Extended postoperative venous thromboembolism prophylaxis after bariatric surgery: a comparison of existing risk-stratification tools and 5-year MBSAQIP analysis.

BACKGROUND: Venous thromboembolism (VTE) is a leading cause of 30-day mortality after metabolic and bariatric surgery (MBS). Multiple predictive tools exist for VTE risk assessment and extended VTE chemoprophylaxis determination. OBJECTIVE: To review existing risk-stratification tools and compare their predictive abilities. SETTING: MBSAQIP database. METHODS: Retrospective analysis of the Metabolic and Bariatric Surgery Accreditation and Quality Improvement Program (MBSAQIP) database was performed (2015-2019) for primary minimally invasive MBS cases. VTE clinical factors and risk-assessment tools were evaluated: body mass index threshold of 50 kg/m2, Caprini risk-assessment model, and 3 bariatric-specific tools: the Cleveland Clinic VTE risk tool, the Michigan Bariatric Surgery Collaborative tool, and BariClot. MBS patients were deemed high risk based on criteria from each tool and further assessed for sensitivity, specificity, and positive predictive value. RESULTS: Overall, 709,304 patients were identified with a .37% VTE rate. Bariatric-specific tools included multiple predictors: procedure, age, race, gender, operative time, length of stay, heart failure, and dyspnea at rest; operative time was the only variable common to all. The body mass index cutoff and Caprini risk-assessment model had higher sensitivity but lower specificity when compared with the Michigan Bariatric Surgery Collaborative and BariClot tools. While the sensitivity of the tools varied widely and was overall low, the Cleveland Clinic tool had the highest sensitivity. The bariatric-specific tools would have recommended extended prophylaxis for 1.1%-15.6% of patients. CONCLUSIONS: Existing MBS VTE risk-assessment tools differ widely for inclusion variables, high-risk definition, and predictive performance. Further research and registry inclusion of all significant risk factors are needed to determine the optimal risk-stratified approach for predicting VTE events and determining the need for extended prophylaxis.

Ambiguous and Incomplete: Natural Language Processing Reveals Problematic Reporting Styles in Thyroid Ultrasound Reports.

OBJECTIVE: Natural language processing (NLP) systems convert unstructured text into analyzable data. Here, we describe the performance measures of NLP to capture granular details on nodules from thyroid ultrasound (US) reports and reveal critical issues with reporting language. METHODS: We iteratively developed NLP tools using clinical Text Analysis and Knowledge Extraction System (cTAKES) and thyroid US reports from 2007 to 2013. We incorporated nine nodule features for NLP extraction. Next, we evaluated the precision, recall, and accuracy of our NLP tools using a separate set of US reports from an academic medical center (A) and a regional health care system (B) during the same period. Two physicians manually annotated each test-set report. A third physician then adjudicated discrepancies. The adjudicated "gold standard" was then used to evaluate NLP performance on the test-set. RESULTS: A total of 243 thyroid US reports contained 6,405 data elements. Inter-annotator agreement for all elements was 91.3%. Compared with the gold standard, overall recall of the NLP tool was 90%. NLP recall for thyroid lobe or isthmus characteristics was: laterality 96% and size 95%. NLP accuracy for nodule characteristics was: laterality 92%, size 92%, calcifications 76%, vascularity 65%, echogenicity 62%, contents 76%, and borders 40%. NLP recall for presence or absence of lymphadenopathy was 61%. Reporting style accounted for 18% errors. For example, the word "heterogeneous" interchangeably referred to nodule contents or echogenicity. While nodule dimensions and laterality were often described, US reports only described contents, echogenicity, vascularity, calcifications, borders, and lymphadenopathy, 46, 41, 17, 15, 9, and 41% of the time, respectively. Most nodule characteristics were equally likely to be described at hospital A compared with hospital B. CONCLUSIONS: NLP can automate extraction of critical information from thyroid US reports. However, ambiguous and incomplete reporting language hinders performance of NLP systems regardless of institutional setting. Standardized or synoptic thyroid US reports could improve NLP performance.

Thyroid Ultrasound Reports: Will the Thyroid Imaging, Reporting, and Data System Improve Natural Language Processing Capture of Critical Thyroid Nodule Features?

