Abstract
Purpose:
Among patients with sickle cell disease (SCD), poor academic attainment in adolescence is associated with a greater frequency of acute care visits and a poorer health-related quality of life in adulthood. We compared the academic performance of high school students with SCD to students without SCD after controlling for race and sex.
Methods:
This retrospective cohort study was constructed based on Tennessee Medicaid program claims data and Tennessee Department of Education Race to the Top educational data from 10/01/2007 to 09/30/2019. For every ninth grade student with SCD identified, eight students without SCD were identified. Academic achievement was measured using rates of proficiency on End-of-Course assessments. The number of days absent per school year, grade retention, and school withdrawal were used as measures of academic attainment.
Results:
The analysis included 498 students with and 3188 students without SCD. After race and sex were adjusted for, students with SCD had English proficiency odds of 0·8 times (95% confidence interval [CI], 0·7–0·9; p<0·001) those of students without SCD. Students with SCD were 1·2 times (95% CI, 1·0–1·5; p=0·045) more likely to be retained in the ninth grade. The effect of SCD on grade retention was mediated by days absent from school (odds ratio for indirect effect, 1·18; 95% CI, 1·1–1·26; p<0·0001).
Discussion:
SCD negatively affects academic performance partly through frequent school absence.
Keywords: sickle cell, academic, education, high school, neurocognitive
Introduction
Sickle cell disease (SCD) is an inherited blood disorder caused by a variant in the hemoglobin gene and can cause multiple medical complications, including vaso-occlusive crises, silent cerebral infarcts, and stroke.1,2 In the United States (US), SCD primarily affects individuals of African descent, occurring in approximately 1 in 400 Black individuals.3 Students with SCD show increasing cognitive deficits as they age,4 which strongly contributes to poor academic achievement.5 Frequently missed school due to hospital visits, unplanned hospitalizations, and chronic pain can further interfere with academic attainment, including progression to the next level of education.6,7 In patients with SCD, poor academic achievement and attainment in adolescence is associated with a greater frequency of acute care visits8 and a poorer health-related quality of life in adulthood.9
Students with SCD perform worse than do siblings and demographically-matched controls on standardized academic achievement measures of math and reading.5 Data suggest that students with SCD are retained in grade10 and withdraw11 from school at a higher rate than their peers, but these studies lacked controls. In the US, Black students obtain lower test scores and are more likely to be retained in grade than their White or Hispanic peers.12
Academic difficulties among students with SCD may become more evident in high school due to worsening cognitive deficits13 and increasing academic14 and medical15 demands. In the general US population, the rates of grade retention spike in the ninth grade (after the transition to high school)16 and are strongly associated with school withdrawal.14 Academic difficulties after the transition to high school are more prominent in male students.17
Previous studies assessing academic performance in students with SCD are limited by methodological shortcomings. Most studies were conducted at a single center/hospital and included a small number of participants. The assessment of academic performance was typically limited to academic achievement measures in elementary or middle school students. No studies have compared the withdrawal or retainment rates of students with SCD to those of controls. Furthermore, race has not been accounted for, which is critical given the prevalence of SCD among Black students.
The primary objective of this study was to compare the academic performance of high school students with SCD to that of students without SCD after accounting for race and sex. Secondarily, we sought to determine if school absences mediated the effect of SCD on school performance. To achieve these objectives, we examined the statewide academic data of all students with SCD who received Medicaid coverage in the state of Tennessee during the study period. We hypothesized that students with SCD have lower rates of academic proficiency, greater grade retention, and greater withdrawal than students without SCD after race and sex are controlled for. We also predicted that the negative effects of SCD on academic achievement and attainment are mediated by school absence.
Methods
This retrospective cohort study included statewide academic and medical data. The cohort was assembled using claims data from the Tennessee Medicaid program (TennCare) and educational data from the Tennessee Department of Education Race to the Top (RTTT) program. RTTT was a statewide education program initiated in 2010 that adopted standardized assessments and data systems to capture student growth in public and charter schools.18 A broad search of enrollees of these two programs was performed to identify patients with SCD in Tennessee.19 Patients were identified by their Medicaid enrollment identification number, and both data sets were then linked by the identification numbers. The study was approved by the institutional review boards at Vanderbilt University Medical Center and the Tennessee Department of Health, and by the Division of TennCare. Initially, all individuals between the ages of 6 and 21 years enrolled in TennCare for any duration during the study period (10/01/2007–09/30/2019) were included. From this cohort, individuals with and without SCD were identified. SCD was defined as a diagnosis of any SCD genotype between 01/01/2007 and 12/31/2018 according to International Classification of Diseases, Ninth or Tenth Revisions (ICD-9, ICD-10). For every ninth grade student identified with SCD, eight control students without SCD within the same birth year were randomly selected. Comorbid conditions, including diabetes, hypertension, seizures, and stroke, were also collected from claims data based on ICD-9 or ICD-10 codes. The presence of these comorbid conditions was defined as one inpatient or at least two outpatient visits 30 days apart occurring at any time during the study period.
Educational variables were identified from RTTT. Academic achievement was measured using End-of-Course (EOC) ninth grade assessments from the following high school courses: English 1, Algebra 1, and Biology. These EOC assessments were chosen for their coverage of a wide range of academic content (e.g., reading, math, and science) and their completion by the largest number of ninth grade students relative to other EOC assessments covering similar domains. The English 1 EOC assessed students’ ability to read, analyze text, and answer text-dependent questions. The Algebra 1 EOC assessed students’ understanding of basic algebraic equations, linear and exponential functions, and probability. The Biology EOC assessed students’ understanding of scientific inquiry and technology in relation to biology. Students take these exams in the semester in which they complete the relevant coursework or at the earliest available test administration. Students enrolled in a course with an associated EOC examination must take the examination to receive credit for the course. Due to changes in test norms and proficiency labels during the study period, students were assigned one of two proficiency levels: proficient or not proficient. Labels including “proficient” and “advanced” were considered proficient, whereas “below proficient,” “below basic,” and “basic” were considered not proficient. “Proficient” performance on an EOC assessment indicates that the student demonstrates mastery in the domain specified by course level content standards. These standards are set by committees of teachers, administrators, and curriculum specialists. A student was considered “retained” if the reported grade level (ninth grade status) did not change for two consecutive school years. A student was considered “withdrew” if the student dropped out, withdrew for health reasons, or died. All others were considered “not retained” or “no withdrawal”. Days absent are the number of days the student was absent for the school year. If a student was retained, the most recent school year’s record of days absent was used. Given the skewness of the number of days absent, this variable was categorized using quartiles.
