Üsküdar University

Idiopathic normal pressure hydrocephalus (iNPH) is a major cause of reversible gait and cognitive impairment in older adults, yet postoperative outcomes remain difficult to predict. Existing prognostic tools are limited by methodological shortcomings, including lack of temporal validation, dependence on subjective radiological markers, insufficient events-per-variable ratios, and absence of transparent graded risk stratification. To address these gaps, we developed the first temporally validated, odds ratio-weighted prognostic scoring system specifically designed for iNPH surgical outcomes.

This single-center retrospective study included 114 consecutive patients treated with ventriculoperitoneal shunting between 2015 and 2024. Candidate predictors were systematically assessed, and the Normal Pressure Hydrocephalus Risk-Prediction Score (NPH-RKP) was constructed using five objective clinical variables routinely available in standard practice: age, iNPH Grading Scale score, Mini-Mental State Examination (MMSE), Fazekas score, and diagnostic delay. Four competing scoring frameworks were compared, and the odds ratio-weighted graded model was selected based on performance and interpretability. The dataset was temporally divided into development (2015-2020, n = 59) and validation (2021-2024, n = 55) cohorts. The primary endpoint was global clinical improvement at 12 months, incorporating gait, cognitive, and urinary metrics.

Multivariable modeling identified five independent predictors (OR range: 2.00-4.14, 95% CI: 1.82-9.42). The resulting 0-12-point NPH-RKP score demonstrated excellent discrimination in the development cohort (AUC = 0.939, 95% CI: 0.881-0.985) with strong calibration (Hosmer-Lemeshow p = 0.511; Brier score = 0.113). Temporal validation revealed retained prognostic accuracy (AUC = 0.824, 95% CI: 0.715-0.933), with expected performance attenuation attributable to evolving clinical practice and improved surgical outcomes. Distinct risk strata yielded sharply separated success rates: low-risk (0-4 points, 91.4%), intermediate-risk (5-7 points, 62.5%), and high-risk (8-12 points, 8.3%). The model achieved a robust EPV of 7.8, outperforming prior iNPH tools while maintaining ease of calculation.

The NPH-RKP provides a transparent, clinically accessible, and methodologically rigorous prognostic framework for predicting iNPH shunt outcomes. Its objective parameters, grading structure, and temporal validation enhance its clinical applicability in routine decision-making and patient counseling. Prospective multicenter validation is warranted to confirm its generalizability across diverse healthcare settings.

This study aimed to evaluate the accuracy, reliability, and comprehensibility of information about surgically-assisted rapid palatal expansion provided by language models based on artificial intelligence (AI).

A cross-sectional content analysis was conducted on the responses to surgically-assisted rapid palatal expansion-related questions by ChatGPT-4 (OpenAI LLC, San Francisco, Calif), Gemini (Alphabet Inc, Mountain View, Calif), and Copilot (Microsoft, Redmond, Wash). In total, 115 questions (categorized into 11 domains) were created by 3 orthodontists and 1 oral and maxillofacial surgeon. The accuracy of the answers generated by the AI language models was independently evaluated by the same experts via a 5-point Likert scale. To test the relationships among categorical variables, when the sample size assumption was met, the Pearson chi-square test was used. However, when the sample size assumption was not met, Fisher's exact test was applied. Analyses were performed in SPSS (version 27; IBM, Armonk, NY).

The responses of the AI types presented a general homogeneous distribution, with no statistically significant difference between the types of AI and the types of responses (P >0.05). Although there were no significant differences, ChatGPT-4 had the highest objectively true rate. In contrast, Gemini produced answers with more balanced accuracy, whereas Copilot had the highest number of false answers.

These findings reveal that the accuracy of AI-supported language models in providing medical information may vary according to subject matter.