Carolinas HealthCare System Research Articles

Who tests for lead and why? A 10-year analysis of blood lead screening, follow-up and CNS outcomes in a statewide US healthcare system

ObjectivesThis study aims to determine (1) which providers in US healthcare systems order lead tests, why and at what frequency and (2) whether current patient population lead levels are predictive...

Effect measure modification of the association between short-term exposures to PM2.5 and hospitalizations by long-term PM2.5 exposure among a cohort of people with Chronic Obstructive Pulmonary Disease (COPD) in North Carolina, 2002–2015

BackgroundApproximately nine million adults in the United States are living with chronic obstructive pulmonary disease (COPD), and positive associations between short-term air pollution exposure and increased risk of COPD hospitalizations in older adults are consistently reported. We examined the association between short-term PM2.5 exposure and hospitalizations and assessed if there is modification by long-term exposure in a cohort of individuals with COPD.MethodsIn a time-referent case-crossover design, we used a cohort of randomly selected individuals with electronic health records from the University of North Carolina Healthcare System, restricted to patients with a medical encounter coded with a COPD diagnosis from 2004–2016 (n = 520), and estimated ambient PM2.5 concentrations from an ensemble model. Odds ratios and 95% confidence intervals (OR (95%CI)) were estimated with conditional logistic regression for respiratory-related, cardiovascular (CVD), and all-cause hospitalizations. Exposures examined were 0–2 and 0–3 day lags of PM2.5 concentration, adjusting for daily census-tract temperature and humidity, and models were stratified by long-term (annual average) PM2.5 concentration at the median value.ResultsWe observed generally null or low-magnitude negative associations with short-term PM2.5 exposure and respiratory-related (OR per 5 µg/m3 increase in 3-day lag PM2.5: 0.971 (0.885, 1.066)), CVD (2-day lag: 0.976 (0.900, 1.058) and all-cause (3 day lag: 1.003 (0.927, 1.086)) hospitalizations. Associations between short-term PM2.5 exposure and hospitalizations were higher among patients residing in areas with higher levels of annual PM2.5 concentrations (OR per 5 µg/m3 in 3-day lag PM2.5 for all-cause hospitalizations: 1.066 (0.958, 1.185)) than those in areas with lower annual PM2.5 concentrations (OR per 5 µg/m3 in 3-day lag PM2.5 for all-cause hospitalizations: 0.914 (0.804, 1.039)).ConcluisonsDifferences in associations demonstrate that people in areas with higher annual PM2.5 exposure may be associated with higher risk of hospitalization during short-term increases in PM2.5 exposure.

Immediate Open Access: The Good, the Bad, and the Impact on Academic Society Publishing.

Arthritis Care & ResearchEarly View Notes from the Field Immediate Open Access: The Good, the Bad, and the Impact on Academic Society Publishing Amr H. Sawalha, Corresponding Author Amr H. Sawalha [email protected] orcid.org/0000-0002-3884-962X University of Pittsburgh, Pittsburgh, Pennsylvania Address correspondence via email to Amr H. Sawalha, MD, at [email protected].Search for more papers by this authorDaniel H. Solomon, Daniel H. Solomon orcid.org/0000-0001-8202-5428 Brigham and Women's Hospital, Harvard Medical School, Boston, MassachusettsSearch for more papers by this authorKelli D. Allen, Kelli D. Allen orcid.org/0000-0002-3886-0677 University of North Carolina & Durham VA Healthcare System, Durham, North CarolinaSearch for more papers by this authorPatricia Katz, Patricia Katz orcid.org/0000-0002-8146-2519 University of California San Francisco, San Francisco, CaliforniaSearch for more papers by this authorEd Yelin, Ed Yelin orcid.org/0000-0002-9229-9980 University of California San Francisco, San Francisco, CaliforniaSearch for more papers by this author Amr H. Sawalha, Corresponding Author Amr H. Sawalha [email protected] orcid.org/0000-0002-3884-962X University of Pittsburgh, Pittsburgh, Pennsylvania Address correspondence via email to Amr H. Sawalha, MD, at [email protected].Search for more papers by this authorDaniel H. Solomon, Daniel H. Solomon orcid.org/0000-0001-8202-5428 Brigham and Women's Hospital, Harvard Medical School, Boston, MassachusettsSearch for more papers by this authorKelli D. Allen, Kelli D. Allen orcid.org/0000-0002-3886-0677 University of North Carolina & Durham VA Healthcare System, Durham, North CarolinaSearch for more papers by this authorPatricia Katz, Patricia Katz orcid.org/0000-0002-8146-2519 University of California San Francisco, San Francisco, CaliforniaSearch for more papers by this authorEd Yelin, Ed Yelin orcid.org/0000-0002-9229-9980 University of California San Francisco, San Francisco, CaliforniaSearch for more papers by this author First published: 11 May 2023 https://doi.org/10.1002/acr.25131 This article is simultaneously published in Arthritis & Rheumatology and ACR Open Rheumatology. 1Amr H. Sawalha, MD: University of Pittsburgh, Pittsburgh, Pennsylvania; 2Daniel H. Solomon, MD, MPH: Brigham and Women's Hospital, Harvard Medical School, Boston, Massachusetts; 3Kelli D. Allen, PhD: University of North Carolina & Durham VA Healthcare System, Durham, North Carolina; 4Patricia Katz, PhD, Ed Yelin, PhD: University of California San Francisco, San Francisco, California. Read the full textAboutPDF ToolsRequest permissionExport citationAdd to favoritesTrack citation ShareShare Give accessShare full text accessShare full-text accessPlease review our Terms and Conditions of Use and check box below to share full-text version of article.I have read and accept the Wiley Online Library Terms and Conditions of UseShareable LinkUse the link below to share a full-text version of this article with your friends and colleagues. Learn more.Copy URL No abstract is available for this article. Early ViewOnline Version of Record before inclusion in an issue RelatedInformation

Associations between long-term fine particulate matter exposure and hospital procedures in heart failure patients.

