumes that the data is categorical, and the relationships between categories are meaningful.
- Homogeneity of Data: The analysis presumes a certain level of homogeneity within the data, meaning that the categorical variables are comparable across different observations.
4.1 Results
Figure 1 presents a MCA biplot that visually represents the relationships between observations and covariates in a reduced dimensional space. The principal dimensions, Dim1 and Dim2, are plotted along the axes and capture the highest variation within the data.
- Principal Dimensions: Dim1 and Dim2 are the principal dimensions that capture the most variation in the data. Dim1 explains the largest portion of variability, followed by Dim2. These dimensions help to simplify complex multivariate data into two primary factors that can be more easily visualized.
- Circular Points: Each circle represents an individual observation. Red points denote children who did not age out of foster care, while blue points indicate children who did age out.
- Ellipses: These highlight the concentration and spread of points within each group, showing the distribution and variability of the observations.
- Proximity of Points: Points that are close to each other on the plot share similar characteristics, while those far apart are dissimilar.
- Black Triangles: These depict the categories of the covariates. Categories that appear close together on the plot tend to co-occur frequently in the dataset, whereas those that are far apart have a lower association or co-occurrence.
This visual representation helps in identifying clusters and associations within the data, providing insights into the relationships between various factors and the likelihood of aging out of foster care.
Figure 1: Multiple Correspondence Analysis
Dim1 and Dim2 capture 4.7% and 4.1% of the variance in data, respectively. Due to the large number of categories, we labeled only a subset of the most relevant ones based on our observations and omitted the rest for ease of readability. In particular we focus on two regions:
- Bottom right: we observe clusters of children who did not age out (red). These clusters are associated with “Pre-adoptive home” current placement and “IVEAA” (Title IV-E Adoption Assistance). These clusters represent children who are in the process of being adopted.
- Middle Top and Top right: we observe that this region has a higher proportion of children who aged out (blue). This region is associated with “Group Home” and “Institution” current placements, a high number of removals and placements, relinquishment, and medical conditions such as emotional disturbance and visual or hearing impairment.
4.2 Variable contributions
Figure 2 consists of two bar charts showing the contribution of variables to Dim-1 and Dim-2. The x-axis lists the variables, and the y-axis represents their contribution percentages.
- In Dim-1, “IsTPR,” “Pre-adoptive home,” and “IVEAA” have the highest contributions, indicating that Dim-1 primarily distinguishes between children who are in the process of being adopted and those who are not.
- Dim-2 includes a combination of different variables, making it challenging to interpret the contribution of individual variables clearly.
Figure 2: Variable Contributions
5. Regression Analysis
Logistic regression is a statistical method used to model the relationship between a binary dependent variable and one or more independent variables. It estimates the probability of an event occurring by fitting a logistic function to the data. Unlike linear regression, which predicts a continuous outcome, logistic regression predicts the likelihood of a particular outcome. The logistic function transforms the linear combination of the predictors into a value between 0 and 1, representing this probability. In this study, logistic regression was used to analyze factors influencing the likelihood of aging out of foster care, identifying key predictors and quantifying their effects on the probability of this outcome. It was chosen specifically for its ability to model binary outcomes and provide interpretable results.
5.1 Model Setup
Based on the Table 1 and Figure 1, it is evident that children currently in the process of being adopted (indicated by “Pre-adoptive home” and “IVEAA” statuses) are highly unlikely to age out of foster care. Consequently, our analysis will concentrate on children who are neither in a “Pre-adoptive home” nor have “IVEAA” status (N = 7,362).
The included covariates are listed in Table 1. We modeled continuous predictors (AgeAtFirstRem, TotalRem, NumPlep) using B-splines to capture non-linear relationships. B-splines, or Basis splines, are a series of piecewise polynomials used in regression analysis to model non-linear relationships by providing a flexible and smooth fit to the data. Additionally, we included interaction terms between Sex and Race variables to investigate potential differences in the likelihood of aging out across different demographic groups.
The dataset was split randomly into training (80%, N = 5889) and testing (20%, N = 1473) sets to evaluate the model’s performance. The logistic regression model was trained on the training set and evaluated on the testing set to assess its predictive accuracy. The model’s performance was evaluated using the area under the receiver operating characteristic curve (AUC-ROC). More details on the model performance are provided in the following sections.
