Predictive Analytics: A Potentially Key Tool in Mitigating Opioid Misuse in Orthopaedic Practice

RESEARCH ANALYSIS

Predictive Analytics: A Potentially Key Tool in Mitigating Opioid Misuse in Orthopaedic Practice

SPENCER RASMUSSEN, BS

Drexel University College of Medicine

SUMMARY POINTS

-       The opioid crisis necessitates the integration of artificial intelligence (AI) and advanced analytics into traditional medical practices to predict, prevent, and treat opioid misuse effectively.

-       Current research focuses on developing machine learning models to predict opioid misuse and addiction, leveraging big data analytics (BDA) and electronic health records (EHR) to enhance patient care.

-       While current research has developed numerous predictive models for opioid misuse, their effectiveness varies, highlighting the need for further refinement and validation to ensure clinical utility.

ANALYSIS

Background

The opioid epidemic remains a critical public health crisis, presenting unique challenges for contemporary physicians. Unlike their predecessors, today's medical professionals will most likely need to integrate artificial intelligence (AI) and advanced analytics into their practice, if they have not already done so. This fusion of technology and traditional medicine is essential for effectively combating a crisis that necessitates all available tools to predict, prevent, and treat opioid misuse.

Since the advent of AI, numerous industries have experienced a rapid proliferation of innovative strategies and models leveraging its vast applications and capabilities. Healthcare is no exception. In a field where precision and accuracy are paramount, AI has swiftly demonstrated its valuable utility (1). Two key components of AI frequently employed in healthcare are big data analytics (BDA) and machine learning (ML). Massive datasets containing patient information are continuously logged into EHRs and utilized to guide patient care. Through BDA and ML, rapid analysis and data-driven decision-making can enhance patient outcomes and overall healthcare performance (2,3). BDA and ML have been cited in various studies for their roles in improving healthcare delivery and services, assisting in patient diagnosis, guiding therapy, predicting disease progression, and detecting early signs of deterioration, among other applications (4-6).

In the context of the opioid crisis, current research aims to develop ML models that mitigate opioid risk by predicting misuse and addiction outcomes. Several specialties and advanced practitioners prescribe opioids at varying rates, with orthopaedics ranking third in average number of opioid prescriptions per prescriber (7). Due to the diverse population of patients that use opioids, numerous predictive models have been formulated, often focusing on specialty-specific patient populations. This analysis aims to delve into the current literature on predictive analytics and examine its potential in identifying orthopaedic patients who are at higher risk for opioid misuse or adverse outcomes. This study also aims to inspire practitioners to recognize the profound benefits of AI and to actively participate in advancing its development.

Findings

In a systematic review published by the Journal of Addiction Medicine, a comprehensive analysis of 41 studies that developed nearly 160 predictive models was conducted (8). This focused review offers a comparative evaluation and commentary on the predictive models' effectiveness in forecasting outcomes such as opioid overdose, opioid use disorder (OUD), and persistent opioid use. Common predictors across these studies included age, sex, mental health diagnosis history, and substance use disorder history. The review categorizes the proposed predictors into six distinct categories: demographics, mental health comorbidities, substance use disorders, physical health comorbidities, characteristics of prescribed opioids, and non-opioid medications prescribed. The studies utilized a range of statistical modeling techniques, such as regression modeling, alongside ML approaches like random forest, neural network, and gradient boosting algorithms. Model performance was predominantly reported via the c-statistic, with performance values ranging from 0.507 to 0.959 (8). The review highlights a significant variance in model efficacy and underscores the high risk of bias and low applicability present across the studies. This systematic review not only underscores the diverse methodologies employed to create predictive models for opioid risk but also reflects the extensive research efforts dedicated to addressing this critical issue (8).

Further research has specifically targeted the field of orthopaedic surgery, where several studies have sought to develop predictive models tailored to specific surgical procedures. These procedure groups include ACL reconstruction, discectomy, cervical spine fusion, total hip arthroplasty, and arthroscopic hip surgery (9-13). The reported model performance values ranged from 0.7 to 0.81, indicating strong internal validity within the respective study populations. Additionally, inter-study variability was apparent when reporting predictors of adverse opioid outcomes. For example, the model used for ACL reconstruction found that pharmacy ordering site was the strongest predictor of opioid addiction (9). In contrast, the model used for lumbar discectomy found that low hemoglobin and high white blood cell counts had high predictive value for adverse opioid outcomes (10). This variability reflects the complexity and diversity of factors influencing opioid misuse in orthopaedic patients.

