The association of fitness and school absenteeism

1 downloads 0 Views 593KB Size Report
Cardiovascular Endurance Run; BMI, Body Mass Index; CDC, Centers for Disease .... Progressive Aerobic Cardiovascular Endurance Run (PACER), muscle ...
Accepted Manuscript The association of fitness and school absenteeism across gender and poverty: A prospective multilevel analysis in New York City middle schools Emily M. D’Agostino, DrPH, Sophia E. Day, MA, Kevin J. Konty, PhD, Michael Larkin, MA, Subir Saha, PhD, Katarzyna Wyka, PhD PII:

S1047-2797(17)30336-8

DOI:

10.1016/j.annepidem.2017.12.010

Reference:

AEP 8339

To appear in:

Annals of Epidemiology

Received Date: 12 June 2017 Revised Date:

8 December 2017

Accepted Date: 21 December 2017

Please cite this article as: D’Agostino EM, Day SE, Konty KJ, Larkin M, Saha S, Wyka K, The association of fitness and school absenteeism across gender and poverty: A prospective multilevel analysis in New York City middle schools, Annals of Epidemiology (2018), doi: 10.1016/ j.annepidem.2017.12.010. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT Fitness, poverty and gender on absenteeism

RI PT

Emily M. D’Agostino, DrPHa; Sophia E. Day, MAb; Kevin J. Konty, PhDb; Michael Larkin, MAc; Subir Saha, PhDc; Katarzyna Wyka, PhDa CUNY Graduate School of Public Health and Health Policy, 55 W 125th Street, New York, NY 10027, USA b NYC Department of Health and Mental Hygiene, Office of School Health, 1 Court Square, 20th Floor Queens, NY 11101, USA c NYC Department of Education, Office of School Wellness, 335 Adams Street, 28th Floor Brooklyn, NY 11201, USA

SC

a

Present address

TE D

1

M AN U

Address for correspondence: Emily D’Agostino, DrPH, MS, MA Miami-Dade Department of Parks, Recreation and Open Spaces1 275 NW 2nd Street Suite 416 Miami, Florida 33128 email: [email protected] Office: 305-755-7938 | Mobile: 646-853-1223| Fax: 305-755-7864

EP

4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27

The association of fitness and school absenteeism across gender and poverty: A prospective multilevel analysis in New York City middle schools

AC C

1 2 3

1

ACCEPTED MANUSCRIPT Fitness, poverty and gender on absenteeism Abstract

29

Purpose

30

One-fifth to one-third of students in high-poverty, urban school districts do not attend school

31

regularly (missing ≥6 days per year). Fitness is shown to be associated with absenteeism,

32

although this relationship may differ across poverty and gender subgroups.

33

Methods

34

Six cohorts of New York City public school students were followed from grades 5-8 during

35

2006/7-2012/13 (n=349,381). Stratified three-level longitudinal generalized linear mixed models

36

were used to test the association of fitness changes and one-year lagged child-specific days

37

absent across gender and poverty.

38

Results

39

In girls attending schools in high/very high poverty areas, greater improvements in fitness the

40

year prior were associated with greater improvements in attendance (p=0.034). Relative to the

41

reference group (>20% decrease in fitness composite percentile scores from the prior year), girls

42

with a large increase in fitness (>20%) demonstrated 10.3% fewer days absent (Incidence Rate

43

Ratio (IRR) 95% CI: 0.834, 0.964), followed by those who had a 10-20% increase in fitness

44

(9.2%, IRR 95% CI: 0.835, 0.987), no change (5.4%, IRR 95% CI: 0.887, 1.007) and a 10-20%

45

decrease in fitness (3.8%, IRR 95% CI: 0.885, 1.045). In girls attending schools in low/mid

46

poverty areas, the fitness-attendance relationship was also positive, but no clear trend emerged.

47

In boys, the fitness-attendance relationship was also positive but not significant in either poverty

48

group.

49

Conclusions

AC C

EP

TE D

M AN U

SC

RI PT

28

2

ACCEPTED MANUSCRIPT Fitness, poverty and gender on absenteeism Fitness improvements may be more important to attendance improvements in high/very high

51

poverty girls compared with low/mid poverty girls, and both high/very high and low/mid poverty

52

boys. Expanding school-based physical activity programs for youth particularly in high-poverty

53

neighborhoods may increase student attendance.

