Machine learning for risk prediction of acute coronary syndrome.

Acute coronary syndrome (ACS) accounts for 1.36 million hospitalizations and billions of dollars in costs in the United States alone. A major challenge to diagnosing and treating patients with suspected ACS is the significant symptom overlap between patients with and without ACS. There is a high cost to over- and under-treatment. Guidelines recommend early risk stratification of patients, but many tools lack sufficient accuracy for use in clinical practice. Prognostic indices often misrepresent clinical populations and rely on curated data. We used random forest and elastic net on 20,078 deidentified records with significant missing and noisy values to develop models that outperform existing ACS risk prediction tools. We found that the random forest (AUC = 0.848) significantly outperformed elastic net (AUC=0.818), ridge regression (AUC = 0.810), and the TIMI (AUC = 0.745) and GRACE (AUC = 0.623) scores. Our findings show that random forest applied to noisy and sparse data can perform on par with previously developed scoring metrics.

Adopting real-time surveillance dashboards as a component of an enterprisewide medication safety strategy.

BACKGROUND: High-alert medications are frequently responsible for adverse drug events and present significant hazards to inpatients, despite technical improvements in the way they are ordered, dispensed, and administered. METHODS: A real-time surveillance application was designed and implemented to enable pharmacy review of high-alert medication orders to complement existing computerized provider order entry and integrated clinical decision support systems in a tertiary care hospital. The surveillance tool integrated real-time data from multiple clinical systems and applied logical criteria to highlight potentially high-risk scenarios. Use of the surveillance system for adult inpatients was analyzed for warfarin, heparin and enoxaparin, and aminoglycoside antibiotics. RESULTS: Among 28,929 hospitalizations during the study period, patients eligible to appear on a dashboard included 2224 exposed to warfarin, 8383 to heparin or enoxaparin, and 893 to aminoglycosides. Clinical pharmacists reviewed the warfarin and aminoglycoside dashboards during 100% of the days in the study period-and the heparinlenoxaparin dashboard during 71% of the days. Displayed alert conditions ranged from common events, such as 55% of patients receiving aminoglycosides were missing a baseline creatinine, to rare events, such as 0.1% of patients exposed to heparin were given a bolus greater than 10,000 units. On the basis of interpharmacist communication and electronic medical record notes recorded within the dashboards, interventions to prevent further patient harm were frequent. CONCLUSIONS: Even in an environment with sophisticated computerized provider order entry and clinical decision support systems, real-time pharmacy surveillance of high-alert medications provides an important platform for intercepting medication errors and optimizing therapy.

Acute coronary syndrome pathways: alignment with a bundled care reimbursement model.

Changes in public policy, population demographics, and market dynamics are spurring innovative approaches to value-based care. Annually, hospitalizations for Acute Coronary Syndromes (ACS) account for a substantial proportion of healthcare expenditures in the United States. Vanderbilt University Medical Center has developed a framework to deliver comprehensive care incorporating inpatient and outpatient care teams for patients with ACS under an episode-based, bundled reimbursement model for a term of 6 months. As such, a value-oriented pathway was created with the goals of (1) optimizing patient outcomes following ACS; (2) minimizing complications from the treatment of ACS; and (3) reducing costs of healthcare related to the treatment of ACS. In a tertiary care academic medical system receiving patients from multiple facilities involving multiple providers, standardization of care by using practice guidelines and evidence-based data coupled with a robust computerized provider order entry system provides a unique opportunity to produce a "best practice" algorithm for treating patients presenting with ACS. Presented in this study are in-hospital and postdischarge care pathways for treating a diverse group of patients presenting with ACS to our institution.

Informatics and evidence-based medicine: prescription for success.

This article reports on the experience of one organization between 2004 and 2009 to develop an effective people-process-technology system to better manage the quality of health care. The creation of this system started with creating a strategic plan for quality and then establishing a structure to implement the plan. The next phase consisted of establishing a number of simultaneous steps that ranged from identifying and leveraging the appropriate informatics tools to the oversight process, and from the implementation team to strategies for working with clinical groups. The outcome as of 2009 is a well established evidence-based quality process and team in place. There are over 450 evidence-based medicine quality sets. More than 52% of all patients are admitted on quality evidence-based medicine pathways and protocols. This article reflects a successful prescription for combining informatics and evidence-based medicine to improve the quality of health care.

Does the specific time of day used to capture data on ventilator-days have an impact on the documented rates of ventilator-associated pneumonia?

