Low sports participation is associated with withdrawn and depressed symptoms in urban, school-age children

Abstract

Background:

The association between sports participation and mental health has not been studied in primary care samples of school-age children, nor in underrepresented minority children. We assessed the relationship between number of sports played and psychiatric symptoms in children ages 6–11 at well-child visits in an urban clinic.

Methods:

Guardians of 206 children (85% Latinx) ages 6–11 completed Child Behavior Checklists (CBCL) in Spanish (66.5%) or English at well-child visits at an urban community health center. We performed linear regression between number of sports played and individual CBCL syndrome scores, and multiple logistic regression with normal (T-score <60) vs. elevated (T-score ≥60) CBCL syndrome scale score as the outcome. We conducted bivariate, multiple logistic regression, and linear regression analyses between low (1 or fewer) vs. high (2 or more) sports participators and subscales of interest.

Results:

Fewer sports played was associated with higher Withdrawn/Depressed CBCL syndrome scale T-scores (p = 0.019), but not with other CBCL syndrome scale scores nor number of syndrome scale elevations (p = 0.638). Low participators had higher odds of an elevated Withdrawn/Depressed T-score (p = 0.033) than high participators.

Limitations:

Our dataset did not contain certain details about sports played, nor information about income and insurance, and our results may not generalize to other populations.

Conclusions:

Playing fewer sports is associated with higher withdrawn/depressed symptoms in urban, predominantly Latinx, school-age children. Therefore, urban school-age children with low sports participation may be at risk for depression, and sports participation might protect against depressive symptoms in childhood.

INTRODUCTION

While sports participation during childhood is clearly important for physical health, growing evidence suggests that sports can benefit mental health as well. Several studies have reported a bidirectional association between higher levels of sports participation or physical activity and improved psychosocial functioning in youth.^1–6 Other studies have found an association between lower levels of physical activity and psychiatric symptoms in adolescents, particularly mood disorders and internalizing symptoms.^7–9 Additional evidence suggests that physical activity may benefit both adults and children with ADHD, depression, and anxiety.^2,10–13

However, there are significant gaps in the evidence linking sports participation with improved psychosocial functioning in children. Most studies do not assess for a range of internalizing and externalizing symptom outcomes, but focus on narrower symptom domains - most commonly depression or ADHD.^10–12 In addition, most studies have been conducted with adolescents as opposed to younger children, and no studies have focused on underrepresented minority children who have sparse access to mental health care^14–16 as well as possible economic barriers to sports participation.^17,18 Furthermore, no studies have examined this association in primary care-based samples, where referrals to community-based sports programs could be an early intervention for children with emerging psychiatric symptoms. Finally, there is little understanding about the importance of number of unique sports played - rather than whether sports are played at all, or how many minutes of physical activity are achieved per day.

Thus, the objective of this study was to assess the relationship between number of unique sports played and psychiatric symptoms in school-age children presenting for well-child visits at an urban primary care clinic. Our goal was to identify whether there was an association between sports participation and different domains of guardian-reported psychiatric symptoms, and whether the association between sports participation and psychiatric symptoms differed based on demographic characteristics. We conducted a cross-sectional, secondary data analysis using an existing dataset collected at well-child visits at an urban community health center serving a predominantly Spanish-speaking population.

METHODS

Population and Setting:

The dataset for this study was collected at the pediatric unit of an urban community health center affiliated with a major academic medical center. Data was collected originally for a study on use of psychiatric screening tools in primary care from 241 children, ages 6–11, whose guardians completed a Child Behavior Checklist (CBCL) for ages 6–18 at well-child visits from January of 2013 until June of 2015, with the assistance of research staff.¹⁹ Guardians were eligible to complete the CBCL if they spoke Spanish or English, and arrived with their child for a well-child visit (i.e. annual routine health visit) on a study recruitment day. Guardians either completed the CBCL themselves by entering their answers directly into the secure Research Electronic Data Capture (REDCap) database, or had it read to them by research staff who entered their answers into the database.

The Partners Healthcare Human Research Committee approved this study. The Boston University Medical Campus Institutional Review Board approved this analysis.