Statistical Analysis
Demographic and health characteristics of the study participants were summarized using appropriate summary statistics such as median, interquartile range, frequency, and percentage. The cohort was then stratified by SCD status (SCD versus No SCD), sex, and race, with summary statistics again calculated for all characteristics. Statistics were compared across SCD status, sex, and race by using Wilcoxon-Mann-Whitney for continuous characteristics (Kruskal-Wallis for race stratification) and chi-squared tests for categorical characteristics (Fisher’s exact for small expected cell counts).
Univariable logistic and multinomial logistic models were run to individually model each outcome as a function of SCD status, sex, or race. Next, each outcome was modeled with a multivariable logistic/multinomial logistic model as a function of SCD status, sex, and race.
A mediation analysis was used to determine whether the number of days absent mediated SCD status–predicted student outcomes, with and without adjustments for sex and race. Days absent were considered binary and were defined as being above or below the median number of days absent for the full cohort. According to the definition of a mediator, the association between the number of days absent and school outcomes was assessed by applying logistic regressions to model each student’s outcome as a function of the number of days absent (binary). Finally, a mediation analysis was applied with SCD status as the primary predictor, days absent (binary) as the possible mediator, and a school outcome as the primary outcome. The mediation analysis was performed with and without adjusting for sex and race. The total, direct, and indirect effects are presented as of odds ratios with 95% confidence intervals (CI). For each analysis, days absent was also evaluated as a possible a moderator (i.e., effect modifier) by using an interaction term in the model. The mediation analysis was performed in R (version 4·3·1) using the ExactMed package.20 This package uses exact parametric regression-based casual mediation, adapted specifically for instances in which the predictors, mediator, and outcome are all binary. All other analyses except the mediation analyses were performed in SAS 9·4. The significance level for all analyses was set at p<0·05.
Results
Sample Characteristics
The analysis included 498 students with and 3188 students without SCD. Table 1 shows the demographic, clinical, and academic characteristics of the cohort stratified by SCD status. There was an even distribution of male and female students in the cohort and across groups (p=0·083). Most of the SCD group identified as Black (95%), whereas only 29% of the No SCD group identified as Black (p<0·001). Only 29% of students with SCD were prescribed hydroxyurea. The history of stroke, seizures, and hypertension were higher in the SCD group (p<0·01 for each parameter; Table 1).
Table 1.
Demographic, clinical, and academic characteristics of study participants, stratified by sickle cell disease status.
| Characteristics, n (%) or mean (SD) median (Q1, Q3) | All | SCD | No SCD | p-value* |
|---|---|---|---|---|
|
| ||||
| N=3686 | n=498 | n=3188 | ||
|
| ||||
| Age | 14·5 (1·9) 15 (14, 15) | 14·8 (1·7) 15 (14, 15) | 14·4 (1·8) 145 (14, 15) | <0·0001 |
| Sex | ||||
| Female | 1843 (50) | 267 (54) | 1576 (49) | 0·083 |
| Male | 1843 (50) | 231 (46) | 1612 (51) | |
| Race | ||||
| Black, non-Hispanic | 1408 (38) | 475 (95) | 933 (29) | <0·0001 |
| Hispanic | 239 (6) | 8 (2) | 231 (7) | |
| Other | 51 (1) | 3 (1) | 48 (2) | |
| White, non-Hispanic | 1988 (54) | 12 (2) | 1976 (62) | |
| Region | ||||
| East | 591 (16) | 24 (5) | 567 (18) | <0·0001 |
| Mid-Cumberland | 803 (22) | 107 (22) | 696 (22) | |
| Northeast | 267 (7) | 8 (2) | 259 (8) | |
| South Central | 218 (6) | 9 (2) | 209 (7) | |
| Southeast | 331 (9) | 31 (6) | 300 (9) | |
| Upper Cumberland | 234 (6) | 0 (0) | 234 (7) | |
| West | 1220 (33) | 317 (64) | 903 (29) | |
| ^ Any hydroxyurea | ||||
| No | 3543(96) | 355 (71) | 3188 (100) | <0·0001 |
| Yes | 143 (4) | 143 (29) | 0 (0) | |
| # Diabetes | ||||
| Yes | 88 (2) | 12 (2) | 76 (2) | 0·972 |
| No | 3598 (98) | 486 (98) | 3112 (98) | |
| # Stroke | ||||
| No | 3659 (99) | 472 (95) | 3187 (100) | <0·0001 |
| Yes | 27 (1) | 26 (5) | 1 (0) | |
| # Seizures | ||||
| No | 3624 (98) | 482 (97) | 3142 (99) | 0·004 |
| Yes | 62 (2) | 16 (3) | 46 (1) | |
| # Hypertension | ||||
| Yes | 98 (3) | 43 (9) | 55 (2) | <0·0001 |
| No | 3588 (97) | 455 (91) | 3133 (98) | |
| Number of health maintenance visits previous 270 days | 1 (3) 0 (0, 1) | 2·6 (6·8) 0 (0, 1) | 0·7 (1·6) 0 (0, 1) | 0·1252 |
|
| ||||
| Academic outcomes | ||||
|
| ||||
| English proficiency | ||||
| Not proficient | 1599 (52) | 269 (66) | 1330 (49) | <0·0001 |
| Proficient | 1500 (48) | 139 (34) | 1361 (51) | |
| Biology proficiency | ||||
| Not proficient | 420 (44) | 71 (61) | 349 (41) | <0·0001 |
| Proficient | 539 (56) | 46 (39) | 493 (59) | |
| Algebra proficiency | ||||
| Not proficient | 1288 (58) | 193 (67) | 1095 (57) | 0·001 |
| Proficient | 918 (42) | 93 (33) | 825 (43) | |
| Above median no. days absent | ||||
| Category 1 (lower 25th percentile) | 781 (21) | 53 (11) | 728 (23) | <0·0001 |
| Category 2 (25th–50th percentile) | 1058 (29) | 108 (22) | 950 (30) | |
| Category 3 (50th–75th percentile) | 883 (24) | 122 (24) | 761 (24) | |
| Category 4 (upper 75th percentile) | 964 (26) | 215 (43) | 749 (23) | |
| Days absent [mean (SD) median (IQR)] | 13·7 (15·2) 10 (4, 18) | 20·5 (19·6) 15 (8, 27) | 12·7 (14·1) 9 (4, 17) | <0·0001 |
| Retained | ||||
| No | 3508 (95) | 454 (91) | 3054 (96) | <0·0001 |
| Yes | 178 (5) | 44 (9) | 134 (4) | |
| Withdrew | ||||
| No | 1472 (95) | 178 (95) | 1294 (95) | 0·884 |
| Yes | 78 (5) | 9 (5) | 69 (5) | |
Comparisons across SCD status were performed with Chi-squared tests or Wilcoxon-Mann-Whitney tests.