Ambient fine particulate matter (PM2.5) contributes to global morbidity and mortality. One way to understand the health effects of PM2.5 is by examining its impact on performed hospital procedures, particularly among those with existing chronic disease. However, such studies are rare. Here, we investigated the associations between annual average PM2.5 and hospital procedures among individuals with heart failure. Using electronic health records from the University of North Carolina Healthcare System, we created a retrospective cohort of 15,979 heart failure patients who had at least one of 53 common (frequency > 10%) procedures. We used daily modeled PM2.5 at 1x1 km resolution to estimate the annual average PM2.5 at the time of heart failure diagnosis. We used quasi-Poisson models to estimate associations between PM2.5 and the number of performed hospital procedures over the follow-up period (12/31/2016 or date of death) while adjusting for age at heart failure diagnosis, race, sex, year of visit, and socioeconomic status. A 1 μg/m3 increase in annual average PM2.5 was associated with increased glycosylated hemoglobin tests (10.8%; 95% confidence interval = 6.56%, 15.1%), prothrombin time tests (15.8%; 95% confidence interval = 9.07%, 22.9%), and stress tests (6.84%; 95% confidence interval = 3.65%, 10.1%). Results were stable under multiple sensitivity analyses. These results suggest that long-term PM2.5 exposure is associated with an increased need for diagnostic testing on heart failure patients. Overall, these associations give a unique lens into patient morbidity and potential drivers of healthcare costs linked to PM2.5 exposure.

Source-Apportioned Particulate Air Pollution and 30-day Readmissions Among Heart Failure Patients

BACKGROUND AND AIM: Fine particulate air pollution (PM2.5) exposure is associated with increased risk of hospital readmissions. This study aims to determine if this association differs according to the attributable source of the PM2.5. METHODS: We used zero-inflated poisson mixed-effects models to associate source-apportioned PM2.5 with the number of 30-day readmissions after HF diagnosis. The study cohort included patients diagnosed with HF who had a hospital visit at a University of North Carolina Healthcare System facility between July 5, 2004, and December 31, 2010. The exposure was the annual average source-apportioned PM2.5 at the date of HF diagnosis. Patient zip code was included as a random effect and models were adjusted for year of diagnosis, sex, race, age, smoking status and 2010 census block group measures of urbanicity, percent receiving public assistance, median income, and median house value. PM2.5 was apportioned into the sources using a Chemical Mass Balance Gas Constrained-Iteration model. Results are presented as the percent change in number of expected readmissions per interquartile range increase in source-apportioned PM2.5 and the associated 95% confidence interval (CI). RESULTS: We observed associations with 30-day readmissions for gasoline (16.89%; 95% CI = 4.82-30.36), diesel (9.87%; 95% CI = 1.72-18.69), and secondary organic carbon (SOC; 20.44%; 95% CI = 8.29-33.95) PM2.5. Associations were greater for study participants in a census block group with an income below the median value. Associations were also greater for Black HF patients compared to white HF patients. CONCLUSIONS: Associations between 30-day readmissions and PM2.5 appear greatest for traffic and SOC-related sources which may be due to a combination of differential toxicity as well as the distribution of these sources amongst HF patients. This abstract does not necessarily represent the views or policies of the US Environmental Protection Agency. KEYWORDS: Heart failure, hospital utilization, particulate matter, air pollution

Chronic Kidney Disease (CKD) and Exposure to Fine Particulate Matter and Ozone: A Cross-Sectional Study Utilizing North Carolina Electronic Healthcare Records

