No Factorial Validity Support For Revised Baseline And Injured Factor Structures Of Response To PCSS

Journal of Athletic Training.. 2013 Jun;


Piland, S. G., K. K. Byon, T. E. Gould, P. R. Curry, J. D. Miles and M. S. Ferrara.



Context: A recent study published in 2012, provided “revised” factor structures for both non-injured (baseline) and injured (concussed) responses to the 22-item, ImPACT Post Concussion Symptom Scale (PCSS). Employing Exploratory Factor Analysis (EFA), the study suggested that both samples of responses maintained four symptom clusters. However, the symptom clusters and construct definitions not only varied from previously reported factor structures, but also between baseline and concussed responses. The strength of inferences drawn from composite scores depends upon the quality and quantity of available validity evidence. Such evidence to support the proposed use of the PCSS composite scores is minimal. However, 4-factor measurement models have been suggested by a few studies employing EFA methods (e.g., PCA). Thus, more rigorous confirmation is warranted. Objective: To confirm the factorial validity of the 4-factor response structures of both non-concussed and concussed athlete responses to the PCSS found within the ImPACT computerized neurocognitive exam. Design: Retrospective Analysis Settings: University Laboratory. Participants: Non-concussed (N=908) and concussed athletes (N=146) enrolled at a southeastern Division I institution. Interventions: Baseline and concussed responses to the 22-item PCSS instrument. Main Outcome Measures: Two separate Confirmatory Factor Analyses (CFAs) (i.e., one CFA with baseline data set and the other CFA with injured sample) were performed to fit the specified PCSS model, for which the baseline was hypothesized as 4-factor with 20/ 22 items (i.e., Cognitive-Sensory, Affective, Vestibular-Somatic, and Sleep-Arousal) and the injured sample as 4-factor with 17/22 items (i.e., Cognitive-Migraine-Fatigue, Affective, Somatic, and Sleep). Several model fit indexes were employed, including CFI, and RMSEA. Results: The first data set (N = 908) with baseline was subject to a CFA. Goodness of fit indices revealed that the 4-factor model did not fit the data well. Goodness of fit indices for the injured sample (N=146) revealed that the injured responses to the 4-factor PCSS model also did not fit the data well. Conclusion: Responses of both non-concussed and concussed athletes failed to support the posited 4-factor measurement model of the PCSS. Such evidence severely reduces the strength of inferences that can be drawn from the clinical use of the four symptom clusters within the PCSS. Clinicians should be cautioned before using a composite score from the 22-item PCSS or its constituent clusters in prognosis

Links to full article: