The Longitudinal Link Between Student Health and Math Achievement Scores

Abstract

This study investigated the relationship between health conditions suffered over time and student scores on the Stanford Achievement Test 9 in Yuma County, Arizona, public grade schools. The majority of children in Yuma County were of Hispanic origin. The poverty and low income status of most of these children placed them at greater risk for health problems. The primary purpose of the study was to ascertain how the rate of math achievement of students who suffered from an illness or injury prior to testing compared to those students who had no known illness or injury. Findings suggested that physical illness influenced math achievement trajectories adversely. Importantly, the achievement gaps associated with several health conditions attenuated over time. However, children who experienced an injury or poisoning, or an endocrine, nutritional, or metabolic disease developed permanent math achievement deficits.

ACKNOWLEDGMENTS

This study was supported by funding from an American Educational Research Association (AERA)/Institute of Education Sciences (IES) postdoctoral fellowship. I wish to thank Professor David C. Berliner, the two anonymous reviewers, Professor William G. Johnson and the Center for Health Information and Research, and the Arizona Department of Education.

Notes

¹There is a small difference between the percentage of Yuma County public school students who had reported or were selected as Hispanic and the July 1, 2003 Census Bureau estimates of the percentage of children 5 to 17 who were Hispanic. This is in part due to the slight difference in the age range of students (second through ninth grade) and the age range of the Census Bureau estimate. In addition, dual racial/ethnic identity could not be reported on the SAT 9 testing forms. The 2003 Census Bureau estimate includes Hispanics of any race. Finally, not all of Yuma County's children attended the public school system.

²The U.S. National Center for Health Statistics defines a chronic illness as one that lasts for a minimum of 3 consecutive months. According to this definition, multiple instances of illness cannot be accurately labeled as chronic illness or injury because it was not known whether the same specific illness or injury lasted for the 3-month duration. Instead, the focus in this study was on students who had multiple insurance claims associated with an illness or injury within the same major disease classification category.

³The models discussed in the main data and methods section treat the various health statuses, FRPL status, and school mobility status as time varying because some students experience changes in these statuses. Specifying the main analytic variables and selected covariates as time varying is simple and requires no special strategy other than subscripting the covariates correctly (Singer & Willett, 2003). However, the treatment of these specific covariates as time varying only allows individuals to move from one trajectory to another as their status changes over time. To capture both intraindividual differences in slopes over time and interindividual differences in the main effects of the time varying covariates over time, addition random effects must be included. Models must jointly include an interaction with time and the time varying covariates, as well as associated random effects (Raudenbush & Bryk, 2002; Singer & Willett, 2003). These models would not converge or they did not converge properly with the more complex covariance structure. The more parsimonious models without the additional covariance parameters are discussed instead.

Note. *p ≤ .0001 (Mean difference between students not in the YCHQ and YCHQ students). YCHQ = Yuma County Health Query; DSD = digestive system diseases; ENM = endocrine, nutritional, or metabolic diseases not including diabetes; IPD = infectious and parasitic diseases; OM = otitis media; IP = injuries and poisonings; NSSO = diseases of the nervous system and sense organs not including OM; LEP = limited English proficient; SPED = special education; FRPL = free and reduced-price lunch.

⁴The percentage of children with asthma in this database is lower than other estimates in Yuma County (5% to 7%) have suggested. The difference can be explained by the following. First, there is a relatively smaller percentage of children represented in the age range (5–9) because the focus of the study is on second through ninth graders. These children typically have the highest asthma prevalence. Older children have lower asthma prevalence rates. Second, the YCHQ database captures health usage information for about 57% of 5- to 9-year-olds in the county and roughly 47% of 10- to 14-year-olds. Therefore, it is possible that some students who have asthma do not have health usage records in the YCHQ database and were treated outside of the county. This probability is higher for older children. Finally, health care utilization by Native American children is underrepresented in this database because many of these children received treatment in the Indian Health Services system.

Note. YCHQ = Yuma County Health Query; LEP = limited English proficient; SPED = special education; FRPL = free and reduced-price lunch; ICC = intraclass correlation coefficient.

^a p < .10.

^b p < .05.

^c p < .01.

^d p < .001.

^e p < .0001.

⁵This was expected. The SAT 9 had an upper and lower test scale score boundary (Grade 9 = 865, Grade 2 = 342).

Note. YCHQ = Yuma County Health Query; LEP = limited English proficient; SPED = special education; FRPL = free and reduced-price lunch; ICC = intraclass correlation coefficient.

^a p < .05.

^b p < .01.

^c p < .001.

^d p < .0001.

Note. YCHQ = Yuma County Health Query; LEP = limited English proficient; SPED = special education; FRPL = free and reduced-price lunch; ICC = intraclass correlation coefficient.

^a p < .05.

^b p < .01.

^c p < .001.

^d p < .0001.

Note. YCHQ = Yuma County Health Query; LEP = limited English proficient; SPED = special education; FRPL = free and reduced-price lunch; ICC = intraclass correlation coefficient.

^a p < .05.

^b p < .01.

^c p < .001.

^d p < .0001.

Note. YCHQ = Yuma County Health Query; LEP = limited English proficient; SPED = special education; FRPL = free and reduced-price lunch; ICC = intraclass correlation coefficient.

^a p < .05.

^b p < .01.

^c p < .001.

^d p < .0001.

Note. YCHQ = Yuma County Health Query; LEP = limited English proficient; SPED = special education; FRPL = free and reduced-price lunch; ICC = intraclass correlation coefficient.

^a p < .05.

^b p < .01.

^c p < .001.

^d p < .0001.

Note. YCHQ = Yuma County Health Query; LEP = limited English proficient; SPED = special education; FRPL = free and reduced-price lunch; ICC = intraclass correlation coefficient.

^a p < .05.

^b p < .01.

^c p < .001.

^d p < .0001.