RI PT

50

54 Key words

56

Poverty Areas; Gender; Schools; School-Age Population; Cardiorespiratory Fitness; Fitness

57

Testing; Youth; Absenteeism

M AN U

58

SC

55

List of abbreviations

60

NYC, New York City; DOE, Department of Education; Fitnessgram, NYC FITNESSGRAM;

61

DOHMH, Department of Health and Mental Hygiene; PACER, Progressive Aerobic

62

Cardiovascular Endurance Run; BMI, Body Mass Index; CDC, Centers for Disease Control and

63

Prevention; ICC, Intraclass Correlation; IRR, Incidence Rate Ratio.

EP AC C

64

TE D

59

3

ACCEPTED MANUSCRIPT Fitness, poverty and gender on absenteeism Introduction

66

Only 25% of youth ages 12-15 meet the recommended daily 60 minutes or more of moderate to

67

vigorous physical activity [1]. Schools in the US have replaced physical education and recess

68

with non-physical instructional time due in part to an increasing emphasis on high-stakes testing

69

[2]. This raises particular concern given a well-established link between school-based physical

70

activity and health [3-7], and potential associations with academics [8-11] and attendance [8, 10-

71

12].

SC

RI PT

65

72

Schools in high-poverty areas, in particular, may hold a critical role in providing youth with safe

74

and accessible physical activity [13, 14]. Environmental factors, including area poverty and the

75

built environment, are shown in the literature to be associated with children’s tendency to

76

participate in school- and neighborhood-based physical activity [13-16]. Neighborhood factors

77

may contribute to opportunities for safe, attractive, and accessible physical activity [15, 17].

78

School contextual factors may also impact community norms and attitudes pertaining to

79

children’s school absences [18, 19]. Given one-fifth to one-third of students in high-poverty,

80

urban school districts do not attend school regularly (missing ≥6 days per year), high-poverty

81

subgroups may benefit from physical activity interventions. However, no literature has examined

82

the potential modification of area poverty on the youth fitness-school absenteeism relationship.

TE D

EP

AC C

83

M AN U

73

84

Disparities in both physical activity and absenteeism also persist across gender. Only 18% of

85

high school girls compared with 37% of boys are reported to meet physical activity guidelines

86

[20]. Numerous studies demonstrate low self-esteem in adolescent girls is significantly

87

associated with both lower physical activity levels [21] and absenteeism [22, 23]. However, no

4

ACCEPTED MANUSCRIPT Fitness, poverty and gender on absenteeism 88

papers were identified that examined gender differences in the fitness-absenteeism relationship.

89

Given this literature, nuanced research on gender differences in the association of fitness and

90

attendance are warranted.

RI PT

91

This study examined gender differences by area poverty on the longitudinal fitness-attendance

93

association in 6 cohorts of New York City (NYC) Department of Education (DOE) middle

94

school students each followed consecutively over 4 years. It was hypothesized that

95

improvements in fitness (cardiorespiratory, muscular endurance, and muscular strength fitness

96

composite percentile scores) would predict lower subsequent absenteeism (one-year lagged days

97

absent) after accounting for potential individual- and school-level confounders, clustering, and

98

time-dependent interactions. It was further hypothesized that higher area poverty and female

99

gender would predict a stronger magnitude of association of fitness and absenteeism.

M AN U

SC

92

TE D

100 Methods

102

Data sources and study population

103

Data for this study were drawn from the NYC FITNESSGRAM (Fitnessgram) dataset jointly

104

managed by NYC Department of Education (DOE) and Department of Health and Mental

105

Hygiene (DOHMH) [24]. It is comprised of annual fitness assessments collected by DOE for

106

approximately 870,000 NYC public school students per year (grades K-12) starting in 2006-07.

107

This study was approved by the City University of New York and DOHMH IRBs and was

108

determined by these boards to be public health surveillance that is not research, and therefore

109

exempt from the requirement for obtaining written informed consent.

