Use of Deep Learning to Identify Peripheral Arterial Disease Cases From Narrative Clinical Notes.

INTRODUCTION: Peripheral arterial disease (PAD) is the leading cause of amputation in the United States. Despite affecting 8.5 million Americans and more than 200 million people globally, there are significant gaps in awareness by both patients and providers. Ongoing efforts to raise PAD awareness among both the public and health-care professionals have not met widespread success. Thus, there is a need for alternative methods for identifying PAD patients. One potentially promising strategy leverages natural language processing (NLP) to digitally screen patients for PAD. Prior approaches have applied keyword search (KWS) to billing codes or unstructured clinical narratives to identify patients with PAD. However, KWS is limited by its lack of flexibility, the need for manual algorithm development, inconsistent validation, and an inherent failure to capture patients with undiagnosed PAD. Recent advances in deep learning (DL) allow modern NLP models to learn a conceptual representation of the verbiage associated with PAD. This capability may overcome the characteristic constraints of applying strict rule-based algorithms (i.e., searching for a disease-defining set of keywords or billing codes) to real-world clinical data. Herein, we investigate the use of DL to identify patients with PAD from unstructured notes in the electronic health record (EHR). METHODS: Using EHR data from a statewide health information exchange, we first created a dataset of all patients with diagnostic or procedural codes (International Classification of Diseases version 9 or 10 or Current Procedural Terminology) for PAD. This study population was then subdivided into training (70%) and testing (30%) cohorts. We based ground truth labels (PAD versus no PAD) on the presence of a primary diagnostic or procedural billing code for PAD at the encounter level. We implemented our KWS-based identification strategy using the currently published state-of-the-art algorithm for identifying PAD cases from unstructured EHR data. We developed a DL model using a BioMed-RoBERTa base that was fine-tuned on the training cohort. We compared the performance of the KWS algorithm to our DL model on a binary classification task (PAD versus no PAD). RESULTS: Our study included 484,363 encounters across 71,355 patients represented in 2,268,062 notes. For the task of correctly identifying PAD related notes in our testing set, the DL outperformed KWS on all model performance measures (Sens 0.70 versus 0.62; Spec 0.99 versus 0.94; PPV 0.82 versus 0.69; NPV 0.97 versus 0.96; Accuracy 0.96 versus 0.91; P value for all comparisons <0.001). CONCLUSIONS: Our findings suggest that DL outperforms KWS for identifying PAD cases from clinical narratives. Future planned work derived from this project will develop models to stage patients based on clinical scoring systems.

Allogenic Vertebral Body Adherent Mesenchymal Stromal Cells Promote Muscle Recovery in Diabetic Mouse Model of Limb Ischemia.

BACKGROUND: Chronic limb threatening ischemia (CLTI) carries a significant risk for amputation especially in diabetic patients with poor options for revascularization. Phase I trials have demonstrated efficacy of allogeneic mesenchymal stromal cells (MSC) in treating diabetic CLTI. Vertebral bone adherent mesenchymal stromal cells (vBA-MSC) are derived from vertebral bodies of deceased organ donors which offer the distinct advantage of providing a 1,000x greater yield compared to that of living donor bone aspiration. This study describes the effects of intramuscular injection of allogenic vBA-MSC in promoting limb perfusion and muscle recovery in a diabetic CLTI mouse model. METHODS: A CLTI mouse model was created through unilateral ligation of the femoral artery in male polygenic diabetic TALLYHO mice. Treated mice were injected with vBA-MSC into the gracilis muscle of the ischemic limb 7 days post ligation. Gastrocnemius or tibialis muscle was assessed post-mortem for fibrosis by collagen staining, capillary density via immunohistochemistry, and mRNA by quantitative real time PCR. Laser Doppler perfusion imaging and plantar flexion muscle testing were performed to quantify changes in limb perfusion and muscle function. RESULTS: Compared to vehicle control, treated mice demonstrated indicators of muscle recovery including decreased fibrosis, increased perfusion, muscle torque, and angiogenesis. PCR analysis of muscle obtained 7- and 30-days post vBA-MSC injection showed an upregulation in expression of MyoD1 (p = 0.03) and MyH3 (p = 0.008) mRNA representing muscle regeneration, VEGF-A (p = 0.002 ; p = 0.004) signifying angiogenesis as well as IL-10 (p < 0.001), T regulatory cell marker Foxp3 (p = 0.04), and M2-biased macrophage marker Mrc1 (CD206) (p = 0.02). CONCLUSIONS: These findings indicate human allogeneic vBA-MSC ameliorate ischemic muscle damage and rescue muscle function. These results in a murine model will enable further studies to develop potential therapies for diabetic CLTI patients.

