The Role of Continuing Disability Reviews in Child Supplemental Security Income Program Participation Patterns

Introduction

The number of child Supplemental Security Income (SSI) recipients has declined since 2013, falling by more than 25 percent through December 2023 (Social Security Administration [SSA] 2024). The agency periodically conducts continuing disability reviews (CDRs) to determine whether SSI recipients remain medically eligible for payments, and the number of CDRs conducted during that period increased substantially—particularly after 2015—perhaps contributing to the decline in child SSI participation.¹ Recent research suggests that the increase in CDR frequency might have reduced duration of payments for affected SSI award cohorts (Hemmeter and others 2021). However, the effects of increased CDR volume on the overall SSI caseload, receipt of payments, and other outcomes are not well understood.

In this article, we analyze the characteristics of and outcomes for children who were removed from the SSI rolls during a recent uptick in the number of CDRs conducted and the role CDRs play in explaining trends in the number of child SSI recipients. Although previous analyses provided insights into outcomes for children and families following a CDR cessation (Deshpande 2016b; Hemmeter and Bailey 2015), they did not consider the substantial increase in CDRs conducted beginning in 2015, when SSA began working through a substantial CDR backlog. In 2002, there was no CDR backlog, but the CDR volume began declining in 2003 because of funding shortfalls and priority shifts. By the end of fiscal year 2014, the CDR backlog had grown to nearly 350,000 cases (SSA 2018).²

This article addresses three research questions:

1. What are the characteristics of children whose SSI payments are ceased because of a CDR?
2. What are the outcomes for children in the years following a CDR cessation?
3. How do CDR cessation patterns affect child SSI caseload trends?

To answer these questions, we relied primarily on descriptive analyses of SSA's administrative data. To answer the first question, we compared the characteristics of children whose payments ceased because of a CDR³ with those of child SSI recipients overall. To answer the second question, we looked at the rates at which children with ceased payments later return to the SSI rolls and at their earnings patterns in the years following cessation. For both questions, we focused on pattern shifts as the CDR frequencies changed. To answer the third question, we conducted a policy simulation exercise to predict the likelihood that children with specific characteristics would have had CDR cessations if the CDR cessation pattern held constant in a given year. That simulation also enables us to account for whether a child might have returned to the SSI rolls following a CDR cessation. We compare CDR cessation patterns from 2008 and 2017. The 2008 cessation cohort was atypically low because of budget constraints and the resulting shift in administrative priorities away from conducting CDRs, which generated a growing CDR backlog. By contrast, the 2017 cessation cohort typified a stable pattern, reflecting the regular processing volume after the backlog had been reduced. Comparing the distinct CDR cessation patterns of the 2008 and 2017 cohorts enabled us to assess how CDR cessations affect overall caseload dynamics. We then used our policy simulation to compare the trends in predicted child SSI participation over that period with actual trends.

We find that most characteristics of children with ceased payments stayed constant over time (except for age at CDR), irrespective of the frequency of cessations. This suggests that, regardless of the volume of CDRs conducted, the profiles of children most at risk of losing SSI payments stay largely the same, with only the number of affected children fluctuating with the rate at which reviews are completed. Children who are expected to improve medically, who live in areas with high socioeconomic deprivation, and who have certain primary diagnoses (such as developmental or respiratory disorders) were disproportionately likely to experience cessation of payments in all years we examined. One notable shift over time was the age at which CDR cessations were likely to occur. From 2003 through 2008, children aged 1–5 more frequently had a CDR cessation than children aged 11–13, whereas from 2011 through 2017, the reverse was true.

For children whose SSI payments ceased, returns to the SSI rolls were infrequent and earnings potential was limited, indicating the many challenges those children face (both in the short term and when they become adults). About 8 percent of children returned to the SSI program within 5 years of payment cessation. Those patterns differ somewhat by the year of cessation—slightly more than 10 percent of children with ceased payments from 2003 through 2008 returned to the program, compared with about 6 percent of children whose payments ceased because of a CDR from 2011 through 2017. Earnings outcomes were also limited. For example, in the fifth year after payment cessation (the year with the highest earnings outcomes), only about 60 percent of youths aged 16 or older at the time of earnings measurement had any earnings. Additionally, in the fifth year after payment cessation, the average annual earnings among this group aged 16 or older at the time of earnings measurement were less than $7,000. These findings suggest that youths whose payments ceased because of a CDR might have struggled to achieve economic self-sufficiency because they could no longer rely on SSI payments and did not earn much money.⁴ That evidence is consistent with findings from Hemmeter, Kauff, and Wittenburg (2009), Deshpande (2016b), and Deshpande and Mueller-Smith (2022).