Definitions of ventilator-associated pneumonia do not note a preferred daily time for obtaining denominator data. We examined collecting data on the number of ventilator-days at different times of day in 7 intensive care units. Rates of ventilator-associated pneumonia did not significantly differ when denominator data were collected at midnight, 8 am, or 4 pm, supporting standard definitions.

Systems initiatives reduce healthcare-associated infections: a study of 22,928 device days in a single trauma unit.

BACKGROUND: "Implementation research" promotes the systematic conversion of evidence-based principles into routine practice to improve the quality of care. We hypothesized a system-based initiative to reduce nosocomial infection would lower the incidence of ventilator-associated pneumonia (VAP), urinary tract infection (UTI), and bloodstream infection (BSI). METHODS: From January 2006 to April 2008, 7,364 adult trauma patients were admitted, of which 1,953 (27%) were admitted to the trauma intensive care unit and comprised the study group. Tight glycemic control was maintained using a computer algorithm for continuous insulin administration based on every 2-hour blood glucose testing. Centers for Disease Control and Prevention definitions of nosocomial infections were used. Evidence-based infection reduction strategies included the following: a VAP bundle (spontaneous breathing, Richmond Agitation-Sedation Scale, oral hygiene, bed elevation, and deep vein thrombosis/stress ulcer prophylaxis), UTI (expert insertion team and Foley removal/change at 5 days), and BSI (maximum barrier precautions, chlorhexidine skin prep, line management protocol). An electronic dashboard identified the at-risk population, and designated auditors monitored the compliance. Infection rates (events per 1,000 device days) were measured over time and compared annually using Fisher's exact test. RESULTS: The study group had 22,928 device exposure days: 6,482 ventilator days, 9,037 urinary catheter days, and 7,399 central line days. Patient acuity, demographics, and number of device days did not vary significantly year-to-year. Annual infection rates declined between 2006 and 2008, and decreases in UTI and BSI rates were statistically significant (p < 0.05). These decreases pushed UTI and BSI rates below Centers for Disease Control and Prevention norms. CONCLUSIONS: Over 28 months, a systems approach to reducing nosocomial infection rates after trauma decreased nosocomial infections: UTI (76.3%), BSI (74.1%), and VAP (24.9%). Our experience suggests that infection reduction requires (1) an evidence-based plan; (2) MD and staff education/commitment; (3) electronic documentation; and (4) auditors to monitor and ensure compliance.

Implementation of a real-time compliance dashboard to help reduce SICU ventilator-associated pneumonia with the ventilator bundle.

BACKGROUND: Ventilator-associated pneumonia (VAP) causes significant morbidity and mortality in critically ill surgical patients. Recent studies suggest that the success of preventive measures is dependent on compliance with ventilator bundle parameters. HYPOTHESIS: Implementation of an electronic dashboard will improve compliance with the bundle parameters and reduce rates of VAP in our surgical intensive care unit (SICU). DESIGN: Time series analysis of VAP rates between January 2005 and July 2008, with dashboard implementation in July 2007. SETTING: Multidisciplinary SICU at a tertiary-care referral center with a stable case mix during the study period. PATIENTS: Patients admitted to the SICU between January 2005 and July 2008. MAIN OUTCOME MEASURES: Infection control data were used to establish rates of VAP and total ventilator days. For the time series analysis, VAP rates were calculated as quarterly VAP events per 1000 ventilator days. Ventilator bundle compliance was analyzed after dashboard implementation. Differences between expected and observed VAP rates based on time series analysis were used to estimate the effect of intervention. RESULTS: Average compliance with the ventilator bundle improved from 39% in August 2007 to 89% in July 2008 (P < .001). Rates of VAP decreased from a mean (SD) of 15.2 (7.0) to 9.3 (4.9) events per 1000 ventilator days after introduction of the dashboard (P = .01). Quarterly VAP rates were significantly reduced in the November 2007 through January 2008 and February through April 2008 periods (P < .05). For the August through October 2007 and May through July 2008 quarters, the observed rate reduction was not statistically significant. CONCLUSIONS: Implementation of an electronic dashboard improved compliance with ventilator bundle measures and is associated with reduced rates of VAP in our SICU.

Closing the loop in practice to assure the desired performance.

A closed loop control process assures that a system performs within control limits. In closed loop control, the system's output feeds back directly to change the system's inputs. We describe an approach to planning and monitoring care that uses closed loop control to assure the desired performance using examples from Vanderbilt University Medical Center's ventilator management initiative. The approach has three components: an explicit end-to-end plan; a record of what is done as it is done; and an instant display of the status of each patient against their plan. The status display provides process control by showing the clinical team where correction is needed while they have time to act prospectively. Plans, displays and performance evolve together iteratively until the desired performance is achieved.