Age-related outcomes of arteriovenous grafts for hemodialysis access.

BACKGROUND: The prevalence of end-stage renal disease spans the spectrum of age. Arteriovenous grafts are viable alternatives for hemodialysis access in patients whose anatomy precludes placement of an arteriovenous fistula. This report describes the age-related outcomes after arteriovenous graft placement in a population-based cohort. METHODS: A retrospective cohort study was conducted of all patients who initiated hemodialysis in the U.S. Renal Data System (2007-2014). The χ2 test, t-test, Kaplan-Meier analysis, log-rank test, and multivariable logistic and Cox regression analyses were employed to evaluate access maturation, interventions, patency, and mortality. RESULTS: Of the 78,341 patients studied, 10,150 (13%) were younger than 50 years, 13,167 (16.8%) were 50 to 59 years, 19,975 (25.5%) were 60 to 69 years, 20,307 (25.9%) were 70 to 79 years, and 14,742 (18.8%) were 80+ years. There was no significant difference in access maturation time for patients in the older age categories compared to patients younger than 50 years. Primary patency at 5 years comparing <50 years vs 50 to 59 years vs 60 to 69 years vs 70 to 79 years vs 80+ years was 12% vs 12% vs 9% vs 9% vs 8% (P < .001). Primary assisted patency at 5 years was 20% vs 21% vs 18% vs 17% vs 14% (P < .001). Secondary patency at 5 years was 36% vs 39% vs 36% vs 30% vs 31% (P < .001). There was no significant difference in primary patency (adjusted hazard ratio [aHR], 1.00; 95% confidence interval [CI], 1.00-1.00; P < .001), primary assisted patency (aHR, 1.00; 95% CI, 1.00-1.00; P < .001), and secondary patency (aHR, 1.00; 95% CI, 1.00-1.00; P = .029) with increasing age. However, there was a decrease in severe prosthetic graft infection requiring graft excision (aHR, 0.99; 95% CI, 0.99-0.99; P < .001) and increase in mortality (aHR, 1.03; 95% CI, 1.03-1.03; P < .001) for the older age categories compared with the younger patients. CONCLUSIONS: In this population-based cohort of hemodialysis patients, there was no significant association between older age and prosthetic graft maturation or patency. However, older age was associated with a decrease in severe graft infection and the expected increase in mortality.

Long-Term Outcomes of Carotid Endarterectomy and Carotid Artery Stenting When Performed by a Single Vascular Surgeon.

OBJECTIVES:: Carotid artery endarterectomy (CEA) and carotid artery stenting (CAS) are 2 effective treatment options for carotid revascularization and stroke prevention. The long-term outcomes of Carotid Revascularization Endarterectomy versus Stenting Trial (CREST) reported similar stroke and death rate between the 2 procedures. This study presents the short- and long-term outcomes of CEA and CAS of all risk patients performed by a single vascular surgeon in a real-world setting. METHODS:: We retrospectively reviewed all patients who underwent CEA and CAS from September 2005 to June 2017 at our institute. Student t test, χ2, and Fisher exact tests were used to compare patient's characteristics. Multivariate logistic, cox regression models and survival analysis were used to compare postoperative and long-term outcomes between the 2 groups. RESULTS:: Over 2000 patients were evaluated for carotid artery stenosis during the study period, and 313 revascularization procedures were performed (CEA: 47%, CAS: 53%). Patients' age (Mean [95% confidence interval, CI] 68.8 [67.2-70.4] vs 69.7 [68.2-71.3], P = .40) was similar between CEA and CAS. Patients who underwent CAS had significantly higher comorbidities (chronic obstructive pulmonary disease [COPD], chronic heart failure [CHF], hyperlipidemia, and prior ipsilateral intervention, all P < .05). No difference was found in 30-day complications after CEA versus CAS including stroke (2.0% vs 1.2%), myocardial infarction (MI; 0.7% vs 1.2%), death (0% vs 1.2%) as well as combined major adverse events (stroke/death/MI; 2.7% vs 3.0%; all P > .05). Overall 7-year survival, stroke-free survival and restenosis-free survival were similar between the 2 groups ( P > .5). Significant predictors of mortality were diabetes (hazard ratio, HR [95% CI]: 2.41 [1.15-5.08]), chronic kidney disease (HR [95% CI]: 4.89 [1.97-12.13]), and COPD (HR [95% CI]: 3.31 [1.43-7.71]; all P values <.05). Statin use was protective with 71% reduction in risk of mortality (HR [95% CI]: 0.29 [0.12-0.67], P = .004). CONCLUSION:: Our experience showed comparable short- and long-term outcomes of CAS and CEA performed for carotid artery stenosis by vascular surgeon. There was no difference between single institutional long-term outcomes and CREST outcomes following CEA and CAS.