Finally, our policy simulations indicate that CDR cessation patterns play an important role in the overall child SSI caseload over time. From 2003 through 2013, that caseload grew by about 365,000. By comparing the patterns between the low and stable cessation cohorts, we find that a stable pattern of CDR volume may have led to slightly more than 220,000 additional cessations. Thus, CDR frequency can explain about 60 percent of the child SSI participation growth during that period. Subsequently, from 2014 through 2021, CDRs explain about two-thirds of the observed program decline: the caseload dropped by 283,000, while the cessation differential between years with stable and low volumes was about 194,000.⁵ Notably, childhood SSI participation would have been more consistent over time in both the low and stable patterns of cessations, compared with actual patterns of cessations. Given that SSA's annual budgets include dedicated program integrity funding to support medical CDRs, SSI redeterminations, and fraud investigations and prosecutions, those results have potential policy implications.

Background

This section provides information about SSI eligibility requirements, child SSI participation rates, CDR frequency criteria, and annual CDR volumes.

SSI provides monthly cash payments to children with disabilities and low income and resources. To qualify as having a disability, a child must have “a medically determinable physical or mental impairment or impairments which result in marked and severe functional limitations” (42 U.S.C. §1382c[C][i]). These limitations must come from an impairment that is expected to last at least 12 months or result in death. Children must also meet a means test, which includes the portion of their parents' income and resources that is considered available to them. Resources available to the child cannot exceed $2,000 (after accounting for certain exemptions such as the value of a residential home). SSI payments are offset by $1 for every $2 in earned income available to a child above $65 and by $1 for every $1 in unearned income above $20, with many exclusions. If the amount due is $0, the child does not receive a payment that month. In 2024, the maximum monthly SSI cash payment for an individual is $943.⁶

SSI recipients face a CDR every few years to determine whether they meet the disability criteria to remain eligible. If SSA determines that a child no longer has a “marked and severe functional limitation” and has no other impairments that meet the definition of disability, then the child's payments are ceased. If SSA determines that a child's medical condition has not improved to that level (or has stayed the same or worsened), or if the child is determined to have a disability because of other impairments, then the child's payments continue.⁷ Families of children whose payments are ceased can appeal the initial CDR decision.

The frequency with which child SSI recipients face CDRs depends on the expectation of their medical improvement, which is assigned when they receive an initial award or undergo a prior CDR.⁸ If medical improvement is expected, a CDR typically occurs within 6 to 18 months of award. If medical improvement is deemed possible, a CDR typically occurs every 3 years. If medical improvement is not expected, a CDR typically occurs every 7 years, although SSA typically does not conduct CDRs for children in that group before their age-18 redeterminations. These “medical diaries,” or assessments of how likely a child's medical condition is to improve, enable SSA to review millions of recipients' eligibility statuses efficiently. From calendar years 2003 through 2017, the share of child SSI recipients for whom medical improvement was deemed possible was about 70 percent, while roughly 15 percent were expected to experience medical improvement and about 15 percent were not expected to experience medical improvement.

Although the medical diaries generally determine CDR frequency, their volumes have varied over the years along with the availability of program integrity funding.⁹ Chart 1 shows that from fiscal year 1998 through fiscal year 2022, the annual number of CDRs conducted varied substantially.¹⁰ From 1998 through 2002, SSA conducted an annual average of 134,934 CDRs. In 2003, SSA began conducting substantially fewer CDRs, which caused a backlog of overdue CDRs to start to accumulate. Then, from 2006 through 2010, SSA conducted only 11,169 CDRs per year on average, fewer than one-tenth as many as were conducted from 1998 through 2002. In 2015, SSA began to reduce the CDR backlog (nearly 350,000 as of 2014) and had eliminated it by the end of 2018 (SSA 2018). CDR decisions resulting in payment cessations followed a similar pattern of a decline after 2003, with particularly low numbers from 2006 through 2010, followed by increases thereafter. In a typical year, the number of initial payment cessations resulting from a CDR is about 35 percent higher than the number of final payment cessations that follow a CDR, indicating that a substantive share of initial cessation decisions are successfully appealed.

Chart 2 shows that child SSI participation increased from the start of the program in 1974 through 2013 (with a dip in the late 1990s) and has since declined. Over the years, legislative changes, Supreme Court decisions, and the COVID-19 pandemic have contributed to changes in the participation rate. For example, the 1990 Sullivan v. Zebley Supreme Court decision loosened the criteria for children—particularly those with mental disorders—to qualify as having a disability (Levere 2021). However, the Personal Responsibility and Work Opportunity Reconciliation Act of 1996 reversed some of the changes that followed the Zebley decision, resulting in more restrictive disability criteria and CDR rules for children. SSA implemented those rules in 1997 and finalized them in 2000. Then, during the Great Recession, a weakening economy likely contributed to enrollment growth (Schmidt and Sevak 2017). The unanticipated decline in program participation from 2014 through 2019 has not yet been extensively studied. Finally, during the COVID-19 pandemic, the decline in child SSI participation was differentially larger than the 2014–2019 decline (Levere, Hemmeter, and Wittenburg 2024a), with school closures potentially explaining a substantial share of the decline (Levere, Hemmeter, and Wittenburg 2024b).