AC C

EP

101

110

5

ACCEPTED MANUSCRIPT Fitness, poverty and gender on absenteeism The Fitnessgram is based on the Cooper Institute’s FitnessGram®, which is demonstrated to

112

have both strong reliability and validity [25]. Fitnessgram performance tests provide a health

113

assessment related to present and future health outcomes. NYC schools are mandated to have

114

≥85% of eligible students complete the Fitnessgram assessment each year. Inclusion criteria for

115

this study included enrollment in a NYC public school for ≥2 consecutive years while in grades

116

6-8 during the study period (2006/07-2012/13) while attending a school that collected

117

Fitnessgram measurements (for Sample Selection Flow Chart see Figure A.1). Student cohorts

118

were defined based on year of initiating grade 6. Fitness-change data from grades 5-6, 6-7 and

119

7-8 was paired with days absent per year for grades 6, 7, and 8, respectively. The final sample of

120

6-8th graders included 349,381 unique students nested in 624 schools (mean school

121

population=541 students; SD=632). Students in 6th, 7th and 8th grades contributed 177,281,

122

220,769, and 186,135 student-years, respectively.

M AN U

SC

RI PT

111

TE D

123 Measures

125

The primary exposure was a categorical variable representing age- and gender-specific percent

126

change in fitness composite percentile scores based on the sum of percentile scores for the

127

Progressive Aerobic Cardiovascular Endurance Run (PACER), muscle strength and endurance

128

(curl-up and push-up) tests [24]. Scores were converted to percentiles to account for expected

129

improvements in performance with increasing age and gender. The fitness variable was

130

categorized as >20% decrease, 10-20% decrease, 20%

131

increase in performance from the year prior consistent with longitudinal research on fitness and

132

academic outcomes drawing from the Fitnessgram dataset [9].

AC C

EP

124

6

ACCEPTED MANUSCRIPT Fitness, poverty and gender on absenteeism 133

The primary outcome for this analysis was child-level number of days absent per year. Annual

134

enrollment and attendance records were matched to their Fitnessgram results by a unique student

135

identifier.

RI PT

136

Gender was based on parent report. Consistent with NYC DOHMH guidelines [26], a school

138

neighborhood’s socioeconomic status was defined according to American Community Survey

139

2007-2012 data as the percentage of households in the school zip code living below the federal

140

poverty threshold (low (30%)

141

area poverty) [27].

M AN U

SC

137

142

Covariates included time, race/ethnicity, place of birth and school size. These covariates are

144

shown in the literature to predict both fitness and absenteeism [8, 28-30]. Time (calendar year) at

145

height and weight measurement was treated as a continuous variable. Race/ethnicity was based

146

on parent report and was grouped into 5 categories: Hispanic, non-Hispanic black, non-Hispanic

147

white, Asian/Pacific Islander, and other (including multiple races). A binary school size variable

148

was based on count of individuals attending each student’s school. Per the literature, schools

149

with less than 400 students were considered small schools, and schools with 400 or more

150

students were considered non-small schools [30].

EP

AC C

151

TE D

143

152

Change in obesity status from the year prior (obese to not obese, consistently not obese,

153

consistently obese, not obese to obese) was also included as a potential confounder based on the

154

literature [8]. BMI is collected annually as part of the Fitnessgram curriculum. Obesity was

155

defined as having a BMI ≥ 95th percentile for youth in the same gender and age group according

7

ACCEPTED MANUSCRIPT Fitness, poverty and gender on absenteeism 156

to CDC guidelines [31]. Change in obesity status category was chosen as a covariate in lieu of

157

changes in BMI percentile to capture meaningful shifts in body composition associated with

158

school outcomes [32].

RI PT

159 Statistical methods

161

Descriptive statistics were computed to summarize sample characteristics. Next, trends in

162

absenteeism (days absent) by fitness, grade, and demographic characteristics were examined.

SC

160

163

Mixed models methodology was used to assess between-school variability in student

165

absenteeism. First, in order to determine the extent of variation in absenteeism at the school-

166

level, unconditional three-level longitudinal linear mixed models with random-intercepts were fit

167

to the data for all students and stratified by gender. The school-level intraclass correlation (ICC)

168

was calculated as the ratio of the variance for the school, divided by the sum of the 3 variance

169

parameter estimates, represented as: σ2school / (σ2student+ σ2school+ σ2ε) . Although univariate

170

distributions for the days absent variable demonstrated a long right-tailed Poisson distribution,

171

the ICC was calculated based on a linear mixed model given the ICC is not well defined for

172

Poisson models [33].

