Trends in Special Education Identification During the COVID-19 Pandemic: Evidence From Michigan

Abstract

We use data from Michigan and an interrupted time series strategy to show how the COVID-19 pandemic impacted special education identifications and discontinuations. We find a substantial decrease in K–5 identifications and discontinuations during the 2019 to 2020 and 2020 to 2021 school years. Identifications fell by 19% and 12% in the first two pandemic years, with smaller but still significant reductions in discontinuations. Districts with remote schooling and Black, Asian, and economically disadvantaged students saw larger decreases in identifications. By 2021 to 2022 rates returned to trend and continued to grow in 2022 to 2023, albeit at a similar rate to pre-pandemic. This suggests some “catch-up” accounting for delayed or missed identifications but likely not enough at this point to overcome the pandemic induced deficit.

Keywords

COVID-19 special education secondary data analysis

Increasing evidence shows that the COVID-19 pandemic negatively impacted student learning (e.g., Goldhaber et al., 2022; Jack et al., 2022; Kilbride et al., 2022; Kogan & Lavertu, 2021; Kuhfeld & Lewis, 2022; Nation’s Report Card, 2024; Sass & Ali, 2022). Policymakers and educators have been particularly worried about how the pandemic affected students with disabilities (SWDs; Government Accountability Office, 2020; National Council on Disability, 2021) as there are indications that the pandemic disrupted schools’ and districts’ abilities to maintain access to services and instructional environments consistent with the needs of SWDs. For example, a report by the American Institutes for Research found that early in the pandemic, between May and September of 2020, surveyed districts overwhelmingly indicated they had difficulties complying with the requirements of the Individuals with Disabilities Education Act (IDEA), with high poverty districts more likely to report challenges (Jackson & Bowden, 2020).

As the pandemic continued to interrupt typical schooling and moved both students with and without disabilities into remote learning contexts that limited in-person interactions between educators and students, the process for referring students who may have a disability for initial special education eligibility determination (SEED) was likely impacted. This is because the SEED process is facilitated by students attending in-person instruction, including observation of the student during learning, and providing evidence that the student has received appropriate instruction prior to placement into special education services. Students are required to be given several weeks of scientific, research-based intervention before being evaluated for special education. School teams could not ensure students received this intervention because in-person attendance for school buildings that were open was significantly lower than attendance in virtual schooling and engagement in virtual schooling was significantly lower than in-person schooling (Darling-Aduana et al., 2022). Attendance rates were even lower for Black students, economically disadvantaged students,¹ and SWDs (Darling-Aduana et al., 2022). Moreover, many special education teachers, school psychologists, and speech-language pathologists had to shift assessments to virtual formats or amend assessment practices to account for face masks and/or social distancing requirements (Brunson McClain et al., 2021; Song et al., 2020). These deviations may have rendered the evaluation results inaccurate or invalid. In addition, special education teachers reported higher rates of burnout and attrition than general education teachers even before the pandemic and there has been an ongoing shortage that increases each teacher’s caseloads (Billingsley & Bettini, 2019). As such, the additional stress of the new pandemic environment and needing to coordinate schedules to ensure that Special Education teachers saw each student on their caseload for their individualized education program (IEP)-guaranteed amount of time likely contributed to delayed assessments. It is therefore unsurprising that anecdotal evidence suggests that identification rates for SWDs dropped during the 2020 to 2021 school year, with districts across the country reporting decreases in SEED evaluations (Association of Psychology Training Clinics, 2020).

The process for conducting the evaluations to discontinue students’ special education services, which we call “discontinuation,” also was likely impacted by the pandemic. To discontinue special education services, best practices suggest that schools conduct a problem-solving process in which a multi-disciplinary team monitors students’ progress toward achieving the goals set out in their IEPs and determines that intensive intervention is no longer needed to maintain the student’s academic achievement (Grimes et al., 2006; Powell-Smith & Ball, 2002). However, there is evidence to suggest that SWDs were not provided with the necessary services or appropriate instruction during the pandemic (e.g., hands-on instruction, differentiated instruction), which may have hampered their full participation and ability to access content, thus impacting their learning growth and achievement—key determining factors for the discontinuation of services (Hurwitz et al., 2021; Sonnenschein et al., 2022).

In this study, we use student-level administrative data from Michigan in an interrupted time series (ITS) framework to investigate how the identification of new SWDs and the discontinuation of services progressed over the course of the pandemic. Michigan is a useful context in which to consider these patterns as it is a diverse state with students from a wide range of economic and racial backgrounds. Further, while all schools in Michigan switched to remote schooling in the spring of 2020, districts implemented a mix of different instructional modalities (fully in-person, hybrid, and fully remote) throughout the 2020 to 2021 academic year (Hopkins et al., 2021). This allows us to examine whether any changes in identification and discontinuation rates for SWDs varied across districts’ instructional modalities. While our results are descriptive in that even our fully specified ITS models cannot fully account for unobserved and time-varying factors, they are nonetheless valuable as they provide the first systematic look at the ways in which special education identifications and discontinuations shifted during the COVID-19 pandemic.

Our results show that there was a substantial decrease in new K–5 identifications and discontinuations during the 2019 to 2020 and 2020 to 2021 pandemic-impacted school years. Overall, new identifications deviated from their long-term trend by 0.77 and 0.49 percentage points in these years, respectively, equivalent to a 19% and 12% reduction in identification rates relative to 2018 to 2019 (the year before the onset of the pandemic). Recovery in 2021 to 2022 and 2022 to 2023 was relatively slow, on the other hand, with new entries increasing to 0.14 and 0.28 percentage points above trend (3% and 7% relative to 2018–2019 rates). Thus, while we find some “catch-up” to account for missed or delayed identifications, there remain fewer total new identifications over the course of the pandemic than what we would have expected based on pre-pandemic trends. Similarly, discontinuations in 2019 to 2020 and 2020 to 2021 decreased by 0.15 and 0.10 percentage points relative to 2018 to 2019 (13% and 8% below the 2018–2019 discontinuation rate of 1.2%, respectively) and have only partially recovered as of 2021 to 2022.² This suggests that during the pandemic, access to, or the discontinuation of, special education services was delayed for a substantial share of SWDs. Finally, we see a substantial increase in exits from Michigan public schools for SWDs during the pandemic though by 2021 to 2022 this has essentially returned to trend.

The largest decreases in identifications occurred for two categories of disabilities: speech and language impairments (SLI) and specific learning disabilities (SLD). As with the overall rates, in 2021 to 2022 and 2022 to 2023, new identifications rose above the trend, but there remained a gap leading to only partial catch-up. The pattern is similar for SLI and other health impairments (OHI) showed little evidence of catch-up. Only autism rates seemed to rebound sufficiently to catch-up (and perhaps even go beyond) pre-pandemic trends.

We further look at how the evolution of identifications and discontinuations during the pandemic differed by student characteristics. First, we show that Black students experienced greater reductions in both new identifications and discontinuations than did White, Asian, and Latino students. Similarly, the changes in identification and discontinuation rates for economically disadvantaged students were significantly larger than for non-economically disadvantaged students. These data provide evidence of the inequitable access to and discontinuation of special education services to Black and low-income students during the pandemic.³ Finally, we consider the role of districts’ instructional modality policies (e.g., fully in-person vs hybrid or fully remote learning) and find that students in districts that were remote for a majority of the 2020 to 2021 school year had a 0.88 percentage point lower likelihood of being newly identified with a disability (21% of the mean identification rate the year prior to the pandemic) in that year, and a 0.25 percentage point lower likelihood of being discontinued than students in districts that were mostly in-person. Despite this gap, we see no evidence these “lost” identifications were offset by higher identification rates among remote districts in 2021 to 2022 or 2022 to 2023, and the increase in exits is too small to account for the gap, suggesting that many students who had remote schooling during 2020 to 2021 and should have special education identifications remain in general education.

This paper proceeds as follows. In section “Relevant Literature,” we motivate this study based on the extant literature that outlines the ways in which the failure to accurately identify students for special education services may be harmful. We also discuss the ways in which retaining students in special education programming when they no longer require it can negatively impact students. The section “Data and Methods” describes the Michigan student-level administrative data and our methods of estimating changes in special education identification and discontinuation rates during the pandemic. The section “Results” describes our results. The section “Discussion and Conclusion” concludes with a discussion of results and implications for policymakers.

Relevant Literature

Any potential pandemic-induced delays in SEED and disruptions to special education services could have substantial deleterious impacts on the short- and longer-term achievement and health outcomes of SWDs. SWDs perform better when they are identified earlier in life, providing students with a “foundation for later learning” which then supports future academic achievement (Peltzman, 1992; Steele, 2004). For example, Lovett et al. (2017) find that students with reading disabilities who first received intervention in first or second grade made gains in literacy almost twice that of children first receiving intervention in third grade and continued to grow at faster rates over the following years. Moreover, early intervention reduces the need for intensive special education services in later grades (Kulkarni & Sullivan, 2019). For instance, Walker et al. (1998) show that students with delayed identification of emotional-behavioral disorders often display patterns of disruptive and externalizing behavior that is unremitting and resistant to treatment by the time they are identified.