The recent declines in child SSI participation have led to targeted outreach to children and families. Concern that SSI was not reaching all eligible children emerged as the number of child SSI recipients began to decline after 2013 (National Academies of Sciences, Engineering, and Medicine 2015). During the COVID-19 pandemic, SSA increased its SSI outreach efforts and designated certain staff as Vulnerable Population Liaisons in field and regional offices to help potentially eligible people apply for payments.¹¹ Such efforts might be most effective in narrow geographic areas where sociodemographic or health claims data indicate many eligible children might not be receiving SSI payments (Levere, Wittenburg, and Hemmeter 2022; Levere and Wittenburg 2024).

A key goal of our research is to identify the extent to which patterns in the frequency of CDRs (and accompanying payment cessations) might have contributed to broader changes in child SSI participation. Simple correlational patterns indicate that CDRs might play an important role in changes to child SSI participation. For example, although the number of new child SSI awards declined each year from fiscal year 2004 through fiscal year 2007 (before the Great Recession), the number of child SSI recipients rose by 9 percent from 2004 through 2007 (SSA 2023c). Those years correspond to the years when the CDR backlog started to grow, resulting in fewer child SSI recipients exiting the program. Similarly, as SSA processed the CDR backlog from fiscal year 2015 through fiscal year 2018, the number of child SSI recipients declined by about 10 percent. Hemmeter and others (2021) showed that the prevalence with which certain award year cohorts face CDRs might explain a substantial share of the difference in program participation across those cohorts.

Data

We used program participation and earnings data from various SSA data files to examine the outcomes for children whose SSI payments ceased because of a medical CDR. First, we identified the population of children with ceased payments from 2003 to 2021¹² using the CDR Waterfall file. We then used the Disability Analysis File to obtain information about those children's characteristics, such as age, date of first SSI payment, medical diary category, primary diagnosis, and ZIP Code and county of residence. The Disability Research File provides information on subsequent returns to the SSI rolls in the 5 years after payment cessation, while the Master Earnings File provides data on earnings in the 5 calendar years following cessation. We included earnings observations only for youths who were aged 16 or older; therefore, the earnings analyses do not include anyone whose payments ceased before age 11.

We used the date of the final CDR decision, rather than the date of the initial CDR decision, to classify the year a child's SSI payments were ceased.¹³ Chart 3 shows that the cessation counts we derive using this methodology differ slightly from those published in the Annual Report of the Supplemental Security Income Program (SSA 2023a), which reflect the year of initial decisions. In particular, Chart 3 shows that in recent years, when cessations increased, this approach led to a slight lag in reporting. Classifying cessations by the year in which payments were officially ceased aligns more closely with the actual changes in the caseload, which is important for our simulation exercise.

To identify the overall child SSI recipient population, we used data from the Disability Analysis File. We collected age, date of first SSI payment, medical diary category, primary diagnosis, and ZIP Code and county of residence for all children. We compare the characteristics of children whose payments were ceased with those of current SSI recipients to gauge whether recipients with certain characteristics differentially face payment cessation. Our policy simulations also rely on estimating a probability of SSI payment cessation and a probability of payment continuation using those characteristics.

Finally, we used American Community Survey 5-year estimates for 2015–2019 on all inputs to the Area Deprivation Index and followed the process described in Singh (2003) to estimate a level of deprivation at the ZIP Code and county levels.¹⁴ Deprivation is expressed as a percentile between 1 and 100 that we divided into quartiles; in this context, it represents a single indicator of the various disadvantages a community faced in 2015–2019. This process mirrors our approach in Levere, Wittenburg, and Hemmeter (2022). In addition, we measured the distribution of race and ethnicity for each ZIP Code or county from the same 5-year estimates.

Methodology

We structured our methodological approach to address our three research questions. The first question addresses the demographic and socioeconomic characteristics of children whose SSI payments were ceased because of a CDR from 2003 through 2021. We divided our study population into annual cohorts based on the calendar year of CDR cessation, enabling us to track trends over the observation period, such as the shares of child SSI recipients with particular primary medical diagnoses who experienced a CDR cessation. Examining longitudinal patterns of children with particular characteristics whose payments ceased allows us to show whether, and how, the characteristics of such children have changed as program funding and CDR frequency have changed. However, the characteristics of child SSI participants overall might have changed over time as well. Therefore, we also benchmarked the characteristics of the study population to the overall child SSI recipient population each year.¹⁵ We divided the share of children with a particular characteristic whose payments ceased by the share of the overall child SSI population with that characteristic. Those descriptive analyses indicate whether certain groups were disproportionately likely to have their SSI payments ceased over time.