5.2 Model Coefficients
Logistic regression outputs several key elements that help in understanding the relationship between the dependent variable and the predictors. The primary output includes the coefficients (or weights) for each predictor variable, which are often exponentiated to obtain odds ratios. Additionally, logistic regression provides confidence intervals and p-values for each predictor, helping to assess their precision and statistical significance.
Figure 3 presents a forest plot summarizing the model coefficients from the logistic regression analysis:
- Odds Ratios (OR): Each predictor is represented by an odds ratio, indicating the strength and direction of the relationship between the predictor and the likelihood of aging out.
- 95% Confidence Intervals (CI): Horizontal lines extending from each point estimate represent the confidence intervals. The length of these lines shows the range within which the true odds ratio is likely to fall.
- Significance:
- Solid circles represent significant predictors (p ≤ 0.05), indicating a statistically significant relationship with the likelihood of aging out.
- Open circles represent non-significant predictors (p > 0.05), indicating no statistically significant relationship.
- Interpretation:
- OR values greater than 1 suggest an increased likelihood of aging out.
- OR values less than 1 suggest a decreased likelihood.
This visual representation allows for a clear and concise interpretation of the impact and significance of each predictor variable in the model.
Figure 3: Coefficient Forest Plot
The logistic regression analysis identified several statistically significant predictors (p ≤ 0.05) for the likelihood of aging out of foster care in the jurisdiction in the western US:
- Removal Reason “Child behavior problems”
- Removal Reason “Caretaker Inability Cope”
- Removal Reason “Relinquishment”
- Emotional Disturbance
- IVEFC – Title IV-E Foster Care Payments
- Current Placement Setting
Although odds ratios (OR) can quantify the effect of each predictor on the likelihood of aging out, changes in odds are less intuitive to interpret compared to changes in probabilities. Therefore, in the following sections, we will translate these effects into changes in probabilities to provide a clearer and more intuitive understanding of the results.
5.3 Effects of Significant Predictors
Using the “empirical” marginalization method, we calculate the marginal average effects by averaging the predicted probabilities of aging out while adjusting for the observed values of all other predictors in the dataset. This approach considers the actual distribution and variability of these predictors as they naturally occur in the dataset, providing a realistic estimation of the average effect of each primary predictor. This ensures that the results are contextually relevant and reflective of real-world conditions. The predicted probabilities are then plotted to visualize the impact of significant predictors on the likelihood of aging out.
Figure 4 highlights the average predicted probabilities of aging out of foster care for statistically significant predictors identified in Figure 3. Key findings include:
- Removal Reasons:
- Child Behavior Problems: Children removed due to child behavior problems have a lower probability of aging out at 14.9%, compared to 21.8% for those not removed for this reason. This lower probability may be because child behavior problems are considered a less severe compared to other removal reasons.
- Caretaker Inability to Cope: Children removed due to caretaker inability to cope have a probability of aging out of 24.1%, compared to 20% for those not removed for this reason.
- Relinquishment: Children removed due to relinquishment show a much higher probability of aging out at 37%, compared to 21.2% for those not removed for this reason.
- Emotional Disturbance:
- Children who are emotionally disturbed have a higher probability of aging out at 24.5%, compared to 20.7% for those who are not emotionally disturbed.
- Title IV-E Foster Care Payments:
- Children receiving Title IV-E foster care payments have a reduced likelihood of aging out at 16.6%, compared to 25.1% for those not receiving these payments.
- Current Placement Settings:
- Supervised Independent Living: This setting has the highest probability of aging out at 52%, but the large confidence intervals suggest insufficient observations for this category.
- Group Homes and Runaways: Excluding supervised independent living, children in group homes and runaway children have high probabilities of aging out at 35.2% and 37.3% respectively.
- Institution: Children in institutions have a probability of aging out at 28.4%.
- Foster Home, Non-relative: Children in non-relative foster homes have a probability of aging out at 25.2%.
- Foster Home, Relative: Children in relative foster homes have a lower probability of aging out at 11.9%.
- Trial Home Visit: Children on a trial home visit have a probability of aging out at 11.2%.