Early identification and services can avert secondary challenges to students’ long-term development that may arise if SWDs are not identified (Ballis & Heath, 2021; Catts, 1991). For example, children with autism often are first diagnosed after reaching school age and engaging with the education system (van’t Hof et al., 2021). Evidence-based interventions for these students are often provided in schools, and the early application of such programs is critical to improved future outcomes (Peters-Scheffer et al., 2011). As another example, low academic achievement (i.e., illiteracy) directly hampers a person’s access and ability to understand health information and to adhere to therapy and medicine schedules. Low academic achievement is also associated with negative societal and crime outcomes, including a greater likelihood of carrying a weapon and bringing weapons to school (Davis et al., 1999; DeWalt et al., 2004; Vaughn & Wanzek, 2014; World Literacy Foundation, 2018).

Pandemic impacts on the discontinuation of special education services could also harm later academic and mental health outcomes for SWDs. If SWDs did not receive the intensity of instruction that they needed during the pandemic and therefore are not making adequate academic progress, they likely will not meet the achievement criteria necessary to discontinue special education services. SWDs who are not discontinued, and thus receive unnecessary special education services for more of their school career, are at risk of poorer future outcomes (Chesmore et al., 2016). For instance, time spent in school receiving special education instruction or related services (e.g., through time receiving speech, physical, or occupational therapy) displaces academic instruction in the general education classroom, potentially hampering educational attainment (Reynolds & Wolfe, 1999; Setren, 2021). Additionally, disability labeling can lead older students to be stigmatized and bullied, which can be deleterious to their mental health (Rose et al., 2009, 2011).

The effects of pandemic-induced disruptions to SEED and the discontinuation of special education services may have varied for students with, or at risk for, different types of disabilities. For example, disabilities like vision or hearing impairments are more often medically diagnosed and may be identified before students reach school age, allowing for services to be put in place prior to their transition to schooling. Alternatively, students with high-incidence disabilities, including SLD or emotional-behavioral disorders (Francis et al., 1996; Losen & Orfield, 2002; Peterson et al., 2013), require measurement of students’ response to instruction or behavioral intervention before special education evaluation to determine if their learning trajectory is due to a disability or to a lack of high-quality instruction (Fletcher et al., 2019; Lewis et al., 2010). Therefore, SEED for disabilities that require measurement of response to instruction, like SLI and SLD, may have been particularly delayed during the COVID-19 pandemic as in-person instruction, along with the opportunities for high-quality face-to-face instruction and evaluation, was limited.

Further, it is possible that disruptions to SEED and the discontinuation of special education services may have been particularly acute for non-White students and students in schools that educate higher proportions of economically disadvantaged students. Historically, Black and Latino students are more likely to be identified for special education services than their White peers but under-identified in schools with larger proportions of non-White students (Artiles et al., 2002; Elder et al., 2021; Losen et al., 2014; Oswald et al., 1999; Sullivan & Bal, 2013). Additionally, students in schools and districts serving larger proportions of economically disadvantaged students are more likely to be identified with an emotional-behavioral disorder (McCoy et al., 2012). Given that, as elsewhere in the country, urban districts in Michigan were more likely to offer only remote instruction throughout the 2020 to 2021 school year (Hopkins et al., 2021), and these same districts serve a large proportion of non-White and economically disadvantaged students, it is likely that SEED for some Black, Latino, and economically disadvantaged students may have been particularly delayed during the pandemic.

Overall, these studies make clear that SWDs could be greatly impacted if the COVID-19 pandemic hampered schools’ and districts’ abilities to identify students for receipt of special education services at the appropriate times or disrupted their abilities to evaluate SWDs for timely discontinuation of services. Importantly, these negative effects on students who should have qualified for special education services and who were not discontinued from special education services could surface in both the short- and the longer term, causing both immediate harm to students’ educational progress as well as later harm to their social, health, and societal outcomes. As such, it is critical that educators and policymakers better understand the potential impacts of the COVID-19 pandemic on student identification and discontinuation from special education services.

Data and Methods

Data

We use administrative student-level data for over 2.9 million unique K–12 Michigan traditional public and charter school students across 4,082 schools, totaling more than 16 million observations between fall 2012 and spring 2023. These data, provided by the Michigan Department of Education (MDE) and the Center for Educational Performance and Information (CEPI), contain demographic information for each student in the panel (i.e., gender, race/ethnicity, economically disadvantaged status, and English learner status) as well as the current grade, school, and district in which each student is enrolled. Important for this study, these data also provide information on special education status (i.e., a student has an IEP) and primary disability identification. We use this information to identify all years in which a student received special education services under an IEP.⁴

Our main outcomes of interest in our analyses are a set of indicators that capture when SWDs first and last receive special education services in Michigan.⁵ Newly identified SWDs are identified by the first year they received special education services (i.e., “newly identified”). We are able to identify new SWDs in each year through 2022 to 2023 school year, thus allowing us to understand not only changes in identification rates during the pandemic but also in the relatively normal post-pandemic school years (2021–2022 and 2022–2023). We identify former SWD discontinuing special education by the last year they received services. We group these students into two categories: students who were discontinued and their special education status indicator turned off (i.e., “discontinued from SWD to general education [GEN]”), and students who stopped receiving services because they left the public school system and their unique identifier is absent in the remaining years of the time series (i.e., “exited Michigan public schools”). It is particularly important to separate these two groups because the COVID-19 pandemic led to public school enrollment declines in both Michigan and across the country (Dee & Murphy, 2021; Musaddiq et al., 2022). As such, we need to draw distinctions between students who exited special education services because they left public schooling in Michigan and students who discontinued special education services because it was determined they no longer needed an IEP.⁶ We rely on 2022 to 2023 data to assess whether or not the student remained an SWD after the 2021 to 2022 school year, thus constraining our assessment of service discontinuation to end with the 2021 to 2022 school year. Note that for the rest of the paper we refer to the combination of the last two groups as “discontinuation” from special education.

To understand how special education identification and discontinuation rates varied by whether districts were teaching remotely throughout the 2020 to 2021 school year, we merge these data with information on district-level monthly remote, hybrid, or in-person learning status collected by MDE for each Michigan school district that was not operating as a “cyber school”⁷ prior to the pandemic.⁸ For our analysis, we assign students to each remote, hybrid, or in-person learning type based on the version most commonly offered by the student’s district throughout the entire 2020 to 2021 school year. Given that districts were able to offer multiple types each month during the 2020 to 2021 school year, it is possible that districts have multiple “most common” settings across remote/hybrid/in-person (e.g., if a district offered both in-person and remote instruction for all 9 months during the school year). For these cases, we assign students to the “most in-person” option offered by districts (i.e., fully in-person is the “most in-person” option, followed by hybrid, and then fully remote instruction).⁹ We note that while this is not an exact measure of the student’s actual exposure, it is akin to an “intention-to-treat” estimate where the availability of in-person options for the student provides a proxy for the actual type of instruction they experience.

Analytic Sample

We present results using three different samples of Michigan students. To explore global changes in the size of the SWD population, we examine trends using the full population of K–12 students across the state. When investigating unadjusted changes in identification and discontinuation rates by grade band and disability, we focus on the approximately 2.3 million students enrolled in kindergarten through eighth grade because almost 93% of newly identified Michigan SWDs in the last full pre-pandemic school year began receiving special education services prior to entering high school. Finally, in our ITS analyses estimating regression-adjusted trends in identification and discontinuation rates, we limit our sample to the nearly 1.9 million students enrolled in kindergarten through fifth grade to avoid conflating our estimates with effects associated with structural school and district switches that students experience when transitioning from elementary to middle school.

Table 1 provides summary statistics for the K–5 sample of students in three representative school years: well prior to the pandemic (2013–2014), the last pre-pandemic school year (2018–2019), and first full school year during the pandemic (2020–2021). This sample includes all students enrolled in a K–5 grade level at a traditional public or charter school, Intermediate School District (ISD),¹⁰ or state-run school for at least one school year during our sample period. The sample does not include students enrolled in private schools. Overall, K–5 enrollment and the total number of SWDs decreased between 2013 to 2014 and 2020 to 2021. It is important to note, however, that the total number of SWDs decreased at a slower rate than did the full population of students, such that the SWD share of the total population increased over the sample period. This is one characteristic of the data that motivates our use of ITS models discussed in the next subsection.¹¹

Table 1

Descriptive Statistics for Analytic Sample; Grades K–5: 2013 to 2014, 2018 to 2019, and 2020 to 2021

Variables	School year (Fall semester)
Variables	2013	2018	2020
Dependent variables (N)
Total students	688,452	653,342	624,155
SWD	94,627	97,128	93,479
SWD percent of total enrollment (%)	13.7	14.9	15.0
Newly identified SWDs	24,571	26,572	23,924
Newly identified SWDs percent of total enrollment (%)	3.6	4.1	3.8
Discontinued SWDs	8,781	8,165	6,753
Discontinued SWDs percent of total enrollment (%)	1.3	1.2	1.1
SWDs exiting public schools	2,543	2,457	3,221
SWDs exiting percent of total enrollment (%)	0.4	0.4	0.5
Student disability types (% of SWDs)
Speech or language impairment	43.4	41.7	43.9
Specific learning disability	21.9	21.0	19.2
Other health impairment	9.9	11.2	11.0
Autism	7.5	9.3	10.3
Cognitive impairment	6.4	5.6	5.4
Emotional impairment	4.0	4.2	3.6
Other disabilities	7.0	6.9	6.6
Student characteristics (% of all students)
Female	48.4	48.4	48.6
Asian	3.1	3.6	3.7
Black	18.0	18.5	18.5
Latino	8.2	8.7	8.7
Other race	4.9	5.8	6.1
Economically disadvantaged	55.1	56.8	56.8
English learner	7.7	8.9	8.4
School characteristics (school-level %)
Average enrollment (N)	329	319	308
Female	46.8	47.0	47.6
Asian	2.6	2.9	3.0
Black	18.3	19.0	18.6
Latino	7.8	8.3	8.4
Other race	5.3	6.2	6.4
Economically disadvantaged	58.0	60.0	59.7
English learner	6.5	7.6	7.3

Note. The “other race” category includes students who identified as “American Indian or Alaskan Native,” “Native Hawaiian or Pacific Islander,” and “two or more races.” SWD = students with disability.

