Machine learning for prediction of sustained opioid prescription after anterior cervical discectomy and fusion

doi:10.1016/j.spinee.2019.01.009

The Spine Journal

Volume 19, Issue 6, June 2019, Pages 976-983

https://doi.org/10.1016/j.spinee.2019.01.009 Get rights and content

Abstract

BACKGROUND CONTEXT

The severity of the opioid epidemic has increased scrutiny of opioid prescribing practices. Spine surgery is a high-risk episode for sustained postoperative opioid prescription.

PURPOSE

To develop machine learning algorithms for preoperative prediction of sustained opioid prescription after anterior cervical discectomy and fusion (ACDF).

STUDY DESIGN/SETTING

Retrospective, case-control study at two academic medical centers and three community hospitals.

PATIENT SAMPLE

Electronic health records were queried for adult patients undergoing ACDF for degenerative disorders between January 1, 2000 and March 1, 2018.

OUTCOME MEASURES

Sustained postoperative opioid prescription was defined as uninterrupted filing of prescription opioid extending to at least 90–180 days after surgery.

METHODS

Five machine learning models were developed to predict postoperative opioid prescription and assessed for overall performance.

RESULTS

Of 2,737 patients undergoing ACDF, 270 (9.9%) demonstrated sustained opioid prescription. Variables identified for prediction of sustained opioid prescription were male sex, multilevel surgery, myelopathy, tobacco use, insurance status (Medicaid, Medicare), duration of preoperative opioid use, and medications (antidepressants, benzodiazepines, beta-2-agonist, angiotensin-converting enzyme-inhibitors, gabapentin). The stochastic gradient boosting algorithm achieved the best performance with c-statistic=0.81 and good calibration. Global explanations of the model demonstrated that preoperative opioid duration, antidepressant use, tobacco use, and Medicaid insurance were the most important predictors of sustained postoperative opioid prescription.

CONCLUSIONS

One-tenth of patients undergoing ACDF demonstrated sustained opioid prescription following surgery. Machine learning algorithms could be used to preoperatively stratify risk these patients, possibly enabling early intervention to reduce the potential for long-term opioid use in this population.

Introduction

Total US expenditures for prescription opioid abuse exceed $78 billion dollars [1]. In 2016, 11 million US citizens misused opioids whereas drug overdose deaths increased exponentially from 1979 to 2016 [1], [2], [3], [4], [5]. The severity of the opioid crisis has led to increased scrutiny of opioid prescribing practices after surgery. Spine surgery, in particular, has been implicated as a particularly high-risk episode for sustained postoperative opioid use [6], [7].

Previous studies in spine surgery have examined large national administrative and insurance claims databases to identify trends in postoperative opioid use and risk factors for opioid use [8], [9], [10], [11], [12]. Although numerous socio-demographic and clinical characteristics have been identified as prognostic factors for sustained opioid use following spine surgery, no predictive algorithms presently exist for risk stratification of patients before an intervention.

The purpose of this analysis was to develop machine learning algorithms for preoperative prediction of sustained opioid prescription after anterior cervical discectomy and fusion (ACDF). Additional aims of this study were to provide explanations of model predictions at the individual patient level and finally to deploy these algorithms as open access applications for dissemination among the health-care community.

Section snippets

Guidelines

The Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis and JMIR guidelines for Developing and Reporting Machine Learning Models in Biomedical Research were followed for this analysis [13], [14].

Data source

Institutional Review Board Approval was granted for retrospective review of electronic health records from two academic medical centers and three community hospitals. Individual patient consent was waived as the study was restricted to retrospective review of

Results

Of 2,737 patients undergoing ACDF, 270 (9.9%) patients were found to meet our criteria for sustained postoperative opioid prescription. Among the population as a whole, 1,440 (52.6%) patients were women and the median age was 51 (interquartile range=44–59) years (Table 1). Before surgery, 695 (25.4%) patients had myelopathy and 992 (36.2%) had radiculopathy. Overall, 2,028 (74.1%) patients had no opioid prescriptions in the year before surgery, 341 (12.5%) had opioid use for less than 180 days

