Jin Wenyi, Liu Zilin, Zhang Yubiao, Che Zhifei, Gao Mingyong
Department of Orthopedics, Renmin Hospital of Wuhan University, Wuhan, China.
Department of Urology, The First Affiliated Hospital of Hainan Medical University, Haikou, China.
Front Med (Lausanne). 2021 Aug 26;8:697649. doi: 10.3389/fmed.2021.697649. eCollection 2021.
Few longitudinal studies have systematically investigated whether or how individual musculoskeletal conditions (IMCs) convey risks for negative psychological health outcomes, and approaches to assess such risk in the older population are lacking. In this Irish nationally representative longitudinal prospective study of 6,715 individuals aged 50 and above, machine learning algorithms and various models, including mediation models, were employed to elaborate the underlying mechanisms of IMCs leading to depression and to develop an IMC-induced negative psychological risk (IMCPR) classification approach. Resultantly, arthritis [odds ratio (95% confidence interval): 2.233 (1.700-2.927)], osteoporosis [1.681 (1.133-2.421)], and musculoskeletal chronic pain [MCP, 2.404 (1.838-3.151)] were found to increase the risk of depression after 2 years, while fracture and joint replacement did not. Interestingly, mediation models further demonstrated that arthritis did not increase the risk of depression; such risk was augmented only when arthritis-induced restrictions of activities (ARA) existed [proportion of mediation: 316.3% (ARA of usual), 213.3% (ARA of social and leisure), and 251.3% (ARA of sleep)]. The random forest algorithm attested that osteoarthritis, not rheumatoid arthritis, contributed the most to depressive symptoms. Moreover, bone mineral density was negatively associated with depressive symptoms. Systemic pain contributed the most to the increased risk of depression, followed by back, knee, hip, and foot pain (mean Gini-Index: 3.778, 2.442, 1.980, 1.438, and 0.879, respectively). Based on the aforementioned findings, the IMCPR classification approach was developed using an interpretable machine learning model, which stratifies participants into three grades. Among the IMCPR grades, patients with a grade of "severe" had higher odds of depression than those with a "mild" [odds ratio (95% confidence interval): 4.055 (2.907-5.498)] or "moderate" [3.584 (2.101-5.883)] grade. Females with a "severe" grade had higher odds of depression by 334.0% relative to those with a "mild" grade, while males had a relative risk of 258.4%. In conclusion, the present data provide systematic insights into the IMC-induced depression risk and updated the related clinical knowledge. Furthermore, the IMCPR classification approach could be used as an effective tool to evaluate this risk.
很少有纵向研究系统地调查个体肌肉骨骼疾病（IMC）是否以及如何带来心理健康负面结果的风险，并且缺乏在老年人群中评估此类风险的方法。在这项针对6715名50岁及以上个体的具有爱尔兰全国代表性的纵向前瞻性研究中，采用机器学习算法和各种模型，包括中介模型，来阐述IMC导致抑郁的潜在机制，并开发一种IMC诱发的负面心理风险（IMCPR）分类方法。结果发现，关节炎[优势比（95%置信区间）：2.233（1.700 - 2.927）]、骨质疏松症[1.681（1.133 - 2.421）]和肌肉骨骼慢性疼痛[MCP，2.404（1.838 - 3.151）]在2年后会增加抑郁风险，而骨折和关节置换则不会。有趣的是，中介模型进一步表明，关节炎本身不会增加抑郁风险；只有当存在关节炎诱发的活动受限（ARA）时，这种风险才会增加[中介比例：316.3%（日常活动ARA）、213.3%（社交和休闲活动ARA）和251.3%（睡眠活动ARA）]。随机森林算法证明，对抑郁症状贡献最大的是骨关节炎，而非类风湿关节炎。此外，骨密度与抑郁症状呈负相关。全身性疼痛对抑郁风险增加的贡献最大，其次是背部、膝盖、臀部和足部疼痛（平均基尼指数分别为：3.778、2.442、1.980、1.438和0.879）。基于上述发现，使用可解释的机器学习模型开发了IMCPR分类方法，该方法将参与者分为三个等级。在IMCPR等级中，“重度”等级的患者患抑郁症的几率高于“轻度”[优势比（95%置信区间）：4.055（2.907 - 5.498）]或“中度”[3.584（2.101 - 5.883）]等级的患者。“重度”等级的女性患抑郁症的几率相对于“轻度”等级的女性高出334.0%，而男性的相对风险为258.4%。总之，本研究数据为IMC诱发的抑郁风险提供了系统的见解，并更新了相关临床知识。此外，IMCPR分类方法可作为评估这种风险的有效工具。