Figure 4: Average Predicted Probability by Significant Predictors
5.4 Non-linear Predictor Effects
Using the same “empirical” marginalization method, we calculated the average predicted probabilities of aging out for non-linear continuous predictors. The plots in Figure 5 show the average predicted probabilities of aging out based on different values of AgeAtFirstRem, TotalRem, and NumPlep.
- Age at First Removal:
- Predicted probabilities of aging out decrease as age at first removal increases. Age at first removal reflects the number of years children spend in foster care. For instance, children first removed at age 0 spend 12-14 years in foster care, while those first removed at age 12 spend only 0-2 years. This suggests that the likelihood of aging out decreases with a shorter length of stay in the foster care system.
- Number of Placements:
- An increased number of placements is associated with a higher probability of aging out.
- Total Number of Removals:
- The effect of the total number of removals on the probability of aging out is very weak.
Figure 5: Average Predicted Probability by Non-linear Predictors
5.5 Model Diagnostics
Model diagnostics are crucial for ensuring the reliability and accuracy of our statistical models. They help us identify and correct issues like multicollinearity and problematic residual patterns. By thoroughly checking these diagnostics, we can strengthen the validity of our findings and ensure our conclusions are statistically sound.
The multicollinearity check assesses the degree to which predictor variables (excluding the interaction terms) in a regression model are correlated with one another. This is typically measured using the Variance Inflation Factor (VIF), where a high VIF indicates a high correlation, suggesting that one predictor can be linearly predicted from others with a substantial degree of accuracy. Addressing collinearity is important as it can inflate standard errors and make it difficult to determine the individual effect of each predictor.
Figure 6 displays the VIF values for each predictor in the logistic regression model. The VIF values are all below 5, indicating that multicollinearity is not a significant concern in our model. This suggests that the predictors are not highly correlated with each other, and the model is not adversely affected by multicollinearity.
Figure 6: Multicollinearity Diagnostics
The binned residuals diagnostic is a method used to evaluate the fit of a logistic regression model by grouping the residuals into bins. This involves plotting the average residuals within each bin against the predicted probabilities. Ideally, the residuals should be randomly scattered around zero, indicating a good model fit. Deviations from this pattern can signal issues such as model misspecification or non-linearity.
Figure 7 displays the binned residuals plot for the logistic regression model. The plot shows that most of the residuals are evenly distributed around zero, indicating that the model is well-specified and captures the relationship between the predictors and the outcome variable effectively.
Figure 7: Residual Diagnostics
5.6 Model Performance
Assessing model performance is essential for validating the accuracy and reliability of our predictions. We use the Receiver Operating Characteristic (ROC) curve and the area under the ROC curve (AUC) to evaluate our logistic regression model on the test set. The ROC curve illustrates the model’s ability to distinguish between classes by plotting the sensitivity (true positive rate) against 1 - specificity (false positive rate) at various threshold settings. The AUC provides a single metric to summarize the overall performance, with higher values indicating better discriminatory power.
Figure 8 displays the ROC curve for the logistic regression model. The AUC value of 0.698 indicates that the model has moderate discriminatory power in distinguishing between children who aged out and those who did not. This suggests that the model performs better than random guessing and provides a reasonable basis for predicting outcomes. Further analysis may explore additional predictors or other modeling approaches to enhance the performance.
Figure 8: Receiver Operating Characteristic (ROC) Curve
5.7 Model Fairness
Assessing model fairness is crucial to ensure equitable outcomes across different demographic groups. We compared the performance of our logistic regression model for White and Black individuals. The reason we chose these two groups is that previous research has shown disparities in the foster care system for these racial groups (Drake et al. 2011). By examining the model’s performance across these demographics, we aim to identify any biases and ensure that our model provides fair and accurate predictions for all individuals.
Figure 9 displays the model fairness evaluation results for White and Black individuals. The ROC curves for both groups are similar and within the confidence intervals, indicating that the model performs consistently across these demographics. This suggests that the model provides fair and unbiased predictions for White and Black individuals, ensuring equitable outcomes across these racial groups.
Figure 9: Receiver Operating Characteristic (ROC) Curve by Race
6 Discussion
This study examined what influences the likelihood of aging out of foster care in the jurisdiction in the western US, using data from the Adoption and Foster Care Analysis and Reporting System (AFCARS). By using logistic regression, we identified key factors that significantly affect foster care outcomes. In this section, we discuss the implications of these findings.