Special education discontinuation rates—either through switching to general education or leaving the public school system—were relatively consistent before and during the pandemic, however, the new first-time identification rate increased in the years leading up to the pandemic (Table 1). Between 2013 to 2014 and 2018 to 2019, students newly identified for special education services increased from 3.6% to 4.1% (a 14% increase). However, during the first full school year of the pandemic, this drops to 3.8%. A much smaller and consistent share of SWDs were either discontinued from SWD to GEN or left the Michigan public school system between 2013 to 2014 and 2020 to 2021. Rates were very similar both before and during the pandemic—slightly more than 1% of all students were discontinued from SWD to GEN each school year, while SWD who left public schools represent only 0.5% of all Michigan students.

The distribution of disabilities within the Michigan SWD population was, for the most part, stable between 2013 to 2014 and 2020 to 2021, with some notable exceptions. SLI, representing nearly half of all SWDs in the state, was the most common primary disability, followed by SLD and OHI. Notably, students identified with autism rose from 7.5% to 10.3% during the sample period, while students identified with an emotional impairment (EI) decreased from 4.0% to 3.6%. These Michigan-specific trends follow national trends in identification (Kauffman & Badar, 2013; Zablotsky et al., 2019).

Other demographic characteristics were relatively stable during our sample period. The proportion of female students remained roughly constant, while the proportions of non-White, economically disadvantaged, and English learner students increased slightly across the state. We find similar patterns in the school-level shares of each student demographic characteristic.

Methods

For our initial analysis, we explore raw trends in enrollment for the full K–12 population, as well as special education identifications and discontinuations for the K–8 population of Michigan students. Our regression analyses, described below, focus only on K–5 students. To understand how special education identification and discontinuation rates in Michigan changed during the COVID-19 pandemic, we use an ITS framework to identify changes in identification and discontinuation patterns specific to each school year directly impacted by the pandemic. We do so because we hypothesize that identification and discontinuation rates may have differed in each pandemic-impacted school year. Specifically, we estimate the following:

\begin{matrix} y_{i s t} = β_{0} + β_{1} {Trend}_{t} + β_{2} 2019_{t} + β_{3} 2020_{t} \\ + β_{4} 2021_{t} + β_{5} 2022_{t} + Ω X_{i s t} + γ_{s} + ε_{i s t}, \end{matrix}

(1)

where $y_{i s t}$ represents an indicator for whether student i in school s is newly identified with a disability, discontinued from SWD to GEN, or was an SWD who left the Michigan public school system in year t. ${Trend}_{t}$ is the time elapsed (i.e., year) since fall 2012. The variables $2019_{t}$ , $2020_{t}$ , $2021_{t}$ , and $2022_{t}$ represent indicators for school years directly impacted by the pandemic, identified by the fall of that school year.¹² $X_{i s t}$ represents a vector of student (i.e., gender, race/ethnicity, economically disadvantaged status, and English learner status) and school (i.e., student shares by gender, race/ethnicity, economically disadvantaged status, English learner status, as well as school urbanicity) characteristics. $γ_{s}$ is a school-fixed effect. The coefficient on ${Trend}_{t}$ captures the change in average special education identification or discontinuation rates over time. For each of the indicators $2019_{t}$ , $2020_{t}$ , $2021_{t}$ , and $2022_{t}$ , $β_{2}, β_{3}, β_{4},$ and $β_{5}$ represent the year-specific net-change in identification or discontinuation rates relative to the long-term linear trend.

Assuming that identification rates were negatively affected by the initial school building closures in spring 2020, as well as the wide spread provision of remote and hybrid instruction during the 2020 to 2021 school year, we would expect a greater “recovery” or “rebound” in identifications in 2021 to 2022 and 2022 to 2023 as schools “catch-up” and work through their backlog of pre-referral interventions and evaluations delayed by the pandemic. For example, if $β_{2} + β_{3} \leq 0$ , this implies, assuming the pre-existing trend would have continued in the absence of the pandemic, that some students who may be eligible for special education services were not evaluated either at the start of the pandemic or during the first full school year following the initial school building closures. In this case, the direction and magnitude of β₄ + β₅ provides some insight into how well recovery efforts have progressed as most Michigan students moved back to in-person schooling by 2021–22. If β₄ + β₅ < | β₂ + β₃ | then the recovery has not yet caught up to trend and some eligible students may have been “missed” for SEED. If β₄ + β₅ ≥| β₂ + β₃ |, this means that schools have likely worked through their backlog of evaluations and students who were eligible for services during the pandemic have been identified by 2021-22.

Given existing evidence that students of color and economically disadvantaged students experienced larger achievement declines during the COVID-19 pandemic (e.g., Jack et al., 2022; Kilbride et al., 2022; Sass & Ali, 2022), it is important to understand whether special education identification could have played a role. Thus, to understand the heterogeneous effects of the pandemic on identification and discontinuation trends for these students, we extend the ITS specification in model (1) to include interactions with indicators for student-level race/ethnicity and economically disadvantaged status. This allows us to test whether specific groups of students were differentially affected by the pandemic and whether those effects varied over time. Specifically, we estimate:

\begin{matrix} y_{i s t} = β_{0} + β_{1} {Trend}_{t} + β_{2} 2019_{t} + β_{3} 2020_{t} \\ + β_{4} 2021_{t} + β_{5} 2022_{t} + ϕ_{1} 2019_{t} \times {Race}_{i t} \\ + ϕ_{2} 2020_{i t} \times {Race}_{i t} + ϕ_{3} 2021_{i t} \times {Race}_{i t} \\ + ϕ_{4} 2022_{i t} \times {Race}_{i t} + Ω X_{i s t} + γ_{s} + ε_{i t} \end{matrix}

(2a)

and

\begin{matrix} y_{i s t} = β_{0} + β_{1} {Trend}_{t} + β_{2} 2019_{t} + β_{3} 2020_{t} \\ + β_{4} 2021_{t} + β_{5} 2022_{t} + θ_{1} 2019_{t} \times {EconDis}_{i t} \\ + θ_{2} 2020_{i t} \times {EconDis}_{i t} + θ_{3} 2021_{i t} \times {EconDis}_{i t} \\ + θ_{4} 2022_{i t} \times {EconDis}_{i t} + Ω X_{i s t} + γ_{s} + ε_{i t} \end{matrix}

(2b)

where ${Race}_{i}$ is a vector of indicators for each race/ethnicity and ${EconDis}_{i t}$ represents an indicator of whether a student was identified as economically disadvantaged in a particular year. All other variables are defined as in model (1). In these models, the coefficients on each of the indicators $2019_{t}$ , $2020_{t}$ , $2021_{t}$ , and $2022_{t}$ capture the post-COVID net changes in identification or discontinuation rates for the reference group of students (White students for race regressions and non-disadvantaged students for economic regressions) relative to the average underlying linear time trend, while the $ϕ$ and $θ$ coefficients reflect the increase or decrease in the trend deviations during the pandemic for each race relative to White students and economically disadvantaged students relative to non-disadvantaged, respectively.

