現今臨床醫療流程中，脊髓刺激治療 (SCS) 需求患者的識別與轉診過程面臨效率瓶頸與漫長等待時間。這主要源於醫師填寫非結構化醫囑資料處理的延遲。並因二分類機器模型在面對高風險臨床情境時，缺乏不確定性溝通機制，限制了人機協作潛力。為解決這些問題，本研究旨在建立一個人機協作模型，以提升醫療品質、臨床判斷的效率、可靠性與安全性。
本研究採用兩階段策略以實現人機協作模型。第一階段，利用電子健康記錄 (EHR) 中的結構化數據，預測目前由醫矚使用自然語言處理 (NLP)所決定的初步診斷結果，旨在縮短因等待醫囑而產生的時間。第二階段，則基於第一階段的結果，開發一個三分類模型（Class 0：不需 SCS；Class 1：需要 SCS；Class 2：轉介人工判斷），以快速區分病患。若模型判斷結果相對明確（Class 0或 Class 1），則進行後續轉診流程；若判斷為模糊（Class 2），則交由專家進行判斷，從而實現有效的人機協作。研究結果顯示，本模型在多方面展現顯著效益：
(1). 患者角度：顯著減少轉診等待時間 。以選定的 SVM 模型為例，僅 18.9% 的病例需人工審閱，且在 22 個實際 SCS 治療病例中，成功識別 15 個，其中 13 個可直接自動分類；
(2). 約診人員角度：大幅減輕約診人員負擔，提升初步篩選效率。模型能快速提供初步診斷，減少等待醫囑時間，並透過 25 個關鍵特徵提供具循證基礎的輔助資訊；
(3). 專科醫師角度：為專科醫師提供高效、可靠且與現行判斷高度一致的初步評估。SVM 模型對 NLP 的相對敏感性可達 93.8%，且其 0/1 分類precision為 72.3%，確保自動判斷的可靠性。Class 2病例轉介機制確保高複雜案例獲得專注。
本研究成功整合結構化數據與人機協作模型，解決了現有痛點，進而提升了整體醫療品質。未來可透過前瞻性臨床試驗，量化其實際效益並持續優化。

In current clinical medical processes, the identification and referral of patients requiring Spinal Cord Stimulation (SCS) treatment face efficiency bottlenecks and prolonged waiting times. This primarily stems from delays in processing physicians filled-out unstructured medical order data. Furthermore, due to the lack of uncertainty communication mechanisms in binary classification machine learning models when dealing with high-risk clinical scenarios, the potential for human-machine collaboration is limited. To address these issues, this study aims to establish a human-machine collaborative model to enhance healthcare quality, as well as the efficiency, reliability, and safety of clinical judgments.
This study adopts a two-stage strategy to implement the human-machine collaborative model. In the first stage, structured data from Electronic Health Records (EHRs) is utilized to predict preliminary diagnostic results currently determined by Natural Language Processing (NLP) of medical orders, aiming to shorten the time generated by waiting for medical orders. The second stage, based on the results of the first stage, involves developing a three-class model (Class 0: no SCS needed; Class 1: SCS needed; Class 2: refer for human judgment) to quickly categorize patients. If the model's judgment is relatively clear (Class 0 or Class 1), the subsequent referral process proceeds; if the judgment is ambiguous (Class 2), it is handed over to experts for decision, thereby achieving effective human-machine collaboration. Research results indicate that this model demonstrates significant benefits in multiple aspects:
 Patient Perspective: Significantly reduces referral waiting times. Taking the selected SVM model as an example, only 18.9% of cases require manual review, and among 22 Actual SCS treatment cases, 15 were successfully identified, with 13 of those directly classified automatically.
 Appointment Scheduler Perspective: Substantially reduces the burden on appointment schedulers and enhances preliminary screening efficiency. The model can quickly provide preliminary diagnoses, reducing the time spent waiting for medical orders, and offers evidence-based supplementary information via 25 key features.
 Specialist Physician Perspective: Provides specialist physicians with efficient, reliable, and highly consistent preliminary evaluations aligned with current judgments. The SVM model's Relative Sensitivity to NLP reached 93.8%, and its 0/1 classification precision was 72.3%, ensuring the reliability of automated judgments. The Class 2 case referral mechanism ensures high-complexity cases receive focused attention.
This study successfully integrated structured data with a human-machine collaborative model, addressing existing pain points and thereby enhancing overall healthcare quality. Future work can involve prospective clinical trials to quantify its actual benefits and enable continuous optimization.

[1] S. P. Cohen, L. Vase, and W. M. Hooten, "Chronic pain: an update on burden, best practices, and new advances," Lancet, vol. 397, no. 10289, pp. 2082-2097, May 29 2021, doi: 10.1016/S0140-6736(21)00393-7.
[2] W. Hauser, F. Wolfe, P. Henningsen, G. Schmutzer, E. Brahler, and A. Hinz, "Untying chronic pain: prevalence and societal burden of chronic pain stages in the general population - a cross-sectional survey," BMC Public Health, vol. 14, p. 352, Apr 13 2014, doi: 10.1186/1471-2458-14-352.