Received treatment at any time during the study period.
Defined as one inpatient or at least two outpatient visits with the diagnosis separated by 30 days occurring at any time during the study period.
Abbreviations: IQR, interquartile range; SCD, sickle cell disease; SD, standard deviation.
After sex stratification, we found no demographic or clinical differences except geographic region (Supplemental Table 1). However, when stratified by race (Table 2), most Black, non-Hispanic students were located in West Tennessee (61%), whereas most Hispanic or Other students were in Mid-Cumberland or West Tennessee (>50% each), and White students were mostly located in East, Mid-Cumberland, or West Tennessee (58%) (p<0·001). This distribution is consistent with our previous findings.21
Table 2.
Demographic, clinical, and academic outcomes of study participants, stratified by race.
| Characteristic, n (%) or mean (SD) median (Q1, Q3) | White, non-Hispanic | Black, non-Hispanic | Hispanic | Other | p-value* |
|---|---|---|---|---|---|
|
| |||||
| n=1988 | n=1408 | n=239 | n=51 | ||
|
| |||||
| Age | 14·4 (1·9) 15 (14, 15) | 6 (1·8) (14, 15) | 14·4 (1·5) 14 (14, 15) | 14·7 (2·5) 15 (14, 15) | 0·0002 |
| Sex | |||||
| Female | 1000 (50) | 714 (51) | 109 (46) | 20 (39) | 0·206 |
| Male | 988 (50) | 694 (49) | 130 (54) | 31 (61) | |
| Region | |||||
| East | 484 (25) | 71 (5) | 31 (13) | 5 (10) | <0·001 |
| Mid-Cumberland | 351 (18) | 344 (25) | 89 (37) | 19 (37) | |
| Northeast | 242 (12) | 13 (1) | 11 (5) | 1 (2) | |
| South Central | 171 (9) | 29 (2) | 13 (5) | 5 (10) | |
| Southeast | 218 (11) | 81 (6) | 26 (11) | 6 (12) | |
| Upper Cumberland | 207 (10) | 4 (0) | 21 (9) | 2 (4) | |
| West | 302 (15) | 858 (61) | 47 (20) | 13 (25) | |
| ^ Any hydroxyurea | |||||
| No | 1986 (100) | 1270 (90) | 237 (99) | 50 (98) | <0·001 |
| Yes | 2 (0) | 138 (10) | 2 (1) | 1 (2) | |
| # Diabetes | |||||
| Yes | 53 (3) | 33 (2) | 1 (0) | 1 (2) | 0·196 |
| No | 1935 (97) | 1375 (98) | 238 (100) | 50 (98) | |
| # Stroke | |||||
| No | 1987 (100) | 1382 (98) | 239 (100) | 51 (100) | <0·001 |
| Yes | 1 (0) | 26 (2) | 0 (0) | 0 (0) | |
| # Seizures | |||||
| No | 1959 (99) | 1379 (98) | 236 (99) | 50 (98) | 0 4766 |
| Yes | 29 (1) | 29 (2) | 3 (1) | 1 (2) | |
| # Hypertension | |||||
| Yes | 30 (2) | 62 (4) | 6 (3) | 0 (0) | <0·001 |
| No | 1958 (98) | 1346 (96) | 233 (97) | 51 (100) | |
| Number of health maintenance visits prior 270 days | 0·7 (1·6) 0 (0, 1) | 1·4 (4·3) 0 (0, 1) | 1 (2·1) 1 (0, 1) | 1·6 (2·6) 0 (0, 2) | <0·001 |
| Academic outcomes | |||||
| English test score proficiency | |||||
| Not proficient | 772 (46) | 691 (59) | 124 (61) | 12 (36) | <0·001 |
| Proficient | 920 (54) | 480 (41) | 79 (39) | 21 (64) | |
| Biology test score proficiency | |||||
| Not proficient | 203 (37) | 184 (57) | 28 (46) | 5 (25) | <0·001 |
| Proficient | 351 (63) | 140 (43) | 33 (54) | 15 (75) | |
| Algebra test score proficiency | |||||
| Not proficient | 642 (54) | 550 (65) | 79 (59) | 17 (59) | <0·001 |
| Proficient | 557 (46) | 294 (35) | 55 (41) | 12 (41) | |
| Above median no. days absent | |||||
| Category 1 (lower 25th percentile) | 411 (21) | 286 (20) | 71 (30) | 13 (25) | <0·001 |
| Category 2 (25th–50th percentile) | 630 (32) | 329 (23) | 82 (34) | 17 (33) | |
| Category 3 (50th–75th percentile) | 483 (24) | 335 (24) | 52 (22) | 13 (25) | |
| Category 4 (upper 75th percentile) | 464 (23) | 458 (33) | 34 (14) | 8 (16) | |
| Days absent [mean (SD) median (IQR)] | 12·7 (14·1) 9 (4, 17) | 16·1 (1·7) 11 (5, 21) | 9·5 (10·4) 7 (3, 13) | 10·5 (11·5) 7 (3, 14) | <0·001 |
| Retained | |||||
| No | 1927 (97) | 1307 (93) | 227 (95) | 47 (92) | <0·001 |
| Yes | 61 (3) | 101 (7) | 12 (5) | 4 (8) | |
| Withdrew | |||||
| No | 817 (95) | 526 (95) | 104 (96) | 25 (96) | 0·882 |
| Yes | 46 (5) | 27 (5) | 4 (4) | 1 (4) | |
Comparisons across race were performed with Chi-squared tests or Wilcoxon-Mann-Whitney tests.
