ORIGINAL RESEARCH Performance of the Maximum Modified Early Warning Score to Predict the Need for Higher Care Utilization Among Admitted Emergency Department Patients 1 1 Corey R. Heitz, MD Department of Emergency Medicine, Wake Forest University School of Medicine, Winston-Salem, 1 John P. Gaillard, MD North Carolina. 1 Howard Blumstein, MD 2 2 Department of Biostatistical Sciences, Wake Forest University School of Medicine, Winston-Salem, Doug Case, PhD 3 North Carolina. Catherine Messick, MD 1 3 Chadwick D. Miller, MD Department of Internal Medicine, Wake Forest University School of Medicine, Winston-Salem, North Carolina. This study was supported in part by the Division of Healthcare Research and Quality of Wake Forest University Baptist Medical Center. Financial support for Dr. Miller is provided by a research scholar award from the Wake Forest University Translational Science Institute. Disclosure: Nothing to report. BACKGROUND: It is uncertain whether ED-calculated risk scores can predict required intensity of care upon hospital admission. This investigation examines whether versions of the Modified Early Warning Score (MEWS) predict high level of care utilization among patients admitted from the ED. METHODS: A retrospective chart review of 299 admissions was implemented. Exclusions prior to abstraction included pediatrics, cardiology, or trauma admissions. Using a data-gathering instrument, abstractors recorded physiologic parameters and clinical variables. Risk scores were calculated electronically. In contrast to the original MEWS, the MEWS Max was calculated using data from the entire ED visit. The primary outcome composite included all-cause mortality and higher care utilization within 24 hours. RESULTS: The final analysis contained 280 participants. 76 (27%) met the composite endpoint of death (n ¼ 1) or higher care utilization (n ¼ 76). The MEWS Max was associated with the composite outcome (OR¼l.6 [95% CI 1.3-1.8] for each one point increase). The MEWS Max had moderate predictive ability (C statistic: MEWS Max 0.73 [0.66-0.79]) but classified 82% of participants as intermediate (10-40%) risk. Inclusion of additional variables slightly improved the predictive ability (C statistic 0.76 [0.69–0.82]) and correctly reclassified 17% of patients as <10% risk. CONCLUSIONS: The MEWS Max has moderate ability to predict the need for higher level of care. Addition of ED length of stay and other variables to MEWS Max may identify patients at both low and high risk of requiring a higher level of care. Journal of Hospital Medicine 2010;5:E46–E52. VC 2010 Society of Hospital Medicine. KEYWORDS: care standardization, early warning scores, emergency department, level of care, patient safety, risk scores. Matching the severity of illness to the appropriate intensity are at risk for deterioration. Such tools for identification of of care is important for the effective delivery of medical inpatients at risk generally use ‘‘single threshold’’ models care. Overtriage to critical care units results in unnecessary triggered by a single abnormal physiologic value, or models resource consumption. Undertriage to the wards may result that combine multiple parameters into a summative in worsening of physiologic parameters1,2 that often go score.6,7 The performance of previously described risk strati- unnoticed or unaddressed for more than 24 hours.3 There- fication tools has generally been to exhibit high sensitivity fore, it is important for emergency department (ED) admis- at the sacrifice of low specificity and discriminatory value.8 sion decisions to be accurate with respect to the level of The value of these models as they apply in the emer- care. Because of the importance of this decision, objective gency department is less well characterized. Because criteria to aid in this decision process, if accurate, would derangements in physiologic parameters are common improve medical care delivery. among ED patients, one might expect that ‘‘single-thresh- Physiologic measurements and procedural interventions old’’ systems would exhibit high sensitivity at the expense of appear to predict the need for a higher level of care among specificity when applied to this population. In contrast, a inpatients.2,4,5 This knowledge has led to the development summative risk score may be better suited for the complex- of tools meant to identify inpatients on general wards who ities of illness in undifferentiated ED patients and offer 2010 Society of Hospital Medicine DOI 10.1002/jhm.552 Published online in wiley InterScience (www.interscience.wiley.com). E46 Journal of Hospital Medicine Vol 5 No 1 January 2010 better discriminatory value in this population. Summative domly selected for further review. Additional criteria were scoring systems have been shown to retain a higher speci- applied at the time the charts were reviewed to exclude ficity as the score increases compared to single-threshold those: without an ED record matching the date of admis- systems.8 sion, without 1 complete set of ED vitals, receiving mechan- The Modified Early Warning Score (MEWS)9 is a predic- ical ventilation at the time of presentation, or patients cur- tive tool for higher level of care that has been tested in the rently receiving hospice or ‘‘comfort care.’’ Charts from the ED setting. This tool produces a summative score using list of 500 were reviewed sequentially until the goal number temperature, respiratory rate, heart rate, level of conscious- of charts had been completed. The number of charts ness, and systolic blood pressure. In a single-site study from reviewed was selected to allow relatively precise 95% confi- the United Kingdom, MEWS, when calculated at the time of dence intervals [CIs] around sensitivity (Æ10%) based on the ED presentation, did not improve decision making over a assumptions of 80% sensitivity and a 20% incidence of the commonly used triage system, exhibiting inadequate sensi- primary outcome. Based on this, the intent was to abstract tivity in identifying patients who would be admitted to the information from 300 patient charts. intensive care unit (ICU).10 However, as a result of the care delivered in the ED, patients’ conditions can change signifi- Study Protocol cantly throughout their stay. Therefore we postulate that the A standardized data abstraction template was created. Data MEWS calculated at a single time in the ED (eg, at the time abstractors included 2 physicians and 2 nurses. Group train- of admission) is not the most accurate predictor of care ing for the abstractors was provided by the primary investi- intensity requirements. gator and included performance review and feedback until The primary objective of this research was to add to the competence was demonstrated. Data abstractors used the 10 literature provided by Subbe et al. by describing the per- paper copy of the ED nursing notes (and physician notes if formance characteristics and discriminatory ability of the clarification required) to abstract data from the medical most abnormal MEWS (MEWS Max) score during the entire record. Abstractors were not aware of the patient’s outcome ED stay in predicting the need for higher levels of care at the time of data abstraction as this information was con- among ED patients presenting to a tertiary care facility in tained in a separate database. During the chart review, and North America. blinded to the abstractors, 25 charts were selected for abstraction by all data abstractors to allow calculation of Patients and Methods interobserver agreement. Study Design Clinical outcomes were determined by referencing hospi- To determine the performance characteristics of the MEWS tal databases and the medical record if clarification was in ED patients, we used a structured explicit retrospective needed. The admission bed location and changes in patient chart review on a random sample of ED patients being location throughout the hospital stay were used to track the admitted to the hospital. need for a higher level of care. The outcome of death was determined by cross-referencing study participants with Study Setting hospital mortality data, and the medical record, if needed. The study was conducted at 1 tertiary care academic medi- cal center in the United States, consisting of 830 beds, Predictor Score Calculation approximately 125 of which provide a higher level of care, Abstracted data were used to calculate the MEWS score defined as intensive care, intermediate care, or acute care. according to the criteria specified in Table 1 at the initial The ED volume in 2005 was 75,000 with an admission rate ED presentation (MEWS Initial), the maximum during the of 20%. In the ED, patients are primarily seen by residents ED stay (MEWS Max), and prior to admission (MEWS who are supervised by board-certified or board-eligible Admit). Parameters not repeated after arrival were carried emergency medicine attendings. forward from the most recent recording. An adaptation of the MEWS score was required by replacing the alert/verbal/ Study Population painful/unresponsive (AVPU) scale to determine the level of All patients presenting to the ED of Wake Forest University consciousness with the Glascow Coma Scale (GCS), a con- 13,14 Baptist Medical Center in 2005 were considered for inclu- version that has been previously described. sion. From these patients, a listing was created of all hospi- tal admissions through the ED in 2005. Because trauma and Clinical Endpoint Definitions and Outcomes cardiology patients have disease-specific risk stratification ‘‘Need for higher level of care’’ was defined as initial admis- tools that are used to guide admission,11,12 they were then sion from the ED or transfer within 24 hours to a nonfloor removed from this list and excluded. Additionally, pediatric bed (acute care, intermediate care unit, or critical care unit). patients were excluded from this listing as the MEWS score Acute care beds at the study hospital have a lower bed-to- relies on vital sign abnormalities, which have varying ranges nurse ratio and more intensive monitoring (beside vs.