Abstract

We have carefully read the study by Lee et al published in your journal, Value in Health, titled “Using Machine Learning to Match Clients and Therapy Providers: Evaluating Clinical Quality and Cost of Care."  This study used a rigorous propensity score matching method to compare the effectiveness of 2 machine learning algorithms in mental health service matching: a “practical algorithm” that solely considers practical factors, such as geographical location and accessibility, and a “value-oriented algorithm” that integrates provider historical clinical outcomes and cost data. The findings revealed that the client group using the value-oriented algorithm, compared with the practical algorithm group, exhibited a similar rate of anxiety symptom improvement but with a 20% reduction in total treatment costs and an average decrease of 2.08 treatment sessions. Both algorithm groups achieved large effect sizes in clinical improvement. This result clearly demonstrates that the value algorithm incorporating historical performance data can significantly improve resource utilization efficiency while maintaining clinical efficacy, providing robust empirical support for addressing the challenges of accessibility and cost-effectiveness in mental health services.