Using Natural Language Processing to Classify Serious Illness Communication with Oncology Patients

One core measure of healthcare quality set forth by the Institute of Medicine is whether care decisions match patient goals. High-quality "serious illness communication" about patient goals and prognosis is required to support patient-centered decision-making, however current methods are not sensitive enough to measure the quality of this communication or determine whether care delivered matches patient priorities. Natural language processing offers an efficient method for identification and evaluation of documented serious illness communication, which could serve as the basis for future quality metrics in oncology and other forms of serious illness. In this study, we trained NLP algorithms to identify and characterize serious illness communication with oncology patients.

603. Evaluating the Effects of Centers for Medicare & Medicaid Services Sepsis Core Measure in a Community Hospital

Background Sepsis is the leading cause of morbidity and mortality in hospitals, accounting for 30% of deaths in the emergency department. In 2001, Rivers et al. found that early goal-directed therapy (EGDT) led to significant mortality benefits, which ultimately prompted United States Centers for Medicare and Medicaid Services (CMS) to mandate EGDT in hospitals through its implementation of sepsis core measures. CMS core measures are intended to facilitate the broad implementation of evidence-based treatment standards, and while voluntary, non-compliance is associated with negative consequences to both quality and financial metrics for participating hospitals. However, while quality measures are implemented to ultimately improve patient care, its effects on the healthcare system can also include negative unanticipated consequences. This study seeks to characterize the effect of the CMS sepsis core measure on sepsis identification, antimicrobial utilization, and nd specific prescribing patterns. Methods This is a retrospective cohort review of 175 randomly selected patients greater than and equal to 18 years of age with admitting diagnosis of sepsis, severe sepsis, and septic shock from January 2013 to December 2018. Medical charts were reviewed for relevant data. Results Comparing ED antibiotic prescribing patterns between pre-and post-Sepsis CMS Core Measures, there was no statistical difference in total antibiotics usage and the initiation of broad antibiotics. There was a decreased time to the first antibiotic, an increase in receiving Normal Saline boluses post-Sepsis CMS Core Measures. Conclusion 1. No significant changes were seen in ED antibiotic prescribing behaviors with regard to volume and spectrum 2. ED time to antibiotic administration was significantly faster after the implementation of CMS Core Measures. Also, there was a significant positive shift in time to fluid bolus, fluid selection, and fluid volume 3. Significantly decreased ICU length of stay after implementation of CMS Core Measures possibly associated with above behavior changes 4. No outcomes benefits (mortality, hospital length of stay) realized after implementation of CMS Core Measures Disclosures All Authors: No reported disclosures

Impact of the Centers for Medicare and Medicaid Services Sepsis Core Measure on Antibiotic Use

Background The Centers for Medicare and Medicaid Services (CMS) implemented a core measure sepsis (SEP-1) bundle in 2015. One element was initiation of broad-spectrum antibiotics within 3 hours of diagnosis. The policy has the potential to increase antibiotic use and Clostridioides difficile infection (CDI). We evaluated the impact of SEP-1 implementation on broad-spectrum antibiotic use and CDI occurrence rates. Methods Monthly adult antibiotic data for 4 antibiotic categories (surgical prophylaxis, broad-spectrum for community-acquired infections, broad-spectrum for hospital-onset/multidrug-resistant [MDR] organisms, and anti–methicillin-resistant Staphylococcus aureus [MRSA]) from 111 hospitals participating in the Clinical Data Base Resource Manager were evaluated in periods before (October 2014–September 2015) and after (October 2015–June 2017) policy implementation. Interrupted time series analyses, using negative binomial regression, evaluated changes in antibiotic category use and CDI rates. Results At the hospital level, there was an immediate increase in the level of broad-spectrum agents for hospital-onset/MDR organisms (+2.3%, P = .0375) as well as a long-term increase in trend (+0.4% per month, P = .0273). There was also an immediate increase in level of overall antibiotic use (+1.4%, P = .0293). CDI rates unexpectedly decreased at the time of SEP-1 implementation. When analyses were limited to patients with sepsis, there was a significant level increase in use of all antibiotic categories at the time of SEP-1 implementation. Conclusions SEP-1 implementation was associated with immediate and long-term increases in broad-spectrum hospital-onset/MDR organism antibiotics. Antimicrobial stewardship programs should evaluate sepsis treatment for opportunities to de-escalate broad therapy as indicated.

