Early Postinjury Screen Time and Concussion Recovery

This study was a planned secondary data analysis of the Advancing Concussion Assessment in Pediatrics (A-CAP) study,¹⁴  a prospective longitudinal cohort study of children and adolescents who sustained a concussion or OI. Participants were recruited from 5 EDs within the Pediatric Emergency Research Canada network from September 2016 to December 2018.

Participants ages 8 to 16 years old who presented to the ED within 48 hours of sustaining a concussion or OI were screened for eligibility. Children and adolescents with concussion were eligible to participate if they had experienced blunt head trauma resulting in 1 or more of the following criteria generally consistent with the World Health Organization definition of mild traumatic brain injury (TBI): observed loss of consciousness, a Glasgow Coma Scale score of 13 to 14, or 1 or more acute signs or symptoms of concussion (eg, confusion, headache). Children and adolescents with OI were eligible if they sustained upper or lower extremity fractures, sprains, or strains because of physical trauma, and had an Abbreviated Injury Scale¹⁵  score of 4 or less.

Exclusion criteria specific to the concussion group included neurologic deterioration, neurosurgical intervention, loss of consciousness >30 minutes or posttraumatic amnesia > 24 hours, and bodily injuries with an Abbreviated Injury Scale score >4. Exclusion criteria for the OI group included head trauma, acute signs and symptoms of concussion, or injury requiring surgical intervention or procedural sedation. Additional exclusion criteria for both groups included hypoxia, hypotension or shock; non-English speaking; previous TBI requiring overnight hospitalization; concussion within the past 3 months; history of severe neurologic or neurodevelopmental disorder; psychiatric hospitalization in the previous year; administration of sedative medicine before ED data collection; alcohol or drug use at time of injury; and legal guardian not present.

The A-CAP study protocol was published previously.¹⁴  In brief, research personnel collected demographic information, injury details, and acute signs or symptoms of concussion in the ED from medical records and medical personnel using a standardized case report form. Follow-up assessments were targeted for 7 days, 3 months, and 6 months postinjury. At each follow-up, participants and their parents each completed questionnaires regarding symptoms and health behaviors. At the first follow-up, parents also completed measures of preinjury symptoms and health behaviors. In addition, participants and their parents completed remote ratings of postconcussion symptoms weekly during the first 3 months and biweekly from 3 to 6 months postinjury. All participants provided written informed consent or assent. The study was approved by each participating institution’s ethics review board.

The Health and Behavior Inventory (HBI),¹⁶  which is recommended as a primary outcome measure in the National Insitute of Neurological Disorders and Stroke Common Data Elements for sport-related concussion,¹⁷  was used as the primary outcome. The self- and parent-proxy versions provide 2 separate scales for cognitive and somatic symptoms, with higher scores reflecting a higher frequency of symptoms.

The Healthy Lifestyle Behaviors Questionnaire (HLBQ) assessed parent-proxy and self-reported pre- and postinjury engagement in health behaviors, including physical activity and rest, cognitive activity and rest, diet, sleep, and screen time. The scale measuring screen time was used as the primary predictor. Each participant and parent rated the frequency and duration that the child or adolescent engaged in screen time over the previous 7 days. A 5:2 weighted average for weekdays and weekends was computed to provide an estimated weekly average.¹⁸  The total weighted average was used in analyses. Supplemental Information provides information on the development and validation of the HLBQ (see Supplemental Document 1).

Descriptive statistics summarized baseline characteristics. To examine the association between screen time and postconcussion symptoms over time, we fit 4 generalized least squares models, 1 for each of the 4 HBI outcome measures: self-reported somatic and cognitive symptoms and parent-reported cognitive and somatic symptoms. We included both self- and parent-reported screen time as predictors, for 3 reasons: (1) the scale that more accurately reflects a participant’s actual screen time is unclear, (2) self- and parent-reported screen time correlated only moderately (Spearman’s r = .60) in our sample, and (3) we evaluated their combined predictive contribution (ie, joint effect) to postconcussion symptoms. Generalized least squares were applied with a continuous first-order autoregressive¹⁹  structure, with nesting of patients within sites, allowing for correlated residuals from repeated measurement of symptoms and to account for potential clustering because of multisite recruitment. A common set of 27 covariates previously shown to be associated with concussion outcomes^10–12  and parent ratings of preinjury screen time and child and parent ratings of other preinjury and postacute lifestyle behaviors (eg, sleep, physical activity) were included in all models to help isolate the effect of postinjury screen time and reduce potential confounding between the risk for poor recovery and screen time use (eg, children with lower risk for prolonged recovery returning to screen time more quickly) (see Supplemental Document 2, Supplemental Table 3).