Finally, while virtually all districts in Michigan operated remotely at the end of spring 2020, there was wide variation in instructional modality during the 2020 to 2021 school year.¹³ As noted above, remote schooling likely limited the ability of schools to provide full evaluations and could have had a disproportionate impact on new identifications and the discontinuation of services. To understand how districts’ 2020 to 2021 instructional modalities are related to identification or discontinuation rates, we again extend model (1) and estimate the following:

\begin{matrix} y_{i j s t} = δ_{0} + δ_{1} {Trend}_{t} + β_{2} 2019_{t} + β_{3} 2020_{t} \\ + β_{4} 2021_{t} + β_{5} 2022_{t} + δ_{1} 2019_{t} \\ \times Modality 2021_{d} + δ_{2} 2020_{i t} \\ \times Modality 2021_{d} + δ_{3} 2021_{i t} \\ \times Modality 2021_{d} + δ_{4} 2022_{i t} \\ \times Modality 2021_{d} + Ω X_{i j s t} + γ_{s} + ε_{i s t} \end{matrix}

(3)

where $Modality 2021$ for district d is a vector of indicators for whether the district offered hybrid or fully remote instructional modalities (fully in-person is the reference category) for a majority of the school year. Again, all other variables are defined as in model (1). In this model, the coefficients on $2019_{t}$ , $2020_{t}$ , $2021_{t}$ , and $2022_{t}$ capture the post-COVID net changes in identification or discontinuation rates for students who attended school in a district that offered fully in-person instruction for a majority of the 2020 to 2021 school year relative to the trend while the $δ_{1}, δ_{2}, δ_{3}$ , and $δ_{4}$ coefficient vectors capture the post-COVID trend deviations for hybrid and remote districts relative to in-person districts.

Results

Descriptive Trends in Overall SWD Population Over Time

To understand overall changes in the Michigan special education population both before and during the pandemic, the bars and line with circular point markers in Figure 1 show how the total SWD population and share of SWDs relative to the total K–12 student population in Michigan, respectively, varied each year of our panel. For the first four 4 years of our sample, the SWD population across the state decreased each year (from 208,783 to 200,890 students) before somewhat stabilizing in the years leading up to the pandemic (in 2018–2019 there were 202,448 SWDs). During the first 2 years of the pandemic, the Michigan SWD population dropped 3.3% to just 195,746 total students. The SWD population then rebounded to near pre-pandemic levels between 2020 to 2021 and 2021 to 2022, increasing by nearly 5,000 students (2.6%). Overall, the total number of SWDs in Michigan decreased from 208,784 to 206,500 students between fall 2012 and spring 2023 (a 1.1% decline).

Even though the K–12 SWD population decreased over the course of our study period, in most years it did so at a slower rate than the full population of students. As a result, the share of Michigan K–12 public school SWD increased from 13.5% to 14.7% between 2012 to 2013 and 2022 to 2023, with a notably larger uptick during the pandemic and in the last year of our sample period, indicating that existing or potential new SWDs were less likely to leave the state’s public schools than were students in general education during the pandemic.

These findings are generally consistent with national trends in special education as well (shown by the line with square point markers in Figure 1). The share of all United States traditional public school students receiving special education services grew slightly faster than the MI SWD population between 2013 and 2014 and from 2019 to 2020. In addition, the proportion of SWDs across the United States increased at about the same rate between 2019 to 2020 and 2020 to 2021 (from 14.4% to 14.5%) relative to Michigan (14.1%–14.2%). Pre-pandemic increasing trends in the SWD population have been observed by other researchers as well, and the SWD population has been growing nationally since 2009 (Durnkin, 2019; Zablotsky et al., 2019).

Descriptive Trends in Overall SWD Identification and Discontinuation Rates Over Time

To understand how patterns in special education identifications, discontinuations, and departures from the school system contributed to these changes in the Michigan SWD population, Figures 2 and 3 document identification and discontinuation rates for the K–8 SWD population between 2012 to 2013 and 2022 to 2023. Figure 2 first shows results for students in all elementary and middle school grade levels, and the remaining figure shows differences by grade band (K–2, 3–5, and 6–8 in the top, middle, and bottom panels of Figure 3, respectively).

Figure 2 shows that Michigan experienced a steady increase in the proportion of newly identified SWDs between 2013 to 2014 and 2018 to 2019, the year prior to the pandemic. There was a sizable decrease in the proportion of newly identified SWDs in the 2019 to 2020 school year (0.46 percentage points), followed by an increase in 2020 to 2021 (0.27 percentage points) that was still below what would have been expected prior to the pandemic. Identifications increased above the pre-pandemic trend in the 2021 to 2022 school year (up 0.59 percentage points), and they continued to rise above prior levels in 2022 to 2023 (up 0.21 percentage points to 3.59%).¹⁴

Figure 2 also shows the rate of SWD discontinuations slowly and steadily declining in the years leading up to the pandemic, including the 2019 to 2020 school year (from 1.20% to 1.07%). However, there was a marked decrease in the proportion of students discontinued from SWD to GEN in 2020 to 2021 and 2021 to 2022 (0.17 percentage points). This proportion appears to return to trend in 2022 to 2023 (1.07%). As expected, we see a sharp increase in the proportion of SWDs exiting Michigan public schools during the pandemic (from 0.36 to 0.51 between 2013–2014 and 2021–2022), and this measure also returns to slightly above trend in 2022 to 2023 (0.41%).

In addition to annual data, we use semester-level data to check whether the timing of SWD identification changes matches the timing of the initial school shutdowns during the pandemic, though the data has several limitations. First, while we can infer SWD status during the fall using student enrollment data collected in October, information on spring data must be inferred from the annual data collected at the end of the school year. That is, we use the difference between the annual data and the fall data to infer whether a student was newly identified in the spring semester. This leaves some measurement error in our determination of when SWD discontinued services during the year as some identifications attributed to the spring semester would have instead been identified during the prior fall semester. Importantly, however, this measurement error ought to be roughly constant across years and thus should not affect our ability to identify a change in identifications due to the pandemic more precisely than with annual data. As such, results presented in Supplemental Appendix Figure A2 and Supplemental Appendix Table A1 (available in the online version of this article) show that the timing of the initial drop in identifications coincides with the onset of the pandemic. Specifically, following the annual trend in Figure 2, we observe a decrease in SWD identifications in spring 2020 whereas in prior years new identifications sharply increased in the spring. At the same time, rates during the fall 2019 semester prior to the pandemic are consistent with prior trends.

As seen in Figure 3, returning to the annual data, general trends in identification, discontinuation, and overall public school exit rates are similar across both the grades K–2 and grades 3 to 5 populations and reflect the averages shown in Figure 2. However, far higher rates of K–2 students are newly identified with a disability than students in grades 3 through 5, reflecting typical variation in the timing of SWD identification. Nonetheless, there is one important difference between the two grade spans during the pandemic: while there was an increase in the proportion of K–2 students identified as SWD in the 2022 to 2023 school year (0.36 percentage point increase to 7.09%), this largely reflects a return to the increasing pre-pandemic trend in identifications, although this appears to have increased above trend in 2022 to 2023. In other words, the increase in identifications in 2021 to 2022 seems in line with where we would have expected K–2 SWD identification rates to be in the absence of the pandemic but does not reflect additional identifications that might be needed to identify students who were “missed” during the pandemic. By contrast, grade 3 through 5 SWD identifications dropped off substantially from 2019 to 2020 (0.50 percentage point decline to 1.44%) before nearly rebounding to pre-pandemic levels in 2020 to 2021 (1.86%) and surpassing them in 2021 to 2022 and 2022 to 2023 (2.30% and 2.44%, respectively).

The bottom panel of Figure 3 shows identification and discontinuation trends for SWDs in sixth through eighth grade in Michigan. As with K–5 students, we see a slightly declining rate of middle school students identified as SWD leading up to the pandemic, decreasing substantially from 2019 to 2020 before rebounding, in this case to slightly above the pre-pandemic trend. Also like earlier grades, the percentage of sixth- through eighth-grade SWDs exiting public middle schools increased from 2019 to 2020 to 2021 to 2022. The percentage of students discontinued from SWD to GEN continued its decreasing trend until 2021 to 2022 (from 0.90 to 0.59%) and returned to trend from 2022 to 2023 (to 0.76%).

Descriptive Trends in SWD Identification Rates Over Time by Disability

There are different criteria by which students qualify for special education services depending on the disability category. In Figure 4, we examine raw K–8 identification trends before and during the pandemic for students with different disabilities. Figure 4 illustrates a slowly increasing trend in most disability categories prior to the pandemic, with a decline in identification during the pandemic and differential recovery in 2020 to 2021, 2021 to 2022, and 2022 to 2023 by disability category. For grades K–8, SLI is the primary diagnosis leading the pandemic “rebound” as these identifications increased beyond the pre-pandemic trend starting in 2021 to 2022.

Supplemental Appendix Figure A3 (available in the online version of this article) shows that SLI is primarily identified in early elementary grades (K–2) and is by far the most common at these grade levels, and thus, the patterns match those in Figure 4.¹⁵ Autism is the next most common disability in early elementary grades. Like identification rates nationally (Zablotsky et al., 2019), autism increased at one of the highest rates between 2013 and 2014 and from 2018 to 2019. The decrease in autism identifications in 2019 to 2020 was smaller than in other categories and the recovery in 2021 to 2022 was higher than pre-pandemic trends would suggest, particularly in grades K–2.

