Data-Driven Healthcare Decision-Making: Applications of Machine Learning in Early Autism Screening

工商管理學系暨商學研究所

撰文者／Anoop Remanan Syamala

The second session of Fall 2025’s Operations Management Seminar I, hosted by Prof. Chia-Wei Guo, occurred on September 25th. The session began with a presentation by Assistant Professor Yu-Hsin Chen from the Department of Industrial Engineering and Engineering Management at National Tsing Hua University, Taiwan. Dr. Chen, educated at Pennsylvania State University and a new faculty member at NTHU, specializes in analytics, optimization, machine learning, and decision-making in health policy. In her presentation, she emphasized her commitment to introducing quantitative methods to healthcare, a field with numerous stakeholders—policymakers, payers, patients, and providers—that is a rich field for impactful operations research.

Dr. Chen’s presentation underscored the pressing need for early screening for autism spectrum disorder (ASD), a topic of paramount importance in the current healthcare landscape. She emphasized that early intervention is a crucial factor in achieving long-term success. The existing screening tools, such as the Modified Checklist for Autism in Toddlers (M-CHAT), fall short in terms of sensitivity and specificity. This discrepancy between universal screening recommendations and actual effectiveness is a critical issue that her work is dedicated to resolving.

Dr. Chen presented a set of studies leveraging extensive health claims data and machine learning algorithms to enhance ASD risk prediction at critical development time points (18, 24, and 30 months). Her earliest research applied predictive modeling to assess individual ASD risk trajectories at multiple time points, highlighting the potential for integrating the models into electronic health record (EHR) systems. A second line of research examined score-based checklists as a feasible primary care screening tool, with particular attention to resource constraints such as diagnostic capacity limits and long waitlists. In this context, ‘operational trade-offs’ refer to the decisions that need to be made when balancing the sensitivity and specificity of the screening tool with the available resources. Collectively, these works emphasized the operational trade-offs between sensitivity, specificity, and system-level resource use.

The seminar also featured Dr. Chen’s 2022 publication in BMJ Health Care Informatics, which demonstrated that random forest models based on split inpatient and outpatient encounter data achieved an AUROC (Area Under the Receiver Operating Characteristic Curve) of 0.834 at 24

months. This value represents the model’s overall performance in distinguishing between ASD and non-ASD cases. The models substantially outperformed existing clinical screening instruments, reaching 96.4% specificity and a 20.5% positive predictive value at a 40% sensitivity. The findings suggest the potential for incorporating such predictive tools into claims databases or EHR systems, thereby enabling continuous ASD risk surveillance and facilitating targeted diagnostic interventions.

Dr. Chen’s research, in addition to its technological advancements, is firmly rooted in the broader context of healthcare operations management. She demonstrated how predictive analytics can revolutionize patient outcomes, health policy, and resource planning. Her study serves as a powerful example of how industrial engineering and operations research-based techniques can be harnessed to tackle complex public health issues, with the potential to revolutionize healthcare operations.

In short, Dr. Chen’s presentation made a compelling case for integrating data-driven predictive modeling into healthcare systems. By linking rigorous machine learning approaches to urgent medical needs, her research not only contributes to the academic discussion in operations management but also provides actionable paths toward enhancing early ASD detection and intervention. The session was a valuable reminder of the necessity of interdisciplinary research in making societal influence and the applied applicability of Dr. Chen’s research to the professional interests of the audience.