TE D

EP

AC C

173

M AN U

164

174

Next, the longitudinal association of change in fitness and lagged number of days absent per year

175

was assessed by fitting stratified negative binomial longitudinal mixed models [34, 35] with

176

random-intercepts, the exposure, child-specific change in fitness from the year prior, and an

177

offset term representing total instructional days per school year. Crude and adjusted models

178

were stratified by school-area poverty (low/mid vs. high/very high poverty) and gender (girls vs.

8

ACCEPTED MANUSCRIPT Fitness, poverty and gender on absenteeism boys) (models 1-8). Negative binomial models were used because data were overdispersed.

180

Statistical significance of the overall association of the exposure, change in fitness and outcome,

181

one-year lagged number of days absent per year was assessed using Type III Tests of Fixed

182

Effects. Exponentiated beta coefficients (incidence rate ratios (IRR)) represented the association

183

of each level of exposure (>20% increase, 10-20% increase, 20% decrease in fitness). A two

185

sided P-value of 20% decrease (19%), 10-20% increase (12%), and 10-20% decrease

197

(12%). There were 624 schools included in the analysis, including 365 (58%) small schools.

EP

AC C

198

TE D

190

9

ACCEPTED MANUSCRIPT

Fitness, poverty and gender on absenteeism

Table 1. Demographic and fitness-change characteristics of in New York City public school students in grades 6-8 across gender and poverty subgroups (Nstudents=349381, Nobs=675266a), 2006/7-2012/13. Males n % 177355 51

High Povertyb

nc,d

nc,d

%e

nc,d

%e

nc,d

59134

34

43462

25

38611

19277 15888 10063

33 27 17

19332 13373 4266

45 31 10

21081 14652 947

13686

23

6331

15

1785

31756 12235 15143

54 21 26

47160 11965 20624 13364 44020 14233 23880

Low Povertyb

Medium Povertyb

High Povertyb

%e

nc,d

%e

nc,d

%e

nc,d

%e

nc,d

%e

23

31422

18

60085

34

45945

26

39899

22

20 21 39

19497 14652 11081

33 24 18

20810 13152 4807

45 29 11

22241 14538 1060

56 36 3

6597

21

14609

24

6979

15

1896

5

TE D

55 6118 38 6484 3 12096 5

Very High Povertyb

22161 13488 7813

51 31 18

22630 13440 2541

59 21905 35 3035 7 6482

70 10 21

31564 12192 16329

53 20 28

22976 14386 8583

50 31 19

23161 14136 2602

58 35 7

EP

30816 18 Overall Race/Ethnicity 6095 20 Hispanic 6619 22 Non-Hispanic Black 11537 37 Non-Hispanic White Asian and/or Pacific 6412 21 Islander Language Spoken at Home 21568 70 English 3139 10 Spanish 6109 20 Other Language Place of Birth 27180 88 US 3628 12 Foreign f Change in Fitness (all years) 9957 16 20% Decrease 6935 11 10-20% Decrease 24291 40 1 race/ethnicity=177. dStudents in 6th, 7th and 8th grades contributed 177,281, 220,769, and 186,135 student-years, respectively. ePercents may not total 100 due to rounding. fChange in fitness composite percentile scores from the year prior was based on PACER (Progressive Aerobic Cardiovascular Endurance Run), Push-up and Curl-up Fitnessgram tests. gChange in obesity status from the year prior was defined according to CDC growth chart-derived norms for gender and age in months based on a historical reference population, and used to compute the BMI percentile for each child. Obesity was defined as having a BMI ≥ 95th percentile for youth in the same gender and age in months group.