[3] H. Tsuji, T. Tetsunaga, T. Tetsunaga, H. Misawa, Y. Oda, S. Takao, K. Nishida, and T. Ozaki, "Factors influencing caregiver burden in chronic pain patients: A retrospective study," Medicine (Baltimore), vol. 101, no. 39, p. e30802, Sep 30 2022, doi: 10.1097/MD.0000000000030802.
[4] A. P. A. Dagnino and M. M. Campos, "Chronic Pain in the Elderly: Mechanisms and Perspectives," Front Hum Neurosci, vol. 16, p. 736688, 2022, doi: 10.3389/fnhum.2022.736688.
[5] S. Hassan, A. Amjad, M. U. Farooq, S. Anwar, and M. I. Ammarullah, "Applying lean production system philosophy to reduce patient waiting time in healthcare services: Simulation-based optimization and validations through experiment," (in English), Aip Adv, vol. 14, no. 9, Sep 1 2024, doi: Artn 09502210.1063/5.0210721.
[6] C. Liddy, L. Cooper, G. Bellingham, T. Deyell, P. Ingelmo, I. Moroz, P. Poulin, A. Singer, G. S. Logan, R. Visca, A. Zahrai, and N. Buckley, "Patient-Reported Wait Times and the Impact of Living with Chronic Pain on their Quality of Life: A Waiting Room Survey in Chronic Pain Clinics in Ontario, Manitoba, and Quebec," Can J Pain, vol. 8, no. 1, p. 2345612, 2024, doi: 10.1080/24740527.2024.2345612.
[7] L. Jiang, Y. L. Huang, J. Fan, C. L. Hunt, and J. S. Eldrige, "Development and Implementation of Automated Referral Triaging System for Spinal Cord Stimulation Procedure in Pain Medicine," J Med Syst, vol. 49, no. 1, p. 14, Jan 21 2025, doi: 10.1007/s10916-025-02148-5.
[8] M. Hobensack, J. Song, D. Scharp, K. H. Bowles, and M. Topaz, "Machine learning applied to electronic health record data in home healthcare: A scoping review," Int J Med Inform, vol. 170, p. 104978, Feb 2023, doi: 10.1016/j.ijmedinf.2022.104978.
[9] A. J. Hamilton, A. T. Strauss, D. A. Martinez, J. S. Hinson, S. Levin, G. Lin, and E. Y. Klein, "Machine learning and artificial intelligence: applications in healthcare epidemiology," Antimicrob Steward Healthc Epidemiol, vol. 1, no. 1, p. e28, 2021, doi: 10.1017/ash.2021.192.
[10] D. P. Tobon, M. S. Hossain, G. Muhammad, J. Bilbao, and A. E. Saddik, "Deep learning in multimedia healthcare applications: a review," Multimed Syst, vol. 28, no. 4, pp. 1465-1479, 2022, doi: 10.1007/s00530-022-00948-0.
[11] M. Ghanem, A. K. Ghaith, V. G. El-Hajj, A. Bhandarkar, A. de Giorgio, A. Elmi-Terander, and M. Bydon, "Limitations in Evaluating Machine Learning Models for Imbalanced Binary Outcome Classification in Spine Surgery: A Systematic Review," Brain Sci, vol. 13, no. 12, Dec 16 2023, doi: 10.3390/brainsci13121723.
[12] V. Kumar, G. S. Lalotra, P. Sasikala, D. S. Rajput, R. Kaluri, K. Lakshmanna, M. Shorfuzzaman, A. Alsufyani, and M. Uddin, "Addressing Binary Classification over Class Imbalanced Clinical Datasets Using Computationally Intelligent Techniques," Healthcare (Basel), vol. 10, no. 7, Jul 13 2022, doi: 10.3390/healthcare10071293.
[13] S. Jayatilake and G. U. Ganegoda, "Involvement of Machine Learning Tools in Healthcare Decision Making," J Healthc Eng, vol. 2021, p. 6679512, 2021, doi: 10.1155/2021/6679512.
[14] M. Nilashi, O. bin Ibrahim, H. Ahmadi, and L. Shahmoradi, "An analytical method for diseases prediction using machine learning techniques," (in English), Comput Chem Eng, vol. 106, pp. 212-223, Nov 2 2017, doi: 10.1016/j.compchemeng.2017.06.011.
[15] S. Mohan, C. Thirumalai, and G. Srivastava, "Effective Heart Disease Prediction Using Hybrid Machine Learning Techniques," (in English), Ieee Access, vol. 7, pp. 81542-81554, 2019, doi: 10.1109/Access.2019.2923707.
[16] C.-T. L. W.-L. C. Y.-L. H. S.-F. Liang, "Improving spinal cord stimulation patient triage: Random forest model with custom," presented at the MedInfo 2025, Taipei, 2025.