To allow for nonlinearity, restricted cubic splines²⁰  were applied to selected continuous variables considering allowable degrees-of-freedom to maintain an approximate 10:1 observation-to-parameter ratio. For self- and parent-reported screen time and preinjury HBI somatic and cognitive symptoms, 4 knots (ie, pivot points) at strategically-spaced quantiles (0.05, 0.35, 0.65, 0.95) were afforded. For age, social deprivation index, material deprivation index, and preinjury parent ratings of HBI somatic and cognitive symptoms, 3 knots were afforded (0.10, 0.5, 0.90). Linearity was assumed for all HLBQ covariates.

For all models, the initial specification featured 2 sets of 3-way interactions inclusive of both linear and nonlinear components of a continuous predictor’s effect. The first set of interactions involved self-reported screen time, group, and time, and the other consisted of parent-reported screen time, group, and time. To simplify the models, we applied a stepwise backward elimination procedure to sequentially remove factors with the largest P value until only P < .05 remained (on the basis of a multiple degrees-of-freedom Wald χ² test, if applicable). As the predictors of primary interest, the main effects of self- and parent-reported screen time were exempted from elimination. Listwise deletion was applied for all models.

To assess the effect of individual predictors and covariates for each model, we report: (1) the Wald χ² and associated P value, (2) Wald χ² minus degrees-of-freedom for evaluating the relative contribution among predictors, and (3) partial effect plots to illustrate the estimated covariate-adjusted relationship between our main predictors of interest and HBI outcomes. Analyses were performed by using R version 4.0.3 (rms package).^20,21 

A total of 967 children and adolescents (58.1% male) were recruited. See Figure 1 for a participant flow diagram. 829 participants completed the postacute assessment. Of those 829, 117 (14%) participants had incomplete covariate data; therefore, 712 participants were included in all 4 final models. Baseline characteristics are summarized in Table 1. There were no significant differences between those who were included or excluded in the final models in age, sex, other demographics, or preinjury history (see Supplemental Document 2, Supplemental Table 4).

Our hypothesis that more early screen time would predict greater symptom severity over time in children and adolescents with concussion relative to those with OI was only partially supported. The interaction between screen time and group was not significant for parent-reported cognitive symptoms or self-reported somatic symptoms (see Table 2). In other words, the association of screen use during the postacute recovery period with these outcomes did not differ by group (see Fig 3). The interaction between self-reported screen time and group was significant for self-reported cognitive symptoms (P = .03) and parent-reported somatic symptoms (P = .01; see Table 2); however, the relationship between screen time and group differences in postconcussion symptoms was not linear. Contrary to our hypothesis, group differences in symptom severity (concussion > OI) were larger at the 25^th and 50^th quantiles of screen time than at the 75^th quantile (see Fig 3). Thus, higher screen time was not uniformly associated with more symptoms in the concussion group relative to the OI group, although group differences were also larger at the highest levels of screen time (eg, 90^th quantile).

In all models, symptom severity was higher in the concussion group than in the OI group while holding all other covariates constant over the first 30 days post injury (See Fig 3). Group differences were negligible after 30 days postinjury regardless of screen time (See Fig 3).

Visual inspection of the relationship between screen time and symptom severity within the concussion group reveals subtle U-shaped distributions in 3 of the 4 models (ie, self-reported somatic, self-reported cogntive, and parent-reported somatic; See Fig 3). Specifically, children and adolescents with concussion who reported the lowest (<25^th quantile) and highest screen time (>90^th quantile) during the first 7 to 10 days postinjury reported more cognitive and somatic symptoms than those in the25^th to 75^th quantile of screen time (See Fig 3). Similarly, parents of children with concussion who reported the lowest or highest levels of early screen time (<25^th and >90^th quantile) reported more somatic symptoms than parents of children and adolescents in the 25th to 75th quantile (See Fig 3). We conducted a supplementary analysis using a validated risk score (“5P”)¹⁰  instead of the full set of 27 covariates, to determine if higher risk for developing persistent symptoms moderated the relationship between screen time and concussion recovery. Children and adolescents with concussion with the worst prognosis had similar recovery trajectories regardless of their early screen time (see Supplemental Document 3, Supplemental Fig 6–9).

For all models (see Fig 2A–D), time since injury, group, and preinjury cognitive and somatic symptom severity were more strongly associated with parent- and self-reported symptom severity than self- and parent-reported screen time. Self- and parent-reported screen time accounted for only 0.6% to 3.5% of the proportion of total predictive ability (ie, individual Wald X²  divided by total Wald X² ) derived from all predictors across the 4 models, compared to 10.3% to 39.6% for time since injury and 2.8% to 14.2% for injury group. Preinjury cognitive symptoms accounted for 60.8% and 19.9% of total predictive ability in models for parent- and self-rated cognitive symptoms. Preinjury somatic symptoms accounted for 21.1% and 9.2% of total predictive ability for parent- and self-rated somatic symptoms.

Several other demographic factors and health behaviors were also important predictors of symptoms regardless of group membership. Female sex predicted higher parent-reported somatic and self-reported somatic and cognitive symptom severity. More postacute napping, more preinjury screen time, and older age also predicted higher self- and parent-reported cognitive symptoms, and more self-reported sleep in the postacute recovery period was associated with fewer self-reported cognitive symptoms.