Discussion

In this study, one-tenth of patients undergoing ACDF demonstrated sustained opioid prescription after surgery. The prospect of sustained opioid prescription following surgery appeared to bedriven by a number of factors, including preoperative opioid prescription, antidepressant use, tobacco use, and Medicaid insurance status. Among the algorithms investigated in this analysis, the stochastic gradient boosting algorithm demonstrated the best discrimination, calibration, and overall performance

Conclusions

In this investigation, one-tenth of patients undergoing ACDF demonstrated sustained prescription opioid prescription following surgery. Machine learning algorithms could be used to stratify patient risk and tailor preoperative management (eg, mental health/social work consultation and/or comanagement with pain medicine service) to reduce the potential for long-term opioid use in this population.

References (40)

N Jain et al.
Chronic preoperative opioid use is a risk factor for increased complications, resource use, and costs after cervical fusion
Spine J
(2018)
AJ Schoenfeld et al.
Outcomes after spine surgery among racial/ethnic minorities: a meta-analysis of the literature
Spine J
(2011)
CY Lu et al.
Charlson and Rx-risk comorbidity indices were predictive of mortality in the Australian health care setting
J Clin Epidemiol
(2011)
AJ Pugely et al.
Opioid use following cervical spine surgery: trends and factors associated with long-term use
Spine J
(2018)
CM Jones
Trends and key correlates of prescription opioid injection misuse in the United States
Addict Behav
(2018)
CS Florence et al.
The economic burden of prescription opioid overdose, abuse, and dependence in the United States, 2013
Med Care
(2016)
ML Barnett et al.
Coupling policymaking with evaluation—the case of the opioid crisis
N Engl J Med
(2017)
E Wood
Strategies for reducing opioid-overdose deaths—lessons from Canada
N Engl J Med
(2018)
RJ Blendon et al.
The public and the opioid-abuse epidemic
N Engl J Med
(2018)
H Jalal et al.
Changing dynamics of the drug overdose epidemic in the United States from 1979 through 2016
Science
(2018)

X Jiang et al.

Chronic opioid usage in surgical patients in a large academic center

Ann Surg

(2017)

CM Brummett et al.

New persistent opioid use after minor and major surgical procedures in US adults

JAMA Surg

(2017)

DB Reid et al.

Effect of Narcotic prescription limiting legislation on opioid utilization following lumbar spine surgery

Spine J

(2018)

AJ Schoenfeld et al.

Sustained preoperative opioid use is a predictor of continued use following spine surgery

J Bone Joint Surg Am Vol

(2018)

AJ Schoenfeld et al.

Risk factors for prolonged opioid use following spine surgery, and the association with surgical intensity, among opioid-naive patients

J Bone Joint Surg Am Vol

(2017)

N Jain et al.

Preoperative chronic opioid therapy: a risk factor for complications, readmission, continued opioid use and increased costs after one- and two-level posterior lumbar fusion

Spine

(2018)

GS Collins et al.

Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD): the TRIPOD statement

BMC Med

(2015)

W Luo et al.

Guidelines for developing and reporting machine learning predictive models in biomedical research: a multidisciplinary view

J Med Internet Res

(2016)

H Clarke et al.

Rates and risk factors for prolonged opioid use after major surgery: population based cohort study

BMJ

(2014)

A Alam et al.

Long-term analgesic use after low-risk surgery: a retrospective cohort study

Arch Intern Med

(2012)

Cited by (94)