BACKGROUND: Critical thyroid nodule features are contained in unstructured ultrasound (US) reports. The Thyroid Imaging, Reporting, and Data System (TI-RADS) uses five key features to risk stratify nodules and recommend appropriate intervention. This study aims to analyze the quality of US reporting and the potential benefit of Natural Language Processing (NLP) systems in efficiently capturing TI-RADS features from text reports. MATERIALS AND METHOD: This retrospective study used free-text thyroid US reports from an academic center (A) and community hospital (B). Physicians created "gold standard" annotations by manually extracting TI-RADS features and clinical recommendations from reports to determine how often they were included. Similar annotations were created using an automated NLP system and compared with the gold standard. RESULTS: Two hundred eighty-two reports contained 409 nodules at least 1-cm in maximum diameter. The gold standard identified three nodules (0.7%) which contained enough information to calculate a complete TI-RADS score. Shape was described most often (92.7% of nodules), whereas margins were described least often (11%). A median number of two TI-RADS features are reported per nodule. The NLP system was significantly less accurate than the gold standard in capturing echogenicity (27.5%) and margins (58.9%). One hundred eight nodule reports (26.4%) included clinical management recommendations, which were included more often at site A than B (33.9 versus 17%, P < 0.05). CONCLUSIONS: These results suggest a gap between current US reporting styles and those needed to implement TI-RADS and achieve NLP accuracy. Synoptic reporting should prompt more complete thyroid US reporting, improved recommendations for intervention, and better NLP performance.

Patterns of opioid use in dialysis access procedures.

BACKGROUND: Overprescription of opioids after surgical procedures is recognized as an important contributor to opioid misuse. Dialysis access procedures are commonly performed outpatient operations with few data or guidelines to inform prescription pain management practices. We sought to characterize opioid pain medication use after dialysis access surgery to promote a conservative approach to postoperative opioid prescriptions. METHODS: We performed a retrospective review of patients who underwent surgical dialysis access procedures from August 2018 through January 2019. Patient-reported opioid use information was captured in a brief questionnaire administered during routine follow-up appointments or phone calls and recorded in the electronic medical record. The procedure, type of intraoperative anesthesia or analgesia, postoperative prescription provided, and patient factors (including age, sex, dialysis type, history of chronic pain, and preoperative opioid or benzodiazepine use) were recorded. All procedures were classified by type (arteriovenous fistula or graft with a short incision [AVF-S], arteriovenous fistula or graft with a long incision [AVF-L], or peritoneal dialysis [PD] catheter), and descriptive statistics were performed using R (R Foundation for Statistical Computing, Vienna, Austria). RESULTS: Eighty-six patients underwent dialysis access procedures in the study time frame, of whom 63 were administered the pain questionnaire and 58 quantified opioid use; 85% of patients received a prescription, but 31% took no opioids and 71% used opioids for ≤2 days. Interquartile ranges (25th-75th percentile) of prescription and consumption quantities for patients who underwent AVF-L procedures were 10 to 28 pills and 2.5 to 20 pills; for patients who underwent AVF-S, quantities were 4.0 to 8.4 pills and 0 to 4.3 pills; and PD quantities were 10 pills and 3.3 to 9 pills. Thirty-one patients (53%) reported receiving more pain medication than they used, which resulted in a median of 8 excess pills per patient with an unused pill interquartile range of 0 to 22 pills for AVF-L procedures, 0 to 4.2 pills for AVF-S procedures, and 1.3 to 6.7 pills for PD procedures. Patients who were prescribed oxycodone or had a repeated operation had significantly increased opioid use. CONCLUSIONS: This investigation of opioid use after surgical dialysis access procedures suggests that most patients use relatively few opioid pills after surgery, which translates into overprescription and leftover medication for >50% of patients. A conservative approach to postoperative prescription guidelines using lower prescription quantities would encourage opioid-related risk reduction while providing adequate postoperative analgesia. Recommended quantities for postoperative prescriptions were generated using the 80th percentile consumed and were 0 to 6 pills for brachiobasilic or brachiocephalic fistulas, 0 to 5 pills for basilic vein transposition, 0 to 5 pills for radiocephalic AVF, 0 to 15 pills for upper arm grafts, and 0 to 10 pills for PD catheter placement.

The optimal dosing scheme for levothyroxine after thyroidectomy: A comprehensive comparison and evaluation.