To address our second research question on the outcomes for children in the years after a CDR cessation, we investigated the share of children with ceased payments that returned to SSI and examined three earnings outcomes for each annual cessation cohort. The outcomes we examined are average annual earnings amount, the prevalence of having any annual earnings, and the prevalence of having earnings above $16,200, which equals the 2022 level that signifies substantial gainful activity for nonblind recipients. The insights this analysis provides about postcessation well-being and financial stability are pivotal for understanding the economic and social ramifications of cessation for children and their families.

Finally, to answer the question of how CDR patterns affect the child SSI caseload overall, we employed policy simulations to explore the hypothetical effects of applying consistent CDR patterns across the study period. We assessed the influence of different CDR cessation rates on the overall trends in child SSI participation, using 2008 (a year with very few CDR cessations) and 2017 (representing a typical year with “stable” CDR patterns) as our base analysis years. This comparison reveals the potential effects of varying CDR frequencies on SSI caseload fluctuations, offering critical perspectives on the policy and programmatic implications of prioritizing (or deemphasizing) CDR workloads. From the models, we estimated both the share of children who would have had payments ceased because of a CDR, and the share of children whose payments had been previously ceased who would have returned to the SSI program.

Our modeling procedure estimates cessations and returns in all years from 2003 through 2021 using a model estimated among a specific base year cohort. The stable cessation cohort (2017) most closely approximates what might happen in a year with no CDR backlog and enabled us to observe postcessation outcomes for at least 5 years.¹⁶ The low-cessation cohort (2008) provides a sharp contrast because few CDRs were conducted—only 4,707 initial decisions and 921 final cessations, compared with 208,500 and 42,402, respectively, in 2017.

Chart 4 shows observed and probabilistic year-to-year patterns of the child SSI caseload. The actual SSI caseload evolves each year because of continuations, cessations, and new applicants (Panel A). Children receiving SSI payments at the start of the year could either continue receiving SSI payments at the start of the next year or have their payments ceased during that year. We distinguish CDR cessations from other program exits—including aging out of the program and having excess resources or income. In addition, new awardees could start to receive SSI payments during the year and thus be added to the rolls by the start of the next year. To understand how CDR frequencies affect SSI caseloads, our estimation procedure holds the CDR cessation count fixed by applying the base year CDR cessation patterns.

Rather than using the observed patterns, our simulation models estimated the likelihood of a CDR cessation in each year, assuming the child CDR cessation rate remained consistent over time. We began with the number of child SSI recipients whose payments were not terminated at the start of the base year and estimated a logistic regression to predict the likelihood of cessation in that year by specific demographic characteristics. We then applied those coefficients to all other years to estimate the probability of a child's SSI payments being ceased, holding the cessation trends constant. In equation 1 (the regression equation), the outcome variable is equal to 1 if the child's payments ceased because of a CDR and is equal to 0 otherwise:

We controlled for sex, age at year end, age at initial SSI entry, duration of SSI payment receipt, medical diary category, and primary disability diagnosis in X _i . With a logistic regression, the predicted value of the outcome based on the coefficients—which represents the probability of cessation—is guaranteed to fall between 0 and 1. This offers an important advantage over a linear probability model—with which the predicted value need not fall between 0 and 1—because predicted values within that range are essential to our models.

We used our policy simulation to estimate the probability of a CDR cessation (Panel B). We started with the sample of child SSI recipients on January 1, 2003, and estimated the probability of CDR cessation in 2003 by applying the coefficients from equation 1, estimated among the base year (2008 or 2017) population, to all child SSI recipients in 2003. By applying the same base year coefficients, the analysis assumes constant CDR cessation patterns over time. The total number of estimated cessations is the sum of the probabilities of cessation across all child recipients. For example, if 1,000 children each have a payment cessation probability of 0.03, we would expect 30 cessations among that group. The likelihood of continuation into 2004 is 1 minus the probability of cessation.

Then, to identify the number of child SSI recipients at the start of 2004, we added newly awarded children and subtracted children who stopped receiving SSI payments for non-CDR reasons during 2003. Thus, for every child receiving SSI payments at the beginning of 2003, we now had the probability that they were still on the SSI rolls at the start of 2004.¹⁷

Finally, we applied the same procedure for all subsequent study years to estimate the number of cessations each year. We applied coefficients from equation 1 estimated in the base year (2008 or 2017) to the children whose payments were not terminated in 2004—increasing their age and payment duration by 1 year relative to those at the start of 2003—to estimate the probability of cessation in 2004 and the probability of continuation into 2005. We then subtracted children whose payments ceased for reasons other than a CDR and added new awardees. The probability of receiving SSI at the start of 2005 is therefore the probability of SSI payment receipt at the start of 2004 multiplied by the probability of continuation into 2005. That procedure, which accounts for the probability of continuation at the beginning of each year, is critical for reliably estimating the number of cessations: The dynamic nature of CDR patterns means that the caseload at the beginning of each year might have shifted had CDR cessation patterns been consistent. For example, if CDR cessations were more frequent, like the 2017 pattern, more children would have had payments ceased in earlier years, meaning they would not have faced a potential CDR cessation in later years. Estimating the flow of cessations each year therefore requires knowing how many children could have faced a cessation at the start of the year—that is, the probability of payment receipt. We repeated this procedure for each year to estimate the probability of payment receipt at the start of each year from 2006 through 2022.