Measures:

All variables in this secondary data analysis were drawn from guardian-report responses on the CBCL for ages 6–18. The CBCL is a well-validated self-report questionnaire to assess past-month emotional, behavioral, and social problems in children, and is available and validated in many languages.²⁰ It consists of two parts. The first part includes items about the child’s life and environment, ranging from guardian demographics to the child’s performance in school, social interactions, and participation in sports and activities. The second part consists of 113 symptom questions whose answers are scored on a 3-point Likert scale from 0–2 (0=not true, 1=somewhat true, or 2=very true). For scoring, symptom questions are grouped into eight empirically-based “syndrome scales,” which indicate different categories of psychiatric/behavior problems. The Likert responses for each question on a scale are added up to produce a raw score. The raw score is then transformed into a T-score using a computer program or conversion table. A T-score of 50 indicates average symptoms compared to other children with the same age and gender. Every 10-point increase in T-score represents one standard deviation from the mean.

To measure sport count, we used the item on the CBCL that asks the guardian to list up to three sports in which their child participates. For this analysis, we counted the number of sports played, as indicated by the guardian. To measure domains of psychiatric symptoms, we used the 8 syndrome scales: Anxious/Depressed; Withdrawn/Depressed; Somatic Complaints; Social Problems; Thought Problems; Attention Problems; Rule-Breaking Behavior; and Aggressive Behavior.

Data Analysis:

Data analysis was performed using Statistical Analytical Systems (SAS) version 9.4.²¹ We generated descriptive statistics for sociodemographic and clinical characteristics. We performed a simple and multiple linear regression between sport count and the individual CBCL syndrome scales. Since age, gender, and ethnicity can contribute to the frequency and severity of symptoms as well as sports participation,^22–26 we controlled for these characteristics in our analyses. We used Cronbach’s alpha to determine the internal consistency of the CBCL in our population. Previously, the CBCL has been used for research in Hispanic/Latinx and Spanish-speaking populations²⁷ and has been found to have high internal consistency on the Total Problems scale and at least acceptable internal consistency on most syndrome subscales when used in a general population of Hispanic/Latino children.²⁸

We also performed a multiple logistic regression by dichotomizing our outcome variable by T-score <60 (normal) vs. ≥60 (elevated) on individual subscales. Although the published cutoff for a clinically elevated CBCL subscale is a T-score of 65, a lower clinical cutoff of 60 (one standard deviation above the mean) has been shown to accurately identify symptomatic children in previous studies.^29,30 A cutoff of 60 could represent a meaningful departure from the mean and may appropriately increase sensitivity as a screening measure.^19,29,30 Age, gender, ethnicity, and language were all included in our model as covariates.

Since number of sports was not normally distributed, we further explored significant associations found in our regression analyses by dichotomizing children into groups of low vs. high participators in sports. To do this, we used the median (1 sport played) as a cut-off, so that low participators were those who played zero or one sport, and high participators played two or more sports. We conducted bivariate analyses to determine the association between low vs. high participation on elevated vs. non-elevated CBCL T-scores for positive results in the regression analyses. We conducted a multiple logistic regression analysis with low vs. high sports participation as a predictor and elevated vs. non-elevated CBCL subscale as outcomes, including covariates as above. Finally, we conducted a multiple linear regression analysis with low vs. high sports participation as a predictor and Withdrawn/Depressed as a continuous outcome. For our model selection, we employed stepwise and backward model selection procedures to determine which predictors to include in our analyses. We checked for confounding by observing a change in our regression coefficients greater than 10%. We checked for effect modification by sociodemographic characteristics.

RESULTS

Demographics:

Of the 241 total enrolled participants, guardians of 206 children completed the CBCL questionnaire. Demographic and clinical characteristics of the final sample are shown in Table 1. Participants were 85% Latinx, and 66.5% of CBCL questionnaires were completed in Spanish. Participants had a mean age of 7.9 (range 6–11) and were almost evenly split by gender (50.5% male, 49.5% female). The most common sport played was soccer (39.3%), followed by basketball (21.4%), swimming (19.4%), and “movement sports” (including dance and gymnastics) (16.5%).

Table 1:

Demographic and Clinical Characteristics.

