Replicating and Extending the Reliability, Criterion Validity, and Treatment Sensitivity of the PANSS10 and PANSS20 for Pediatric Trials

Abstract

Objective

Pediatric studies of schizophrenia have relied on the 30-item Positive and Negative Syndrome Scale (PANSS30) as a primary outcome measure. There have been many efforts to create shorter versions of it to reduce costs and burden. The present aim is to conduct a confirmatory investigation of the reliability and validity of 10- and 20-item abbreviated versions developed in a United States–based National Institute of Mental Health (NIMH) pediatric sample that reflects the 5-factor structure underlying the PANSS, adding more detailed examination of patient-level score reproducibility and extending the examination to a large, placebo-controlled, international pediatric trial.

Method

We applied the same psychometric and treatment sensitivity analyses as in Findling et al. (2023) to an adolescent schizophrenia paliperidone randomized placebo-controlled trial (RCT), accessed via the Yale Open Data Access (YODA) secure data environment (described in Singh et al., 2011). Analyses included confirmatory factor analyses, graded response models, ω reliability coefficients, tests of convergent criterion validity, sensitivity to change, and Bland–Altman plots to evaluate score reproducibility.

Results

Using the paliperidone RCT dataset, with N = 201 participants between the ages of 12 and 17 years (mean age = 15.40, SD = 1.53 years; 59% male), the PANSS 10- or 20-item vs 30-item versions had similar average interitem correlations (0.11-0.15); ω^Total reliabilities of 0.78 to 0.89 with reliability >0.80 across patient presentations from mild residual symptoms to severe pathology; correlations of 0.92 and 0.98 with the 30-item total; partial eta-squared (η²) values for time, treatment, and time by treatment; and also correlations with Clinical Global Impression (CGI) severity and Children’s Global Assessment Scale (CGAS) ratings. Per-item scores differed by 0.04 points on average on the PANSS10 and by 0.01 points for the PANSS20 vs the PANSS30, all not significant.

Conclusion

Results replicated reliability and validity findings for the PANSS10 and PANSS20 short forms in an international pediatric randomized placebo-controlled trial. Findings extend prior work by being the first to apply modern reliability models (ω) for multi-factor composites, also using Bland–Altman methods to evaluate patient-level score reproducibility. Scores based on the PANSS10 or PANSS20 reproduce traditional scores with high fidelity and low bias, offering substantial savings in terms of time, cost, and burden, especially when used for tracking progress or outcomes.

Plain language summary

A secondary analysis of data from a global clinical trial involving 201 young patients with schizophrenia found that shorter, 10 and 20-item versions of the standard 30-item schizophrenia assessment tool (PANSS) are equally reliable and valid. The abbreviated versions correlated >.9 with the full-scale scores, with no significant bias when comparing scores. This finding suggests that mental health assessments for young patients can be significantly shortened without losing accuracy, greatly reducing the time and effort required from both families and clinicians. This advancement in psychiatric evaluation could make mental health care less burdensome for families dealing with schizophrenia.

Clinical guidance

• Consider using the 10-item or 20-item PANSS for regular monitoring and assessment of pediatric schizophrenia patients. These shorter versions maintain high reliability and can streamline the evaluation process, making it more efficient.

• Patient and family engagement: Use the abbreviated PANSS versions to engage more effectively with patients and their families. The reduced length of the assessment can alleviate the stress and fatigue associated with longer evaluations, potentially improving patient cooperation and the quality of the data collected.