BACKGROUND: Patients often struggle to attain euthyroidism after thyroidectomy, and multiple dosing schemes have been proposed to supplant the standard weight-based approach for initial levothyroxine dosing after thyroidectomy. The objectives of this study were to review the literature for existing levothyroxine dosing schemes and compare estimation accuracies with novel schemes developed with machine learning. METHODS: This study retrospectively analyzed 598 patients who attained euthyroidism after total or completion thyroidectomy for benign disease. A scoping review identified existing levothyroxine dosing schemes. Thirteen machine learning algorithms estimated euthyroid dose. Using 10-fold cross-validation, we compared schemes by the proportion of patients having a predicted dose within 12.5 µg/day of their euthyroid dose. RESULTS: Of 264 reviewed articles, 7 articles proposed retrospectively implementable dosing schemes. A novel Poisson regression model proved most accurate, correctly predicting 64.8% of doses. Incorporating 7 variables, Poisson regression was significantly more accurate than the best scheme in the literature (body mass index/weight based) that correctly predicted 60.9% of doses (P = .031). Standard weight-based dosing (1.6 µg/kg/day) correctly predicted 51.3% of doses, and the least effective scheme (age/sex/weight based) correctly predicted 27.4% of doses. CONCLUSION: Using readily available variables, a novel Poisson regression dosing scheme outperforms other machine learning algorithms and all existing schemes in estimating levothyroxine dose.

Does Surgical Margin Impact Recurrence in Noninvasive Intraductal Papillary Mucinous Neoplasms?: A Multi-institutional Study.

OBJECTIVE: The relevance of margin positivity on recurrence after resection of intraductal papillary mucinous neoplasms (IPMNs) is poorly defined and represents one reason controversy remains regarding optimal surveillance recommendations. METHODS: Patients undergoing surgery for noninvasive IPMN at 8 academic medical centers from the Central Pancreas Consortium were analyzed. A positive margin was defined as presence of IPMN or pancreatic intraepithelial neoplasia. RESULTS: Five hundred two patients underwent surgery for IPMN; 330 (66%) did not have invasive cancer on final pathology and form the study cohort. Of these, 20% harbored high grade dysplasia. A positive margin was found in 20% of cases and was associated with multifocal disease (P = 0.02). The majority of positive margins were associated with low grade dysplasia. At a median follow-up of 36 months, 34 (10.3%) patients recurred, with 6.7% developing recurrent cystic disease and 3.6% developing invasive cancer. On multivariate analysis, margin positivity was not associated with recurrence of either IPMN or invasive cancer (P > 0.05). No association between margin status and development of recurrence at the margin was found. Only 6% of recurrences developed at the resection margin and median time to recurrence was 22 months. Of note, 18% of recurrences occurred > 5 years following surgery. CONCLUSION: Margin positivity after resection for noninvasive IPMNs is primarily due to low grade dysplasia and is not associated with developing recurrence in the remnant pancreas or at the resection margin. Long-term surveillance is required for all patients, as a significant number of recurrences developed over 5 years after the index operation.

Bariatric surgery barriers: a review using Andersen's Model of Health Services Use.

Severe obesity affects nearly 20 million adults in the United States and is associated with significant morbidity and mortality. Bariatric surgery is the most effective treatment for weight loss and resolution of obesity-related co-morbidities. Of adults with severe obesity,<1% undergo bariatric surgery annually. Both contextual (health system, clinicians, and community) and individual factors contribute to the underutilization of bariatric surgery. In this review, we summarize potential barriers to undergoing bariatric surgery within the framework of Andersen's Behavioral Model of Health Services Use.

Machine learning to identify multigland disease in primary hyperparathyroidism.

BACKGROUND: 20%-25% of patients with primary hyperparathyroidism will have multigland disease (MGD). Preoperatative imaging can be inaccurate or unnecessary in MGD. Identification of MGD could direct the need for imaging and inform operative approach. The purpose of this study is to use machine learning (ML) methods to predict MGD. METHODS: Retrospective review of a prospective database. The ML platform, Waikato Environment for Knowledge Analysis, was used, and we selected models for (1) overall accuracy and (2) preferential identification of MGD. A review of imaging studies was performed on a cohort predicted to have MGD. RESULTS: 2010 patients met inclusion criteria: 1532 patients had single adenoma (SA) (76%) and 478 had MGD (24%). After testing many algorithms, we selected two different models for potential integration as clinical decision-support tools. The best overall accuracy was achieved using a boosted tree classifier, RandomTree: 94.1% accuracy; 94.1% sensitivity, 83.8% specificity, 94.1% positive predictive value, and 0.984 area under the receiver operating characteristics curve. To maximize positive predictive value of MGD prediction, a rule-based classifier, JRip, with cost-sensitive learning was used and achieved 100% positive predictive value for MGD. Imaging reviewed from the cohort of 34 patients predicted to have MGD by the cost-sensitive model revealed 39 total studies performed: 28 sestamibi scans and 11 ultrasounds. Only 8 (29%) sestamibi scans and 4 (36%) ultrasounds were correct. CONCLUSIONS: ML methods can help distinguish MGD early in the clinical evaluation of primary hyperparathyroidism, guiding further workup and surgical planning.