[17] S. Sheikhalishahi, R. Miotto, J. T. Dudley, A. Lavelli, F. Rinaldi, and V. Osmani, "Natural Language Processing of Clinical Notes on Chronic Diseases: Systematic Review," JMIR Med Inform, vol. 7, no. 2, p. e12239, Apr 27 2019, doi: 10.2196/12239.
[18] A. M. Flores, F. Demsas, N. J. Leeper, and E. G. Ross, "Leveraging Machine Learning and Artificial Intelligence to Improve Peripheral Artery Disease Detection, Treatment, and Outcomes," Circ Res, vol. 128, no. 12, pp. 1833-1850, Jun 11 2021, doi: 10.1161/CIRCRESAHA.121.318224.
[19] J. Adler-Milstein, N. Aggarwal, M. Ahmed, J. Castner, B. J. Evans, A. A. Gonzalez, C. A. James, S. Lin, K. D. Mandl, M. E. Matheny, M. P. Sendak, C. Shachar, and A. Williams, "Meeting the Moment: Addressing Barriers and Facilitating Clinical Adoption of Artificial Intelligence in Medical Diagnosis," NAM Perspect, vol. 2022, 2022, doi: 10.31478/202209c.
[20] I. Habli, T. Lawton, and Z. Porter, "Artificial intelligence in health care: accountability and safety," Bull World Health Organ, vol. 98, no. 4, pp. 251-256, Apr 1 2020, doi: 10.2471/BLT.19.237487.
[21] D. Saraswat, P. Bhattacharya, A. Verma, V. K. Prasad, S. Tanwar, G. Sharma, P. N. Bokoro, and R. Sharma, "Explainable AI for Healthcare 5.0: Opportunities and Challenges," (in English), Ieee Access, vol. 10, pp. 84486-84517, 2022, doi: 10.1109/Access.2022.3197671.
[22] J. Y. Ng, H. Cramer, M. S. Lee, and D. Moher, "Traditional, complementary, and integrative medicine and artificial intelligence: Novel opportunities in healthcare," Integr Med Res, vol. 13, no. 1, p. 101024, Mar 2024, doi: 10.1016/j.imr.2024.101024.
[23] J. Šķilters, J. Pokrotnieks, and A. Derovs, "Towards A Human-AI Hybrid Medicine: Future Medicine—A Hybrid System Where AI Complements Instead of Replaces Humans," in Proceedings of the Latvian Academy of Sciences, 2024, vol. 78, no. 4: De Gruyter Poland, pp. 233-238.
[24] M. P. Pacaux-Lemoine, D. Trentesaux, G. Z. Rey, and P. Millot, "Designing intelligent manufacturing systems through Human-Machine Cooperation principles: A human-centered approach," (in English), Comput Ind Eng, vol. 111, pp. 581-595, Sep 2017, doi: 10.1016/j.cie.2017.05.014.
[25] H. Yuan, L. Kang, Y. Li, and Z. Fan, "Human‐in‐the‐loop machine learning for healthcare: current progress and future opportunities in electronic health records," Medicine Advances, vol. 2, no. 3, pp. 318-322, 2024.
[26] H. Lin, J. T. Han, P. P. Wu, J. Y. Wang, J. Tu, H. Tang, and L. N. Zhu, "Machine learning and human-machine trust in healthcare: A systematic survey," (in English), Caai T Intell Techno, vol. 9, no. 2, pp. 286-302, Apr 2024, doi: 10.1049/cit2.12268.
[27] O. Asan, A. E. Bayrak, and A. Choudhury, "Artificial Intelligence and Human Trust in Healthcare: Focus on Clinicians," J Med Internet Res, vol. 22, no. 6, p. e15154, Jun 19 2020, doi: 10.2196/15154.
[28] G. C. Saha, S. Kumar, A. Kumar, H. Saha, T. Lakshmi, and N. Bhat, "Human-AI collaboration: Exploring interfaces for interactive machine learning," Tuijin Jishu/Journal of Propulsion Technology, vol. 44, no. 2, p. 2023, 2023.
[29] F. Orhan and M. N. Kurutkan, "Predicting total healthcare demand using machine learning: separate and combined analysis of predisposing, enabling, and need factors," BMC Health Serv Res, vol. 25, no. 1, p. 366, Mar 12 2025, doi: 10.1186/s12913-025-12502-5.
[30] M. Hung, E. S. Hon, E. Lauren, J. Xu, G. Judd, and W. Su, "Machine Learning Approach to Predict Risk of 90-Day Hospital Readmissions in Patients With Atrial Fibrillation: Implications for Quality Improvement in Healthcare," Health Serv Res Manag Epidemiol, vol. 7, p. 2333392820961887, Jan-Dec 2020, doi: 10.1177/2333392820961887.
[31] J. K. Kim, "Enhancing Patient Flow in Emergency Departments: A Machine Learning and Simulation-Based Resource Scheduling Approach," (in English), Appl Sci-Basel, vol. 14, no. 10, May 2024, doi: ARTN 426410.3390/app14104264.