Our models also account for the potential return to SSI participation of children who experienced a prior CDR cessation. Specifically, among children with payments ceased in the base year, we estimated a separate logistic regression using equation 2, which differs from equation 1 only in the outcome variable being an indicator for returning to SSI in the 5 years following CDR cessation.

The controls include the same demographic characteristics used in equation 1, measured at the time of cessation. We can therefore apply the coefficients to estimate the probability of returning to the SSI program each year, conditional on payments being ceased.¹⁸ We can then sum those probabilities across the population to estimate the total number of child recipients who had experienced a CDR cessation and would have returned to the SSI program each year. Thus, the probability of being a recipient at the start of each year is the sum of (1) the probability of having been a recipient for the entirety of that year and (2) the probability of returning to the SSI program in that year after experiencing a CDR cessation in a prior year.

We then used those estimates to create a counterfactual caseload. First, we calculated the net reduction in the caseload resulting from CDR cessations in each year under a given model, which is the total number of cessations minus the total number of program returns. Next, we calculated the actual net reduction in the caseload from CDR cessations. To estimate the counterfactual caseload under a given model, we modified the year-to-year changes in child SSI participation by replacing the actual net reduction because of CDRs with the model-based net reduction because of CDRs.¹⁹ That process isolates the role of CDRs from all other programmatic factors contributing to changes in the caseload (new applications, other cessations, and other factors) by holding all other year-to-year changes fixed. One limitation of that approach is that our analysis does not control for or otherwise incorporate economic, sociocultural, environmental, medical, or other factors. As a result, we do not predict non-CDR-related changes in SSI program participation, such as those stemming from outreach efforts or the overall decline in childhood poverty in recent years.

In addition, the differential pattern between the low cessation (2008) cohort and the stable cessation (2017) cohort enables us to estimate the share of SSI caseload fluctuations that stem from CDR cessation patterns. The difference in net cessations between those two models represents the change in the number of SSI recipients that can be attributed to the variance in CDR frequencies. Benchmarking that number to the total change in the SSI caseload each year, or over a multiyear period, indicates what share of SSI caseload fluctuations stems from CDR cessations alone.

Results

We arrange our findings in the order of our three research questions.

Research Question 1: What Are the Characteristics of Children Whose SSI Payments Are Ceased Because of a CDR?

A child SSI recipient's medical diary category is an important predictor of CDR cessation. As noted earlier, the three medical diary categories are medical improvement expected, medical improvement possible, and medical improvement not expected. In a typical calendar year from 2003 through 2021, medical improvement was deemed possible for about 70 percent of children with ceased payments, about 30 percent were expected to experience medical improvement, and a negligible share were not expected to experience medical improvement (Chart 5, Panel A). The share of children with ceased payments who were expected to experience medical improvement was elevated from 2012 through 2014, coinciding with the initial increase in CDR frequency shown in Chart 1.

When we benchmark those trends versus the shares of all child SSI recipients in each medical diary category, it becomes evident that those expected to experience medical improvement constitute a disproportionately high share of children with ceased payments each year (Chart 5, Panel B). As noted earlier, about 15 percent of child SSI recipients are expected to experience medical improvement. With those children making up about 30 percent of the annual cessation (Chart 5, Panel A), they are twice as likely to have their payments ceased as the average child SSI recipient. By contrast, the share of child SSI recipients with medical improvement possible who had their payments ceased aligns closely with their representation in the overall child SSI recipient population. Given an expectation of medical improvement, which causes more frequent CDRs, it is predictable that the rates of CDR cessation are differentially higher among this group. Charts 6–8 and the accompanying discussion use the benchmarked trends as shown in Chart 5, Panel B. That is, the results are expressed as the share of children with a given characteristic whose payments ceased (such as 30 percent who are expected to experience medical improvement in the example above) divided by the share of all child SSI recipients with that characteristic (15 percent in the example above; thus, 30 divided by 15 equals the benchmarked proportion of 2).²⁰

Chart 6 shows that age patterns of children with ceased payments have been mostly stable since 2010, despite the large increase in CDR frequency during the later part of the period. In the early part of the period, children aged 1–5 were disproportionately more likely than those in other age groups to have their payments ceased. However, this pattern began changing in 2009. In all years after 2010, the youngest (1–5) and oldest (14–17) age groups were least likely to have SSI payments ceased, while children aged 6–10 and 11–13 faced differentially higher cessation rates.