Background and aims: Chronic kidney disease (CKD) affects more than 38 million people in the United States, predominantly those over 65 years of age. While CKD etiology is complex, recent research suggests associations with certain environmental exposures. Additional studies are needed to understand the strength of associations. We examine estimated glomerular filtration rate (eGFR) using a random sample of North Carolina electronic healthcare records (EHRs) and potential relationships with PM₂.₅ and O₃. Methods: Patient data came from a random sample of 7,065 EHRs within the EPA CARES resource, with recorded serum creatinine concentrations. Patients were seen at a University of North Carolina Healthcare System affiliated hospital or clinic from 2004-2017. We estimated eGFR using CKD-EPI equations. PM₂.₅ data comes from a hybrid model using 1x1 km grids and O₃ data from CMAQ 12x12 km grids. Exposures were annual average PM₂.₅ and O₃ based on the creatinine lab test date. We used multiple linear regression to estimate eGFR per IQR increase of PM₂.₅ & O₃. We adjusted for patient sex, race, age, comorbidities, and 2010 census block group measures of sociodemographic and economic factors. Results: Patients averaged 55.3 (SD: 16.2) years of age, with 58.3% female. There were 1,001 patients (14.2%) diagnosed with CKD, identified by ICD-9 & 10 codes. Mean concentrations for the study period of PM₂.₅ and O₃ were 9.92 (IQR: 1.61) µg/m³ and 40.20 (IQR: 2.53) ppb respectively. eGFR decreased 8.45 mL/min/1.73m² (95% CI: 7.92, 8.98) per IQR increase of PM₂.₅ and a more modest decrease of 0.66 mL/min/1.73m² (95% CI: -.08, 1.40) per IQR of O₃. Conclusions: Annual average PM₂.₅, and to a lesser extent O₃, were associated with lower (poorer) eGFR. Future work will examine the relationship between air pollution and onset of CKD and impaired renal function. This abstract does not reflect EPA policy.

Associations Between Hospital Readmissions and Proximity to Brownfields in North Carolina

Background: Contaminated sites known as brownfields contribute to air, water, and land pollution and also prevent the beneficial use of lands by the community. The health effects surrounding brownfields have been understudied which can impair efforts to quantify the benefits of their remediation. Methods: Our study cohort was composed of 21,776 heart failure patients diagnosed between 2004 and 2016 at a University of North Carolina Healthcare System affiliated hospital or clinic. We examined the associations between brownfields and hospital readmissions. We used proximity to the nearest brownfield as our exposure based on the primary residence at the time of heart failure diagnosis. Zero-inflated Poisson models were used to associate distance to the nearest brownfield with readmissions while adjusting for age, race, sex, and socioeconomic status indicators based on the 2010 US Census. In sensitivity analyses we examined log transformed distance, restricted to only those brownfields within 2 km of an individual, and further adjusted for county-level indicators of access to healthcare. Results: A 1 km increase in distance to the nearest brownfield was associated with a 1.29% (95% confidence interval = 0.43-2.14%) decrease in 7-day readmissions as well as a 0.55% decrease (95% confidence interval = 0.20-0.90%) in 30-day readmissions. Associations with 7-day readmissions remained when restricting to brownfields within 2km, in the log-transformed distance models, and after adjusting for county-level indicators of access to healthcare. Associations with 30-day readmissions were substantially attenuated under all sensitivity analyses. Conclusion: Proximity to brownfields is associated with increases in hospital visits and readmissions among heart failure patients. These associations may reflect exposure to chemicals present in the brownfields as well as land disinvestment. Understanding the health risks surrounding brownfields may help communities to better protect their environmental health. This abstract does not necessarily represent the views or policies of the US EPA.

Association between long-term and short-term PM2.5 exposure and hospitalizations among a cohort of COPD patients, North Carolina, 2002-2015

BACKGROUND: Approximately nine million adults in the United States are living with Chronic Obstructive Pulmonary Disease (COPD). Previous research has reported associations between short-term air pollution exposures and increased risk of hospitalization for COPD in older adults. This study examined associations individually between short-term and long-term PM2.5 exposure and hospitalizations, and whether the association between short-term PM2.5 exposure is modified by long-term PM2.5 exposure in a cohort of individuals with COPD. METHODS:The study population consists of a cohort of randomly selected individuals with electronic health records from the University of North Carolina Healthcare System that had a COPD diagnosis from 2002-2015 (n=1,429). Estimated ambient PM2.5 concentrations are from a previously validated ensemble model. Poisson regression was used to estimate the Risk Ratios (RR (95%CI)) of cardiovascular disease (CVD) and all-cause hospitalizations with 0-2 day lags of PM2.5, annual average PM2.5 concentrations, and their interactions. Models were adjusted for individual race, and area-level urbanicity, education, unemployment, and proportion on public assistance. RESULTS: Among the 1,429 COPD patients in our cohort there were 4,302 inpatient hospital visits recorded, including 3,559 CVD-related visits. The mean short-term PM2.5 concentration was 9.62 µg/m3 (IQR 6.77-11.45); the mean long-term PM2.5 concentration was 9.83 µg/m3 (IQR 8.73-10.34). We observed a null association with CVD (RR=1.01 (1.00, 1.02) and a small association with all-cause (RR=1.03 (1.02, 1.04)) hospitalizations per unit increase of short-term PM2.5. There were positive associations between long-term exposure and CVD hospitalizations (RR=1.12 (1.09, 1.16) and all-cause hospitalizations (RR=1.20 (1.17, 1.23)). There was no evidence of interaction between long- and short-term PM2.5 exposures with CVD hospitalizations (RR=1.19 (1.07, 1.32), ICR 0.06) and all-cause hospitalizations (RR=1.21 (1.14, 1.29), ICR -0.02) CONCLUSIONS: Our preliminary results show associations between short- and long-term PM2.5 and CVD and all-cause hospitalizations among COPD patients, but no evidence of interaction between exposure periods.

Associations between PFAS occurrence and multimorbidity as observed in an electronic health record cohort.