Supplemental Appendix Figures A4 and A5 (available in the online version of this article) demonstrate that SLD is the most common disability in grades 3 to 5 and 6 to 8 and had the largest decrease in new identifications during the pandemic among these grades. Like SLI, following a sharp drop in 2019 to 2020, SLD exceeded pre-pandemic levels for both grade bands by 2021 to 2022. OHI, which is a diverse set of disabilities typically dominated by students with attention deficit hyperactivity disorder (ADHD), is generally the next most common diagnostic category to be identified in both grade bands, and again in this case there was a drop off during the pandemic followed by a robust recovery. Two other smaller categories—cognitive impairment (CI; the Michigan category for students with intellectual disabilities) and EI, however, also saw notable declines in both grade levels but recovery appears to be a bit slower in grades 3 through 5 than for other categories. Notably, though, EI rates were already falling leading into the pandemic, so this relatively slow recovery for the earlier of the two grade bands may simply be the continuation of a longer trend.

ITS Results of Deviations From Pre-Pandemic Trends in Identification and Discontinuation Rates

Our next set of results provides estimates from our ITS models examining changes in identification, discontinuation, and public school exit rates during school years directly impacted by the pandemic. Figure 5 summarizes baseline results for all three outcomes while Figure 6 shows how trends differed across student characteristics and instructional modality policies during the pandemic (identifications, discontinuations, and school exit trends for each student subgroup are shown in the top, middle, and bottom panels of Figure 6, respectively).¹⁶ In the latter figure, we show bar graphs that provide the magnitude and confidence intervals of the estimates from our ITS model on the coefficients for the three pandemic-impacted years (2019–2020, 2020–2021, and 2021–2022) and, for new identifications, the most recent school year (2022–2023).

Figure 5 shows a marked decrease in identifications and discontinuations during the pandemic, while SWDs leaving the public school system increased. The identification rate decreased by 0.77 percentage points from 2019 to 2020, with a slightly smaller decrease relative to pre-trend of 0.49 percentage points in 2020 to 2021 (19% and 12% below the 2018 to 2019 identification rate of 4.1%, respectively). By 2021 to 2022, the estimate is significant and positive, although this is not substantially different from the predicted trend line, indicating that overall identifications have marginally surpassed the prior trend rather than exhibiting a complete “catch-up” through additional identifications.¹⁷ Supplemental Appendix Table A2 (available in the online version of this article) provides coefficient estimates for the underlying regression (column 1) and indicates that the total percent of students whose identifications are delayed or missed during the pandemic, assuming the long-term trend would have continued $(β_{2} + β_{3})$ equals a total pandemic drop of 1.26 percentage points. Thus, the fact that in 2021 to 2022 and 2022 to 2023 identification rates barely returned to trend indicates that many students may still have residual delays in identification following the initial years of the pandemic.

Similarly, discontinuation rates dropped by 0.15 and 0.10 percentage points in 2019 to 2020 and 2020 to 2021 (13% and 8% below the 2018–2019 discontinuation rate of 1.2%, respectively). Discontinuations in 2021 to 2022 increased above pre-pandemic levels, although we find little if any evidence of a “catch-up” for these outcomes. Nonetheless, the results suggest that many students are either not receiving special education services when they should or continuing to receive services when they would have attended and engaged in enough instruction to be discontinued from special education services.

Finally, SWDs exited the Michigan public school system at much higher rates during the pandemic than pre-pandemic school years. In 2019 to 2020 and 2020 to 2021, SWDs were 0.27 and 0.14 percentage points more likely to leave public schools, respectively, compared to pre-pandemic trends (65% and 35% above the 2018–2019 exit rate of 0.4%). In 2021 to 2022 public school exit rates for SWDs returned to pre-pandemic trends.

The top panel of Figure 6 shows how new identifications changed by race/ethnicity, economically disadvantaged status, and remote/hybrid/in-person instruction. Note that each of these groupings is estimated in three separate regression models. Before addressing specific subgroups, we highlight three key overall patterns. First, there are statistically significant reductions in identifications for every subgroup from 2019 to 2020 and 2020 to 2021. Second, all subgroups returned to trend in 2021 to 2022 except for Black students. Third, all subgroups appear to have made substantial rebounds in 2022 to 2023 to identifications above prior trends. Nevertheless, no subgroup has observed the recovery in 2021 to 2022 to be large enough to offset the reductions in identifications during the first 2 years of the pandemic.

Black and Latino students both experienced larger declines in identification rates than White students in 2019 to 2020 and 2020 to 2021. However, the magnitude of the reductions is much larger for Black students. The total deviation from trend for Black student identifications was −1.18 and −1.35 percentage points, respectively, in the first two pandemic-impacted years. This is compared to just −0.65 and −0.21 percentage points for White students, −0.52 and −0.68 percentage points for Asian students, and −0.76 and −0.42 percentage points for Latino students.

The story for economically disadvantaged students is similar to that seen for Black students. Non-disadvantaged students experienced a 0.65 percentage point decline in identifications from 2019 to 2020. This shrunk to −0.19 and then turned positive at 0.27 percentage points in 2020 to 2021 and 2021 to 2022, respectively. Further, this increased by 0.10 percentage points from 2022 to 2023. Economically disadvantaged students experienced considerably greater decreases in identification rates during the pandemic than did their wealthier peers (−0.85, −0.71, 0.05, and 0.40 percentage points in these 4 years, respectively).

The last set of comparisons in the panel shows that new identifications were −0.36 percentage points below trend in 2020 to 2021 for districts where in-person instruction was offered for the majority of the same year, with only slightly lower rates in districts that mostly offered hybrid instruction. However, the reduction in new identifications was more than two times larger for districts that operated remotely for the majority of the year (−0.88 percentage points).¹⁸ This reflects a negative deviation from trend of 21% off the 2018 to 2019 average identification rate. Nonetheless, districts on average returned to trend by 2021 to 2022 regardless of remote or other instruction. However, since they only return to trend, there is a substantial remaining gap relative to identification that likely would have occurred in the absence of the pandemic. Some of this could be explained by exits from Michigan public schools but in Supplemental Appendix Table A5 (available in the online version of this article) we see that remote exits were only 0.12 percentage points higher than in-person during the 2019 to 2020 school year and no-different afterward. Thus, it appears that at least some students who were in districts that were primarily remote from 2020 to 2021 who we would expect to have a special education identification, remain unidentified.

The middle panel of Figure 6 looks at differences in discontinuation rates during the pandemic by race/ethnicity, economically disadvantaged status, and instructional modality. We find many of the same trends as those shown in the top panel. Again, Black (−0.25 and −0.26 percentage points from 2019 to 2020 and from 2020 to 2021, respectively) and economically disadvantaged students (−0.20 and −0.17 percentage points) experienced the largest declines in discontinuation rates relative to their respective peers. Additionally, other than White students, Black and economically disadvantaged students were the only two sociodemographic subgroups to experience statistically significant declines in discontinuations in both 2019 to 2020 and 2020 to 2021. Similar to the previous results, students in districts that offered in-person instruction for a majority of the 2020 to 2021 school year saw the smallest reductions in the propensity to be discontinued from SWD to GEN (−0.14 and −0.04 percentage points in 2019–2020 and 2020–2021), followed by those students in mostly hybrid (−0.17 and −0.16 in 2019–2020 and 2020–2021) and remote districts (−0.12 and −0.29 percentage points in 2019–2020 and 2020–2021).

Finally, the bottom panel of Figure 6 shows the pandemic-induced changes in SWDs’ propensity to leave Michigan public schools during the pandemic by demographics and district-level instruction modality. We find that Black and Latino SWDs were less likely than White SWDs to exit public schools in both 2019 to 2020 (0.16 and 0.23 percentage points for Black and Latino students, respectively) and 2020 to 2021 (0.10 and 0.11 percentage points). Exit trends for economically disadvantaged students and their more advantaged peers were generally similar across years. Lastly, the exit propensity of students in districts that offered in-person instruction for a majority of the 2020 to 2021 school year increased in both 2019 to 2020 and 2020 to 2021, by 0.28 and 0.13 percentage points, respectively. Students in remote districts were significantly more likely to leave than in-person students after 2020 to 2021 (0.04 percentage points, respectively).

Discussion and Conclusion

The COVID-19 pandemic had substantial negative impacts on K–12 students’ developmental, physical, and mental health. While the pandemic’s effects were worse for students of color, economically disadvantaged students, and those who attended school remotely, there is little evidence detailing the scope of the pandemic’s impact on SWDs. Specifically, little is known about the ways in which SEED and the discontinuation of services were affected. Given the well-documented importance of early and appropriate service provision for SWDs for their long-term mental and physical health and the ways in which early academic achievement serves as a protective factor for social and health outcomes later in life, it is imperative that we understand the scope of this problem so that policymakers can direct the necessary resources toward SWDs.

Overall, our results indicate that Michigan students were less likely to be identified with a disability and less likely to be discontinued from receiving special education services during the height of the COVID-19 pandemic. For some disabilities like SLI and autism, new identifications in 2021 to 2022 and 2022 to 2023 exceeded those expected based on pre-pandemic trends, indicating that there was some “catch-up” of the backlog of non-identified students. In other cases, identifications merely returned to trend suggesting that some students may simply never have been identified even if they should have been. Given the increasing trend in identifications from 2022 to 2023, it is possible that identifications will continue to increase from 2023 to 2024.