Received treatment at any time during the study period.
Defined as one inpatient or at least two outpatient visits with the diagnosis separated by 30 days occurring at any time during the study period.
Abbreviations: IQR, interquartile range; SCD, sickle cell disease; SD, standard deviation.
Associations with Academic Achievements and Attainment
Univariable Analysis
At the univariable level, SCD was associated with all academic outcomes except withdrawal rates. Students with SCD had significantly lower odds of proficiency on the English (OR, 0·5; 95% CI, 0·4–0·6; p<0·001), Algebra (OR, 0·6; 95% CI, 0·5–0·8; p<0·001), and Biology (OR, 0·5; 95% CI, 0·3–0·7; p<0·001) assessments compared to those without SCD (Table 3). They also had higher odds of being absent more days compared to their counterparts. Specifically, students with SCD were 3·9 times (95% CI, 2·9–5·4; p<0·001) more likely to be in the most extreme category (>75th percentile) of number of days absent compared to those without SCD. Students with SCD were 2·2 times (95% CI, 1·5–3·1; p<0·001) as likely to be retained.
Table 3.
Univariable models of the effects of sickle cell disease status, sex, and race on academic performance.
| Outcomes | Predictors | Unadjusted odds ratio (95% confidence interval)* | p-value |
|---|---|---|---|
|
| |||
| English proficiency (Not proficient, reference) | SCD vs No SCD (ref) | 0·5 (0·4, 0·6) | <0·001 |
| Sex: Male vs Female (ref) | 0·7 (0·6, 0·8) | <0·001 | |
| Race: Hispanic vs Black (ref) | 0·9 (0·7, 1·2) | 0·579 | |
| Race: Other vs Black (ref) | 2·5 (1·2, 5·2) | 0·012 | |
| Race: White vs Black (ref) | 1·7 (1·5, 2) | <0·001 | |
| Algebra proficiency | SCD vs No SCD | 0·6 (0·5, 0·8) | 0·001 |
| Sex: Male vs Female | 1 (0·8, 1·2) | 0·778 | |
| Race: Hispanic vs Black | 1·3 (0·9, 1·9) | 0·164 | |
| Race: Other vs Black | 1·3 (0·6, 2·8) | 0·469 | |
| Race: White vs Black | 1·6 (1·4, 1·9) | <0·001 | |
| Biology proficiency | SCD vs No SCD | 0·5 (0·3, 0·7) | <0·001 |
| Sex: Male vs Female | 0·9 (0·7, 1·2) | 0·475 | |
| Race: Hispanic vs Black | 1·5 (0·9, 2·7) | 0·119 | |
| Race: Other vs Black | 3·9 (1·4, 11·1) | 0·009 | |
| Race: White vs Black | 2·3 (1·7, 3) | <0·001 | |
| Retainment (yes vs no, reference) | SCD vs No SCD | 2·2 (1·5, 3·1) | <0·001 |
| Sex: Male vs Female | 1·6 (1·2, 2·1) | 0·004 | |
| Race: Hispanic vs Black | 0·7 (0·4, 1·3) | 0·226 | |
| Race: Other vs Black | 1·1 (0·4, 3·1) | 0·856 | |
| Race: White vs Black | 0·4 (0·3, 0·6) | <0·001 | |
| Withdrawal (yes vs no, reference) | SCD vs No SCD | 0·9 (0·5, 1·9) | 0·884 |
| Sex: Male vs Female | 1·6 (1, 2·6) | 0·044 | |
| Race: Hispanic vs Black | 0·7 (0·3, 2·2) | 0·598 | |
| Race: Other vs Black | 0·8 (0·1, 6) | 0·811 | |
| Race: White vs Black | 1·1 (0·7, 1·8) | 0·71 | |
| Days absent (category 2 vs 1, reference) ^ | SCD vs No SCD | 1·6 (1·1, 2·2) | 0·0108 |
| Sex: Male vs Female | 1 (0·8, 1·2) | 0·824 | |
| Race: Hispanic vs Black | 1 (0·7, 1·4) | 0·9825 | |
| Race: Other vs Black | 1·1 (0·5, 2·4) | 0·734 | |
| Race: White vs Black | 1·3 (1·1, 1·6) | 0·0052 | |
| Days absent (category 3 vs 1) ^ | SCD vs No SCD | 2·2 (1·6, 3·1) | <0·001 |
| Sex: Male vs Female | 1 (0·8, 1·2) | 0·8075 | |
| Race: Hispanic vs Black | 0·6 (0·4, 0·9) | 0·0186 | |
| Race: Other vs Black | 0·9 (0·4, 1·9) | 0·6929 | |
| Race: White vs Black | 1 (0·8, 1·2) | 0·975 | |
| Days absent (category 4 vs 1) ^ | SCD vs No SCD | 3·9 (2·9, 5·4) | <0·001 |
| Sex: Male vs Female | 0·9 (0·7, 1) | 0·1128 | |
| Race: Hispanic vs Black | 0·3 (0·2, 0·5) | <0·001 | |
| Race: Other vs Black | 0·4 (0·2, 0·9) | 0·0358 | |
| Race: White vs Black | 0·7 (0·6, 0·9) | 0·0006 | |
Unadjusted odds ratios with 95% confidence intervals were calculated from logistic regressions modeling each outcome as a function of SCD status, sex, and race separately.
Category 1 = lower 25th percentile, category 2 = 25th – 50th percentile, category 3 = 50th – 75th percentile, and category 4 = upper 75th percentile in days absent.
Abbreviations: ref, reference; SCD, sickle cell disease.