Pharmacist-led transitions-of-care program reduces 30-day readmission rates for Medicare patients in a large health system

Purpose This report describes the growth and development of the Pharmacy Transitions of Care (PTOC) program at a Florida health system and examines its impact on 30-day readmission rates for Medicare core-measure patients. Summary BayCare Health System is a large not-for-profit community health system with 15 hospitals in central Florida. In 2015, the PTOC program was developed to integrate 2 pharmacists into the transitions-of-care space to reduce readmissions, enhance patient care, and improve medication safety. The PTOC program focuses on traditional Medicare beneficiaries 65 years of age or older with the goal of preventing 30-day readmissions. The service model includes integration of a pharmacist into the discharge medication reconciliation process, as well as postacute care telephonic follow-up. Data and outcomes have been carefully tracked since program inception and consistently demonstrate a reduction in 30-day readmissions, with a 63% relative risk reduction during the beginning phases of the program and a ratio of observed to expected readmissions of 0.77. As a result, in less than 3 years the PTOC program has grown from 2 to 23 pharmacists and is a key component of BayCare Health System’s patient care strategy. Conclusion Medication reconciliation, clinical interventions, and patient education by pharmacists after hospital discharge reduced 30-day readmission rates for Medicare core-measure patients across a large health system. The adaptability of this program to other health systems and hospitals of varying size to achieve similar outcomes is valuable to share with the profession.

Abstract WP456: Using An Electronic Dashboard To Improve Communication On Stroke Core Measure Compliance

Background and Purpose: Several years ago, an excel spreadsheet was created to improve real time communication about stroke core measure compliance to all the inpatient units. This daily spreadsheet was sent to the nursing leadership to be shared with the nurse or MD to ensure the stroke orders were carried out and patient education provided. The process to review each chart daily for compliance with the core measures, enter it into the spreadsheet, send it to the nursing leadership daily took approximately 3 hours or more a day. This was not an efficient use of time. We decided to create an electronic dashboard that would extract all of the stroke core measure information from the electronic patient record automatically, eliminating manual abstraction, thus making the process more efficient while allowing effective communication throughout the organization. Methods: Our IT department, working alongside our Stroke Care Team, developed data extraction processes pertinent to Stroke Core Measures to allow for an hourly extraction into a Tableau dashboard. This software allows for logic to be applied to the various elements to evaluate the successful completion of each core measure for patients currently admitted to our institution for Stroke-related diagnoses and conditions. Results: The key stake holders for each unit review the dashboard which is refreshed hourly to identify areas that have not met the measure. After the review, the leadership implements a plan to improve attainment for these measures. The dashboard has also allowed for correction of documentation in real time but has limitations because it depends on the data within the electronic medical record and the programing to extract that data. The team communicates with our IT department to correct the issues with the data displayed in tableau. Conclusions: The tableau dashboard is an effective and efficient tool providing a snapshot of the current attainment of the primary and comprehensive stroke core measures. The implementation of this dashboard has increased awareness of the stroke core measure compliance and has improved communication among the clinical staff.

Abstract TP362: Stroke Order Set Utilization Improves Performance Measures and Decreases Length of Stay

Introduction: Stroke quality programs aim to ensure patients receive evidence based care by measuring patient outcomes and adherence to core measures. Site specific stroke order sets embedded in EHRs augment core measure performance. The true benefit of stroke order sets on core measure performance is still an open question. We hypothesize that utilizing stroke specific order sets increases compliance to core measures and subsequently decrease length of stay (LOS) and readmissions. Methods: A retrospective cohort study was conducted, and included 1095 stroke patients discharged between May 1, 2017 and April 30, 2018. Hospital data was extracted from The Joint Commission stroke registry and supplemented with administrative data. The primary outcome was core measure compliance and was analyzed using Chi-square and Cochran-Mantel-Haenszel tests. Results: The majority of stroke patients (1009, 92%) had a stroke admission order set. Between the order set and non-order set groups there were significant differences in age ( p =.03), stroke type ( p <.001), and EHR system ( p =0.002). The order set group had a marginal decrease in LOS (days) compared with the non-order set group, 3 and 3.9, respectively ( p =.06). Unplanned readmissions within 30-days did not differ between groups ( p =.16). For ischemic stroke, phase specific order set usage showed significantly higher core measure compliance: venous thromboembolism prophylaxis (STK1) (94.0% vs 6.0%, p =.01), antithrombotic by end of hospital day two (STK5) (96.9% vs 73.3%, p <.001), discharged on statin medication (STK6) (99.8% vs 97.5%, p =.006), stroke education (93.3% vs 47.1%, p <.001) (STK8), and national institute of health stroke scale within 12 hours of arrival (CSTK1) (95.8% vs 44.1%, p <.001). The hemorrhagic stroke population showed no significant differences between order set usage and core measure compliance. Conclusion: Use of stroke specific order sets decreased overall LOS and increased compliance to STK1, STK5, STK6, STK8, and CSTK1 core measures.