This page was translated automatically from Turkish. Read the original

Bu sayfa, 26. Dünya Psikiyatri Kongresi'nde sunulan çalışmanın yapay zekâ desteğiyle hazırlanmış Türkçe özetidir. Çalışmanın sahibi sunan yazarlardır; atıf için orijinal kaynağa başvurun.

SO094

Development and validation of a model predicting improvement following electroconvulsive therapy: a nationwide study

DEVELOPMENT AND VALIDATION OF A PREDICTION MODEL OF IMPROVEMENT AFTER ELECTROCONVULSIVE THERAPY: A NATION-WIDE STUDY

The model developed using nationwide registry data confirms known predictors of ECT response, but its discrimination is insufficient to support decision-making at the individual level.

The study aims to develop a clinical prediction model based on demographic and clinical information readily available before the decision to administer electroconvulsive therapy (ECT). The Swedish National Quality Register for ECT (Q-ECT) was linked to other national registers, identifying 16,553 individuals (mean age 54 years, 59% women) who completed an acute course of ECT for a major depressive episode between 2011 and 2021. The data were randomly divided into training (50%) and test (50%) sets. Response was defined by dichotomizing clinician-rated CGI Improvement scores, with “much improved” or “very much improved” indicating response; of 29 candidate predictors, those selected in at least 90% of bootstrap LASSO iterations were included in a logistic regression model, and a random forest classification model was also developed as an alternative. The abstract reports that 75.8% of the sample responded to ECT. Discrimination in the test set was limited: the area under the curve (AUC) was 0.67 (95% confidence interval [CI], 0.66 to 0.68) for improvement, 0.69 (0.68 to 0.70) for self-reported remission, and 0.66 (0.65 to 0.68) for the random forest model; Nagelkerke R² values were 9.8% and 13.5%, respectively, while calibration was excellent (intercept −0.0007, slope 0.99). In the final model, adjusted odds ratios (ORs) for improvement were as follows: age 70 versus age 30, 2.18 (1.90 to 2.51); no personality disorder diagnosis in the past year, 1.86 (1.54 to 2.24); current psychotic depression, 1.65 (1.39 to 1.96); a history of psychotic depression, 1.76 (1.51 to 2.05); and fewer than four hospital admissions in the past year, 1.32 (1.11 to 1.56). In internal and external validation, discrimination varied somewhat across centers but remained highly stable across calendar years. In conclusion, the model confirms the importance and universality of known predictors and can inform clinicians and patients about the likelihood of response, but it cannot be used to predict the probability of response at the individual level; more detailed biological or phenotypic data are required to achieve adequate performance.

Tüm WCP 2026 özetleri