Males had lower odds of English proficiency and greater odds of being retained or withdrawing compared to females (p<0·05 for each; Table 3). White students were more likely than Black students to be proficient in English, Algebra, and Biology (p<0·001 for each; Table 3). White students were also less likely to be retained compared to Black students (p<0·001); no differences across race for withdrawal were observed.
Multivariable Analysis
In the fully adjusted models, SCD status remained a significant predictor for English proficiency, retainment, and days absent (Table 4). After adjusting for sex and race, we found that students with SCD had English proficiency odds of 0·8 times (95% CI, 0·7–0·9; p<0·001) those of students without SCD. Similarly, students with SCD were 1·2 times (95% CI, 1·0–1·5; p=0·045) more likely to be retained after adjustments. After sex and race adjustments, as with the univariable models, students with SCD were significantly more likely to be absent than were their counterparts, with increasing odds for a greater number of days absent (p<0·001 for each).
Table 4.
Academic outcomes as a function of sickle cell disease status, sex, and race.
| Outcome | Predictors | Adjusted odds ratio (95% confidence interval)* | p-value |
|---|---|---|---|
|
| |||
| English proficiency (Not proficient, reference) | SCD vs No SCD (ref) | 0·8 (0·7, 0·9) | <0·001 |
| Sex: Male vs Female (ref) | 0·8 (0·8, 0·9) | <0·001 | |
| Race: Hispanic vs Black (ref) | 0·6 (0·5, 0·8) | 0·001 | |
| Race: Other vs Black (ref) | 1·9 (1·1, 3·2) | 0·023 | |
| Race: White vs Black (ref) | 1·1 (0·9, 1·4) | 0·244 | |
| Algebra proficiency | SCD vs No SCD | 0·9 (0·8, 1·1) | 0·242 |
| Sex: Male vs Female | 1·0 (0·9, 1·1) | 0·765 | |
| Race: Hispanic vs Black | 1·0 (0·7, 1·4) | 0·979 | |
| Race: Other vs Black | 1·0 (0·6, 1·8) | 0·957 | |
| Race: White vs Black | 1·2 (1·0, 1·5) | 0·069 | |
| Biology proficiency | SCD vs No SCD | 0·9 (0·7, 1·1) | 0·186 |
| Sex: Male vs Female | 0·9 (0·8, 1·1) | 0·432 | |
| Race: Hispanic vs Black | 0·8 (0·5, 1·2) | 0·31 | |
| Race: Other vs Black | 2 (0·9, 4·4) | 0·076 | |
| Race: White vs Black | 1·1 (0·8, 1·6) | 0·426 | |
| Retainment (yes vs no, reference) | SCD vs No SCD | 1·2 (1, 1·5) | 0·045 |
| Sex: Male vs Female | 1·3 (1·1, 1·5) | 0·003 | |
| Race: Hispanic vs Black | 0·9 (0·6, 1·6) | 0·835 | |
| Race: Other vs Black | 1·5 (0·7, 3·2) | 0·347 | |
| Race: White vs Black | 0·6 (0·4, 0·8) | 0·003 | |
| Withdrawal (yes vs no, reference) | SCD vs No SCD | 1 (0·7, 1·5) | 0·973 |
| Sex: Male vs Female | 1·3 (1, 1·6) | 0·044 | |
| Race: Hispanic vs Black | 0·8 (0·3, 2·1) | 0·709 | |
| Race: Other vs Black | 0·9 (0·2, 3·9) | 0·84 | |
| Race: White vs Black | 1·2 (0·7, 2·3) | 0·51 | |
| Days absent (Category 2 vs 1, reference) ^ | SCD vs No SCD | 2·1 (1·4, 3) | 0·0001 |
| Male vs Female | 1 (0·9, 1·2) | 0·7348 | |
| Race: Hispanic vs Black | 1·2 (0·8, 1·7) | 0·3937 | |
| Race: Other vs Black | 1·3 (0·6, 2·8) | 0·48 | |
| Race: White vs Black | 1·6 (1·3, 2) | <0·001 | |
| Days absent (Category 3 vs 1) ^ | SCD vs No SCD | 2·5 (1·7, 3·6) | <0·001 |
| Male vs Female | 1 (0·9, 1·3) | 0·6754 | |
| Race: Hispanic vs Black | 0·8 (0·5, 1·1) | 0·1967 | |
| Race: Other vs Black | 1 (0·5, 2·3) | 0·9444 | |
| Race: White vs Black | 1·3 (1, 1·6) | 0·0434 | |
| Days absent (Category 4 vs 1) ^ | SCD vs No SCD | 3·9 (2·7, 5·4) | <0·001 |
| Male vs Female | 0·9 (0·7, 1·1) | 0·2212 | |
| Race: Hispanic vs Black | 0·4 (0·3, 0·7) | 0·0002 | |
| Race: Other vs Black | 0·5 (0·2, 1·3) | 0·1838 | |
| Race: White vs Black | 1·1 (0·8, 1·3) | 0·648 | |
Adjusted odds ratios with 95% confidence intervals were calculated from logistic regressions modeling each outcome.
Category 1 = lower 25th percentile, category 2 = 25th – 50th percentile, category 3 = 50th – 75th percentile, and category 4 = upper 75th percentile in days absent.
Abbreviations: ref, reference; SCD, sickle cell disease.
Mediation Analysis
Students with the greatest number of days absent from school, particularly those who missed more than the median number of days (10), had substantially lower odds of proficiency for any academic category and were significantly more likely to be retained or withdraw (p<0·0001; Supplemental Table 2). We observed similar patterns when we further adjusted for sex and race (data not provided; available upon request).
Table 5 and Supplemental Table 3 show the adjusted and unadjusted mediation analysis results, respectively, for the direct, indirect, and total effects of SCD on academic achievement and attainment with days absent as a possible mediator. When sex and race were adjusted for in the mediation analysis, the total, direct, and indirect effects on English proficiency were significant (p<0·01 for each). For algebra and biology proficiency scores, the total effect was not significant (Supplemental Figure 1). For retainment, both the total and indirect effects were significant, suggesting that the impact of SCD status on retainment was primarily mediated by the number of days absent (p<0·05). Consistent with the unadjusted analysis, SCD status did not have a total or direct effect on withdrawal, but did have an indirect effect via days absent (p<0·001; Supplemental Figure 2).