Background:Per and polyfluoroalkyl substances (PFAS) are associated with health outcomes ranging from cancer to high cholesterol. However, there has been little examination of how PFAS exposure might impact the development of multiple chronic diseases, known as multimorbidity. Here, we associated the presence of one or more PFAS in water systems serving the zip code of residence with chronic disease and multimorbidity.Methods:We used data from the unregulated contaminant monitoring rule 3 to estimate exposure to PFAS for a random sample of 10,168 patients from the University of North Carolina Healthcare System. The presence of 16 chronic diseases was determined via. their electronic health records. We used a logistic regression model in a cross-sectional study design to associate the presence of one or more PFAS with multimorbidity. Models were adjusted for age, race, sex, smoking status, socioeconomic status, and 20 county-level confounders.Results:There were four PFAS found in water systems that served at least one zip code represented in our patient data: PFOA, PFHpA, PFOS, and PFHxS. Exposure to any PFAS was associated with a odds ratio of 1.25 for multimorbidity (95% confidence interval = 1.09, 1.45). Among the chronic diseases with at least 300 cases, we observed associations with dyslipidemia, hypertension, ischemic heart disease, and osteoporosis.Conclusion:Exposure to PFAS is associated with a range of chronic diseases as well as multimorbidity. Accounting for the joint impacts of PFAS on multiple chronic conditions may give an increasingly clear picture of the public health impacts of PFAS.

Experiences of COVID-19 infection in North Carolina: A qualitative analysis.

It has been demonstrated that marginalized populations across the U.S. have suffered a disproportionate burden of the coronavirus disease 2019 (COVID-19) pandemic, illustrating the role that social determinants of health play in health outcomes. To better understand how these vulnerable and high-risk populations have experienced the pandemic, we conducted a qualitative study to better understand their experiences from diagnosis through recovery. We conducted a qualitative study of patients in a North Carolina healthcare system's registry who tested positive for COVID-19 from March 2020 through February 2021, identified from population-dense outbreaks of COVID-19 (hotspots). We conducted semi-structured phone interviews in English or Spanish, based on patient preference, with trained bilingual study personnel. Each interview was evaluated using a combination of deductive and inductive content analysis to determine prevalent themes related to COVID-19 knowledge, diagnosis, disease experience, and long-term impacts. The 10 patients interviewed from our COVID-19 hotspot clusters were of equal distribution by sex, predominantly Black (70%), aged 22-70 years (IQR 45-62 years), and more frequently publicly insured (50% Medicaid/Medicare, vs 30% uninsured, vs 20% private insurance). Major themes identified included prior knowledge of COVID-19 and patient perceptions of their personal risk, the testing process in numerous settings, the process of quarantining at home after a positive diagnosis, the experience of receiving medical care during their illness, and difficulties with long-term recovery. Our findings suggest areas for targeted interventions to reduce COVID-19 transmission in these high-risk communities, as well as improve the patient experience throughout the COVID-19 illness course.

Associations between short-term exposure to PM2.5 and cardiomyocyte injury in myocardial infarction survivors in North Carolina

ObjectiveShort-term ambient fine particulate matter (PM2.5) is associated with adverse cardiovascular events including myocardial infarction (MI). However, few studies have examined associations between PM2.5 and subclinical cardiomyocyte damage outside of...

Robotic hepatectomy is a feasible minimally-invasive modality for experienced liver surgeons

Presenter: Benjamin Motz MD | Carolinas HealthCare System Background: Modern liver surgery has seen increasing success over the last several decades due to advances in perioperative anesthesia care and surgical technique. One such improvement has been the development of minimally-invasive approaches to hepatectomy, which is associated with lower surgical morbidity, shorter hospital length of stay (LOS) and patient-reported outcomes. Although laparoscopic liver resection continues to be associated with a steep learning curve, more recently, the development of robotic approaches has reduced the technical limitations of these complex procedures. Herein we present outcomes and learning curve experience following the development of a high-volume robotic liver surgery program at a major hepatobiliary surgery center. Methods: A retrospectively maintained registry of robotic hepatectomies performed at a single high-volume tertiary care center was queried for all procedures performed between September 1, 2012 when the robotic hepatobiliary program was started and September 30, 2020. All operations were performed by either of two surgeons with significant experience in hepatobiliary surgery: JBM and DV. DV began his robotic liver surgery practice in the midst of the study period. Patient charts were retrospectively reviewed and data was collected, including patient demographics, preoperative characteristics, intraoperative details, and postoperative outcomes. Once cases were identified, they were then divided into five epochs of equal size (n=53) based on date of surgery in order to identify trends in outcomes related to experience. Statistical comparisons were performed using ANOVA or χ2 test, as appropriate. All data collection and chart review was approved by the Institutional Review Board. Results: During the study period, 265 patients underwent robotic hepatectomy. 152 (57.4%) were female, and average age was 58 years. 101 (38.1%) were major hepatectomies involving ≥3 segments. Case volume has steadily increased over time, before decreasing slightly in 2020 (Figure 1). Intraoperative outcomes remained similar between the five epochs. Average EBL over the study period was 435mL for all hepatectomies, 555 mL for major hepatectomies, 301cc for minor hepatectomies, and there was no significant improvement or decrement over time (p=0.57). Average operative time was 196 minutes, and there was no significant difference in operative time between epochs (p=0.90). Postoperative outcomes also remained similar over time. Median hospital LOS was 3 days and has remained stable between the five epochs (p=0.91). Finally, 54 patients (20.4%) suffered a complication, with the majority being minor and only 12 major complications (Clavien-Dindo Grade ≥3). The highest Clavien-Dindo Complication Grade within 30 days of surgery was not significantly different between epochs (p=0.23). Conclusion: Robotic hepatectomy is a safe and feasible option for most liver resections, including complex major resections. Over the development period of a high-volume robotic liver surgery program, there were no significant changes in major intraoperative and postoperative outcomes over time, despite the fact that one of the surgeons was going through his learning curve during the study period. This supports the hypothesis that robotic techniques may reduce the learning curve for minimally-invasive liver surgery, opening minimally-invasive options for experienced hepatobiliary surgeons who typically perform open liver surgery.

