Automated Matching of Patients to Clinical Trials: A Patient-Centric Natural Language Processing Approach for Pediatric Leukemia.

PURPOSE: Matching patients to clinical trials is cumbersome and costly. Attempts have been made to automate the matching process; however, most have used a trial-centric approach, which focuses on a single trial. In this study, we developed a patient-centric matching tool that matches patient-specific demographic and clinical information with free-text clinical trial inclusion and exclusion criteria extracted using natural language processing to return a list of relevant clinical trials ordered by the patient's likelihood of eligibility. MATERIALS AND METHODS: Records from pediatric leukemia clinical trials were downloaded from ClinicalTrials.gov. Regular expressions were used to discretize and extract individual trial criteria. A multilabel support vector machine (SVM) was trained to classify sentence embeddings of criteria into relevant clinical categories. Labeled criteria were parsed using regular expressions to extract numbers, comparators, and relationships. In the validation phase, a patient-trial match score was generated for each trial and returned in the form of a ranked list for each patient. RESULTS: In total, 5,251 discretized criteria were extracted from 216 protocols. The most frequent criterion was previous chemotherapy/biologics (17%). The multilabel SVM demonstrated a pooled accuracy of 75%. The text processing pipeline was able to automatically extract 68% of eligibility criteria rules, as compared with 80% in a manual version of the tool. Automated matching was accomplished in approximately 4 seconds, as compared with several hours using manual derivation. CONCLUSION: To our knowledge, this project represents the first open-source attempt to generate a patient-centric clinical trial matching tool. The tool demonstrated acceptable performance when compared with a manual version, and it has potential to save time and money when matching patients to trials.

Using Machine Learning to Predict Likelihood and Cause of Readmission After Hospitalization for Chronic Obstructive Pulmonary Disease Exacerbation.

Background: Chronic obstructive pulmonary disease (COPD) is a leading cause of hospital readmissions. Few existing tools use electronic health record (EHR) data to forecast patients' readmission risk during index hospitalizations. Objective: We used machine learning and in-hospital data to model 90-day risk for and cause of readmission among inpatients with acute exacerbations of COPD (AE-COPD). Design: Retrospective cohort study. Participants: Adult patients admitted for AE-COPD at the University of Chicago Medicine between November 7, 2008 and December 31, 2018 meeting International Classification of Diseases (ICD)-9 or -10 criteria consistent with AE-COPD were included. Methods: Random forest models were fit to predict readmission risk and respiratory-related readmission cause. Predictor variables included demographics, comorbidities, and EHR data from patients' index hospital stays. Models were derived on 70% of observations and validated on a 30% holdout set. Performance of the readmission risk model was compared to that of the HOSPITAL score. Results: Among 3238 patients admitted for AE-COPD, 1103 patients were readmitted within 90 days. Of the readmission causes, 61% (n = 672) were respiratory-related and COPD (n = 452) was the most common. Our readmission risk model had a significantly higher area under the receiver operating characteristic curve (AUROC) (0.69 [0.66, 0.73]) compared to the HOSPITAL score (0.63 [0.59, 0.67]; p = 0.002). The respiratory-related readmission cause model had an AUROC of 0.73 [0.68, 0.79]. Conclusion: Our models improve on current tools by predicting 90-day readmission risk and cause at the time of discharge from index admissions for AE-COPD. These models could be used to identify patients at higher risk of readmission and direct tailored post-discharge transition of care interventions that lower readmission risk.

Comparing Bleeding and Thrombotic Rates in Spine Surgery: An Analysis of 119 888 Patients.

STUDY DESIGN: Retrospective, database review. OBJECTIVES: Examine the utilization rate of postoperative deep vein thrombosis (DVT) prophylaxis and compare the incidence and severity of bleeding and thrombotic complications in elective spine surgery patients. METHODS: We utilized PearlDiver, a national orthopedics claims database. All patients who underwent elective spine surgery from 2007 to 2017 were included. Patients were stratified by the presence of DVT prophylaxis drug codes, then by comorbidities for postoperative bleeding/thrombosis. The severity of all bleeding and thrombotic complications in each cohort was studied, including the incidence of complications requiring operative washout, diagnosis of pulmonary embolism, intensive care unit admission, and mortality. RESULTS: A total of 119 888 patients were included. The majority of patients (118 720, >99%) were not administered postoperative DVT chemoprophylaxis while a minority of patients (1168) were. The overall rates of bleeding and thrombotic complications within the population not receiving DVT prophylaxis were 1.96% and 2.45%, respectively (P < .001). The incidence of surgical intervention for a wound washout was 0.62% compared with 1.05% for pulmonary embolism (P < .001). Intensive care unit admission rates related to a wound washout procedure or pulmonary embolism also significantly differed (0.07% vs 0.34%, P < .001). There were no observed differences in mortality. When controlling for patient comorbidity, patients with atrial fibrillation, cancer, or a prior history of thrombotic complications experienced the greatest increased risks of postoperative thrombosis. CONCLUSIONS: DVT prophylaxis is not routinely utilized following elective spine procedures. We report that there exist specific populations which may receive benefit from these practices, although further study is necessary to determine optimal prevention strategies for both thrombotic and bleeding complications in spine surgery.

