The Problem
Lower back pain (LBP) remains the leading cause of years of life lived with a disability (YLDS) worldwide and activity limitation in the world.1 Despite innovations in therapeutic and diagnostic modalities, the impact of LBP on patients has continued to grow at an alarming rate.2 The financial cost of LBP is estimated to be $50-100 billion annually, with a significant proportion of the cost associated with indirect costs such as work absenteeism.3,4 Thus, the burden of LBP has been established as a significant burden on economic productivity and patient morbidity. This fiscal cost, as well as the burden of disability, make the efficient diagnosis and management of LBP a priority for both providers and healthcare systems worldwide.
Our Solution
To address this gap, Duke University’s Departments of Neurosurgery and Orthopaedic Surgery, along with the Duke Institute for Health Innovation (DIHI), formed a interdisciplinary team to develop a machine learning model ensemble for the triage of new LBP patients using electronic health record (EHR) data. Our goal is to support a more standardized and effective intake process for LBPpatients and ultimately achieve a higher rate of positive clinical outcomes for our Duke patients. This project holds the promise of significantly improving the lives of patients with LBP and the efficiency of our healthcare system overall.
Our study cohort included data from 137,915 patients. The average age was 56.47 (SD: 27.21). 40% of the patients were male (n = 55,340), 63% were white (n = 87,556), and 91% of patients were not Hispanic (n = 125,074). The cohort was made up of 3,148 patients who received surgery within 90 days of their index encounter. The remaining patients were divided into sub-groups by their index intervention provider type: physical medicine and rehab (“PM&R”: 8,239 patients), physical therapy (“PT”: 4,889 patients), and primary care (“PCP”: 52,886 patients). A total of 3 models were developed for each cohort predicting the use of opioids at 90, 180, and 365 days (3, 6, and 12 months), respectively.
Outcomes
For the cohort who received early surgery, the prevalence of opioid use was 40%, 29%, and 20% at 90, 180, and 365 days, respectively. The model’s performance in predicting opioid use showed an AUPRC of 0.66 and an AUROC of 0.62 at 90 days. At 180 days, the AUPRC was 0.47, and the AUROC was 0.67. By 365 days, the AUPRC had decreased to 0.35, while the AUROC improved to 0.68. Similar performance patterns were seen in the non-surgery groups. The results can be seen in Table 1 on page 18 of Impact Volume 25.
Next Steps
We are expanding the model prediction ouctomes to include pain reduction and emergent healthcare utilization, which will provide additional data points to support improved patient scheduling decision-making. We plan to pilot the LBP triage solution in the Duke University Health System (DUHS) Spine Center and select primary care sites to evaluate its performance. The clinical workflow utilizes a patient prediction calculation application while a patient is routed at the spine center. At the conclusion of the pilot period, we will evaluate the solution’s impact on the surgical consult-to-surgery rate and the clinical outcomes of prolonged opioid use, reduction in pain, and emergent hospitalizations.
References
- GBD 2015 Disease and Injury Incidence and Prevalence Collaborators. Global, regional, and national incidence, prevalence, and years lived with disability for 310 diseases and injuries, 1990-2015: a systematic analysis for the Global Burden of Disease Study 2015. Lancet Lond Engl. 2016;388(10053):1545-1602. doi:10.1016/S0140-6736(16)31678-6
- Vos T, Allen C, Arora M, et al. Global, regional, and national incidence, prevalence, and years lived with disability for 310 diseases and injuries, 1990–2015: a systematic analysis for the Global Burden of Disease Study 2015. The Lancet. 2016;388(10053):1545-1602. doi:10.1016/S0140-6736(16)31678-6
- Hoy D, March L, Brooks P, et al. The global burden of low back pain: estimates from the Global Burden of Disease 2010 study. Ann Rheum Dis. 2014;73(6):968-974. doi:10.1136/annrheumdis-2013-204428
- Fatoye F, Gebrye T, Mbada CE, Useh U. Clinical and economic burden of low back pain in low- and middle-income countries: a systematic review. BMJ Open. 2023;13(4):e064119. doi:10.1136/bmjopen-2022-064119