Demographics	N = 206
Age: Mean ± SD	7.9±1.4
Gender: n(%)
Male	104 (50.5%)
Female	102 (49.5%)
Child ethnicity: n(%)
Hispanic/Latinx	175 (85.0%)
Not Hispanic/Latinx	31 (15.0%)
Guardians’ primary language: n(%)
English	69 (33.5%)
Spanish	137 (66.5%)
CBCL Syndrome Scale T-score ≥ 60: n(%)
Anxious/Depressed	30 (14.6%)
Withdrawn/Depressed	45 (21.8%)
Somatic Complaints	45 (21.8%)
Social Problems	38 (18.4%)
Thought Problems	25 (12.1%)
Attention Problems	38 (18.4%)
Rule-Breaking Behavior	33 (16.0%)
Aggression	33 (16.0%)
Number of Sports Playeda: n(%)
0	36 (17.5%)
1	84 (40.8%)
2	41 (19.9%)
3	45 (21.8%)
Frequency of Specific Sportsb: n(%)
Soccer	81 (39.3%)
Basketball	44 (21.4%)
Swimming	40 (19.4%)
Movement (e.g. dance, gymnastics)	34 (16.5%)
Road Sports (e.g. bike/scooter riding)	30 (14.6%)
Running	25 (12.1%)
Football	17 (8.3%)
Baseball	16 (7.8%)
Otherc	8 (2.6%)

SD=Standard Deviation

CBCL=Child Behavior Checklist for ages 6–18

^aCBCL specifies over the prior 6 months

^bTotal percentages do not add to 100% because many children played 2 or 3 sports

^cIncludes one each of: hockey, lacrosse, volleyball, tennis, boating, bowling, fishing, and golf

In our sample, Cronbach’s alpha ranged from 0.65–0.86 for the eight subscales and 0.94 for the ‘Total Problems Scale.’ All values are in the range of acceptable to very good, consistent with earlier use of the CBCL in a child population.³¹

Association between number of sports and CBCL syndrome scales

A simple linear regression between number of sports and each of the syndrome scale T-scores found that a higher number of reported sports played was associated with lower scores on the CBCL Withdrawn/Depressed subscale (β-estimate = −1.093; t-statistic= −2.43; p-value = 0.016) (Table 2). The other CBCL syndrome subscale T-scores were not significantly associated with number of sports (all p > 0.05). In a simple linear regression where the predictor was the number (from 0 to 8) of elevated CBCL subscales (T-score ≥ 60), number of sports was also not significantly associated with number of subscale elevations (β-estimate = 0.064; t-statistic = 0.471; p-value = 0.638).

Table 2.

Association (unadjusted) between number of sports and CBCL syndrome scale T-scores.

CBCL Syndrome Scale	β-estimate	t-value	p-value
Anxious/Depressed	0.358	0.872	0.384
Withdrawn/Depressed	−1.094	−2.434	0.016**
Somatic Complaints	0.634	1.581	0.115
Social Problems	0.164	0.394	0.694
Thought Problems	0.401	1.053	0.294
Attention Problems	0.321	0.631	0.529
Rule-Breaking Behavior	0.368	0.921	0.358
Aggressive Behavior	0.364	0.838	0.403

CBCL=Child Behavior Checklist for ages 6–18

^**p < 0.01

Results from the multiple linear regression analysis indicated that there was a collective significant association between Withdrawn/Depressed subscale and higher number of sports, age, and ethnicity (F(5,200) = 4.93; R² = 0.109; p-value = 0.003). When individual predictors were examined, we found a higher number of sports remained significantly associated with lower scores on the Withdrawn/Depressed subscale after controlling for demographic characteristics (β-estimate = −1.044; t-statistic = −2.37; p-value = 0.0186) (Table 3). Age and ethnicity were also significantly associated with the Withdrawn/Depressed subscale. Older children had higher Withdrawn/Depressed subscale scores (β-estimate = 0.958; t-statistic = 3.05; p-value = 0.0026), and Latinx children had higher Withdrawn/Depressed subscale scores compared to non-Latinx children (β-estimate = 3.158; t-statistic = 2.10; p-value = 0.0368).

Table 3.

Association between the CBCL Withdrawn/Depressed Syndrome Scale T-Score and number of sports adjusting for sociodemographic factors.

Predictora	β-estimate	t-value	p-value
Number of Sports	−1.044	−2.37	0.019*
Age	0.958	3.05	0.003**
Gender (male)	1.527	1.72	0.088
Ethnicity (Hispanic/Latinx)	3.158	2.10	0.037*
Guardian’s Language (English)	−0.028	−0.02	0.980

^aOutcome is the Withdrawn/Depressed Syndrome Scale T-Score.

CBCL=Child Behavior Checklist for ages 6–18

^*:p < 0.05

^**:p < 0.01

In the multiple logistic regression analysis controlling for age, gender, ethnicity, and guardian language, number of sports was not significantly associated with elevation of any CBCL syndrome scale T-score using ≥60 as the cut-off, including for the Withdrawn/Depressed syndrome (Wald χ² = 2.715; p-value = 0.099) (Table 4). Older children had higher odds of an elevated Withdrawn/Depressed scale (Wald χ² = 8.586; Adjusted Odds Ratio [aOR] [95% CI] = 1.451 [1.131–1.860]; p-value = 0.003).

Table 4.

Association between Withdrawn/Depressed Syndrome Scale elevation and number of sports.

Predictora	Wald χ2b	aORc	95% CId	p-value
Number of Sports	2.715	0.735	0.509–1.060	0.099
Age	8.587	1.451	1.131–1.860	0.003**
Gender (male)	0.007	0.970	0.483–1.947	0.931
Ethnicity (Hispanic/Latinx)	2.303	0.280	0.054–1.449	0.129
Guardian’s Language (English)	0.328	1.310	0.520–3.298	0.567

^aOutcome is Withdrawn/Depressed Syndrome Scale Elevation (T-score ≥60)

^bWald χ² = adjusted test statistics

^caOR = Adjusted Odds Ratio

^d95% CI= 95% Confidence Interval

^*:p < 0.05

^**:p < 0.01

In a bivariate analysis of non-participators (0 sports) vs. participators (any sports), there was no significant association between sports participation and Withdrawn/Depressed T-score elevation. However, in a bivariate analysis of low participators (0 or 1 sports) vs. high participators (2 or more sports), low participators had twice the odds of having elevated Withdrawn/Depressed T-scores than high participators (χ² = 5.385; OR [95% CI] = 2.339 [1.127–4.853]; p-value = 0.020).

In a logistic regression model using low vs. high sports participation as the predictors, low participators had higher odds of an elevated Withdrawn/Depressed syndrome scale than high participators (Wald χ² = 4.400; aOR [95% CI] = 2.240 [1.054–4.760]; p-value= 0.033) (Table 5). Older children had significantly higher odds of an elevated Withdrawn/Depressed syndrome scale (Wald χ² = 8.255; aOR [95% CI] = 1.441 [1.123–1.849]; p-value = 0.004) adjusting for sociodemographic characteristics. In a multiple linear regression model with the same predictors (F(5, 200) = 5.32; R² = 0.12; p-value = 0.0001), low participators had significantly higher Withdrawn/Depressed T-scores compared to high participators (β-estimate = 2.448; t-statistic = 2.72; p-value = 0.0071.)(Table 6) Older children (β-estimate = 0.935; t-statistic = 2.99; p-value = 0.0031) and Hispanic/Latinx children (β-estimate = 3.146; t-statistic = 1.50; p-value = 0.036) had significantly higher Withdrawn/Depressed T-scores in this model.

Table 5.

Association between Withdrawn/Depressed Syndrome Scale elevation and low sports participation.

Predictora	Wald χ2b	aORc	95% CId	p-value
Low Sport Participation (0–1)	4.400	2.240	1.054–4.760	0.036*
Age	8.256	0.694	0.541–0.890	0.004**
Gender (male)	0.0008	0.990	0.491–1.996	0.978
Ethnicity (Hispanic/Latinx)	2.298	0.280	0.054–1.452	0.130
Guardian’s Language (English)	0.2999	1.296	0.540–1.452	0.584

^aOutcome is Withdrawn/Depressed Syndrome Scale Elevation (T-score ≥60)

^bWald χ² = adjusted test statistics

^caOR = Adjusted Odds Ratio

^d95% CI= 95% Confidence Interval

^*:p < 0.05

^**:p < 0.01

Table 6.

