Authors

  1. Caplan, Leslie J. PhD
  2. Ivins, Brian MPS
  3. Poole, John H. PhD
  4. Vanderploeg, Rodney D. PhD
  5. Jaffee, Michael S. MD, MC, USAF
  6. Schwab, Karen PhD

Abstract

Objective: To evaluate alternative models of symptom clusters for the 22-item Neurobehavioral Symptom Inventory.

 

Participants: Three military samples, including 2 nonclinical samples (n = 2420, n = 4244) and 1 sample of individuals with recent head injury (n = 617).

 

Methods: In the first sample, exploratory factor analysis of Neurobehavioral Symptom Inventory responses was performed with tests of significant factors and model fit. In the other 2 samples, confirmatory factor analysis evaluated the fit of 3 models: 2- and 3-factor models based on the initial exploratory factor analysis, and a 9-factor model based on prior research.

 

Main Outcome Measures: The exploratory factor analysis used 2 tests for the number of factors: Parallel Analysis and Minimum Average Partial test. Confirmatory factor analysis models were evaluated using 2 measures of model fit, Root Mean Square Error of Approximation and Comparative Fit Index.

 

Results: Postconcussive symptoms can be described accurately by the 9 factors. However, the model of 3 intercorrelated factors, reflecting cognitive, affective, and somatic/sensory symptoms, fits the data more parsimoniously with little loss in model fit.

 

Conclusion: Although the 9-cluster result from prior research provides a valid description of the relations among items of the inventory, a 3-factor model, consisting of somatic/sensory, affective, and cognitive factors, provides nearly as good a fit to the data, with greater parsimony. We encourage clinicians and researchers to conceptualize the Neurobehavioral Symptom Inventory in terms of 3 coherent clusters of symptoms rather than as 22 individual items.