Session Information
Date: Sunday, October 7, 2018
Session Title: Huntington's Disease
Session Time: 1:45pm-3:15pm
Location: Hall 3FG
Objective: To develop a model for prediction of suicidal ideation or suicidal behaviour in Huntington disease gene expansion carriers (HDGEC) based on Enroll-HD data using machine learning approach (MLA).
Background: Suicidal ideation and suicidal behaviour are frequently reported, severe features in HDGEC. So far, no suicidality prediction models have been developed using MLA.
Methods: We used the third Enroll-HD study periodic dataset (PDS3). The Columbia-Suicide Severity Rating Scale (C–SSRS) was used for the assessment of suicidal ideation/behaviour. HDGECs with either having no suicidal ideation or with presence of ‘passive’ suicidal ideations [state 1] at the 1st visit, who at the annual follow-up visit (FUP) either stayed in state 1 or worsened to ‘active’ suicidal ideations and/or suicidal behaviour [state 2] were included into our analysis. The PBAs scale was used to assess behavioural symptoms. Prediction algorithm was based on Boosted Trees (implementation from XGBoost Library for Python) and contained 48 variables from the PDS3. For further analysis we also used Fisher Exact test, Mann–Whitney U-test, and Holm method.
Results: Out of 8,714 subjects from the PDS3 only 377 HDGEC (114 pre-manifest; 161 males; median age 50 [20;78]; median nCAG=43 [38;65]) had state 1 at the 1st visit and either state 1 or state 2 at the FUP. At FUP, 61 HDGEC worsened to state 2 and 316 remained in state 1. Sixty four percent of the HDGECs who remained in state 1 at FUP were accurately classified (probability as having state 2 < 30%). HDGEC who worsened to state 2 were correctly predicted (probability of being classified as having state 2 > 60%) in 37.7% cases. We then compared HDGEC in state 2 at FUP who were poorly (probability <30%; 31 subjects) and well (probability >60%; 23 subjects) classified and found significant difference in the PBAs total scores for depression, anxiety, aggression, and apathy. Well classified HDGEC had more severe scores. Further regression analysis failed to show significant linear relationship of those features with probability of being classified by the algorithm as subject in state 2 at FUP.
Conclusions: Our model for suicidality prediction in HDGEC showed relatively moderate accuracy. Further research is needed to understand the risk for development of suicidal ideation/behaviour in HDGECs without/less severe behavioural symptoms.
To cite this abstract in AMA style:
Y. Seliverstov, A. Borzov, E. van Duijn, B. Landwehrmeyer, M. Belyaev. Prediction of suicidality in Huntington disease: Analysis of Enroll-HD data using machine learning approach [abstract]. Mov Disord. 2018; 33 (suppl 2). https://www.mdsabstracts.org/abstract/prediction-of-suicidality-in-huntington-disease-analysis-of-enroll-hd-data-using-machine-learning-approach/. Accessed October 6, 2024.« Back to 2018 International Congress
MDS Abstracts - https://www.mdsabstracts.org/abstract/prediction-of-suicidality-in-huntington-disease-analysis-of-enroll-hd-data-using-machine-learning-approach/