Chart 7 shows that children with certain primary medical diagnoses were much more likely to face payment cessation than children with selected other diagnoses.²¹ Among children with mental disorders (Panel A), those with developmental disorders were consistently more likely to have payments ceased prior to 2021. Those with intellectual disorders and autism spectrum disorders were less likely to have a CDR cessation. In the later years, the pattern slightly shifted, as children with other mental disorders²² became slightly more likely than children with developmental disorders²³ to have a CDR cessation. Among nonmental diagnoses (Panel B), children with respiratory system diseases were more likely to have SSI payments ceased while children with nervous system and sense organ diseases were less likely to have SSI payments ceased.

Finally, Chart 8 shows that children with ceased payments disproportionately lived in areas of high socioeconomic deprivation. For example, 36.8 percent of children with ceased payments in 2017 lived in ZIP Codes in the fourth quartile of socioeconomic deprivation (indicating worse economic outcomes). Yet only 30.7 percent of all child SSI recipients lived in such ZIP Codes. Thus, the proportional cessation rate was 1.2 for the fourth quartile in 2017. By contrast, 14.1 percent of children with ceased payments in 2017 lived in ZIP Codes in the first (or most advantaged) quartile of socioeconomic deprivation, compared with 19.4 percent of all child SSI recipients living in those ZIP Codes (a 0.7 proportional cessation rate). The general patterns hold true whether measuring socioeconomic deprivation at the ZIP Code or county level.²⁴ We explore whether appeals might play a role in this finding: In theory, one might expect that children in areas with lower socioeconomic deprivation might be more likely to appeal an initial cessation because of greater knowledge of the appeal process or greater resources to go through that process. However, we found that, in each of the four socioeconomic deprivation quartiles, the percentage of children with ceased payments who appealed that cessation was roughly equal to the percentage of the overall child SSI population, indicating no differential patterns in appeals (results available on request).

Research Question 2: What Are the Outcomes for Children in the Years Following a CDR Cessation?

Among children whose payments were ceased, a modest share returned to the SSI program within 5 years of the cessation (Chart 9).²⁵ On average during the study period, 1.7 percent of children returned to SSI in the first year after a cessation, 5.6 percent returned in the first 3 years, and 8.0 percent returned in the first 5 years. The SSI return rates began to drop with the 2009 cessation cohort. For the 2011 and later cessation cohorts, the average annual rates of return were 1.2 percent after 1 year, 4.1 percent after 3 years, and 5.7 percent after 5 years.²⁶ Interestingly, even as the frequency of CDRs substantially increased starting in 2015, the SSI return rates did not change noticeably.

Our analysis of postcessation earnings focuses exclusively on youths who had reached working age at the time of measurement, specifically those aged 16 or older. This analysis helps us to understand the economic outcomes of former child SSI recipients as they transition into the workforce. Given the typical progression of career development, we anticipate an increase in earnings as those youths age, particularly when comparing earnings 5 years after cessation with earnings in the first year after cessation.

Chart 10 shows that earnings increased in the years following cessation, but it also suggests that the earnings potential for former child SSI recipients was fairly limited. For the 2017 cessation cohort, average annual earnings increased from $2,716 in the first calendar year after cessation to $9,608 5 years after cessation (Panel A). Correspondingly, the shares of former recipients with any postcessation annual earnings (Panel B) and with substantive earnings amounts (more than $16,200, the 2022 annualized substantial gainful activity amount; Panel C) wavered in the early cessation cohorts but increased with the passage of time for the later cohorts. The effects of the Great Recession are evident: earnings amounts 3 years after cessation were lowest for the 2007 and 2008 cohorts, while earnings amounts 5 years after cessation were lowest for the 2005 and 2006 cohorts—aligning with earnings accrued in calendar years 2010 and 2011. Interestingly, it seems the COVID-19 pandemic had minimal effects on those cohort-based patterns, as 2020 calendar year earnings aligned closely with adjacent years.

Research Question 3: How Do CDR Cessation Patterns Affect Child SSI Caseload Trends?

As described earlier, we estimated the counterfactual caseload over time based on the likelihood that each child SSI recipient would have payments ceased because of a CDR each year. Then for those with payments ceased, we estimated the likelihood that they would return to the SSI program. By summing those probabilities, we estimated the number of total cessations and program returns that would have occurred in each calendar year if the cessation and return probabilities had followed the patterns of an individual annual cessation cohort over time—namely, if the rate of CDR cessations had been constant (rather than the variation in CDR frequency that occurred over the years). Finally, to obtain the counterfactual child SSI caseload, we replaced the observed net reduction from CDRs in each year (cessations minus returns) with the model-based prediction.

