Reporting guidelines
The study used the STrengthening the Reporting of OBservational studies in Epidemiology (STROBE) guidelines for reporting of cohort studies. STROBE is endorsed by a growing number of biomedical journals and aims to improve reporting standards of observational studies [25].
Study design
The study was a retrospective, observational cohort design that evaluated subjects after exposure to a clinically diagnosed concussion based on mechanism of injury, symptom report, and clinical presentation as defined by the international consensus statement on concussion in sport [23]. Follow-up data were captured at a number of time-points including baseline (post-exposure), follow-up and discharge. The study protocol (ID Pro00058398) was approved by the institutional review board at the medical center where the study was conducted.
Setting
All data were captured at a sports concussion clinic that operates within the Duke Sports Sciences Institute (DSSI). The DSSI is a multidisciplinary facility with medical and rehabilitation management options for sports related injury and non-injury management. The concussion clinic is staffed by four primary care sports medicine physicians trained in concussion management. All individuals with concussion were evaluated by one of these physicians and clinical decision making was based on established protocols supported by the literature. The evaluation required the individual to complete subjective questionnaires including the Post-Concussion Symptom Scale, The Neck Disability Index, and the Dizziness Handicap Index. The clinical examination was comprised of standardized concussion tests including the Standardized Assessment of Concussion (SAC) for cognitive assessment, King-Devick Test for visual assessment, Vestibular-Ocular Motor Screen (VOMS) for visual (smooth pursuit and convergence) and vestibular (saccades, VOR, VMS) assessment, and the Balance Error Scoring System (BESS) for balance assessment. Cervical testing was considered positive if neck pain was present at rest, with palpation, or with active motion. Standard clinical concussion care was provided to all individuals based on their initial presentation.
Participants
Participants included 163 adolescent and college aged individuals who self-presented or were referred for evaluation at the sports concussion clinic at the Duke Sports Sciences Institute between July 1, 2013 and January 1, 2015. Eligibility required a medical diagnosis of sports-related concussion and the willingness to receive a battery of questionnaires, examinations, and follow up. Follow up occurred through January 22, 2015 as indicated by standard of care.
Variables
Descriptive characteristics included age, sex, Neck Disability Index score (NDI), Dizziness Handicap Index score (DHI), Post-Concussion Symptom Scale (PCSS), and time from concussion incident to clinical examination. The NDI is a 10-item questionnaire that identifies the presence of neck pain with daily activities such as personal care, reading, lifting, driving, sleeping, and work and is scored as a percentage disability [26]. The DHI asks the subject to rate self-perceived dizziness or imbalance difficulties during 25 daily activities which may be impacted by vestibular involvement [27]. The Post-Concussion Symptom Scale is a subjective rating of the severity of 22 common post-concussive symptoms scored on a 7 point (Likert) scale with the maximum value being 132. The PCSS is a commonly used self-assessment rating of concussion symptoms that has been shown to be predictive of concussive injuries. The symptoms on the PCSS can be categorized into 4 domains which include physical, cognitive, emotional, and sleep disturbances [28, 29].
Predictor variables included the presence or absence of 1) headache, 2) dizziness, 3) neck pain, 4) cognitive impairments, 5) photophobia, 6) phonophobia, and 7) vision disturbances as a primary symptom. The above variables were determined by subjective history taken at the beginning of the evaluation and were considered positive if individual endorsed presence of the symptom during the acute phase of the concussion. Predictors also included a series of standard clinical examinations for cognitive testing, cervical screening, visual testing, balance testing, and vestibular testing (Table 1). Clinical examination data were calculated as ‘positive’ or ‘negative’. An additional predictor was “time to examination”, which was the difference between the first examination date and the original injury date. Time to examination was dichotomized by median values to improve the understanding of the analyses and to examine the necessity of a sub-classification of time to examination.
The outcome variable for this study was time to clearance. Time to clearance was calculated by subtracting the numbers of days from the date of clearance by the date of examination. The resulting value was in ‘days’. Time to clearance was dichotomized by median values to improve the understanding of the analyses and to create odds ratios for each predictor variable. Those with days above the median were defined as “delayed recovery” whereas those with days below the median were defined as “non-delayed recovery”.
Table 1 Clinical examination tests utilized at initial evaluation Bias
To decrease risk of bias the statistician was different than the clinicians and database stewards of the study.
Study size
For simple univariate multinomial or logistic regression, Homer and Lemeshow have recommended a minimum observation-to-variable ratio of 10:1, but cautioned that a number this low will likely overfit a model [30]. Their preferred observation-to-variable ratio is 20:1 for the multivariate modeling, thus an appropriate number for multivariate modeling would range 120 to 240 if all predictor variables were eligible for the final analysis [30].
Statistical methods
All analyses were performed using SPSS version 20.0 (IBM Corp. Armonk, NY, USA). Subject characteristics, including means, standard deviations, and frequencies were reported for age, gender, disability, dizziness, symptom statuses and time to examination. Descriptive characteristics were reported in raw values. All patient report data were categorized as present or absent.
Univariate logistic regression analyses were performed for each of the predictor variables using the outcome variable of time to clearance. Nuisance variables for history of concussion, LD/ADHD and migraines were included in each regression. Logistic regression analysis was used because the pass rates were not normally distributed, could not be appropriately log-transformed, and failed to meet the assumptions of a linear regression analysis. For each univariate analysis, individual P-values, odds ratios (ORs) and 95% confidence intervals (CIs), were reported.
Associations in the univariate analyses with P-values ≤0.05 were considered in a distinct hierarchical multivariate predictive model which included the full sample. After assessment of collinearity, multivariate analyses (backwards stepwise regression) were used to define the best predictors of delayed recovery. Secondary analyses included multivariate modeling for two subgroups: 1) delayed time to examination and 2) early time to examination, to determine the influence of examination timing on predictive outcomes.
Availability of data and materials
The datasets analyzed during the current study are not publicly available due to institutional restrictions regarding the accessibility of private health information. A limited dataset, with HIPAA identifiers removed, may be available from the corresponding author on reasonable request.