Laparoscopic-assisted microwave ablations for hepatic malignancies: a 10-year retrospective database review

Presenter: Joshua Davis MD | Carolinas HealthCare System Background: Microwave ablation (MWA) is a safe and established method for achieving locoregional control of both primary and secondary hepatic malignancies. This technology has expanded the indications for liver-directed therapy for a broad range of cancers, but long-term data on oncologic outcomes are lacking. Although CT-guided microwave ablation is the most commonly used localization modality, a laparoscopic approach utilizing intraoperative ultrasound offers the additional benefit of evaluating the abdominal cavity and liver for unidentified tumors with minimal added morbidity. This study reports the 10-year use of laparoscopic-assisted microwave ablation in a high-volume hepatobiliary practice at a large tertiary care center. Methods: A prospectively maintained registry of laparoscopic-assisted MWAs performed in a single high-volume center was retrospectively queried for all patients who underwent laparoscopic MWA from January 1, 2010 through December 31, 2019. Patient demographics, intraoperative details, and postoperative outcomes were collected. Follow up imaging was analyzed to determine rates of incomplete ablation as identified on the first computed tomography scan performed postoperatively (typically within 4-6 weeks), and locoregional recurrence. Comparative statistics was performed as appropriate. Results: Overall, 777 separate ablation procedures for a total of 1317 lesions within the 10-year time interval was performed. Patients were 33% female; 76% Caucasian, and 18% African American. The majority of patients had an ASA score of III (76%, n=540) and average BMI of 29kg/m2. There was an average number of 1.7 lesions ablated per patient surgery with an average lesion diameter of 2.4 cm (range: < 0.5 to 8.0 cm). Hepatocellular carcinoma (58%, n=397) was the most frequent lesion ablated, followed by colorectal metastasis (20%, n=135). Patients had a median intraoperative blood loss of 50 mL and the majority of patients had no major 30-day morbidity or mortality (96% of patients with highest Clavien-Dindo classification < 3). Median length-of-stay was 1 hospital day (range 0 to 46 days). At postoperative follow up, 5.7% of patients had an incomplete ablation. The local recurrence rate was 18.2% with median time to local recurrence of 8.9 months (range: 0.7 to 90 months). The regional recurrence rate was 48.9% with median time to regional recurrence of 7.5 months (range: 0.5 to 103 months). Conclusion: Laparoscopic MWA is a viable treatment for locoregional control of both primary and secondary liver lesions. This surgery is performed in a minimally invasive fashion with assistance of intraoperative ultrasound to aid in visualization of subclinical lesions. This is an established safe technique with low associated morbidity and mortality and short associated length of stay. Additionally, locoregional recurrence in this study is comparable to historical controls of liver resection, suggesting comparable efficacy with this approach.

Comparing laparoscopic hepatectomy and propensity-matched open hepatectomy: a single center investigation

Presenter: Jordan Robinson MD, MPH | Carolinas HealthCare System Background: The benefits associated with minimally invasive surgery are well-documented in the surgical literature; however, the degree of benefit in hepatic resection has yet to be precisely delineated. Patient- and disease-related factors such as comorbid conditions and the nature and extent of a patient’s disease confound the clinical impact of laparoscopy when compared with open hepatectomy. This study aims to investigate the effects of minimally invasive hepatic resection through comparison of outcomes with a propensity-matched open surgical cohort. Methods: A retrospectively maintained database of hepatectomies performed at our institution between January 2006 and December 2019 was queried to identify all laparoscopic and open cases. Data collected includes patient demographics, comorbidities, diagnosis, as well as intraoperative and post-operative outcomes. Open hepatectomies were propensity-matched based on factors influencing surgical risk including age, gender, smoking history, the presence of type II diabetes, history of prior surgery, benign vs. malignant disease and the extent of hepatic resection. Outcomes for the two cohorts were subsequently assessed and included operative estimated blood loss (EBL), hospital length of stay (LOS), ICU LOS, post-operative complications and mortality. Results: A total of 731 hepatectomies were performed during the study period. In all, 273 open hepatectomies and 458 laparoscopic hepatectomies were performed. After propensity matching (nearest neighbor algorithm) there were a total of 251 laparoscopic and 253 open hepatectomies. The conversion to open rate for laparoscopic hepatectomies was 9.6%. Length of stay was notably longer for open hepatectomy with a mean of 10.51 (+/- 10.25) compared to 5.78 (+/- 6.73; p<0.0001). Mortality (29.1% vs 51.4%; p<0.0001) and complication rates (38.6% vs. 63.2%; p<0.0001) were also notably greater in the open group. Major resection (54 (21.5%) vs 101 (39.9%)) and disease recurrence (29 (11.6%) vs. 46 (18.2%)) were more likely with open surgery. Of note, comparison of laparoscopic cohorts including and excluding conversion to open noted a higher proportion of major resection (41(18.7%) versus 54 (21.5%)) in the cohort unmatched for open conversion (0 conversions vs. 24 conversions (9.6%)). Conclusion: Open surgery is more likely for patients undergoing major hepatic resection which is associated with longer length of stay, and greater frequency of complications, disease recurrence and mortality. Laparoscopic cases overall are associated with decreased LOS, complications and mortality compared with open surgery. In laparoscopic cases, major resection markedly increased the rate of open conversion and predominantly account for the observed differences in outcomes. The sample size of laparoscopic major resections was inadequate to assess the magnitude of effect for conversion to open for laparoscopic cases and major resection in the open and laparoscopic cohorts.

