Abstract
BACKGROUND: Patient-reported outcome (PRO) questionnaires are essential tools for evaluating symptom burden and quality of life in patients with chronic rhinosinusitis with nasal polyps (CRSwNP). Instruments such as NOSE, SNOT-22, CRS-PRO, and NPQ are commonly used; however, the capability to translate scores between these instruments remains largely unexplored, limiting cross-study comparisons and continuity in patient care. OBJECTIVE: To develop and validate predictive models quantifying relationships between widely utilized PRO questionnaires in CRSwNP and to assess their practical implications for clinical management and integration into research. METHODS: In this observational cross-sectional study, 200 patients with CRSwNP completed the NOSE, SNOT-22, CRS-PRO, and NPQ questionnaires. Pairwise predictive models were constructed using linear and Random Forest regression methods. Model performance was evaluated through metrics such as R² and mean squared error (MSE). Model validity was ensured using the Durbin-Watson, Breusch-Pagan, Shapiro-Wilk, and variance inflation factor (VIF) tests. Clinical subgroup analyses based on variables such as asthma and prior nasal surgery were also conducted. RESULTS: Strong correlations among questionnaires were observed (r=0.61-0.87). Linear regression models demonstrated high predictive accuracy, notably for SNOT-22 predicting NPQ (R²=0.76), NPQ predicting CRS-PRO (R²=0.76), and CRS-PRO predicting SNOT-22 (R²=0.74). Random Forest models showed minor performance enhancements (ΔR²≤0.03). Subgroup analyses indicated increased predictive precision in patients with asthma or previous nasal surgery. These predictive models enable clinicians to interpret scores across different instruments confidently, optimizing patient management decisions, particularly in monitoring treatment responses and longitudinal follow-ups. RESULTS: Predictive modeling among PRO questionnaires in CRSwNP is both feasible and clinically impactful. These models facilitate the translation of scores between instruments, thus enhancing clinical decision-making, streamlining patient assessments, and supporting data harmonization in multicentric and longitudinal studies. Future research should pursue external and longitudinal validations to ensure broader applicability and reliability of these predictive tools.