Chart 11 shows that the cessation cohort has substantial implications for the likelihood of a child being removed from the SSI program by a CDR. Each of the three medical diary categories is represented by one of the chart's panels. The red line in each panel shows the actual cessation rate for that category. Children expected to experience medical improvement had the highest cessation rates, followed by children for whom medical improvement was possible. Almost no one who was not expected to experience medical improvement had payments ceased in any year, consistent with SSA practice of generally not conducting CDRs for this population. Patterns over time match the frequency of child CDRs shown in Chart 3, with peaks in the years before and after the Great Recession, and almost no cessations from 2007 through 2010. The dark blue line represents the predicted probability of cessation using the 2017 cohort as the base year (stable cessations) in place of the observed cessation rate. Those probabilities are consistently higher than the light blue line representing the predicted probability of cessation using the 2008 cohort as the base year (low cessations). Both of the predicted probabilities exhibit smoother trends in cessation rates over time.

Chart 12 shows that the cessation cohort does not have as much of an effect on the likelihood of returning to the SSI program after a child CDR cessation. Return rates were slightly higher for children not expected to experience medical improvement. That group also had noisier observed rates of return because of the small sample of children who had payments ceased. However, return rates are more stable than cessation rates—both in the actual rates of return over time and in the difference between the patterns using the 2008 and 2017 cessation cohorts as the base year.

Chart 13 shows the net reduction in the child SSI caseload attributable to CDRs. The red line shows the actual net reduction, and the light blue and dark blue lines respectively show the predicted net reductions using the 2008 (low cessations) and 2017 (stable cessations) models. The net contribution to the caseload from CDRs depends heavily on the selected model, although with consistent CDR volumes, it would follow a stable pattern. Under a low cessation model, there would be few cessations and thus few returns. Under a stable cessation model, there would be many more cessations—roughly 25,000 to 30,000 per year. Both models offer substantially less fluctuation than the actual pattern. One reason the 2017 model leads to a smaller net caseload reduction than the actual net reduction is that many children would have had their payments ceased earlier, leading to a lower caseload by 2017. The negative numbers indicate that CDRs on net contribute to a reduction in the caseload; there are always more children whose payments have ceased than children who returned to SSI following a previous cessation.

Taken together, the policy simulations suggest that CDR cessation patterns play an important role in the overall dynamics of the child SSI caseload over time (Chart 14). The actual caseload in 2021 consisted of around 1.04 million child SSI recipients, which reflects the varying CDR policies during that period. If SSA had consistently applied a stable CDR cessation policy, the 2021 child SSI caseload might have been only 0.92 million. If SSA had consistently applied a low cessation policy, the 2021 child SSI caseload might have been 1.36 million. Thus, the net child SSI caseload difference between a low cessation policy and a stable cessation policy is roughly 400,000.

Although the effect of CDR volume in any single year is relatively minor, when cumulated over an extended period, CDRs can explain an important share of the change in the child SSI caseload. For example, from 2002 through 2013, the number of child SSI recipients increased by about 406,000. Comparing the 2017 (stable cessations) and 2008 (low cessations) models' patterns, we estimate that a stable cessation pattern might have netted nearly 250,000 more cessations. Thus, low CDR volumes can explain about 60 percent of enrollment growth from 2002 through 2013.²⁷ Increased CDR volume in the subsequent period, 2013 through 2021, may account for two-thirds of that period's program participation decline: The caseload fell by 283,000, while the differential between the stable and low CDR cessation models is about 194,000.

Interestingly, the probability of SSI return plays a minimal role in the simulations. For both base year cohorts, if we had not adjusted for program return probability, the estimated value in the final year (which allows for the greatest number of possible returns over time) would be essentially unchanged. For example, the estimated number of SSI recipients shown in Chart 14 in 2021 from the low cessation model is 1.363 million; if we had not accounted for the probability of return, the estimated number in 2021 would have been 1.382 million. For the stable cessation model, the estimate would have been only 3,000 recipients lower if we did not adjust for program-return probability.

Finally, Chart 15 shows that the results are mostly not sensitive to the year chosen to represent the stable cessation cohort. When we use 2003 or 2019 as the base year instead of the 2017 cessation cohort, we still find that the child SSI caseload would have been substantially smaller, with subsequently smaller declines in the caseload than have been observed. Comparing the estimates with the 2008 low cessation cohort, all three cohorts would still lead us to the conclusion that CDR cessation patterns can explain a substantial share of the changes in child SSI program participation. With the 2017 base year, CDR volume could explain 61 percent of the caseload growth from 2002 through 2013 and 68 percent of the caseload decline from 2014 through 2021. With the 2019 base year, CDR volume could explain 57 percent and 62 percent, respectively, and with the 2003 cessation cohort, CDR volume could explain 44 percent and 47 percent, respectively.

Conclusion

We investigated the role that the increase in the frequency of CDRs had on recent changes in the child SSI recipient population. We found that cessation rates by children's primary diagnoses and other demographic characteristics were stable, though the ages of children at the time of their CDR cessations shifted noticeably around 2010. That finding is not unexpected, as children aged 11–13 were presumably more likely to be overdue for a CDR because of a backlog in the earlier period. By contrast, the backlog also would likely contain cases of child SSI recipients that would be more consistent across other characteristics. Children who lived in ZIP Codes with higher levels of socioeconomic deprivation were also more likely to have SSI payments ceased by a CDR.