Use Of Deep Learning To Predict Tumor Response To Neoadjuvant Therapy In Pancreatic Adenocarcinoma: Pure And Hybrid Modelling

Presenter: Michael Watson MD | Carolinas HealthCare System Background: Neoadjuvant chemotherapy is being increasingly utilized for the treatment of pancreatic adenocarcinoma and is associated with improved survival. However, determining response to neoadjuvant therapy is difficult with preoperative imaging. New advances in artificial intelligence and deep learning techniques allow for novel analysis of images. Our hypothesis is that a deep learning model can be used to predict tumor response to neoadjuvant therapy. Methods: We identified patients with pancreatic adenocarcinoma undergoing neoadjuvant therapy that underwent planned Whipple’s procedure between November 2009 and January 2018. Surgical pathology report was reviewed to determine tumor treatment effect, with grade 0-2 defined as treatment response and grade 3-4 defined as no treatment response. Only patients with adequate imaging and pathologically determined treatment effect were included for analysis. Computed tomography (CT) scans after neoadjuvant therapy and prior to surgery were reviewed. All axial slices of the tumor were obtained with 5mm spacing. These slices were converted to an image file for analysis. These images were used to create a deep learning model to predict tumor response to neoadjuvant therapy. Image augmentation using Gaussian blurring was used to increase the number of images for model creation and validation. Accuracy of the model was determined by area under the curve (AOC) and Brier statistics. Results: 91 patients appropriate for image analysis (had appropriate CT imaging and tumor treatment effect documented by pathology report) were identified. Images were divided based on treatment response (333 images) or no treatment response (443 images) and image augmentation was used to artificially increase the image sample size. A “training set” of images consisting of 80% of patient images were used for creation of the deep learning model. Images from 73 patients were run through a 5-layer Convolutional Neural Network (CNN) with a flattened layer and two fully connected layers (LeNet model). Model initial training and validation accuracy were achieved to 100% with a loss function of < 0.02. The “testing” set of images consisted of images from 20% of patients. Images from the remaining 16 patients were then processed through the internally validated model. The model had an AOC of 0.7383 (p<0.001) and Brier statistic of 0.2347. Using the same procedure, a model was created with patients found to have greater than 10% reduction in CA 19-9 level after neoadjuvant chemotherapy (n=58). This improved the model with AOC improved to 0.7846 (p<0.0001) with Brier statistic of 0.1735. Conclusion: A deep learning model can be used to predict response to neoadjuvant chemotherapy for patients with pancreatic adenocarcinoma. Model accuracy is further improved when combined with another indicator of tumor response (hybrid model). With further model improvement and increased patients for analysis, an application can be created to predict tumor response to neoadjuvant chemotherapy so as to safely obviate the need for subjective clinical interpretation.

The gallbladder that nobody wants: A prediction tool to assess the difficulty of cholecystectomy

Presenter: Maria Baimas-George MD, MPH | Carolinas HealthCare System Background: Cholecystectomy is a common procedure yet can have significantly varied outcomes based on patient physiology and comorbidities. As providers move towards fee-for-value models, it is important to risk stratify patients for appropriate referral and cost prediction. We analyzed differences in comorbidities, outcomes, and cost associated with cholecystectomy by acute care (ACS) versus hepatobiliary (HPB) surgical practices to create a predictive tool for difficult gallbladders. Methods: Patients were retrospectively identified through procedure codes from 2008 through 2015. Exclusion criteria included: (1) cholecystectomy performed at time of or as part of another procedure; (2) abdominal trauma requiring cholecystectomy; (3) ICU admission or preoperative vasopressor requirement. Cholecystectomies referred to HPB were identified and matched to general surgery cholecystectomies by demographics including sex, age, and surgery date. A regression model was created and trained to create a scoring system of independent risk factors. Results: There were 1160 cholecystectomies performed of which 736 were subjectively referred to HPB surgeons from general surgeons. Of these, 122 HPB cholecystectomies met inclusion criteria and were case matched to 126 general surgery cholecystectomies. In the HPB subset, there was higher burden of comorbid disease, different indications for surgery, and a significantly lower projected 10-year survival (87.4 % ACS vs. 68.5% HPB, p < .0001). Median surgical and hospital lengths of stay were longer in HPB patients (1 vs. 4 days, p < .0001; 2 vs. 5 days, p < .0001) as were 30-day and 90-day readmission rates (5.6% vs. 13.1%, p = .040; 7.9% vs. 20.5%, p = .005 respectively). Median cost was higher in the HPB population including operative supply cost (969.42 vs. 1920.66, p < .0001) and total cost of care (7340.66 vs. 19338.05, p < .0001). A predictive scoring system for a difficult gallbladder was constructed using five factors including history of liver disease, ASA, and alkaline phosphatase level. A score of ≥ 5 was determined to be the best cut-off value for the high-risk difficult gallbladder group and a phone application was created to assist providers in quick identification of such cases. Conclusion: Although often regarded as a simple procedure, a cholecystectomy in a complicated patient can become anything but, leading to a difficult operation, significantly longer hospital stays, and higher costs. The hybrid use of an objective scoring system along with a subjective clinical impression can help providers appropriately plan for high-risk patients with management of pre-operative patient expectations and consideration of an HPB consultation or transfer. This predictive tool highlights a population that should have a different reimbursement tool as the use of procedure codes to appropriately explain disparities in outcomes and costs is insufficient. The incorporation of comorbid disease and risk factors with such clinical metrics needs to be addressed for operative planning as well as categorization and reimbursement of surgical practices.

Effect of hepatic cyst size on postoperative outcomes: size is irrelevant

Presenter: Michael Watson MD | Carolinas HealthCare System Background: Simple hepatic cysts may present with a wide range of sizes and symptomatology. There is no formal diagnostic schema for simple hepatic cysts, however previous studies have used various cut-offs to define a “giant” hepatic cyst. There have been few studies of the clinical implications of simple hepatic cyst size. Methods: Patients undergoing treatment of simple hepatic cysts at our institution between January 2008 to July 2019 were identified by retrospective record review. Demographic data, preoperative symptoms, cyst characteristics, surgical details, and postoperative outcomes were determined from retrospective record review. Cyst size was defined as the greatest measured diameter on axial imaging. If multiple cysts were present, the largest cyst size was used for analysis. Linear regression for continuous variables and logistic regression for categorical variables was used to determine correlations between cyst size and postoperative outcomes. Results: 120 patients with simple hepatic cysts were identified and included in our analysis. The average cyst size was 12.0±5.5 cm, 85.8% of patients were symptomatic, and pain (83.3%) was the predominant symptom. 114 (95%) of patients were treated with minimally invasive surgery, mean estimated blood loss (EBL) was 193±413 mL, mean operative time was 123±59 min, conversion to open rate was 1.8%, and the rate of simultaneous hepatectomy was 23.3%. Rate of reoperation for recurrence was 4.2%. Regression analysis considering size as a continuous variable demonstrated that cyst size was weakly correlated with operative time (R2=0.0683; p=0.0056), however it was not associated with EBL (p=0.4874), need for hepatectomy (p=0.4094), length of stay (p=0.1305), complications (p=0.2943), Clavien-Dindo grade of complications (p=0.4072), or reoperation for recurrence (p=0.2488). Regression analysis was then carried out to determine correlation between cyst percentile size and postoperative outcomes. Using cutoffs at the 10th, 25th, 50th, 75th, and 90th percentiles, there was no correlation between these cutoffs and operative time. There were no correlations between any percentile cut-offs and EBL, need for hepatectomy, length of stay, complications, or Clavien-Dindo grade of complications Conclusion: For surgical therapy of simple hepatic cysts there is a weak correlation with cyst size and operative time, however there is no correlation between clinically relevant patient outcomes and cyst size. This demonstrates that for patients with simple hepatic cysts, the size of the cyst does not predict patient outcomes. Traditional definitions of “giant” simple hepatic cysts are irrelevant.

CT scan is more accurate in predicting viable disease after microwave ablation then tace in patients undergoing locoregional treatment in hepatocellular carcinoma

Presenter: Maria Baimas-George MD, MPH | Carolinas HealthCare System Background: The use of locoregional therapies such as laparoscopic tumor ablation or intra-arterial chemotherapy in hepatocellular carcinoma (HCC) to achieve local control, reduce transplant list dropout, downstage, or reduce recurrence is common practice. Assessing tumor response after treatment, however, remains difficult and unreliable without an explant specimen. We sought to determine the accuracy of imaging in treatment success after locoregional therapy in HCC. Methods: Patients with HCC who underwent either microwave ablation (MWA) and/or transarterial chemoembolization (TACE) prior to hepatectomy or transplantation were retrospectively identified from 2008 to 2018. Computed tomography (CT) scans after treatment and prior to resection were evaluated for characteristics consistent with viable tumor. The explant specimen was identified, and tumor response was assessed through pathologic evaluation. The accuracy of the pre-operative CT scan readings was compared with the pathologic findings of the explant specimen. Results: Ninety-one patients were identified who underwent locoregional treatment prior to resection or transplant. Of these, 14 patients underwent laparoscopic MWA, 46 underwent TACE, and 31 underwent both treatments in a neoadjuvant fashion. The ability of CT imaging to predict tumor viability was assessed by treatment modality. For MWA, CT scans demonstrated a 75% positive predictive value (PPV) and a 67% negative predictive value (NPV). This correlated with a 50% sensitivity and 86% specificity. For TACE, CT scans demonstrated an 87% positive predictive value (PPV) and a 25% negative predictive value (NPV). This correlated with an 82% sensitivity and 33% specificity. For MWA-TACE, CT scans demonstrated an 80% positive predictive value (PPV) and a 39% negative predictive value (NPV). This correlated with a 42% sensitivity and 78% specificity. Conclusion: Imaging to guide response to locoregional therapies in HCC has variable accuracy. There is a high probability of finding tumor given a positive CT scan after all three treatment modalities, whereas a negative CT scan is unreliable. Specifically, CT imaging after MWA demonstrates a high specificity and will likely rule out any residual tumor and CT imaging after TACE is highly sensitive and will likely find residual tumor.

