Open AccessS105. VALIDATING THE PREDICTIVE ACCURACY OF THE NAPLS-2 PSYCHOSIS RISK CALCULATOR IN A CLINICAL HIGH-RISK SAMPLE FROM THE SHARP (SHANGHAI AT RISK FOR PSYCHOSIS) PROGRAM
Author(s) -
Tianhong Zhang,
Huijun Li,
Li-Hua Xu,
Yingying Tang,
HuiRu Cui,
Junjie Wang,
Chunbo Li,
Kristen A. Woodberry,
Daniel I. Shapiro,
Margaret A. Niznikiewicz,
Martha E. Shenton,
Matcheri S. Keshavan,
William S. Stone,
Jijun Wang,
Robert W. McCarley,
Larry J. Seidman
Publication year - 2018
Publication title -
schizophrenia bulletin
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 3.823
H-Index - 190
eISSN - 1745-1707
pISSN - 0586-7614
DOI - 10.1093/schbul/sby018.892
Subject(s) - calculator , receiver operating characteristic , psychosis , psychology , risk assessment , prodrome , psychiatry , sample size determination , schizophrenia (object oriented programming) , clinical psychology , medicine , statistics , computer security , mathematics , computer science , operating system
Background The present study aims to validate the predictive accuracy of the NAPLS-2 psychosis risk calculator in a clinical high-risk (CHR) sample from the SHARP (ShangHai At Risk for Psychosis) program in Shanghai, China using comparable inclusion/exclusion criteria and assessments. Methods Three hundred CHR individuals were identified by the Chinese version of the Structured Interview for Prodromal Symptoms. Of these, 228 (76.0%) completed neuro-cognitive assessments at baseline and 199 (66.3%) had at least a one-year follow-up assessment. The latter group was used in risk calculation. Six key predictors (baseline age, unusual thoughts and suspiciousness, symbol coding and verbal learning test performance, functional decline and family history of psychosis) were entered into the NAPLS-2 model to generate a psychosis risk estimate for each case. The area under the receiver operating characteristic curve (AUC) was used to test the effectiveness of this discrimination. Results The NAPLS risk calculator showed moderate discrimination of subsequent transition to psychosis in the SHARP sample with an AUC of 0.631 (p = 0.007). Whether discriminating either transition or poor treatment/clinical outcomes, the AUC of the model increased to 0.754 (p < 0.001). A risk estimate of 30% or higher had moderate sensitivity (53%) and excellent specicity (86%) for prediction of poor treatment/clinical outcome. Discussion The NAPLS-2 risk calculator largely generalizes to a Shanghai CHR sample but is meaningfully improved when predicting an individual’s poor clinical outcome as well as conversion. Our findings provide a critical step in the implementation of CHR risk calculation in China.