The dip in SWD identifications and discontinuations in both the 2019 to 2020 and 2020 to 2021 school years is likely due to pandemic-related interruptions to the 2019 to 2020 school year which likely disrupted the initial evaluation processes and re-evaluation of students for discontinuation that occurred during a typical school year. As noted earlier, the SEED process often relies on students attending school so that educators can observe student learning, can ensure that students have received appropriate instruction and that students have attended several weeks of pre-referral intervention prior to placement in special education services. Moreover, discontinuation from special education services is hampered by lower attendance and engagement rates in instruction that may have occurred due to COVID-19 spread in in-person learning and lack of engagement in virtual learning. Further, the decrease in discontinuations from special to general education may be driven by a variety of similar factors that disrupted special education services for SWDs and created a lack of evidence (e.g., absence of progress monitoring data) to discontinue special education services in the following years. On top of these logistical challenges, and to put it plainly, educators and families were living through a global pandemic and many important events were curtailed simply because other things—such as physical safety and mental health—were more pressing at the time.

In addition to the possibility that declines were driven by under-identification in schools, there are several other potential mechanisms that could have caused the observed decrease. For example, students with an unidentified disability may have been disproportionately likely to leave the school system. These students may have transferred to private schools or participated in homeschooling, and it is not clear whether subsequent increases in identification could be due to these students re-entering the public school system.

Regarding the discontinuation of services, another possible mechanism for the observed decline is that students may not have been able to complete the required assessments to exit SWD status. Similarly, schools might have slowed the delivery of existing services, which could have prevented students from making enough progress to leave. While we are unable to directly observe this possibility, it is important to note that these and other mechanisms, in addition to school-side factors, likely contributed to decreased identifications and discontinuations during the pandemic.¹⁹

Even if districts eventually “catch-up” in providing identifications to all students who were unable to receive an appropriate identification during the pandemic, the sharp declines from 2019 to 2020 and from 2020 to 2021 raise significant concerns for these students as the initial receipt of special education services would have been delayed by at least 1 to 2 years. Given the importance of early intervention, this delay in receiving intensive intervention could have significant impacts on long-term student achievement and additional outcomes. Future research should continue to follow these students to determine how this delayed identification relates to long-term outcomes.

Additional research should also explore why identifications for some disabilities were differently impacted compared to others. For K–5 students, disabilities that are often diagnosed when students reach school age (e.g., SLI and ADHD within OHI) decreased at a steeper rate than those disabilities that are often detected before school entry (e.g., CI and the seven disabilities included in our “other” grouping). There are several potential explanations for these differences in identification. For example, some disabilities (e.g., hearing impairments, vision impairments, and CI) require a shorter testing battery in the evaluation to meet diagnostic conditions and do not require a period of 6 to 8 weeks of specialized instruction for a diagnosis (Individuals with Disabilities Education Improvement Act [IDEA], 2004), and, thus, could have been identified outside of the educational environment (e.g., remote testing) even during the pandemic. Other disabilities, like autism, rely on behavioral criteria that may allow for initial identification in a clinical or medical setting, which may then provide parents and schools with preliminary data to pursue an IEP rather than waiting for data to appear in the educational environment. Potentially due to this, we see a more muted change in autism rates relative to other disabilities. An alternative explanation is that autism is primarily behaviorally based, includes impairments in reciprocal social interactions and communication, as well as a restricted range of interests or repetitive behavior, and must adversely affect educational performance in academic, behavioral, or social domains. Thus, rather than relying solely on the impact on academic performance, the expanded ability to also consider behavioral and social performance may have facilitated the evaluation team’s ability to identify a student with autism within the restrictive pandemic environment.

Alternatively, disabilities like SLI and SLD rely on the evaluation of academic performance which was severely disrupted for all students during the pandemic. For SLI, there is the additional issue that the pandemic itself may have generated language delays due to factors such as masking and limited ability to see teachers’ faces during remote schooling sessions. This would be consistent with the patterns we see where the “catch-up” in SLI could also be due to an increase in underlying incidence.

With respect to SLD, other disabilities have clearer criteria; the criteria to qualify for SLD is not the same across schools and interpretation of “unexpected low achievement” would necessarily change in the context of disrupted schooling. The combination of no standardized criteria and uncertainty about what constitutes unexpected low achievement likely exacerbated variability in diagnosis. Given the universal disruption, it was likely more difficult for schools to identify those students who truly had an SLI/SLD versus those students who were struggling because of the change in learning modality.

The diagnosis of SLD is further compounded by the inability of some popular SEED diagnostic procedures to detect SLD before grade 3 (Miciak & Fletcher, 2020). Research suggests that some students who are diagnosed with SLD in later elementary grades (3 through 5) are often initially diagnosed with SLI in early elementary grades (K through 2) because the SLI diagnostic criteria are “easier” to meet (Georgan et al., 2023). In the current study, the greater recovery rate of SLI in grades K through 2 combined with the greater recovery rates of SLD in grades 3 through 5 provide some correlational evidence that schools are attempting to provide some services to SWDs more rapidly. The current analysis does not provide causal evidence for this interpretation and future research is needed in this area. However, it does provide consistent policy implications that state and SEED stakeholders consider the limitations of diagnostic criteria that delay identification of SLD to later elementary grades (i.e., patterns of strengths and weaknesses) and adopt processes that allow for earlier detection of SLD in early elementary grades (i.e., response to intervention and hybrid models; Miciak & Fletcher, 2020). Improving policy that identifies SLD in earlier grade levels fits with other policy initiatives in Michigan and other states to identify dyslexia (one of the categories of SLD) in earlier grades where research shows that instruction is most effective (e.g., Hall et al., 2023).

Critically, the delays in identification also differed by several key student variables (e.g., race/ethnicity, economically disadvantaged status, instructional modality), demonstrating the intersectionality of disability with other disadvantaged groups. Black students, in particular, experienced considerably larger reductions in identifications during the first two pandemic-impacted school years relative to their White, Latino, and Asian peers. We also found no indication that the recovery was any faster for Black students. We find similar results for economically disadvantaged students relative to their more advantaged peers. While we do not know why identifications fell more for these groups, it is nonetheless consistent with other evidence that marginalized groups have suffered more during the pandemic in terms of health and achievement (Goldhaber et al., 2022; Kilbride et al., 2022; Kuhfeld & Lewis, 2022).

Finally, we consider how delays in identification vary with the extent to which districts offered in-person schooling during the 2020 to 2021 school year. As noted above, evaluations for many conditions require in-person assessments, and remote instruction likely constrained educators’ abilities to administer and interpret such assessments, as well as provide the 6 to 12-week intervention phase of the special education identification process. Further, even if in-person evaluations are not strictly required for identification, teachers in remote or hybrid settings were likely hampered in their ability to observe potential disabilities and refer students. On the other hand, parents may have had better insights into their own children’s learning behaviors when they were learning at home relative to inside a school building, providing them with greater opportunity to raise concerns to their schools. Though we stress that these estimates are not causal, given the choice of a district to return to in-person education is likely related to many other factors, we nonetheless find that students in districts that were remote for a majority of the 2020 to 2021 school year were three times less likely to be identified with a disability during that year relative to their peers in districts that were in person for the majority of the school year. From 2022 to 2023, identifications for previously remote districts (all districts in Michigan returned to in-person instruction in 2021–2022) reverted to trend but it does not appear that the entire backlog has been addressed. Thus, it is possible that if identifications do not continue to increase from 2023 to 2024, some of these students may remain permanently unidentified.

Given these findings, it is important that policymakers are cognizant of students whose identifications may have been delayed or missed altogether. To ensure that these students receive the services they need to address their disabilities, districts will need resources to expand their screening efforts, expand the quality of their pre-referral instruction and intervention provision, and update SLD identification practices to make sure that instruction to SWD accounts for the time lost from the delayed identifications. Further research should investigate how the recovery in special education identification rates progresses over the next few years and what interventions can be applied.

Given the discrepancies in identification rates across students by race/ethnicity and economic disadvantage status, it will be imperative for school districts and state agencies to pay particular attention to students who are more at risk for delayed or missed identification. Since late identification can lead to academic and behavioral challenges in schooling, other policies adjacent to but not specifically about special education should be considered in light of the increased probability that lower-income and Black students were less likely to be identified for necessary services on time. For instance, educators and policymakers may wish to consider changes to discipline policies and to programs that aid with socio-emotional learning to address what may be increased behavioral challenges among groups of students who were inadequately served during the pandemic.²⁰

Finally, we would like to note that something that the pandemic continued to highlight is the challenges associated with delivering an evidence-based intervention implemented with fidelity for 6 to 12 weeks or more, as a critical part of the special education identification process. The intervention process of special education identification is considered best practice (Fletcher et al., 2019) and has been part of the federal IDEA law and state laws since 2004 (Zirkel & Thomas, 2010). However, schools have not been successful in implementing the intervention component because it requires funding, resources, and unique training for schools to coordinate universal assessment, and specialized intervention from trained staff; therefore, it is likely that a large number of schools do not implement intervention (D. Fuchs & Fuchs, 2017). Indeed, only 51% of school psychologists report using intervention in their evaluation process some at the time (Benson et al., 2020). However, it is not possible to know how many schools or individual student evaluations use an intervention process because there is no systematic data collection at the regional or state level about the use of the intervention process in Special Education identification. Addressing these issues will require funds, resources, and a new data system.