MELD score does not increase after hepatectomy in cirrhotic patients

Presenter: Maria Baimas-George MD, MPH | Carolinas HealthCare System Background: With prevalence of hepatic cirrhosis increasing, there is a subsequent increase in patients needing liver resections as it is the mainstay of treatment for many hepatobiliary neoplasms. However, because of the reduced ability of a cirrhotic liver to tolerate surgical insult and mass reduction, resection requires consideration between oncologic benefit and risk of impaired hepatic reserve as liver function is expected to deteriorate post-operatively. To quantify this risk, we evaluated change in MELD and MELD-Na scores in cirrhotic patients with hepatocellular carcinoma (HCC) after liver resection. Methods: A retrospective review identified all cirrhotic patients with HCC who underwent a hepatic resection between 2010 to 2017 at a single tertiary-care institute. The MELD scores for each patient were calculated pre-operatively, immediately post-operative, and the first time point 3 months after surgery. Patients were stratified by major (>3 segment resection) and minor hepatectomy and by MELD/MELD-Na score (MELD: < 10 and 10; MELD-Na < 12 and 12) and the change after 3 months was evaluated. Results: Thirty-five patients were identified. Of these, 20 patients underwent a minor hepatectomy and 15 underwent a major hepatectomy. In the minor hepatectomy cohort, low-MELD patients (n=14) demonstrated an increase of 0.6 in MELD score 3 months post-operatively whereas the high-MELD cohort (n=6) demonstrated a decrease of 0.9. When stratified by MELD-Na, low-MELD-Na (n=14) demonstrated an increase of 0.2 whereas the high-MELD-Na (n=6) had a decrease of 1.0 after 3 months. In the major hepatectomy cohort, low-MELD patients (n=13) demonstrated an increase in MELD of 3.9 whereas high-MELD (n=2) had a decrease of 0.5 after 3 months. When stratified by MELD-Na, low-MELD-Na (n=9) demonstrated an increase of 4.9 whereas high-MELD-Na (n=6) had a decrease of 1.0. Conclusion: Liver function is expected to deteriorate post-operatively in cirrhotic patients undergoing hepatic resection. We demonstrate that MELD and MELD-Na scores either stay the same or mildly improve after 3 months post-operatively in patients undergoing minor hepatectomies, suggesting that either there is a adequate hepatic reserve present or the liver may repopuluate healthy hepatocytes.

The aberrant common hepatic artery identified during a Whipple; lessons learned from a difficult case

Presenter: Michael Genz MD | Carolinas HealthCare System Background: We report a case of a replaced common hepatic artery originating from the superior mesenteric artery with an anterior course identified during a pylorus preserving pancreaticoduodenectomy. The overall incidence of celiac vascular anatomic aberrancies ranges from 22% - 48%. To our knowledge, there has been only one case report describing this specific anatomic variant that was identified during a planned pancreaticoduodenectomy. Methods: Our patient is a 56-year-old male who presented with obstructive jaundice and found to have a 3cm mass in the head of the pancreas that was found to be adenocarcinoma on fine needle aspiration. He was taken to surgery for a planned open pancreaticoduodenectomy. Intra-operatively, the common hepatic artery was found to be absent from its usual location, along the superior border of the pancreas originating from the celiac trunk. Instead, we identified the common hepatic artery taking off from the superior mesenteric artery, travelling just anterior to the neck of the pancreas and bifurcating into the right and left hepatic arteries at the superior border of the pancreas. This was confirmed on intraoperative ultrasound. The location of the common hepatic artery was found to be sitting anatomically in the area usually occupied by the gastroduodenal artery. We were safely able to dissect the pancreatic head and neck off the common hepatic artery without injury. We then completed the reconstruction, performing a duct-to-duct pancreaticojejunostomy, hepaticojejunostomy and duodenojejunostomy. Results: Cross sectional imaging on post-operative day six demonstrated expected contrast enhancement of the common hepatic, right and left hepatic arteries, confirming no injury had been made during the case. He was eventually discharged on post-operative day elven after resolution of a post-operative ileus. Conclusion: We feel this case stresses the importance of performing a thorough dissection prior to ligating named vessels during a pancreaticoduodenectomy. Failure to correctly identify this type of vascular aberrancy could lead to a disastrous outcome for a patient with this variant.