The Development of Psychiatric Non-Suicidal Self-Injury Risk Assessment Scale

Two rounds of Delphi consultation were conducted with twenty experts who had a bachelor's degree or higher, senior titles, and more than ten years of experience in clinical psychiatry, nursing, or education. The experts rated the importance of each issue on a Likert scale (1: strongly disagree to 5: strongly agree) and made appropriate change suggestions.

• The authority coefficient (Cr) describes the expert's authority, and a value greater than 0.7 is deemed reliable. The formula is Cr=(Ca + Cs)/2;
• Kendall's coefficient of co-ordination (Kendall's W) and a chi-square test were used to assess expert opinion coordination, with statistical significance at p < 0.05;
• Item screening criteria include a mean item importance assignment of ≥ 3.50 and CV of ≤ 0.25. Expert opinions are combined to determine the scale's items;
• The questionnaire's recovery rate serves as a gauge for the experts' motivation.

The reliability was assessed using Cronbach α, test-retest reliability, inter-observer reliability, and split-half reliability methods. (1)Internal consistency: The higher the Cronbach α coefficient, the greater the reliability of the scale internal consistency.Cronbach α ≥ 0.9 indicates very good reliability, while 0.8 ≤ Cronbach α< 0.9 indicates high reliability. Cronbach α above 0.6 indicates adequate reliability; (2) Test-retest reliability: The closer the retest correlation coefficient tends to 1, the higher the retest reliability; (3)Inter-observer reliability: Inter-rater reliability approaches 1, signifying greater reliability of the scale. Inter-rater reliability should be at least > 0.6, and > 0.75 indicates that the scale possesses relatively high reliability. (4)Split-half reliability: Split-half reliability coefficients greater than 0.70 for the scale as a whole and for each dimension are widely regarded as good.

• CVI: The I-CVI was the number of experts who rated 4-5 divided by the total number of experts; the S-CVI was separated into two parts: S-CVI/Ave and S-CVI/Ua; I-CVI > 0.78 and S-CVI ≥ 0.9 indicate strong content validity；
• Construct validity: The maximum variance approach was used in exploratory factor analysis to extract KMO values, Bartlett's test of sphericity values, component matrix (factor loading values), initial eigenvalues, percentage of variance for each component, and cumulative variance contribution ratio;
• Criterion validity: The Barrett Impulsivity Scale (BIS-11) was used as a validity scale to assess the validity of the scale's correlational validity. The higher the correlation coefficient, the better the validity.

The Receiver Operator Characteristic (ROC) was used to characterize the subject and the maximum value of the Youden Index (YI) was used as a threshold to find the optimal diagnostic threshold.

Based on the optimal diagnostic threshold, true-positive patients were screened and the degree of risk of non-suicidal self-harm in patients with mental disorders was graded using the interquartile method, defining P0-P25 as grade I, P25-P75 as grade II, and P75-P100 as grade III.

Criterion validity. The 30 items in the revised Chinese version of the Barratt Impulsivity Scale are divided into three categories: Motor Impulsivity, Cognitive Impulsiveness and Non-Planning Impulsiveness. Each item is scored on a Likert scale (not, rarely, sometimes, often, always), with the Planning and Cognitive subscales reverse scored and the Motor subscale positive. The total scores range from 30 to 150, with higher scores denoting higher levels of impulsivity and high scores on the three subscales representing hyperactivity, inattention, and lack of planning.