Bibliometric analysis of machine learning trends and hotspots in arthroplasty literature over 31 years
2024, Journal of Orthopaedics
Artificial intelligence has demonstrated utility in orthopedic research. Algorithmic models derived from machine learning have demonstrated adaptive learning with predictive application towards outcomes, leading to increased traction in the literature. This study aims to identify machine learning arthroplasty research trends and anticipate emerging key terms.
Published literature focused on machine learning in arthroplasty from 1992 to 2023 was selected through the Web of Science Core Collection of Clarivate Analytics. Following that, bibliometric indicators were attained and brought in to perform an additional examination using Bibliometrix and VOSviewer to identify historical and present patterns within the literature.
A total of 235 documents were obtained through bibliometric sourcing based on machine learning applications within the arthroplasty literature. Thirty-four countries published articles on the topic, and the United States was demonstrated to be the largest global contributor. Four hundred-five institutions internationally contributed articles, with Harvard Medical School and the University of California system as the most relevant institutes, with 75 and 44 articles produced, respectively. Kwon YM was the most productive author, while Haeberle HS and Ramkumar PN were the most impactful based on h-index. The Thematic map and Co-occurrence visualization helped identify both major and niche themes present in the scientific databases.
Machine learning in arthroplasty research continues to gain traction with a growing annual production rate and contributions from international authors and institutions. Institutions and authors based in the United States are the leading contributors to machine learning applications within arthroplasty research. This research discerns trends that have occurred, are presently ongoing, and are emerging within this field, aiming to inform future hotspot development.
An Integrated Voice Recognition and Natural Language Processing Platform to Automatically Extract Thoracolumbar Injury Classification Score Features From Radiology Reports
2024, World Neurosurgery
Many predictive models for estimating clinical outcomes after spine surgery have been reported in the literature. However, implementation of predictive scores in practice is limited by the time-intensive nature of manually abstracting relevant predictors. In this study, we designed natural language processing (NLP) algorithms to automate data abstraction for the thoracolumbar injury classification score (TLICS).
We retrieved the radiology reports of all Mayo Clinic patients with an International Classification of Diseases, 9th or 10th revision, code corresponding to a fracture of the thoracolumbar spine between January 2005 and October 2020. Annotated data were used to train an N-gram NLP model using machine learning methods, including random forest, stepwise linear discriminant analysis, k-nearest neighbors, and penalized logistic regression models.
A total of 1085 spine radiology reports were included in our analysis. Our dataset included 483 compression, 401 burst, 103 translational/rotational, and 98 distraction fractures. A total of 103 reports had documented an injury of the posterior ligamentous complex. The overall accuracy of the random forest model for fracture morphology feature detection was 76.96% versus 65.90% in the stepwise linear discriminant analysis, 50.69% in the k-nearest neighbors, and 62.67% in the penalized logistic regression. The overall accuracy to detect posterior ligamentous complex integrity was highest in the random forest model at 83.41%. Our random forest model was implemented in the backend of a web application in which users can dictate reports and have TLICS features automatically extracted.
We have developed a machine learning NLP model for extracting TLICS features from radiology reports, which we deployed in a web application that can be integrated into clinical practice.
Identifying Modifiable Cost Drivers of Outpatient Unicompartmental Knee Arthroplasty With Machine Learning
2023, Journal of Arthroplasty
Implementing tools that identify cost-saving opportunities for ambulatory orthopaedic surgeries can improve access to value-based care. We developed and internally validated a machine learning (ML) algorithm to predict cost drivers of total charges after ambulatory unicompartmental knee arthroplasty (UKA).
We queried the New York State Ambulatory Surgery and Services database to identify patients who underwent ambulatory, defined as <24 hours of care before discharge, elective UKA between 2014 and 2016. A total of 1,311 patients were included. The median costs after ambulatory UKA were $14,710. Patient demographics and intraoperative parameters were entered into 4 candidate ML algorithms. The most predictive model was selected following internal validation of candidate models, with conventional linear regression as a benchmark. Global variable importance and partial dependence curves were constructed to determine the impact of each input parameter on total charges.
The gradient-boosted ensemble model outperformed all candidate algorithms and conventional linear regression. The major differential cost drivers of UKA identified (in decreasing order of magnitude) were increased operating room time, length of stay, use of regional and adjunctive periarticular analgesia, utilization of computer-assisted navigation, and routinely sending resected tissue to pathology.
We developed and internally validated a supervised ML algorithm that identified operating room time, length of stay, use of computer-assisted navigation, regional primary anesthesia, adjunct periarticular analgesia, and routine surgical pathology as essential cost drivers of UKA. Following external validation, this tool may enable surgeons and health insurance providers optimize the delivery of value-based care to patients receiving outpatient UKA.
III.
Opioid prescription trends after ambulatory anterior cervical discectomy and fusion
2023, Spine Journal