Table 5.
Days absent mediating the effect of sickle cell disease on academic performance after race and sex are adjusted for.*
| Outcomes | Summary of effects | Adjusted odds ratio (95% confidence interval) | p-value |
|---|---|---|---|
|
| |||
| English proficiency (Not proficient, reference) | a Direct effect | 0·65 (0·51, 0·83) | 0·0006 |
| b Indirect effect | 0·96 (0·93, 0·99) | 0·006 | |
| c Total effect | 0·63 (0·49, 0·8) | 0·0002 | |
| Algebra proficiency | Direct effect | 0·89 (0·66, 1·2) | 0·4424 |
| Indirect effect | 0·94 (0·91, 0·98) | 0·002 | |
| Total effect | 0·84 (0·62, 1·13) | 0·2421 | |
| Biology proficiency | Direct effect | 0·83 (0·54, 1·3) | 0·4231 |
| Indirect effect | 0·89 (0·82, 0·96) | 0·0051 | |
| Total effect | 0·74 (0·47, 1·16) | 0·1851 | |
| Retained (yes vs no, reference) | Direct effect | 1·29 (0·87, 1·93) | 0·2085 |
| Indirect effect | 1·18 (1·1, 1·26) | <0·0001 | |
| Total effect | 1·52 (1·02, 2·28) | 0·0408 | |
| Withdrew (yes vs no, reference) | Direct effect | 0·8 (0·36, 1·78) | 0·5774 |
| Indirect effect | 1·25 (1·11, 1·4) | 0·0001 | |
| Total effect | 0·99 (0·44, 2·23) | 0·986 | |
Each outcome is modelled via logistic regression with sickle cell disease status as the primary predictor (no sickle cell disease, the reference) and days absent (binary defined as above and below the median, latter as the reference) as the mediator, with adjustment for sex and race.
The direct effect is defined as the odds of positive school outcomes among students with sickle cell disease (SCD) compared to those same odds among students without SCD, regardless of days absent.
The indirect effect is defined as impact of SCD on school outcomes, after being mediated by days absent (> 10 days absent).
The total effect is defined as the odds of proficiency/not being retained/not withdrawing among students with SCD compared to the same odds among students without SCD, incorporating both direct and indirect pathways and calculated as the product of the indirect and direct odds ratios.
Discussion
Students with SCD experience the combined effects of living with a chronic disease and exposure to racial inequities.22 This retrospective cohort study of Tennessee Medicaid recipients revealed that ninth grade students with SCD had poorer academic performance than their same-grade peers when race was controlled for. The negative effect of SCD on academic performance may be due in part to extensive amounts of missed school. SCD also directly affected academic achievement when missed school was controlled for, possibly due to the neurocognitive deficits associated with SCD.
We observed racial disparities in academic attainment outcomes within this cohort of Medicaid recipients. Black ninth grade students missed more school and were more likely to be retained in grade than their White peers when SCD status and sex were controlled for. These disparities have been previously documented,23 and proposed mechanisms include low socioeconomic status in the home and broader community environment,24 underfunded schools,25 and structural racism.26 Within the Black population, students with stronger group pride and more positive views on how society views Black individuals have better academic outcomes.12
Consistent with our hypothesis, we observed a significant additional effect of SCD on academic achievement and attainment after controlling for race and sex. These findings are consistent with previous studies that report poorer academic achievement among students with SCD compared to sibling27 or demographically matched28 controls. Unexpectedly, after race and sex were controlled for, the effects of SCD on academic achievement were isolated to a measure of English proficiency and were not observed on measures of math and biology. The use of a dichotomous proficiency scale may have limited our observation of variability within each measure. Furthermore, the achievement assessments used in the current study captured a narrow range of content associated with a specific course (e.g., algebra) rather than cumulative knowledge in an academic domain (e.g., mathematics). Fewer ninth grade students completed algebra and biology assessments, suggesting that some students may not have enrolled in these courses until later in high school. All previous studies measured achievement using individually administered achievement tests (e.g., Woodcock Johnson Achievement Tests)29 that capture knowledge across the span of all grades.
The number of days absent from school was higher among students with SCD after race and sex were controlled for. Students with SCD often miss a large amount of school due to frequent hospital visits and recurrent pain crises.6,7 As our data show, frequently missed school can lead to missed content and worse performance on formal achievement tests. A greater number of absences was also positively associated with the likelihood of grade retention and withdrawal. After controlling for race and sex, we found that SCD status both directly and indirectly (through school absence) affected English proficiency. In contrast, the effect of SCD on grade retention was primarily mediated by days absent. Our data suggest that missed school is a primary mechanism through which SCD can negatively influence academic performance and should therefore be a target of interventions.
Poor academic attainment in high school is associated with significant negative consequences in adulthood. Students who are retained in high school are much less likely to graduate high school30 or attend post-secondary education.31 Individuals who do not graduate high school are more likely to experience depressive symptoms,32 lack health insurance coverage,32 and display lower adult earnings33 than their peers who graduated. Among patients with SCD, those with only a high-school education visit the emergency department three times more often than those with a post-secondary education.8
Interventions to improve academic performance in students with SCD must address the sociodemographic factors that are linked to the disease. Recent studies have highlighted the potential benefits of cultural pride reinforcement to limit racial disparities among Black youth.34 More broadly, earlier academic intervention is more beneficial than later intervention in reducing racial disparities in academic performance.35 Our data suggest that limiting school absences is an important intervention target for students with SCD. A wide range of primarily school-based attendance interventions exist for the general population, but the evidence supporting their effectiveness is limited.36 Furthermore, it is unknown whether the mechanisms supporting these interventions are relevant to the SCD population, given that these students’ absences are often medically related. Among adolescent students with higher medically-related absences, a public health intervention was successful in reducing rates of absenteeism.37 The intervention involved increasing communication between parents and schools, referral to and assessment by a health care provider, and development of a collaborative attendance plan. Future work should assess similar interventions in students with SCD.