This study sought to assess the association between screen time during the first 7 to 10 days postinjury and postconcussion symptom severity over the following 6 months in those with concussion or OI. After controlling for other factors associated with screen time and symptom persistence, greater screen time was not consistently associated with increased postconcussion symptoms. The association between screen time and postconcussion symptoms differed between the concussion and OI groups for only 2 of 4 symptom measures; and for these outcomes, higher screen time was not consistently associated with more symptoms or larger group ifferences (ie, concussion versus OI) in symptoms. Moreover, when present, group differences as a function of screen time were limited to the first 30 days postinjury of the 6-month follow-up period. Although the Macnow et al⁵  clinical trial demonstrated that reducing screen time in the first 48 hours postinjury may offer short-term benefits, small group differences in our sample after 30-days postinjury suggest that early screen time is not detrimental for long-term recovery in children and adolescents with concussion.

A novel aspect of the current study was the examination of concussion recovery in relationship to the full naturalistic range of screen time. The screen time abstinence group in Macnow et al (2021) spent an average of 45 minutes per day on screens compared to 3.5 hours per day in the unrestricted screen time group, indicating a comparison of less to more screen time, rather than none to some. This comparison, while important, provides little insight into the association of the full naturalistic range of screen time with concussion recovery. In the current study, the association between screen time and symptom severity was not linear. Participants in the concussion group with low screen time (<25^th percentile) reported more postconcussion symptoms relative to those in the OI group than did those with moderate screen time (25–75^th percentile). This finding has at least 2 possible explanations. First, a child or adolescent with more severe acute symptoms may self-limit their screen time if it is bothersome or their health care provider or parent may be more likely to restrict their screen use. However, we statistically adjusted for acute symptom severity and other risk factors for protracted recovery, making this explanation less likely. Second, screen time is often used to interact with friends and family or for other tasks like school work. Undue restrictions from screen time after concussion could prevent access to social support or disrupt routines and increase emotional distress,²²  worsening symptoms. In healthy young adults, separation from mobile devices (eg, texting, answering phone calls) can lead to increased physiologic arousal²³  and higher self-reported anxiety,^23–25  especially for heavy phone users.²⁵ 

The “goldilocks” effect we observed in our sample, whereby moderate levels of screen time were associated with less pronounced group differences in symptoms compared to low and high screen use, is not unique to children and adolescents with concussion. The association between screen time and well-being is similarly nonlinear in healthy adolescents.²⁶  In the digital age, too little screen time may limit meaningful social connections or disrupt routines, whereas too much may disrupt sleep or interfere with engagement in other important activities, such as recreation.²⁶  As is true for the return to physical activity after concussion,¹³  moderation in screen time may be a helpful principle in concussion management.

Consistent with previous studies, greater preinjury cognitive and somatic symptoms,²⁷  adolescent age,^10,12  and female sex^12,28  were associated with more severe postconcussion symptoms over time. These variables had a larger effect on symptom severity than screen time. Other lifestyle behaviors also predicted worse symptoms, including more postacute napping, less sleep, and more preinjury screen time.

Our study has several limitations. We only measured the duration of screen time. We did not assess the timing (eg, concentrated versus dispersed), quality (eg, active versus passive), or nature (eg, watching television versus social connection) of use. These may be important moderators of the association between screen time and postconcussion symptoms. Second, when parent and self-reports of screen time were discrepant, we could not verify which was more accurate. We used the more conservative estimate of screen time when a large discrepancy existed and truncated values when the sum of reported screen time and sleep duration was >24 hours. Only a small proportion of participants met these criteria (9% for self-report and 7.3% for parent-report). Third, because recruitment occurred in EDs, the results may not generalize to children or adolescents who do not seek care or seek alternative care after injury. We also do not have detailed data on the timing of follow-up care, which could impact recovery. Fourth, children and adolescents who enrolled in the current study may differ from those who did not enroll leading to selection bias. Finally, future work is needed to understand if certain individuals, such as those with vestibulocular dysfunction, are more sensitive to screen use postconcussion.

High screen time in the first 7 to 10 days postinjury was not strongly associated with worse symptoms after concussion, especially after the first 30 days postinjury, and low screen time was also associated with relatively more postconcussion symptoms. Moreover, other pre- and postinjury risk factors were more important than screen time in predicting postconcussion symptoms. These findings support advising moderation rather than blanket restrictions in screen time, especially beyond the first 48 hours, for children and adolescents postconcussion.

A-CAP

Advancing Concussion Assessment in Pediatrics

ASHS

Adolescent Sleep Hygiene Scale

CRSP

Children’s Report of Sleep Patterns

ED

emergency department

HBI

Health and Behavior Inventory

HBSC

Healthy Behaviors in School-aged Children

HLBQ

Healthy Lifestyle Behaviours Questionnaire

OI

orthopedic injury

TBI

traumatic brain injury