Assessing long-term test-retest reliability of the CPT-IP in schizophrenia

Abstract

Background: The Continuous Performance Test-Identical Pairs version (CPT-IP) is a well-established measure of sustained attention, and its more challenging versions are particularly suited to detect subtle processing deficits in patients with schizophrenia. However, while there are few longitudinal samples for the CPT-IP, no study has addressed stability for more than two month in patients with schizophrenia. Assessing long-term test-retest reliability of the CPT-IP would facilitate the ability of clinicians to draw conclusions from studies involving interventions as long term cognitive or pharmacological treatments. The present study assessed 12 month test-retest reliability for the two most challenging versions of CPT-IP (4-digit and shapes) in a matched sample of clinically stable schizophrenia outpatients and healthy controls.

Methods: Fifty clinically stable schizophrenia outpatients and 50 healthy controls were assessed with the CPT-IP for the 4-digit and shape conditions. From these, 40 patients and 47 controls were reassessed with an average interval of 12.3 months between test sessions. Test-retest reliability was analyzed with Pearson correlations and results were compared with previous data involving healthy controls and short-term studies in patients with schizophrenia.

Results: Especially d' and hit rate discriminated well between patients with schizophrenia and healthy controls for both CPT-IP conditions and at both test sessions. Healthy controls demonstrated sufficient long term test-retest correlations of d', hit rate and reaction time for both the 4-digit and shape conditions. However, in schizophrenia patients, long-term reliability correlations were at best moderate for d' and hit rate only.

Conclusions: The current study provides further evidence that d' and hit rate yield consistent cross-sectional discrimination sensitivity. At best moderate long-term test-retest reliability of d' in schizophrenia outpatients may be not sufficient for practical use of this measure in long term clinical trials.