Discussion

Due to the absence of external validation, the previously discussed predictive analytics models currently lack substantial clinical utility for orthopaedists. The variability in predictors utilized in each model underscores the inherent complexity of opioid misuse, presenting a significant challenge for researchers aiming to develop a viable tool. The complex nature of opioid misuse requires multifaceted modeling approaches, complicating model development. Future advancements in predictive models are anticipated to facilitate the initiation of rigorous testing and validation processes by healthcare providers, potentially leading to their integration into clinical decision-making frameworks. The refinement of predictive analytics models for opioid misuse holds significant promise in aiding healthcare providers to make more precise and individualized clinical decisions regarding opioid prescriptions, thereby ensuring that each patient’s specific needs are appropriately addressed. The continuous evolution and validation of these models are essential for enhancing their clinical applicability and effectiveness in guiding treatment strategies.

REFERENCES

1.              Poalelungi DG, Musat CL, Fulga A, et al. Advancing Patient Care: How Artificial Intelligence Is Transforming Healthcare. J Pers Med. 2023;13(8):1214. doi:10.3390/jpm13081214

2.              Del Giorgio Solfa F, Simonato FR. Big Data Analytics in Healthcare: Exploring the Role of Machine Learning in Predicting Patient Outcomes and Improving Healthcare Delivery. Int J CIM. 2023;3(1):1-9. doi:10.54489/ijcim.v3i1.235

3.              Javaid M, Haleem A, Pratap Singh R, Suman R, Rab S. Significance of machine learning in healthcare: Features, pillars and applications. International Journal of Intelligent Networks. 2022;3:58-73. doi:10.1016/j.ijin.2022.05.002

4.              Pollettini JT, Panico SRG, Daneluzzi JC, Tinós R, Baranauskas JA, Macedo AA. Using Machine Learning Classifiers to Assist Healthcare-Related Decisions: Classification of Electronic Patient Records. J Med Syst. 2012;36(6):3861-3874. doi:10.1007/s10916-012-9859-6

5.              Abdali-Mohammadi F, N. Meqdad M, Kadry S. Development of an IoT-based and cloud-based disease prediction and diagnosis system for healthcare using machine learning algorithms. IJ-AI. 2020;9(4):766. doi:10.11591/ijai.v9.i4.pp766-771

6.              Berger JS, Haskell L, Ting W, et al. Evaluation of machine learning methodology for the prediction of healthcare resource utilization and healthcare costs in patients with critical limb ischemia—is preventive and personalized approach on the horizon? EPMA Journal. 2020;11(1):53-64. doi:10.1007/s13167-019-00196-9

7.              Guy GP, Zhang K. Opioid Prescribing by Specialty and Volume in the U.S. Am J Prev Med. 2018;55(5):e153-e155. doi:10.1016/j.amepre.2018.06.008

8.              Song SL, Dandapani HG, Estrada RS, Jones NW, Samuels EA, Ranney ML. Predictive Models to Assess Risk of Persistent Opioid Use, Opioid Use Disorder, and Overdose. Journal of Addiction Medicine. 2024;18(3):218. doi:10.1097/ADM.0000000000001276

9.              Anderson AB, Grazal CF, Balazs GC, Potter BK, Dickens JF, Forsberg JA. Can Predictive Modeling Tools Identify Patients at High Risk of Prolonged Opioid Use After ACL Reconstruction? Clinical Orthopaedics and Related Research®. 2020;478(7):00. doi:10.1097/CORR.0000000000001251

10.           Karhade AV, Ogink PT, Thio QCBS, et al. Development of machine learning algorithms for prediction of prolonged opioid prescription after surgery for lumbar disc herniation. The Spine Journal. 2019;19(11):1764-1771. doi:10.1016/j.spinee.2019.06.002

11.           Karhade AV, Ogink PT, Thio QCBS, et al. Machine learning for prediction of sustained opioid prescription after anterior cervical discectomy and fusion. The Spine Journal. 2019;19(6):976-983. doi:10.1016/j.spinee.2019.01.009

12.           Karhade AV, Schwab JH, Bedair HS. Development of Machine Learning Algorithms for Prediction of Sustained Postoperative Opioid Prescriptions After Total Hip Arthroplasty. The Journal of Arthroplasty. 2019;34(10):2272-2277.e1. doi:10.1016/j.arth.2019.06.013

13.           Grazal CF, Anderson AB, Booth GJ, Geiger PG, Forsberg JA, Balazs GC. A Machine-Learning Algorithm to Predict the Likelihood of Prolonged Opioid Use Following Arthroscopic Hip Surgery. Arthroscopy: The Journal of Arthroscopic & Related Surgery. 2022;38(3):839-847.e2. doi:10.1016/j.arthro.2021.08.009