Several study limitations exist. We did not have access to individual-level metrics of socioeconomic status or more detailed information about the schools that students attended. Without this information, we were unable to fully separate the interrelated contributions of race, socioeconomic status, and school factors on academic performance. Not all students completed achievement assessments in algebra and biology, potentially limiting our ability to detect group differences. All students included were receiving Medicaid, which may have attenuated the influence of socioeconomic status on academic performance but limited the generalizability of findings to students living in low-income households. Despite lacking granular sociodemographic information, we identified specific groups of students within Tennessee who require academic intervention. We were unable to determine whether specific treatment or disease modifiers contributed to poor academic outcomes in students with SCD. Hydroxyurea potentially attenuates the detrimental impacts of SCD on academic performance by reducing the frequency of pain episodes38 and limiting neurocognitive deficits.39 However, the benefits of hydroxyurea treatment are confounded by disease severity,40 preventing the controlled examination of treatment effects in our data set. Future examination of the role of disease-modifying treatments on academic performance is needed. Finally, the presented data were cross-sectional. Future work should examine the trajectory of academic performance among students with SCD to assess trends in performance and identify causal factors.
In conclusion, this population-based study enabled the examination of the unique impact of SCD on academic performance across diverse school districts in the state of Tennessee. The detailed academic achievement and attainment outcomes provide a comprehensive view of the academic profile of ninth grade students with SCD and identify potential targetable treatment mechanisms. The use of government data to measure population-level functional outcomes associated with a chronic disease may be foundational for future intervention studies.
Supplementary Material
Acknowledgements:
This work was supported by the Centers for Disease Control and Prevention’s National Center on Birth Defects and Developmental Disabilities (Grant number NU58DD000035: Sickle Cell Data Collection Program). A.M.H. was supported by K23HL166697 (National Heart, Lung, and Blood Institute) during the time of this study. A.A.K. was supported by K12HL137942 and K24HL148305 (National Heart, Lung, and Blood Institute) during the time of the study. A.D.W was supported by K01DA051683 (National Institute on Drug Abuse) during the time of the study.
Abbreviations
- SCD
Sickle Cell Disease
- RTTT
Race to the top
- TennCare
Tennessee Medicaid Program
- EOC
End of course
- CI
Confidence interval
- OR
Odds ratio
Footnotes
Conflicts of Interest: Jane Hankins and Allison King receive royalties from UpToDate.
Data Sharing:
The supporting data are maintained by the state of Tennessee. Data requests can be submitted to data.health@tn.gov.
References
- 1.Schatz J, McClellan CB. Sickle cell disease as a neurodevelopmental disorder. Mental retardation and developmental disabilities research reviews. 2006;12(3):200–207. [DOI] [PubMed] [Google Scholar]
- 2.Yawn BP, Buchanan GR, Afenyi-Annan AN, et al. Management of sickle cell disease: summary of the 2014 evidence-based report by expert panel members. Jama. 2014;312(10):1033–1048. [DOI] [PubMed] [Google Scholar]
- 3.Hassell KL. Population Estimates of Sickle Cell Disease in the U.S. American Journal of Preventive Medicine. 2010;38(4):S512–S521. doi: 10.1016/j.amepre.2009.12.022 [DOI] [PubMed] [Google Scholar]
- 4.Prussien KV, Siciliano RE, Ciriegio AE, et al. Correlates of Cognitive Function in Sickle Cell Disease: A Meta-Analysis. Journal of Pediatric Psychology. 2020;45(2):145–155. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Heitzer AM, Hamilton L, Stafford C, et al. Academic Performance of Children With Sickle Cell Disease in the United States: A Meta-Analysis. 10.3389/fneur.2021.786065. Front Neurol. 2021;12:786065. doi:10.3389/fneur.2021.786065 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Crosby LE, Joffe NE, Irwin MK, et al. School performance and disease interference in adolescents with sickle cell disease. Physical disabilities: education and related services. 2015;34(1):14. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Schwartz LA, Radcliffe J, Barakat LP. Associates of school absenteeism in adolescents with sickle cell disease. Pediatric blood & cancer. 2009;52(1):92–96. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Jonassaint C, Beach MC, Haythornthwaite J, et al. The association between educational attainment and patterns of emergency department utilization among adults with sickle cell disease. International journal of behavioral medicine. 2016;23(3):300–309. [DOI] [PubMed] [Google Scholar]
- 9.Knisely MR, Pugh N, Kroner B, et al. Patient-reported outcomes in sickle cell disease and association with clinical and psychosocial factors: Report from the sickle cell disease implementation consortium. American Journal of Hematology. 2020;95(9):1066–1074. doi: 10.1002/ajh.25880 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.King AA, Rodeghier MJ, Panepinto JA, et al. Silent cerebral infarction, income, and grade retention among students with sickle cell anemia. American journal of hematology. 2014;89(10):E188–E192. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Harris KM, Dadekian JN, Abel RA, et al. Increasing Educational Attainment in Adolescents with Sickle Cell Disease. Social Work in Public Health. 2019;34(6):468–482. doi: 10.1080/19371918.2019.1629142 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Chavous TM, Bernat DH, Schmeelk-Cone K, Caldwell CH, Kohn-Wood L, Zimmerman MA. Racial Identity and Academic Attainment Among African American Adolescents. Child Development. 2003;74(4):1076–1090. doi: 10.1111/1467-8624.00593 [DOI] [PubMed] [Google Scholar]
- 13.Wang W, Enos L, Gallagher D, et al. Neuropsychologic performance in school-aged children with sickle cell disease: a report from the Cooperative Study of Sickle Cell Disease. The Journal of pediatrics. 2001;139(3):391–397. [DOI] [PubMed] [Google Scholar]