In conclusion, the pandemic-impacted students in many different dimensions. SWDs were at particularly high risk of poor outcomes from the educational disruption that resulted. Using data from Michigan, we show that many students were delayed in their access to special education services through new identifications and IEPs and through discontinuation to general education. While this analysis looks at a key input into these students’ educational progress, it is likely that the sharp reductions in identifications during the pandemic—which created a backlog that as of 2022 to 2023 has not been fully worked through—impacted academic and behavioral development. It is incumbent upon future research to investigate the impacts of these outcomes.

Supplemental Material

sj-pdf-1-epa-10.3102_01623737241274799 – Supplemental material for Trends in Special Education Identification During the COVID-19 Pandemic: Evidence From Michigan

Supplemental material, sj-pdf-1-epa-10.3102_01623737241274799 for Trends in Special Education Identification During the COVID-19 Pandemic: Evidence From Michigan by Bryant G. Hopkins, Matthew Guzman, Scott A. Imberman, Adrea J. Truckenmiller, Katharine O. Strunk and Marisa H. Fisher in Educational Evaluation and Policy Analysis

Supplemental Material

sj-pdf-2-epa-10.3102_01623737241274799 – Supplemental material for Trends in Special Education Identification During the COVID-19 Pandemic: Evidence From Michigan

Supplemental material, sj-pdf-2-epa-10.3102_01623737241274799 for Trends in Special Education Identification During the COVID-19 Pandemic: Evidence From Michigan by Bryant G. Hopkins, Matthew Guzman, Scott A. Imberman, Adrea J. Truckenmiller, Katharine O. Strunk and Marisa H. Fisher in Educational Evaluation and Policy Analysis

Footnotes

Acknowledgements

We greatly appreciate feedback from audiences at the Association for Public Policy Analysis and Association for Education Finance and Policy annual meetings. This research result used data structured and maintained by the Michigan Education Research Institute’s (MERI) Michigan Education Data Center (MEDC). MEDC data is modified for analysis purposes using rules governed by MEDC and are not identical to those data collected and maintained by the Michigan Department of Education (MDE) and/or Michigan’s Center for Educational Performance and Information (CEPI). Results, information, and opinions solely represent the analysis, information, and opinions of the author(s) and are not endorsed by, or reflect the views or positions of, grantors, the Institute of Education Sciences, the U.S. Department of Education, MDE, CEPI, or any employee thereof.

Authors’ Note

Bryant G. Hopkins is now affiliated to Bates-White Consulting. Katharine O. Strunk is now affiliated to University of Pennsylvania.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The research reported here was supported by the Institute of Education Sciences, U.S. Department of Education, through Grant R305B200009 to Michigan State University. This work was also supported by grants received from the State of Michigan and private philanthropy to the Education Policy Innovation Collaborative.

ORCID iDs

Scott A. Imberman

Adrea J. Truckenmiller

Marisa H. Fisher

Notes

Authors

BRYANT G. HOPKINS, PhD, is a senior economist in the Mass Torts practice at Bates White Economic Consulting. His research focuses on student outcomes following the COVID-19 pandemic.

MATTHEW GUZMAN, BA, is a PhD student at Michigan State University. His research focuses on career and technical education and student labor market outcomes.

SCOTT A. IMBERMAN, PhD, is a professor of economics and education policy at Michigan State University. His research focuses on the economics of education with specific interests in special education students, school choice, the economic returns to higher education, and the intersection of health and education markets.

ADREA J. TRUCKENMILLER, PhD, is an associate professor of special education at Michigan State University. Her research focuses on the evidence-based instructional practices in reading, writing, and language and the school systems decisions that facilitate outcomes in those areas.

KATHARINE O. STRUNK, PhD, is the Dean and George and Diane Weiss Professor of Education in the Graduate School of Education at the University of Pennsylvania. Her research focuses on working with state and district education agencies to effectively use data to inform and improve policy and practice, particularly focused on policies related to the educator workforce, accountability, and instruction.

MARISA H. FISHER, PhD, is an associate professor of special education at Michigan State University. Her research examines the psychosocial outcomes of individuals with intellectual and developmental disabilities and the implementation of educational interventions to improve those outcomes.

References

Artiles

A. J.

Harry

Reschly

D. J.

Chinn

P. C.

(2002). Over-identification of students of color in special education: A critical overview. Multicultural Perspectives, 4(1), 3–10.

Association of Psychology Training Clinics. (2020). The Association of Psychology Training Clinics Assessment Workgroup, assessment training considerations during COVID-19. https://www.aptc.org/images/files/aptc_awg_recommendations-final.pdf

Ballis

Heath

(2021). The long-run impacts of special education. American Economic Journal: Economic Policy, 13(4), 72–111.

Bassok

Michie

Mayaris Cubides-Mateus

Doromal

B. J.

Kiscaden

(2020). The divergent experiences of early educators in schools and child care centers during COVID-19: Findings from Virginia. Ed Policy Works at the University of Virginia.

Billingsley

Bettini

(2019). Special education teacher attrition and retention: A review of the literature. Review of Educational Research, 89(5), 697–744.

Brunson McClain

Roanhorse

T. T.

Harris

Heyborne

Zemantic

P. K.

Azad

. (2021). School-based autism evaluations in the COVID-19 era. School Psychology, 36(5), 377–387.

Catts

H. W.

(1991). Early identification of reading disabilities. Topics in Language Disorders, 12(1), 1–16.

Chesmore

A. A.

S.-R.

Reynolds

A. J.

(2016). Childhood placement in special education and adult well-being. The Journal of Special Education, 50(2), 109–120.

Darling-Aduana

Woodyard

H. T.

Sass

T. R.

Barry

S. S.

(2022). Learning-mode choice, student engagement, and achievement growth during the COVID-19 pandemic. AERA Open, 8, 23328584221128035.

10.

Davis

T. C.

Byrd

R. S.

Arnold

C. L.

Auinger

Bocchini

J. A.

(1999). Low literacy and violence among adolescents in a Summer Sports Program. Journal of Adolescent Health, 24, 403–411.

11.

Dee

T. S.

Murphy

(2021). Patterns in the pandemic decline of public school enrollment. Educational Researcher, 50(8), 566–569.

12.

DeWalt

D. A.

Berkman

N. D.

Sheridan

Lohr

K. N.

Pignone

M. P.

(2004). Literacy and health outcomes: A systematic review of the literature. Journal of General Internal Medicine, 19(12), 1228–1239.

13.

Durnkin

M. S.

(2019). Increasing prevalence of developmental disabilities among children in the US: A sign of progress?. Pediatrics, 144(4), e20192005.

14.

Elder

T. E.

Figlio

D. N.

Imberman

S. A.

Persico

C. L.

(2021). School segregation and racial gaps in special education identification. Journal of Labor Economics, 39(Suppl 1), 151–197.

15.

Fletcher

J. M.

Lyon

G. R.

Fuchs

L. S.

Barnes

M. A.

(2019). Learning disabilities: From identification to intervention (2nd ed.). Guilford Press.

16.

Francis

D. J.

Shaywitz

S. E.

Stuebing

K. K.

Shaywitz

B. A.

Fletcher

J. M.

(1996). Developmental lag versus deficit models of reading disability: A longitudinal, individual growth curves analysis. Journal of Educational Psychology, 88(1), 3.

17.

Fuchs

L. S.

(2017). Critique of the national evaluation of response to intervention: A case for simpler frameworks. Exceptional Children, 83(3), 255–268.

18.

Fuchs

L. S.

Fuchs

(2009). On the importance of a unified model of responsiveness to intervention. Child Development Perspectives, 3(1), 41–43.

19.

Georgan

W. C.

(2022). What’s in a name? Investigating labels and legislation in education for children with developmental language disorder. Harvard University ProQuest Dissertations Publishing.

20.

Georgan

W. C.

Archibald

L. M.

Hogan

T. P.

(2023). Speech/Language impairment or specific learning disability? Examining the usage of educational categories. Journal of Speech, Language, and Hearing Research, 66(2), 656–667.

21.

Goldhaber

Kane

T. J.

McEachin

Morton

Patterson

Staiger

D. O.

(2022). The consequences of remote and hybrid instruction during the pandemic (No. w30010). National Bureau of Economic Research.

22.

Government Accountability Office. (2020). Distance learning: Challenges providing services to K-12 English learners and students with disabilities during COVID-19 (GAO-21-43). https://www.gao.gov/products/gao-21-43

23.

Grimes

Kurns

Tilly

W. D.

III (2006). Sustainability: An enduring commitment to success. School Psychology Review, 35(2), 224–244.

24.

Hall

Dahl-Leonard

Cho

Solari

E. J.

Capin

Conner

C. L.

Kehoe

K. F.

(2023). Forty years of reading intervention research for elementary students with or at risk for dyslexia: A systematic review and meta-analysis. Reading Research Quarterly, 58(2), 285–312.