AC C

EP

TE D

M AN U

a

RI PT

2407 50299 7299 1941

SC

Obese to not obese Consistently not obese Consistently obese Not obese to obese Attending Small Schools (20% decrease 22573 12.02(12.36) 10-20% decrease 10526 11.63(11.91) 20% increase 23836 10.93(11.24)

n 22532 16414 55658 17418 23946

Grade 8 (nobs=213482)

n 18416 12475 40911 12386 17602

Mean(SD) 12.85(13.06) 11.40(11.69) 10.51(10.79) 10.49(11.04) 10.79(10.85) Mean(SD) 14.87(15.40) 13.81(14.73) 13.62(14.37) 12.64(13.39) 12.99(14.31)

Tabulated mean estimates based on days absent per year. bNmale students=177355, Nfemale students=172026, Nlow/mid c poverty students=181457, Nhigh/very high poverty students=167,917. Observations account for 1-3 years of fitness-change per d student. Based on percentage of households in the school zip code living below the federal poverty threshold (low (30%) area poverty) drawing from the e American Community Survey (ACS) 2007-2012. Nmissing Area Poverty=52. fChange in fitness composite percentile scores from the year prior was based on PACER (Progressive Aerobic Cardiovascular Endurance Run), Push-up and Curl-up Fitnessgram tests.

199

TE D

a

n 17493 13707 48590 14108 17794

SC

Low/Mid Povertyde (nobs=355951) Change in Fitnessf n >20% decrease 20949 10-20% decrease 10937 20% increase 29954

Grade 7 (nobs=257201)

RI PT

Grade 6 (nobs=204583)

Twenty six percent and 22% of students attended a school in a high- or very high-poverty area,

201

respectively. Table 2 shows higher absenteeism with increasing grade level and increasing

202

poverty for students who had a decrease or no improvement vs. increase in fitness from the year

203

prior. Students in the 6th grade show inverse dose-response trends between fitness and

204

absenteeism, although higher grades do not show this trend.

AC C

205

EP

200

206

Table 3 shows higher absenteeism with decreasing fitness and increasing grade level across both

207

genders. Boys showed more mean days absent than girls across all grades, although girls

12

ACCEPTED MANUSCRIPT Fitness, poverty and gender on absenteeism

Table 3. Mean in absenteeisma for New York City public school students in grades 6-8, by level of fitness-change from the previous year across gender (Nstudents=349381b, Nobs= 675318c) Grade 7 (nobs=257203)

Grade 8 (nobs=213526)

22053 10446 29804 11509 27202

10.23(10.78) 9.75(10.57) 9.55(10.66) 9.32(10.06) 9.14(9.86)

23640 15492 50242 15536 22613

11.32(12.49) 10.30(11.77) 9.61(11.34) 9.40(10.91) 9.53(10.50)

Boys (nobs=340982) Change in Fitnessd >20% decrease 10-20% decrease 20% increase

n 21471 11017 32325 12174 26588

Mean (SD) 11.06(11.76) 10.60(11.20) 10.43(11.07) 10.08(10.65) 9.97(10.39)

n 23035 16023 51272 16401 22949

Mean (SD) 11.82(12.80) 10.94(12.00) 10.52(11.98) 10.02(11.04) 10.13(11.13)

n 17982 13129 44723 13593 18300

M AN U

a

17934 13065 44795 12904 17101

13.51(13.77) 12.33(13.19) 11.60(12.28) 11.27(12.15) 11.80(12.85)

SC

Girls (nobs=334336) Change in Fitnessd >20% decrease 10-20% decrease 20% increase

RI PT

Grade 6 (nobs=204589)

Mean (SD) 14.27(14.93) 12.76(13.36) 12.27(13.00) 11.71(12.32) 11.97(12.65)

Tabulated mean estimates based on days absent per year. bNmale students=177355, Nfemale students=172026, Nlow/mid poverty c students=181457, Nhigh/very high poverty students=167917. Observations account for 1-3 years of fitness-change per student. d Change in fitness composite percentile scores from the year prior was based on PACER (Progressive Aerobic Cardiovascular Endurance Run), Push-up and Curl-up Fitnessgram tests.

showed greater increases in absenteeism with increasing grade and decreasing fitness. The overall

209

mean absentee rate across all schools (n=624) based on the unconditional longitudinal

210

model intercept was 12 days per year. ICC estimates demonstrated a large degree of clustering at

211

the school-level (9% for models including all students; 10% for gender-stratified models; see

212

Table A.1).

EP

AC C

213

TE D

208

214

Longitudinal association of fitness-change, absenteeism and school-area poverty across gender

215

Crude models examining the fitness-absenteeism association stratified by school-area poverty

216

and gender showed significant associations between change in fitness and lagged absenteeism for

217

all poverty levels across both genders (p20% Increase 10-20% Increase 20% Decrease

IRR (model 1)e,f

BOYS (nobs=340982) Change in Fitnessd >20% Increase 10-20% Increase 20% Decrease

IRR (model 3)e,h

95% CI lower upper 0.821 0.912 0.831 0.936 0.850 0.933 0.888 1.003 . .