Peroral endoscopic myotomy is effective and safe in non-achalasia esophageal motility disorders: an international multicenter study.

BACKGROUND AND STUDY AIMS: The efficacy of per oral endoscopic myotomy (POEM) in non-achalasia esophageal motility disorders such as esophagogastric junction outflow obstruction (EGJOO), diffuse esophageal spasm (DES), and jackhammer esophagus (JE) has not been well demonstrated. The aim of this international multicenter study was to assess clinical outcomes of POEM in patients with non-achalasia disorders, namely DES, JE, and EGJOO, in a large cohort of patients. PATIENTS AND METHODS: This was a retrospective study at 11 centers. Consecutive patients who underwent POEM for EGJOO, DES, or JE between 1/2014 and 9/2016 were included. Rates of technical success (completion of myotomy), clinical response (symptom improvement/Eckardt score ≤ 3), and adverse events (AEs, severity per ASGE lexicon) were ascertained . RESULTS: Fifty patients (56 % female; mean age 61.7 years) underwent POEM for EGJOO (n = 15), DES (n = 17), and JE (n = 18). The majority of patients (68 %) were treatment-naïve. Technical success was achieved in all patients with a mean procedural time of 88.4 ±â€Š44.7 min. Mean total myotomy length was 15.1 ±â€Š4.7 cm. Chest pain improved in 88.9 % of EGJOO and 87.0 % of DES/JE ( P  = 0.88). Clinical success was achieved in 93.3 % of EGJOO and in 84.9 % of DES/JE ( P  = 0.41) with a median follow-up of 195 and 272 days, respectively. Mean Eckardt score decreased from 6.2 to 1.0 in EGJOO ( P  < 0.001) and from 6.9 to 1.9 in DES/JE ( P  < 0.001). A total of 9 (18 %) AEs occurred and were rated as mild in 55.6 % and moderate in 44.4 %. CONCLUSION: POEM is effective and safe in management of non-achalasia esophageal motility disorders, which include DES, JE, and EGJOO.

An international, multicenter, comparative trial of EUS-guided gastrogastrostomy-assisted ERCP versus enteroscopy-assisted ERCP in patients with Roux-en-Y gastric bypass anatomy.

BACKGROUND AND AIMS: ERCP is challenging in patients with Roux-en-Y gastric bypass (RYGB) anatomy. EUS-guided gastrogastrostomy (GG) creation is a promising novel technique to access the excluded stomach to facilitate conventional ERCP. We aimed to compare procedural outcomes and adverse events (AEs) between EUS-guided GG-assisted ERCP (EUS-GG-ERCP) and enteroscopy-assisted ERCP (e-ERCP) in patients with RYGB. METHODS: Patients with RYGB anatomy who underwent EUS-GG-ERCP or e-ERCP between 2014 and 2016 at 5 tertiary centers were included. The primary outcome was technical success of ERCP, defined as successful cannulation of the selected duct with successful intervention as intended. Secondary outcomes included total procedural time (in the EUS-GG group, total procedural time included EUS-GG creation plus ERCP procedure time), length of hospital stay, and rate/severity of AEs graded according to the American Society for Gastrointestinal Endoscopy lexicon. RESULTS: A total of 60 patients (mean age, 57.2 ± 13.2; 75% women) were included, of whom 30 (50%) underwent EUS-GG-ERCP and 30 (50%) underwent e-ERCP (double-balloon enteroscope ERCP, 19; single-balloon enteroscope ERCP, 11). The technical success rate was significantly higher in the EUS-GG-ERCP versus the e-ERCP group (100% vs 60.0%, P < .001). Total procedure time was significantly shorter in patients who underwent EUS-GG-ERCP (49.8 minutes vs 90.7 minutes, P < .001). Postprocedure median length of hospitalization was shorter in the EUS-GG group (1 vs 10.5 days, P = .02). Rate of AEs was similar in both groups (10% vs 6.7%, P = 1). CONCLUSIONS: EUS-GG-ERCP may be superior to e-ERCP in patients with RYGB anatomy in terms of a higher technical success and shorter procedural times and offers a similar safety profile.

EUS-guided gastroenterostomy: a multicenter study comparing the direct and balloon-assisted techniques.