25.

Hopkins

Kilbride

Strunk

(2021, May). Instructional delivery under Michigan districts’ extended COVID-19 learning plans—May update. Education Policy Innovation Collaborative. https://epicedpolicy.org/ecol-reports/

26.

Hurwitz

Garman-McClaine

Carlock

(2021). Special education for students with autism during the COVID-19 pandemic: “Each day brings new challenges.” Autism, 26(4), 889–899.

27.

Individuals with Disabilities Education Improvement Act. (2004). CFR. 34 CFR §300.309. https://sites.ed.gov/idea/regs/b/d/300.309/a1400

28.

Jack

Halloran

Okun

Oster

(2022). Pandemic schooling mode and student test scores: Evidence from US school districts. American Economic Review: Insights, 5(2), 173–190.

29.

Jackson

Bowden

(2020). National survey of public education’s response to COVID-19: Spotlight on students with disabilities (Research Brief). American Institutes for Research. https://www.air.org/sites/default/files/COVID-Survey-Spotlight-on-Students-with-Disabilities-FINAL-Oct-2020.pdf

30.

Kauffman

J. M.

Badar

(2013). How we might make special education for students with emotional or behavioral disorders less stigmatizing. Behavioral Disorders, 39(1), 16–27. https://doi.org/10.1177/019874291303900103

31.

Kilbride

Hopkins

Strunk

(2022, April). Michigan’s Fall 2021 benchmark assessments. Education Policy Innovation Collaborative. https://epicedpolicy.org/wp-content/uploads/2022/04/Benchmark_Report_April2022.pdf

32.

Kogan

Lavertu

(2021, January 27). The COVID-19 pandemic and student achievement on Ohio’s third-grade English language arts assessment. https://glenn.osu.edu/covid-19-pandemic-and-student-achievement-ohios-third-grade-english-language-arts-assessment

33.

Kuhfeld

Lewis

(2022). Student achievement in 2021–2022: Cause for hope and continued urgency. NWEA. https://www.nwea.org/uploads/2022/07/Student-Achievement-in-2021-22-Cause-for-hope-and-concern.researchbrief-1.pdf

34.

Kulkarni

Sullivan

A. L.

(2019). The relationship between behavior at school entry and services received in third grade. Psychology in the Schools, 56, 809–823.

35.

Lewis

T. J.

Jones

S. E. L.

Horner

R. H.

Sugai

(2010). School-wide positive behavior support and students with emotional/behavioral disorders: Implications for prevention, identification and intervention. Exceptionality, 18(2), 82–93.

36.

Losen

D. J.

Hodson

Martinez

(2014). Disturbing inequities: Exploring the relationship between racial disparities in special education identification and discipline. Journal of Applied Research on Children, 5(2), 15.

37.

Losen

D. J.

Orfield

(2002). Racial inequity in special education. Harvard Education Publishing Group.

38.

Lovett

Frijters

J. C.

Wolf

Steinbach

K. A.

(2017). Early intervention for children at risk for reading disabilities: The impact of grade at intervention and individual differences on intervention outcomes. Journal of Educational Psychology, 109(7), 889–914.

39.

McCoy

Banks

Shevlin

(2012). School matters: How context influences the identification of different types of special educational needs. Irish Educational Studies, 31(2), 119–138.

40.

Miciak

Fletcher

J. M.

(2020). The critical role of instructional response for identifying dyslexia and other learning disabilities. Journal of Learning Disabilities, 53(5), 343–353.

41.

Musaddiq

Stange

Bacher-Hicks

Goodman

(2022). The pandemic’s effect on demand for public schools, homeschooling, and private schools. Journal of Public Economics, 212. The Nation’s Report Card. Retrieved February 2, 2022, from https://www.nationsreportcard.gov/

42.

National Center for Education Statistics. (2023). Digest of education statistics. https://nces.ed.gov/programs/digest/

43.

National Council on Disability. (2021). The impact of COVID-19 on people with disabilities. https://ncd.gov/progressreport/2021/2021-progress-report

44.

Nation’s Report Card. (2024). NAEP long-term trend assessment results: Reading and Mathematics. https://www.nationsreportcard.gov/ltt/?age=9

45.

Oswald

D. P.

Coutinho

M. J.

Best

A. M.

Singh

N. N.

(1999). Ethnic representation in special education: The influence of school-related economic and demographic variables. The Journal of Special Education, 32(4), 194–206.

46.

Peltzman

B. R.

(1992). Guidelines for early identification and strategies for early intervention of at-risk learning disabled children. Education Resources Information Center.

47.

Peterson

C. A.

Wall

Jeon

H.-J.

Swanson

M. E.

Carta

J. J.

Luze

G. J.

Eshbaugh

(2013). Identification of disabilities and service receipt among preschool children living in poverty. The Journal of Special Education, 47(1), 28–40.

48.

Peters-Scheffer

Didden

Korzilius

Sturmey

(2011). A meta-analytic study on the effectiveness of comprehensive ABA-based early intervention programs for children with Autism Spectrum Disorders. Research in Autism Spectrum Disorders, 5(1), 60–69.

49.

Powell-Smith

K. A.

Ball

P. L.

(2002). Best practices in reintegration and special education exit decisions. In Thomas

Grimes

(Eds.), Best practices in school psychology IV (pp. 535–552). National Association of School Psychologists.

50.

Reynolds

A. J.

Wolfe

(1999). Special education and school achievement: An exploratory analysis with a central-city sample. Educational Evaluation and Policy Analysis, 21(3), 249–269.

51.

Rose

C. A.

Espelage

D. L.

Monda-Amaya

L. E.

(2009). Bullying and victimization rates among students in general and special education: A comparative analysis. Educational Psychology, 29(7), 761–776.

52.

Rose

C. A.

Monda-Amaya

L. E.

Espelage

D. L.

(2011). Bullying perpetration and victimization in special education: A review of the literature. Remedial and Special Education, 32(2), 114–130.

53.

Sass

Ali

(2022). Student achievement growth during the COVID-19 pandemic: Spring 2022 update. Georgia Policy Labs. https://scholarworks.gsu.edu/cgi/viewcontent.cgi?article=1040&context=gpl_reports

54.

Setren

(2021). Targeted vs. general education investments: Evidence from special education and English language learners in Boston charter schools. Journal of Human Resources, 56(4), 1073–1112.

55.

Song

S. Y.

Wang

Espelage

D. L.

Fenning

Jimerson

S. R.

(2020). COVID-19 and school psychology: Adaptations and new directions for the field. School Psychology Review, 49(4), 431–437.

56.

Sonnenschein

Stites

M. L.

Grossman

J. A.

Galczyk

S. H.

(2022). “This will likely affect his entire life”: Parents’ views of special education services during COVID-19. International Journal of Educational Research, 112, 101941.

57.

Steele

M. M.

(2004). Making the case for early identification and intervention for young children at risk for learning disabilities. Early Childhood Education Journal, 32(2), 75–79.

58.

Sullivan

A. L.

Bal

(2013). Disproportionality in special education: Effects of individual and school variables on disability risk. Exceptional Children, 79(4), 475–494.

59.

Van’t Hof

Tisseur

van Berkelear-Onnes

van Nieuwenhuyzen

Daniels

A. M.

Deen

Hoek

H. W.

Ester

W. A

. (2021). Age at autism spectrum disorder diagnosis: A systematic review and meta-analysis from 2012 to 2019. Autism, 25(4), 862–873.

60.

Vaughn

Wanzek

(2014). Intensive interventions in reading for students with reading disabilities: meaningful impacts. Learning Disabilities Research and Practice, 29(2), 46–53.

61.

Walker

H. M.

Kavanagh

Stiller

Golly

Severson

H. H.

Feil

E. G.

(1998). First step to success: An early intervention approach for preventing school antisocial behavior. Journal of Emotional and Behavioral Disorders, 6(2), 66–80.

62.

Walker

H. M.

Shinn

M. R.

(2002). Structuring school-based interventions to achieve integrated primary, secondary, and tertiary prevention goals for safe and effective schools. Interventions for Academic and Behavior Problems II: Preventive and Remedial Approaches, 44, 1–25.

63.

World Literacy Foundation. (2018). The economic and social cost of illiteracy: A white paper by the World Literacy Foundation. https://worldliteracyfoundation.org/wp-content/uploads/2021/07/TheEconomicSocialCostofIlliteracy-2.pdf

64.

Zablotsky

Black

L. I.

Maenner

M. J.

Schieve

L. A.

Danielson

M. L.

Bitsko

R. H.

Blumberg

S. J.

Kogan

M. D.

Boyle

C. A.

(2019). Prevalence and trends of developmental disabilities among children in the United States: 2009–2017. Pediatrics, 144(4), e20190811. https://doi.org/10.1542/peds.2019-0811

65.

Zirkel

P. A.

Thomas

L. B.

(2010). State laws for RTI: An updated snapshot. Teaching Exceptional Children, 42(3), 56–63.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.08 MB

0.15 MB

0.61 MB