Corke Michelle, Mullin Katherine, Angel-Scott Helena, Xia Shelley, Large Matthew
School of Psychiatry, University of New South Wales, Australia.
South Eastern Sydney Local Health District and School of Medicine, University of Notre Dame, Australia.
BJPsych Open. 2021 Jan 7;7(1):e26. doi: 10.1192/bjo.2020.162.
Suicide prediction models have been formulated in a variety of ways and are heterogeneous in the strength of their predictions. Machine learning has been a proposed as a way of improving suicide predictions by incorporating more suicide risk factors.
To determine whether machine learning and the number of suicide risk factors included in suicide prediction models are associated with the strength of the resulting predictions.
Random-effect meta-analysis of exploratory suicide prediction models constructed by combining two or more suicide risk factors or using clinical judgement (Prospero Registration CRD42017059665). Studies were located by searching for papers indexed in PubMed before 15 August 2020 with the term suicid* in the title.
In total, 86 papers reported 102 suicide prediction models and included 20 210 411 people and 106 902 suicides. The pooled odds ratio was 7.7 (95% CI 6.7-8.8) with high between-study heterogeneity (I2 = 99.5). Machine learning was associated with a non-significantly higher odds ratio of 11.6 (95% CI 6.0-22.3) and clinical judgement with a non-significantly lower odds ratio of 4.7 (95% CI 2.1-10.9). Models including a larger number of suicide risk factors had a higher odds ratio when machine-learning studies were included (P = 0.02). Among non-machine-learning studies, suicide prediction models including fewer risk factors performed just as well as those including more risk factors.
Machine learning might have the potential to improve the performance of suicide prediction models by increasing the number of included suicide risk factors but its superiority over other methods is unproven.
自杀预测模型已通过多种方式制定，其预测强度各不相同。机器学习被提议作为一种通过纳入更多自杀风险因素来改善自杀预测的方法。
确定机器学习以及自杀预测模型中包含的自杀风险因素数量是否与所得预测的强度相关。
对通过组合两个或更多自杀风险因素或使用临床判断构建的探索性自杀预测模型进行随机效应荟萃分析（国际前瞻性系统评价注册库登记号CRD42017059665）。通过检索2020年8月15日前在PubMed上索引的标题中带有“suicid*”一词的论文来查找研究。
总共86篇论文报告了102个自杀预测模型，纳入了20210411人以及106902例自杀案例。合并优势比为7.7（95%置信区间6.7 - 8.8），研究间异质性较高（I² = 99.5）。机器学习与优势比略高（11.6，95%置信区间6.0 - 22.3）但无统计学意义相关，临床判断与优势比略低（4.7，95%置信区间2.1 - 10.9）但无统计学意义相关。当纳入机器学习研究时，包含更多自杀风险因素的模型具有更高的优势比（P = 0.02）。在非机器学习研究中，包含较少风险因素的自杀预测模型与包含较多风险因素的模型表现相当。
机器学习可能有潜力通过增加纳入的自杀风险因素数量来提高自杀预测模型的性能，但其相对于其他方法的优越性尚未得到证实。