Opioid utilization has been well studied for inpatient anterior cervical discectomy and fusion (ACDF). However, the amount and type of opioids prescribed following ambulatory ACDF and the associated risk of persistent use are largely unknown.
To characterize opioid prescription filling following single-level ambulatory ACDF compared with inpatient procedures.
Retrospective cohort study.
Opioid-naive patients who underwent ambulatory (no overnight stay) or inpatient single-level ACDF from 2011 to 2019 were identified from a national insurance database.
Rate, amount, and type of perioperative opioid prescription.
Opioid-naive patients who underwent ambulatory (no overnight stay) or inpatient single-level ACDF from 2011 to 2019 were identified from a national insurance database. Perioperative opioids were defined as opioid prescriptions 30 days before and 14 days after the procedure. Rate, amount, and type of opioid prescription were characterized. Multivariable analyses controlling for any differences in demographics and comorbidities between the two treatment groups were utilized to determine any association between surgical setting and persistent opioid use (defined as the patient still filling new opioid prescriptions >90 days postoperatively).
A total of 42,521 opioid-naive patients were identified, of which 2,850 were ambulatory and 39,671 were inpatient. Ambulatory ACDF was associated with slightly increased perioperative opioid prescription filling (52.7% vs 47.3% for inpatient procedures; p<.001). Among the 20,280 patients (47.7%) who filled perioperative opioid prescriptions, the average amount of opioids prescribed (in morphine milligram equivalents) was similar between ambulatory and inpatient procedures (550 vs 540, p=.413). There was no association between surgical setting and persistent opioid use in patients who filled a perioperative opioid prescription, even after controlling for comorbidities, (adjusted odds ratio, 1.15, p=.066).
Ambulatory ACDF patients who filled perioperative opioid prescriptions were prescribed a similar amount of opioids as those undergoing inpatient procedures. Further, ambulatory ACDF does not appear to be a risk factor for persistent opioid use. These findings are important for patient counseling as well as support the safety profile of this new surgical pathway.
Does a High Postoperative Opioid Dose Predict Chronic Use After ACDF?
2023, World Neurosurgery
The purpose of this study is to determine if increased postoperative prescription opioid dosing is an isolated predictor of chronic opioid use after anterior cervical diskectomy and fusion (ACDF).
A retrospective cohort analysis of patients undergoing ACDF for degenerative diseases from 2016–2019 at a single institution was performed. Preoperative and postoperative opioid and benzodiazepine prescriptions, including morphine milligram equivalents (MMEs) and duration of use, were obtained from the Pennsylvania Prescription Drug Monitoring Program. Univariate analysis compared patient demographics and surgical factors across groups on the basis of postoperative opioid dose (high: MME ≥90, low: MME <90) and chronicity of use (chronic: ≥120 days or >10 prescriptions). Logistic regressions identified predictors of high opioid dose and chronic use.
A total of 385 patients were included. Preoperative opioid tolerance and tobacco use were associated with high postoperative opioid dose and chronic usage. Younger age correlated with high-dose prescriptions. Increased body mass index and preoperative benzodiazepine use were associated with chronic opioid use. Chronic postoperative opioid use correlated with high-dose prescriptions, change in opioid prescribed, private pay scripts, and more than 1 prescriber and pharmacy. Logistic regression identified high postoperative opioid dose, opioid tolerance, increased body mass index, and no prior cervical surgery as predictors of chronic opioid use. Regression analysis determined younger age, increased medical comorbidities, and opioid tolerance to be predictors for high MME prescriptions.
High postoperative opioid dose independently predicted chronic opioid use after ACDF regardless of preoperative opioid tolerance.
Implementation of Machine Learning to Predict Cost of Care Associated with Ambulatory Single-Level Lumbar Decompression
2022, World Neurosurgery
With the emergence of the concept of value-based care, efficient resource allocation has become an increasingly prominent factor in surgical decision-making. Validated machine learning (ML) models for cost prediction in outpatient spine surgery are limited. As such, we developed and internally validated a supervised ML algorithm to reliably identify cost drivers associated with ambulatory single-level lumbar decompression surgery.
A retrospective review of the New York State Ambulatory Surgical Database was performed to identify patients who underwent single-level lumbar decompression from 2014 to 2015. Patients with a length of stay of >0 were excluded. Using pre- and intraoperative parameters (features) derived from the New York State Ambulatory Surgical Database, an optimal supervised ML model was ultimately developed and internally validated after 5 candidate models were rigorously tested, trained, and compared for predictive performance related to total charges. The best performing model was then evaluated by testing its performance on identifying relationships between features of interest and cost prediction. Finally, the best performing algorithm was entered into an open-access web application.
A total of 8402 patients were included. The gradient-boosted ensemble model demonstrated the best performance assessed via internal validation. Major cost drivers included anesthesia type, operating room time, race, patient income and insurance status, community type, worker’s compensation status, and comorbidity index.
The gradient-boosted ensemble model predicted total charges and associated cost drivers associated with ambulatory single-level lumbar decompression using a large, statewide database with excellent performance. External validation of this algorithm in future studies may guide practical application of this clinical tool.