BACKGROUND: EUS-guided gastroenterostomy (EUS-GE) is a developing modality in the management of gastric outlet obstruction (GOO) with several technical approaches, including the direct and balloon-assisted techniques. The aim of this study was to compare the direct with the balloon-assisted modality while further defining the role of EUS-GE in GOO. METHODS: This multicenter, retrospective study involved consecutive patients who underwent EUS-GE with the direct or balloon-assisted technique for GOO (January 2014 to October 2016). The primary outcome was technical success. Secondary outcomes were success (ability to tolerate at least a full fluid diet), procedure time, and rate/severity of adverse events (AEs). RESULTS: A total of 74 patients (44.6% women; mean age 63.0 ± 11.7 years) underwent EUS-GE for GOO (direct gastroenterostomy, n = 52; balloon-assisted gastroenterostomy, n = 22). GOO was of malignant and benign etiology in 66.2% and 33.8% of patients, respectively. Technical success was achieved in 94.2% of the direct and 90.9% of the balloon-assisted approach (P = .63). Mean procedure time was shorter with the direct technique (35.7 ± 32.1 minutes vs 89.9 ± 33.3 minutes, P < .001). The clinical success rate was 92.3% for the direct technique and 90.9% for the balloon-assisted modality (P = 1.00), with a mean time to oral intake of 1.32 ± 2.76 days. The AE rate was 6.8% with only 1 severe AE noted. Rate of AEs, postprocedure length of stay, need for reintervention, and survival were similar between the 2 groups. CONCLUSIONS: EUS-GE is effective and safe in the management of GOO. The direct technique may be the preferred method given its shorter procedure time when compared with the balloon-assisted approach. Prospective trials are needed to confirm these findings.

Concomitant Removal of Gastric Band and Gastric Bypass: Analysis of Outcomes and Complications from the ACS-NSQIP Database.

BACKGROUND: Conversion of laparoscopic adjustable gastric banding (LGB) to laparoscopic Roux-en-Y gastric bypass (LRYGB) is an established procedure. However, multiple reports have indicated higher morbidity and mortality rates associated with this operation, especially when performed as a single-staged procedure. PURPOSE: We sought to compare mortality and morbidity of LRYGB vs. LRYGB with concomitant gastric band removal (LRYGB/LGBR). METHODS: Data from the American College of Surgeons' National Surgical Quality Improvement Program (ACS-NSQIP) database was obtained for the time period of 2008 to 2014 using CPT codes for LRYGB and LGBR. Sepsis was the primary outcome measure with overall morbidity as a secondary outcome. Bivariate and multivariate analyses were carried out using SAS (Statistical Analysis System). RESULTS: During the study period, 64,866 patients had primary LRYGB and 1212 had LRYGB/LGBR. On bivariate analyses, mean operative time was lower for patients undergoing LRYGB rather than LRYGB/LGBR (132.88 ± 56.29 vs. 177.72 ± 70.21 min, p < 0.001). There was no statistically significant difference in the rate of postoperative mortality (0.16 vs. 0.08 %, p > 0.999), sepsis (0.78 vs. 0.74 %, p = 0.87), or other postoperative outcomes such as return to the operating room, wound infection, and venous thromboembolism. The odds ratio (OR) for sepsis remained not significant (OR = 0.74; 95 % confidence interval (CI) = (0.38-1.45)) after multivariate analysis. CONCLUSION: LRYGBP/LGBR is not associated with a higher morbidity and mortality compared to LRYGB alone. The data implies that a one-step revisional procedure is appropriate when converting a failed gastric band to LRYGB.

Prevalence of vitamin D deficiency in adults presenting for bariatric surgery in Lebanon.

BACKGROUND: Vitamin D deficiency is common among obese patients presenting for bariatric surgery in Europe and North America. The prevalence of vitamin D deficiency in this patient population in Lebanon and the Middle East has not been studied. OBJECTIVES: The aim of this study was to determine the rate of vitamin D deficiency in a cohort of patients presenting for bariatric surgery in Lebanon. SETTING: American University of Beirut Medical Center, Beirut, Lebanon. METHODS: Data was extracted from a prospective database of patients presenting for bariatric surgery at the American University of Beirut Medical Center from July 2011 until June 2014. The prevalence of vitamin D deficiency was determined using established cut-offs followed by analysis of the relationship between low vitamin D and certain patient characteristics. RESULTS: More than two thirds of all patients (68.9%) were vitamin D deficient (≤19.9 ng/mL), whereas 22.6% had insufficient levels (20-29.9 ng/mL) and only 8.6 % had sufficient levels (≥30 ng/mL). Vitamin D levels were inversely associated with BMI>50 kg/m(2). Low vitamin D levels were also correlated with younger age, male gender, lack of physical exercise, and nonsunny season. No association was shown between 25-hydroxyvitamin D deficiency and type 2 diabetes mellitus, cardiovascular disease, osteoarticular disease, hypertension, or depression. CONCLUSION: Vitamin D deficiency is prevalent among patients with Class II or Class III obesity presenting for bariatric surgery in Lebanon. These findings emphasize the need for careful attention when evaluating patients before bariatric surgery and the importance of providing patients with adequate supplementation.