TE D

0.866 0.882 0.890 0.944 .

M AN U

SC

Table 4. Crude longitudinal association of fitness-change and absenteeism in New York City public school students in grades 6-8 (Nstudents=349381a, Nobs=675266b, 624 schools) across school area poverty and gender. Low/Mid School-Area Povertyc High/Very High School-Areac (nobs=355951) Poverty (nobs=319315)

AC C

EP

0.877 0.885 0.916 0.936 .

95% CI lower upper 0.817 0.941 0.818 0.958 0.860 0.974 0.864 1.015 . .

IRR (model 2)e,g 0.897 0.915 0.966 0.971 .

IRR (model 4)e,i 0.891 0.906 0.962 0.971 .

95% CI lower upper 0.847 0.950 0.856 0.979 0.919 1.016 0.909 1.037 . .

95% CI lower upper 0.819 0.969 0.823 0.997 0.894 1.036 0.882 1.069 . .

Abbreviation: IRR, Incidence rate ratios. aNmale students=177355, Nfemale students=172026, Nlow/mid poverty b students=181457, Nhigh/very high poverty students=167917. Observations account for 1-3 years of fitness-change per c student. Based on percentage of households in the school zip code living below the federal poverty threshold (low (30%) area poverty) drawing from the American Community Survey (ACS) 2007-2012. dChange in fitness composite percentile scores from the year prior was based on PACER (Progressive Aerobic Cardiovascular Endurance Run), Push-up and Curl-up Fitnessgram tests. eExponentiated beta coefficients (IRR) represented the association of each level of exposure (>20% increase, 10-20% increase, 20% decrease in fitness). P-values were calculated using adjusted negative binomial longitudinal mixed models; Statistical significance of number of days absent per year was assessed using Type III Tests of Fixed Effects. A two sided P-value of 20% Increase 10-20% Increase 20% Decrease

RI PT

Table 5. Adjusteda longitudinal association of fitness-change and absenteeism in New York City public school students in grades 6-8 (Nstudents=349381b, Nobs=675266c, 624 schools) across school-area poverty and gender. Low/Mid School-Area Povertyd High/Very High School-Area (model 5) Povertyd (model 6) IRRf,g 95% CI IRRf,h 95% CI Girls (nobs=334336)

upper 0.975 0.862 0.979 1.039 .

IRRf,j

0.893 0.898 0.941 0.957 .

95% CI

lower 0.786 0.775 0.840 0.826 .

upper 1.016 1.041 1.060 1.110 .

225

AC C

EP

TE D

Abbreviation: IRR, Incidence rate ratios. aAdjusted for individual-level race/ethnicity, grade, calendar year, change in obesity status from the year prior, place of birth (US (not NYC), NYC, or foreign), group-level school size, and including interactions (Grade*Ethnicity and Grade*Place of Birth). bNmale students=177355, Nfemale students=172026, Nlow/mid poverty students=181457, Nhigh/very high poverty students=167917. c Observations account for 1-3 years of fitness-change per student. dBased on percentage of households in the school zip code living below the federal poverty threshold (low (30%) area poverty) drawing from the American Community Survey (ACS) 2007-2012. eChange in fitness composite percentile scores from the year prior was based on PACER (Progressive Aerobic Cardiovascular Endurance Run), Push-up and Curl-up Fitnessgram tests. fExponentiated beta coefficients (IRR) represented the association of each level of exposure (>20% increase, 10-20% increase, 20% decrease in fitness). P-values were calculated using adjusted negative binomial longitudinal mixed models; Statistical significance of number of days absent per year was assessed using Type III Tests of Fixed Effects. A two sided P-value of 20% Increase

10-20% Increase

20% decrease in fitness)

Note. The graph illustrates the relationship between fitness change and absenteeism in girls by

517

school-area poverty. As shown above, relative to girls attending schools in high/very high

518

poverty areas who had a large decrease in fitness (>20% from the year prior), those who had a

519

large increase (>20%), small increase (10-20%),