View all citing articles on Scopus

: FDA device/drug status: Not applicable.

: Author disclosures: AVK: Nothing to disclose. PTO: Nothing to disclose. QCBST: Nothing to disclose. MLDB: Nothing to disclose. TDC: Consulting: GE Healthcare (A, paid directly to institution), NuVasive (C, paid directly to institution), K2M (A, paid directly to institution), Bio2 (A, paid directly to institution). SHH: Nothing to disclose. JM: Nothing to disclose. WCP: Nothing to disclose. AJS: Nothing to disclose. CMB: Wolters Kluwer (A), Elsevier (B). Consulting: United Health Care (B). Other Office: NASS/The Spine Journal (D). Fellowship Support: OMEGA (D, Paid directly to institution/employer). JHS: Nothing to disclose.

View full text

Clinical StudyMachine learning for prediction of sustained opioid prescription after anterior cervical discectomy and fusion

Abstract

BACKGROUND CONTEXT

PURPOSE

STUDY DESIGN/SETTING

PATIENT SAMPLE

OUTCOME MEASURES

METHODS

RESULTS

CONCLUSIONS

Introduction

Section snippets

Guidelines

Data source

Results

Discussion

Conclusions

Spine J

Spine J

J Clin Epidemiol

Spine J

Addict Behav

The economic burden of prescription opioid overdose, abuse, and dependence in the United States, 2013

Med Care

Coupling policymaking with evaluation—the case of the opioid crisis

N Engl J Med

Strategies for reducing opioid-overdose deaths—lessons from Canada

N Engl J Med

The public and the opioid-abuse epidemic

N Engl J Med

Changing dynamics of the drug overdose epidemic in the United States from 1979 through 2016

Science

Chronic opioid usage in surgical patients in a large academic center

Ann Surg

New persistent opioid use after minor and major surgical procedures in US adults

JAMA Surg

Effect of Narcotic prescription limiting legislation on opioid utilization following lumbar spine surgery

Spine J

Sustained preoperative opioid use is a predictor of continued use following spine surgery

J Bone Joint Surg Am Vol

Risk factors for prolonged opioid use following spine surgery, and the association with surgical intensity, among opioid-naive patients

J Bone Joint Surg Am Vol

Preoperative chronic opioid therapy: a risk factor for complications, readmission, continued opioid use and increased costs after one- and two-level posterior lumbar fusion

Spine

Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD): the TRIPOD statement

BMC Med

Guidelines for developing and reporting machine learning predictive models in biomedical research: a multidisciplinary view

J Med Internet Res

Rates and risk factors for prolonged opioid use after major surgery: population based cohort study

BMJ

Long-term analgesic use after low-risk surgery: a retrospective cohort study

Arch Intern Med

Clinical Study
Machine learning for prediction of sustained opioid prescription after anterior cervical discectomy and fusion