Delirium is associated with increased mortality and morbidity, prolonged hospital stays, and a heightened risk of developing dementia, while its complete prevention remains challenging. Therefore, an efficient and time-saving screening tool is essential for the timely detection of delirium, enabling prompt intervention and reducing the associated healthcare burden. The original 13-item Delirium Observation Screening Scale (DOSS) is widely used and well-validated, but its relatively high workload leaves room for simplification.

This study aimed to develop and validate a shortened version of the DOSS (s-DOSS) to reduce screening burden while maintaining diagnostic accuracy.

The s-DOSS was developed in two phases. First, the content validity index (CVI) of the original 13-item DOSS was assessed through ratings from 14 delirium experts. Second, psychometric properties were assessed using 16,110 routine DOSS assessments from hospitalized patients. Item response theory parameters, classical test theory indices, and inter-item correlations were examined to inform item retention or deletion. A refined factor structure of the s-DOSS was subsequently evaluated. Finally, psychometric validation of the s-DOSS was conducted via secondary analysis of a prospective study, comparing its diagnostic performance to the Confusion Assessment Method (CAM) in 113 paired patient assessments.

Four items were removed from the original DOSS. “Attention to conversation” was excluded due to conceptual redundancy, a low collected item-total correlation, and limited information contribution. The items ‘remembers events’, ‘pulls tubes’, and ‘sudden emotional’ were removed because of poor content validity, redundancy or local dependence, and weak internal coherence with the remaining items. The resulting 9-item s-DOSS demonstrated an average scale-level CVI (S-CVI/Ave) of 0.833 and a Cronbach's α of 0.853. Using a cutoff point of two, s-DOSS achieved a sensitivity of 90.9% (95% CI: 62–98), a specificity of 92.2% (95% CI: 85–96), and an overall diagnostic accuracy of 92.0% (95% CI: 86–96). The area under the ROC curve (AUC) was 0.935 (95% CI: 0.822–1.000).

The first validation of the 9-item s-DOSS suggests that it is a reliable and efficient tool for delirium screening, maintaining high diagnostic accuracy while reducing workload. Further adequately powered, multi-site validation studies are required to confirm the robustness of the cut-off and diagnostic performance across clinical settings.