Diagnostic performance of the Philippine Children's Medical Center clinical criteria for pediatric anti-NMDAR encephalitis: a retrospective validation in a national referral center.

Autoimmune encephalitis cases continue to rise globally. Early recognition, prompt diagnosis and adequate treatment are crucial in improving patient outcomes. However, high cost and limited availability of antibody testing remain a challenge for timely definitive diagnosis. This study evaluated the diagnostic performance of the clinical criteria for pediatric anti-NMDAR (N-methyl-D-aspartate receptor) encephalitis using the Philippine Children's Medical Center (PCMC) Algorithm. A retrospective diagnostic accuracy study was conducted among 160 patients aged 0 to < 19 years admitted at PCMC from 2011 to 2025 with a clinical suspicion of autoimmune encephalitis. Subjects underwent cerebrospinal fluid (CSF) anti-NMDAR antibody testing: 119 tested positive and 41 tested negative. Clinical data were obtained and analyzed to evaluate PCMC algorithm's diagnostic performance using sensitivity, specificity, predictive values and likelihood ratios. Receiver operating characteristic (ROC) analysis and logistic regression were utilized to determine optimal symptom combinations. Of the 160 subjects, abnormal behavior or cognitive dysfunction was the most common symptom. Among CSF anti-NMDAR-positive cases, seizures, movement disorder, and speech dysfunction occurred more frequently than in antibody-negative cases. The number of PCMC clinical symptoms demonstrated acceptable discrimination for anti-NMDAR encephalitis, with an AUC of 0.752 (95% CI 0.666-0.831). A threshold of ≥ 3 symptoms showed high sensitivity but low specificity [sensitivity 94.96% (95% CI 89.35-98.13); specificity 36.59% (95% CI 22.12-53.06)], whereas ≥ 4 symptoms improved specificity with lower sensitivity [sensitivity 63.87% (95% CI 54.55-72.47); specificity 73.17% (95% CI 57.06-85.78)]. Multivariable analysis identified abnormal behavior or cognitive dysfunction, speech dysfunction, seizures, and movement disorder as independent predictors. Symptom combinations including movement disorder showed stronger diagnostic performance. Abnormal behavior plus movement disorder achieved sensitivity of 72.27% (95% CI 63.32-80.08) and specificity of 68.29% (95% CI 51.91-81.92), while seizures plus movement disorder achieved sensitivity of 65.55% (95% CI 56.28-74.02) and specificity of 70.73% (95% CI 54.46-83.87). The triad of abnormal behavior, seizures, and movement disorder provided higher specificity [82.93% (95% CI 67.94-92.85)] with moderate sensitivity [56.30% (95% CI 46.91-65.37)].Overall, the study demonstrates that symptom burden and specific clinical patterns provide strong predictive ability for anti-NMDAR encephalitis in pediatric patients. Among all combinations tested, abnormal behavior, seizures, and movement disorder emerged as the most practical triad, offering the best balance between sensitivity and specificity. It demonstrated high positive predictive value with reasonable specificity for early diagnosis in resource-limited settings. Two- and three-symptom combinations derived from the parsimonious predictors demonstrated variable diagnostic performance. Combinations including movement disorder consistently achieved the highest overall accuracy and likelihood ratios, while those without it performed less well. Meanwhile, the four-symptom model, despite its high specificity, lacked sufficient sensitivity and accuracy, limiting its value for routine diagnostic use.