University of Trier, Germany.
University of Trier, Germany.
Behav Res Ther. 2019 Sep;120:103438. doi: 10.1016/j.brat.2019.103438. Epub 2019 Jul 8.
In this study, a computer-based feedback, decision and clinical problem-solving system for clinical practice will be described - the Trier Treatment Navigator (TTN). The paper deals with the underlying research concepts related to personalized pre-treatment recommendations for drop-out risk and optimal treatment strategy selection as well as personalized adaptive recommendations during treatment. The development sample consisted of 1234 patients treated with cognitive behavioral therapy (CBT). Modern statistical machine learning techniques were used to develop personalized recommendations. Drop-out analyses resulted in seven significant predictors explaining 12.0% of variance. The prediction of optimal treatment strategies resulted in differential prediction models substantially improving effect sizes and reliable improvement rates. The dynamic failure boundary reliably identified patients with a higher risk for no improvement or deterioration and indicated the usage of clinical problem-solving tools in risk areas. The probability to be reliably improved for patients identified as at risk for treatment failure was about half of the probability for other patients (35% vs. 62.15%; χ = 82.77, p < .001). Results related to the computer-based feedback system are discussed with regard to the implication for clinical applications as well as clinical training and future research possibilities.
本研究描述了一种基于计算机的反馈、决策和临床问题解决系统，即特里尔治疗导航仪（TTN）。本文涉及与个性化预处理建议相关的基础研究概念，这些建议涉及脱落风险和最佳治疗策略选择，以及治疗过程中的个性化适应性建议。开发样本包括 1234 名接受认知行为治疗（CBT）的患者。现代统计机器学习技术用于制定个性化建议。脱落分析得出了七个显著的预测因子，解释了 12.0%的方差。最佳治疗策略的预测产生了差异化的预测模型，显著提高了效果大小和可靠的改善率。动态失效边界可靠地识别出改善或恶化风险较高的患者，并提示在风险区域使用临床问题解决工具。被识别为治疗失败风险较高的患者的可靠改善概率约为其他患者的一半（35%对 62.15%； χ = 82.77，p <.001）。与基于计算机的反馈系统相关的结果，讨论了其对临床应用以及临床培训和未来研究可能性的影响。