- 14.Neild RC. Falling off track during the transition to high school: What we know and what can be done. The Future of Children. 2009:53–76. [DOI] [PubMed] [Google Scholar]
- 15.Porter JS, Wesley KM, Zhao MS, Rupff RJ, Hankins JS. Pediatric to adult care transition: Perspectives of young adults with sickle cell disease. Journal of pediatric psychology. 2017;42(9):1016–1027. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Warren JR, Hoffman E, Andrew M. Patterns and Trends in Grade Retention Rates in the United States, 1995–2010. Educational Researcher. 2014;43(9):433–443. doi: [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Sum A, Harrington P. The Hidden Crisis in the High School Dropout Problems of Young Adults in the US: Recent Trends in Overall School Dropout Rates and Gender Differences in Dropout Behavior. 2003; [Google Scholar]
- 18.Boser U Race to the Top: What Have We Learned from the States So Far? A State-by-State Evaluation of Race to the Top Performance. Center for American Progress. 2012; [Google Scholar]
- 19.Eckrich MJ, Wang WC, Yang E, et al. Adherence to transcranial Doppler screening guidelines among children with sickle cell disease. Pediatric Blood & Cancer. 2013;60(2):270–274. doi: 10.1002/pbc.24240 [DOI] [PubMed] [Google Scholar]
- 20.Package ‘ExactMed’. Caubet M, Samoilenko M, Gervais J, Lefebvre G, Caubet MM. 2023. [Google Scholar]
- 21.Plaxco AP, Hankins JS, Davis R, et al. Descriptive epidemiology of sickle cell disease in Tennessee: population-based estimates from 2008 to 2019. Frontiers in Hematology. 2023;2doi: 10.3389/frhem.2023.1277548 [DOI] [Google Scholar]
- 22.Boghossian NS, Greenberg LT, Saade GR, et al. Association of Sickle Cell Disease With Racial Disparities and Severe Maternal Morbidities in Black Individuals. JAMA Pediatrics. 2023;177(8):808. doi: 10.1001/jamapediatrics.2023.1580 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Assari S, Mardani A, Maleki M, Boyce S, Bazargan M. Black-White Achievement Gap: Role of Race, School Urbanity, and Parental Education. Pediatric Health, Medicine and Therapeutics. 2021;Volume 12:1–11. doi: 10.2147/phmt.s238877 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Henry DA, Betancur Cortés L, Votruba-Drzal E. Black–White achievement gaps differ by family socioeconomic status from early childhood through early adolescence. Journal of Educational Psychology. 2020;112(8):1471. [Google Scholar]
- 25.Reardon SF. School Segregation and Racial Academic Achievement Gaps. RSF: The Russell Sage Foundation Journal of the Social Sciences. 2016;2(5):34–57. doi: 10.7758/rsf.2016.2.5.0330775444 [DOI] [Google Scholar]
- 26.Merolla DM, Jackson O. Structural racism as the fundamental cause of the academic achievement gap. Sociology Compass. 2019/06/01 2019;13(6):e12696. doi: 10.1111/soc4.12696 [DOI] [Google Scholar]
- 27.Nabors NA, Freymuth AK. Attention deficits in children with sickle cell disease. Perceptual and motor skills. 2002;95(1):57–67. [DOI] [PubMed] [Google Scholar]
- 28.Bills SE, Katz T, McNeil J, Schatz J. Does Obstructive Sleep Apnea Increase Cognitive Deficits in Pediatric Sickle Cell Disease? Journal of the International Neuropsychological Society. 2019;25(09):922–930. doi: 10.1017/s1355617719000730 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Schrank FA, Mather N, McGrew KS. Woodcock-Johnson IV Tests of Achievement. Riverside; 2014. [Google Scholar]
- 30.Larsen MF, Valant J. The long-term effects of grade retention: Evidence on persistence through high school and college. Journal of Research on Educational Effectiveness. 2024;17(4):615–646. [Google Scholar]
- 31.Fine JG, Davis JM. Grade retention and enrollment in post-secondary education. Journal of school psychology. 2003;41(6):401–411. [Google Scholar]
- 32.Topitzes J, Godes O, Mersky JP, Ceglarek S, Reynolds AJ. Educational Success and Adult Health: Findings from the Chicago Longitudinal Study. Prevention Science. 2009;10(2):175–195. doi: 10.1007/s11121-009-0121-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Watts TW. Academic Achievement and Economic Attainment: Reexamining Associations Between Test Scores and Long-Run Earnings. AERA Open. 2020;6(2):233285842092898. doi: 10.1177/2332858420928985 [DOI] [Google Scholar]
- 34.Jones SCT, Neblett EW. Racial–Ethnic Protective Factors and Mechanisms in Psychosocial Prevention and Intervention Programs for Black Youth. Clinical Child and Family Psychology Review. 2016/06/01 2016;19(2):134–161. doi: 10.1007/s10567-016-0201-6 [DOI] [PubMed] [Google Scholar]
- 35.Duncan GJ, Dowsett CJ, Claessens A, et al. School readiness and later achievement. Developmental Psychology. November 2007. 2015/05/10/2007;43(6):1428. doi: 10.1037/0012-1649.43.6.1428 [DOI] [PubMed] [Google Scholar]
- 36.Maynard BR, McCrea KT, Pigott TD, Kelly MS. Indicated Truancy Interventions: Effects on School Attendance among Chronic Truant Students. Campbell Systematic Reviews. 2012;8(1):1–84. doi: 10.4073/csr.2012.10 [DOI] [Google Scholar]
- 37.Vanneste YTM, Mathijssen JJP, van de Goor IAM, Rots–de Vries CMC, Feron FJM. Addressing medical absenteeism in pre-vocational secondary students: effectiveness of a public health intervention, using a quasi-experimental design. BMC Public Health. 2016;16:1–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Agrawal RK, Patel RK, Shah V, Nainiwal L, Trivedi B. Hydroxyurea in Sickle Cell Disease: Drug Review. Indian Journal of Hematology and Blood Transfusion. 2014;30(2):91–96. doi: 10.1007/s12288-013-0261-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Heitzer AM, Longoria J, Okhomina V, et al. Hydroxyurea treatment and neurocognitive functioning in sickle cell disease from school age to young adulthood. Br J Haematol. Oct 2021;195(2):256–266. doi: 10.1111/bjh.17687 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.McGann PT, Ware RE. Hydroxyurea therapy for sickle cell anemia. Expert Opinion on Drug Safety. 2015;14(11):1749–1758. doi: 10.1517/14740338.2015.1088827 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The supporting data are maintained by the state of Tennessee. Data requests can be submitted to data.health@tn.gov.