Association between Withdrawn/Depressed Syndrome Scale T-score and low sports participation.

Predictora	β-estimate	t-value	p-value
Low Sports Participation (0–1)	2.448	2.72	0.007*
Age	0.935	2.99	0.003**
Gender (male)	1.463	1.65	0.10
Ethnicity (Hispanic/Latinx)	3.145	1.45	0.037*
Guardian’s Language (English)	−0.012	−0.01	0.991

^aOutcome is the Withdrawn/Depressed Syndrome Scale T-Score.

CBCL=Child Behavior Checklist for ages 6–18

^*:p < 0.05

^**:p < 0.01

DISCUSSION

We found that among urban, predominantly Latinx children ages 6–11 seen in primary care, higher Withdrawn/Depressed syndrome scale scores on the CBCL were associated with fewer sports played. This important finding is consistent with the limited existing literature that children who participate in sports have fewer depressive symptoms. Our findings are also consistent with the more extensive literature in adolescents and adults that has found an association between sports and depressive symptoms.^{1–5,9,11,13,32} Furthermore, we did not find an association between sports participation and any of the other seven psychiatric symptom domains measured with the CBCL in our sample. These findings are novel because, to our knowledge, this is the first study to examine the association between mental health symptoms and guardian-reported sports participation at well-child visits and in an urban, Latinx, predominantly Spanish-speaking sample of children. This is also one of few studies on this topic conducted with school-age children in the United States (U.S.).^2,21 and one of very few to look at such a broad range of psychiatric symptoms using a well-validated, broad-band psychometric tool. Our findings suggest that asking about sports participation could be a helpful additional measure of psychosocial functioning and a point for early intervention in primary care.

Prior research has examined the association between sports participation or physical activity and psychiatric symptoms in both general population surveys and samples of children with psychiatric disorders.^1,7,9,10,32 However, this is the first study to examine this question at well-child visits, reflecting a clinical population including concerned guardians and children with early signs of psychiatric symptoms. Half of the patients in our study had at least one CBCL subscale elevation, suggesting that guardians will readily report symptoms at well-child visits and may be receptive to early, low-level interventions. With the rise of pediatric medical homes intended to serve patients by connecting them with community-based resources, referring patients who present with depressive symptoms and social withdrawal to sports programs from primary care may be quite feasible. Our results also suggest that adding a short measure of functioning to primary care psychiatric screens (e.g. asking about participation in after school activities including sports) could be important for intervention planning.

The chance to play sports may not be available to many children for economic reasons. Participation in sports can be expensive due to costs for enrollment, equipment, and travel for families. Our sample was from an urban community health center serving predominantly low-income, publicly-insured Hispanic and other immigrant children. In the U.S., Hispanic youth have lower rates of sports participation compared to non-Hispanic white and non-Hispanic black children^23,33–35. Low socioeconomic status (SES) is also associated with lower rates of sports participation^17,36 as well as higher rates of depression.^37–39 It is certainly possible that our results are confounded by SES (i.e. that the lowest SES children are both more likely to experience depression and less likely to participate in sports). However, we also wonder if lack of access to sports could be one contributor to increased depression in low SES children,^40–42 via the mechanism of decreased opportunity for socialization and confidence-building. Future research including socioeconomic variables will be important to explore these associations. Furthermore, creative solutions to solving the gap in access to sports for low SES children - such as health insurers contributing to the cost of sports participation - may be important for reducing socioeconomic disparities in child health.³³

