The troy sleep scale: clinical validation and establishment of a screening cut-off

Sleep complaints are common and clinically consequential, highlighting the need for brief multidimensional screening tools. The Troy Sleep Scale (TSS) was previously developed in a non-clinical sample; this study aimed to validate the TSS in a clinical sample and estimate a screening threshold based on subjective sleep-complaint status. This case–control psychometric study included 434 adults, comprising 272 participants with subjective sleep complaints and 162 healthy controls. Participants completed the 11-item TSS, which includes three subscales, and established comparator measures of sleep disturbance, daytime sleepiness, and anxiety/depression symptoms. Reliability was assessed using internal consistency, item analyses, and 2–4-week test–retest reliability in a subsample (n = 172). Structural validity was examined with ordinal confirmatory factor analysis based on polychoric correlations. Construct validity was evaluated through convergent, discriminant, known-groups, and covariate-adjusted analyses. ROC/AUC analysis with Youden’s index was used to estimate a sample-derived screening threshold, and item response theory (IRT) assessed item-level functioning. Internal consistency was excellent (α = 0.952; ω = 0.950), with high corrected item–total correlations (0.717–0.833). Test–retest reliability was moderate for the total score (ICC = 0.629) and good for subscales (ICC = 0.728–0.803). Ordinal CFA supported the three-factor structure with excellent fit (SRMR = 0.028; robust RMSEA = 0.032; robust CFI = 0.998; robust TLI = 0.997) and high standardized loadings. TSS scores correlated significantly with comparator sleep measures (r = 0.631–0.838; Holm-corrected p < 0.004), and associations remained significant after controlling for HADS. Known-groups validity was strong, including after covariate adjustment (partial η²=0.236), and discriminant validity was supported (HTMT = 0.762–0.870). ROC analysis showed excellent discrimination (AUC = 0.931); a threshold of ≥ 24 yielded sensitivity = 0.873 and specificity = 0.851. IRT analyses further supported item-level performance. The TSS demonstrated robust reliability, strong construct validity, and promising screening utility in this clinical sample. The provisional sample-derived threshold of ≥ 24 may support screening and triage of sleep complaints, but requires external validation against rigorous clinical reference standards.