6.1 Significant Predictors and Their Implications
- Removal Reason “Child Behavior Problems”:
- Findings: Children removed due to child behavior problems have a lower probability of aging out at 14.9%, compared to 21.8% for those not removed for this reason. This lower probability may be because child behavior problems are considered less severe compared to other removal reasons.
- Implications: While children with behavior problems have a lower likelihood of aging out, further research is needed to verify this claim and understand the underlying factors contributing to this outcome.
- Removal Reason “Caretaker Inability to Cope”:
- Findings: Children removed due to their caretaker’s inability to cope show a higher likelihood of aging out.
- Implications: Enhancing support services for at-risk families, such as comprehensive parenting programs and mental health services, can prevent foster care entries due to caretaker issues. Proactive support may help maintain family unity and stability, reducing the incidence of children aging out.
- Emotional Disturbance:
- Findings: Emotional disturbance significantly increases the probability of aging out.
- Implications: Increasing access to mental health services for children in foster care is essential. Early intervention, continuous support, and therapeutic services can help manage emotional disturbances, thereby improving the chances of achieving permanency.
- IVEFC – Title IV-E Foster Care Payments:
- Findings: Title IV-E foster care payments are associated with a reduced likelihood of aging out.
- Implications: Ensuring that adequate funding is provided is important. Such financial support should aim to provide comprehensive services that promote stable placements and address the diverse needs of foster families and children.
- Current Placement Setting:
- Findings: Children in group homes or runaway situations have a higher probability of aging out.
- Implications: Developing and implementing programs focused on finding stable, long-term placements for children currently in group homes is important. Additionally, enhancing monitoring and support mechanisms to prevent runaways and ensure children are placed in family-like settings whenever possible will help reduce the likelihood of aging out.
7 Limitations
While this study provides valuable insights, it also has certain limitations:
- Geographical Limitation: The analysis is based on data from the jurisdiction in the western US, which may not be generalizable to other states or regions due to varying policies, demographic factors, and support systems.
- Cohort-Specific Focus: The study focuses on a specific cohort of children aged 12-14 at the start of the 2016 Fiscal Year, potentially excluding the experiences and outcomes of younger or older children in foster care.
- Scope of Predictors: The study primarily considers a limited set of predictors, which might not capture the full spectrum of factors influencing the likelihood of aging out of foster care.
Refences
Ahn, Eunhye, Yolanda Gil, and Emily Putnam-Hornstein. 2021. “Predicting Youth at High Risk of Aging Out of Foster Care Using Machine Learning Methods.”Child Abuse & Neglect 117 (July): 105059. https://doi.org/10.1016/j.chiabu.2021.105059.
Children’s Bureau Administration on Children, Youth and Families. 2023. “AFCARS Foster Care File, 6-Month Periods (FY2016A – 2022B).”https://doi.org/10.34681/yjxr-zz92.
Drake, Brett, Jennifer M. Jolley, Paul Lanier, John Fluke, Richard P. Barth, and Melissa Jonson-Reid. 2011. “Racial Bias in Child Protection? A Comparison of Competing Explanations Using National Data.”Pediatrics 127 (3): 471–78. https://doi.org/10.1542/peds.2010-1710.
Dworsky, Amy, Laura Napolitano, and Mark Courtney. 2013. “Homelessness During the Transition From Foster Care to Adulthood.”American Journal of Public Health 103 (S2): S318–23. https://doi.org/10.2105/AJPH.2013.301455.
Fowler, Patrick J., Katherine E. Marcal, Jinjin Zhang, Orin Day, and John Landsverk. 2017. “Homelessness and Aging Out of Foster Care: A National Comparison of Child Welfare-Involved Adolescents.”Children and Youth Services Review 77 (June): 27–33. https://doi.org/10.1016/j.childyouth.2017.03.017.
Husson, François, Sébastien Lê, and Jérôme Pagès. 2017. “Multiple Correspondence Analysis.” In Exploratory Multivariate Analysis by Example Using R. Computer Science and Data Analysis Series. Boca Raton, Fla.: CRC Press.
Lê, Sébastien, Julie Josse, and François Husson. 2008. “FactoMineR: An r Package for Multivariate Analysis.”Journal of Statistical Software 25 (1): 1–18.
Leave a Reply