The rate at which children with payments ceased because of a CDR returned to the SSI program was lower among the post-2008 cessation cohorts. We were not able to isolate a single cause of that decrease, but two factors are consistent with the observed trends. First, childhood CDR volume was not particularly high before 2008. The resulting CDR backlog may have included a substantial number of cases involving children with relatively less severe disabilities, who then experienced post-2008 CDR cessations as SSA emphasized backlog reduction. Having less severe disabling conditions, these children would be less likely to return to the SSI program thereafter.²⁸ However, returns to the SSI program did not increase as the frequency of CDRs subsequently increased. Second, earnings 5 years after cessation were relatively higher for the cohorts whose payments were ceased after 2008 (the postrecession economic recovery may have contributed). Thus, there might have been less of a perceived need for SSI payments.²⁹ Those two explanations are not mutually exclusive, and more research is needed to confirm whether either is correct or whether other factors explain the trend.³⁰

Our results indicate that CDRs can explain a notably large portion of the SSI caseload dynamics during the study period. The frequency of CDRs can explain about 60 percent of the increase in child SSI participation from 2002 through 2013: If CDR volume had followed a more stable pattern, the number of child SSI recipients would have been substantially lower. In addition, CDR frequency can explain two-thirds of the decline in program enrollment from 2013 through 2021. Without the increase in CDRs that occurred during that period, the caseload would have been roughly unchanged. These findings indicate that CDR policies that would have maintained consistent cessation counts over time would have led to a more stable pattern of child SSI participation. Expectations about when CDRs will be conducted might be especially important for families in planning for the possibility of losing payments. Even though SSA conducted age-18 redeterminations consistently during the study period, few families correctly anticipate the potential payment cessation for their child (Deshpande and Dizon-Ross 2023).

Given the variation in childhood CDR frequency even among the low-volume years, an unanticipated loss in payments would be that much more difficult for families. Deshpande and Dizon-Ross (2023) also showed that families tend not to change their behavior even when they obtain accurate information about the likelihood of payment cessation, which may further complicate matters for them. Still, two large demonstration projects that sought to support young SSI recipients as they transition to adulthood, the Youth Transition Demonstration and Promoting Readiness of Minors in SSI, found that combining program information on redeterminations with vocational rehabilitation or similar services can enhance the recipient's human capital investment; however, the effect seems to be short-lived (Fraker and others 2014; Patnaik and others 2022). SSA already informs all child recipients aged 14-17 about the age-18 redetermination through an annually mailed brochure noting the high likelihood of losing payments.³¹ Including additional information about childhood CDRs in the award notification or other program communications could potentially avert any parental expectations that SSI eligibility is permanent.

We found that a substantial fraction of the trends in child SSI participation can be attributed to the increase in program integrity funding for CDRs from 2014 through 2018. Although the volume of CDRs has increased since the early 2000s, the stable rates of cessation across demographic characteristics, even as CDR volume varied, indicates that CDRs are targeted consistently. The main determinant of who faces a CDR is the medical diary category. In its fiscal year 2020 budget request, the agency proposed expanding the number of diary categories from three to four to enable it “to conduct CDRs more frequently for those medical impairments that are expected or likely to improve” (SSA 2019, 35). Our results suggest that such a change likely would have lowered the number of child SSI recipients, extending recent trends in declining program enrollment that continued during the COVID-19 pandemic (Levere, Hemmeter, and Wittenburg 2024a).

Even with consistent targeting, children with ceased payments are not necessarily receiving the supports they need to be self-sufficient. As noted earlier, we found low earnings levels for children whose SSI payments were ceased as they transitioned to adulthood, which correlates with poor adult economic outcomes (Patnaik and others 2022; Luecking and Leggett 2009; Fraker and others 2014). Although SSA determines that such children's impairments are no longer sufficiently severe to qualify for SSI, most of those children still face substantial barriers to full participation in educational settings or the labor force. Because SSI eligibility is linked with Medicaid eligibility, cessation of the former means loss of the latter for many children. However, state Medicaid offices typically look for ways to retain a child's Medicaid eligibility, perhaps based on limited incomes. Although there is substantial overlap in SSI and Medicaid eligibility (Levere and Wittenburg 2024), additional supports might be necessary for many children to fully participate in society. SSA is currently conducting the Beyond Benefits Study,³² which examines the supports needed for adults whose SSI payments were ceased by a medical CDR. SSA is also planning a new survey of children that might highlight their potentially unique needs.³³

Studies of CDR policies and practices should also address potential equity concerns, given that children from high-deprivation areas are more likely than others to have their SSI payments ceased. The role of economic, medical, environmental, or social factors in payment cessations is beyond the scope of this study. Illuminating the childhood SSI experience will help policymakers understand whether children and families are prepared for CDRs, have the necessary resources when undergoing a CDR, and can weather a payment cessation.