One important strength of our study was that using the CBCL, we were able to examine 8 different domains of psychiatric symptoms, allowing us to look at the association between sports participation and all major categories of psychiatric symptoms in children. We found an association only between sports participation and the Withdrawn/Depressed symptom domain, but not the other symptom domains (Anxious/Depressed; Somatic Complaints; Social Problems; Thought Problems; Attention Problems; Rule-Breaking Behavior; and Aggressive Behavior). The Withdrawn/Depressed subscale includes items related to social withdrawal, isolative behavior, shyness, anhedonia, and poor energy, and is distinct from the Anxious/Depressed subscale which includes fear and worry items, self-consciousness, guilt and suicidal thinking. It is possible that the interactive nature of most sports participation may impart benefits particularly for isolative and anergic symptoms in this young age group. The reverse also makes clinical sense: symptoms of social withdrawal, anhedonia, and low energy may affect a child’s willingness to leave the house and engage in interactive activities such as sports. While other studies have shown an association between sports and depressive symptoms in youth,^25,43,44 we show that this association may be specific to certain depressive symptoms in this young age group. Other research has connected sports participation with prosocial behaviors in adolescence,^43,44 and our work suggests that the connection between social isolation and sports participation begins in grade school. This has important implications for intervention building and clinical practice. Importantly, we also did not see an association between sports participation and the Attention Problems subscale, suggesting that children who present with ADHD symptoms are not lacking for these activities compared to peers. Furthermore, despite the fact that physical activity can be beneficial for ADHD symptoms, our sample of well children does not show that children who participate in more sports receive any benefit for their attention.^{10, 12}

Another strength of our study is that for our logistic regression analyses, we examined the association between sports participation and symptoms using a cut-off on the CBCL that is well-supported in the literature but lower than the suggested clinical cut-off, and thus were able to demonstrate an association with sports and emerging withdrawn/depressive symptoms that may not have yet reached a clinical threshold. For a well-child population, where detecting early warning signs and utilizing preventive or non-pharmacological treatment modalities may be helpful, we felt that this was appropriate. Previous studies have utilized a lower CBCL cut-off for screening purposes, and have found associations between T-scores below 60 and subsequent psychiatric diagnoses on the CBCL, particularly in internalizing domains.^26,27,42

The fact that children who played one or fewer sports had twice the odds of elevated Withdrawn/Depressive syndrome scales engages the attention given emerging literature shows that playing multiple team sports in childhood has benefits over concentrating on just one team sport (early specialization).^46–48 While most of these studies have been surveys of elite athletes and/or refer to concerns related to overuse injuries, the emotional toll of early sports specialization (including related to social isolation and lack of flexibility) is also a concern, as discussed in recent guidelines on youth sport participation published by the American Academy of Pediatrics (AAP).²⁵ The downside of sports specialization may be supported further by the fact that we found no significant difference in T-score elevation between children who played no sports and children who played one or more sports. Further research will be important in this area.

We found that Hispanic/Latinx ethnicity predicts a higher Withdrawn/Depressed T-score, but not a Withdrawn/Depressed T-score elevation. We think this is likely due to loss of power to demonstrate an effect at the clinical range T-scores, given that we have a well child sample with a low rate of positive scores which is 85% Hispanic/Latinx.

Our study has a few limitations. Given that this was a secondary data analysis, we did not have the detail about sports participation that would allow for a more rigorous analysis, such as hours played, organized vs. informal participation, team vs. individual sport, etc. We also did not collect income or insurance data and therefore cannot comment on the relationship between SES on sports participation and withdrawn/depressive symptoms in youth. Future studies should consider including measures of SES as a covariate as well as measures of sedentary behaviors (e.g. screen time) which may be inversely related to sports participation and also affect mental health.^49,50 Our sample may have been underpowered to detect associations between sports participation and other symptoms, especially since this was a secondary data analysis. Finally, given that this was a predominantly Latinx sample from an urban community health center affiliated with a major academic medical center, our findings may not generalize to other populations.

Conclusion:

Despite these limitations, we found that participation in fewer sports is associated with higher CBCL Withdrawn/Depressed subscale scores - but not other syndrome scale scores - in school-age children at well-child visits in an urban community health center serving primarily Latinx patients. Future studies should clarify the directionality of this link, and examine whether facilitating multi-sports participation could be a feasible, appropriate, and effective early intervention in primary care for children across all SES groups who present with social withdrawal and depressive symptoms, including social isolation, low energy, and anhedonia. Future studies should also confirm our finding that other psychiatric symptoms in children are likely not related to level of sports participation.

Highlights:

• Participation in fewer sports was associated with more child depressive symptoms.

• Number of sports played was not associated with other psychiatric symptoms.

• Low (vs. high) participators had 2x odds of elevated withdrawn/depressed scores.

• Urban school-age children with low sports participation may be at risk for depression

• Sports participation might protect against depressive symptoms in childhood