Benchmarks of Duration and Magnitude of Opioid Consumption After Common Spinal Procedures: A Database Analysis of 47,823 Patients.

STUDY DESIGN: A retrospective cohort study performed in a nationwide insurance claims database. OBJECTIVE: This study aimed to examine duration and magnitude of postoperative opioid prescriptions following common spinal procedures. SUMMARY OF BACKGROUND DATA: Postoperative opioid prescription practices vary widely among providers and procedures and standards of care are not well-established. Previous work does not adequately quantify both duration and magnitude of opioid prescription. METHODS: Forty seven thousand eight hundred twenty three patients with record of any of four common spinal procedures in a nationwide insurance claims database were stratified by preoperative opioid use into three categories: "opioid naive," "sporadic user," or "chronic user," defined as 0, 1, or 2+ prescriptions filled in the 6 months preceding surgery. Those with record of subsequent surgery or readmission were excluded. Duration of opioid use was defined as the time between the index surgery and the last record of filling an opioid prescription. Magnitude of opioid use was defined as milligram morphine equivalents (MME) filled by 30 days post-op, converted to 5 mg oxycodone pills for interpretation. RESULTS: Opioid naive patients were less likely than chronic opioid users to fill any opioid prescription after surgery (63-68% naive vs. 91-95% chronic, P < 0.001), and when they did, their prescriptions were smaller in magnitude (76-91 pills naive vs. 127-152 pills chronic). One year after surgery, 15% to 18% of opioid naive and 50% to 64% of chronic opioid users continued filling prescriptions. CONCLUSION: Opioid naive patients use less postoperative opioids, and for a shorter period of time, than chronic users. This study serves as a normative benchmark for examining postoperative opioid use, which can assist providers in identifying patients with opioid dependence. Importantly, this work calls out the high risk of opioid exposure, as 15% to 18% of opioid naive patients continued filling opioid prescriptions 1 year after surgery. LEVEL OF EVIDENCE: 3.

Benchmarks of Duration and Magnitude of Opioid Consumption After Total Hip and Knee Arthroplasty: A Database Analysis of 69,368 Patients.

BACKGROUND: Opioid prescribing after orthopedic surgeries varies widely, and there is little consensus establishing proper standards of care. This retrospective cohort study examines opioid prescribing trends following total hip (THA) and knee (TKA) arthroplasty and evaluates preoperative opioid use as a predictor of duration and magnitude of postoperative opioid use. METHODS: Patients who underwent THA or TKA in a nationwide insurance database were stratified by preoperative opioid use. Naive, sporadic, and chronic users were defined as 0, 1, or 2+ prescriptions filled 6 months before surgery. Patients were excluded for readmission or subsequent surgery. Duration of opioid use was defined as time between the procedure and the last opioid prescription record, and magnitude of opioid use was defined as quantity of pills filled by 30 days postop. RESULTS: Naive patients were less likely than chronic users to fill any opioid prescription after surgery (THA: 61.5% naive vs 90.4% chronic, TKA: 72.0% naive vs 95.9% chronic), and they obtained fewer pills (THA: 73 pills naive vs 126 pills chronic, TKA: 86 pills naive vs 126 pills chronic, 5-mg oxycodone equivalent). Between 10% (THA) and 13% (TKA) of naive and between 47% (THA) and 62% (TKA) of chronic users continued opioid use at 1 year postop. CONCLUSION: Chronic users obtain more opioids postoperatively and continue filling prescriptions for longer than naive patients. This work benchmarks norms regarding opioid use and furthermore these data highlight the powerful effect of opioid exposure during surgery as 10%-13% of naive patients continued opioids at 1 year postop.

Cages in ACDF are Associated With a Higher Nonunion Rate Than Allograft: A Stratified Comparative Analysis of 6130 Patients.

STUDY DESIGN: A retrospective database review. OBJECTIVE: The purpose of this study was to analyze the rate of nonunion in patients treated with structural allograft and intervertebral cages in anterior cervical discectomy and fusion (ACDF). SUMMARY OF BACKGROUND DATA: Existing literature consists primarily of single-center studies with inconsistent findings. METHODS: We performed a retrospective analysis of 6130 patients registered in the PearlDiver national database through Humana Insurance from 2007 to 2016. All ACDF patients with anterior plating who were active in the database for at least 1 year were included in the study. Patients with a fracture history within 1 year of intervention, past arthrodesis of hand, foot, or ankle, or a planned posterior approach were excluded from the study. Patients were stratified by number of levels treated, tobacco use, and diabetic condition. Nonunion rates of structural allograft and intervertebral cage groups after 1 year were compared using Chi-squared analyses. RESULTS: Four thousand sixty-three patients were included in the allograft group, while 2067 were included in the cage group. Overall nonunion rates were significantly higher in the cage group (5.32%) than in allograft group (1.97%) (P < 0.01). When controlling for confounders, increased rates of nonunion were consistently observed in the cage group, achieving statistical significance in 25 of the 26 analyses. CONCLUSION: The increased rate of nonunion associated with intervertebral cages may suggest the superiority of allograft over cages in ACDF. LEVEL OF EVIDENCE: 3.