User login
Predictors of Clinically Significant Echocardiography Findings in Older Adults with Syncope: A Secondary Analysis
Syncope, defined as a transient loss of consciousness and postural tone followed by complete, spontaneous return to neurological baseline, accounts for over 1 million (or approximately 1%) of all emergency department (ED) visits per year in the United States (US).12 Given the breadth of etiologies for syncope, including certain life-threatening conditions, extensive diagnostic evaluation and hospitalization for this complaint is common.3-7 The estimated costs of syncope-related hospitalizations are over $2.4 billion annually in the US.8
The 2011 American College of Cardiology Foundation appropriate use criteria for echocardiography state that syncope is an appropriate indication for transthoracic echocardiography (TTE) even when there are no other symptoms or signs of cardiovascular disease.9 This broad recommendation may be appropriate since a finding of severe valvular disease would generally merit consultation with a cardiothoracic surgeon to assess the potential for surgical intervention.10 However, routine use of echocardiogram in all syncope patients could result in increased healthcare costs, patient discomfort, and incidental findings of unclear significance, while rarely changing diagnosis or management.11,12
In an attempt to reduce potentially unnecessary TTE testing, several studies have tried to identify patients at very low risk of structural heart disease.13-17 These investigations suggest that TTE is not indicated in syncope patients with a normal ECG and a normal cardiac exam. However, this literature is limited by retrospective study design and/or small sample sizes. The 2017 American Heart Association/American College of Cardiology/Heart Rhythm Society syncope guidelines recommend TTE for a patient in whom structural heart disease is suspected, but they are not explicit about how to make this determination. 18 Thus, it is still unclear which syncope patients require TTE since a standardized approach to assessing risk of clinically significant findings on TTE has not yet been rigorously developed.
The objective of this study was to develop a risk-stratification tool to identify older adults at very low risk of having a major, clinically significant finding on rest TTE after presenting to the ED with syncope or near-syncope. Using clinical, ECG, and cardiac biomarker data, we created the ROMEO (Risk Of Major Echocardiography findings in Older adults with syncope) score to help optimize resource utilization for syncope.
METHODS
Study Design and Setting
We conducted a large, multicenter, prospective, observational cohort study of older adults who presented to an ED with syncope or near-syncope (ClinicalTrials.gov identifier: NCT01802398). The study was approved by the institutional review boards at all sites and written informed consent was obtained from all participating subjects. The study was conducted at 11 academic EDs across the US (See Appendix Table 1).
Study Population
Patient inclusion criteria for eligibility were age ≥60 years with a complaint of syncope or near-syncope. Syncope was defined as transient loss of consciousness, associated with postural loss of tone, with immediate, spontaneous, and complete recovery. Near syncope was defined as the sensation of imminent loss of consciousness. Patients were excluded if their symptoms were thought to be due to intoxication, seizure, stroke, head trauma, or hypoglycemia. Additional exclusion criteria were the need for medical intervention to restore consciousness (eg, defibrillation), new or worsening confusion, and inability to obtain informed consent from the patient or a legally authorized representative.
This analysis included only patients who received a TTE during the index visit (either in the ED, observation unit, or while admitted to the hospital). This dataset was also used for other analyses addressing questions relevant to the ED management of syncope.
Measurements
All patients underwent a standardized history, physical examination, laboratory, and 12-lead ECG testing. Trained research assistants (RA) directly queried patients about symptoms associated with the syncopal episode. Data on the patient’s past medical history, medications, and physical examination findings were collected prospectively from treating providers.
Research staff obtained blood samples for testing at a core laboratory (University of Rochester, Rochester, NY). Two assays were performed using the Roche Elecsys platform: N-terminal pro B-type natriuretic peptide (NT-proBNP) and the 5th generation high-sensitivity cardiac troponin T (hs-TnT). NT-proBNP was classified as abnormal above a cutoff of 125 pg/mL. Hs-TnT was classified as abnormal above the 99th percentile for a reference population (14 pg/mL). Although hs-TnT was not approved by the U.S. Food and Drug Administration (FDA) at the time of the study, we anticipated that this assay would receive approval and be integrated into future standard of care (FDA approval was granted in January 2017). Rest TTEs were ordered at the discretion of the treating providers.
Outcome Measures
The primary outcome for this secondary analysis was a major, clinically significant finding on TTE.13,14,16,19 These included severe aortic stenosis (<1 cm2), severe mitral stenosis, severe aortic/mitral regurgitation, reduced ejection fraction (defined either quantitatively as less than 45% or qualitatively as “severe left ventricular dysfunction”), hypertrophic cardiomyopathy with outflow tract obstruction, severe pulmonary hypertension, right ventricular dysfunction/strain, large pericardial effusion, atrial myxoma, or regional wall motion abnormalities.
All echocardiogram reports were independently reviewed by two research physicians. Discrepant reviews were resolved by the research physicians and two of the study investigators (BS, CB). Of note, all the TTEs obtained were formal echocardiographic studies, not bedside ultrasonography performed by the emergency physician.
Candidate Predictors
Potential candidate predictors were identified through a prior expert panel process.20,21 Candidate predictors included age, sex, abnormal heart sounds, exertional syncope, shortness of breath, chest pain, near-syncope, family history of sudden cardiac death, high (>180 mm Hg) or low (<90 mm Hg) systolic blood pressure, abnormal ECG, elevated hs-TnT, elevated NT-proBNP, and history of the following: hypertension, cardiac dysrhythmia, renal failure, diabetes, congestive heart failure (CHF), and coronary artery disease (CAD).
The first obtained ECG was abstracted by one of five research study physicians blinded to all clinical data. Research study physicians demonstrated high interrater reliability (kappa > 0.80) in distinguishing normal from abnormal ECGs in a training set of 50 ECGs. Abnormal ECG interpretations included nonsinus rhythms (including paced rhythms), multiple premature ventricular complexes, sinus bradycardias (≤40 bpm), ventricular hypertrophies, short PR segment intervals (<100 milliseconds [ms]), axis deviations, first degree blocks (>200 ms), complete bundle branch blocks, Brugada patterns, Wolff-Parkinson-White patterns, abnormal QRS duration (>120 ms) or abnormal QTc prolongations (>450 ms), and Q/ST/T segment abnormalities suggestive of acute or chronic ischemia.
Statistical Analysis
We calculated descriptive statistics for each predictor variable, stratified by the presence or absence of TTE findings. Chi-square and t-tests were used to test associations between categorical or continuous variables and TTE findings using a significance level of 0.05 and 2-sided hypothesis testing. To identify a robust set of predictors of the primary outcome, we used multivariate logistic regression with the LASSO (Least Absolute Shrinkage and Selection Operator) to fit a parsimonious model.22 The LASSO selects variables and shrinks the associated coefficients to avoid overfitting.23-25 We then used a bootstrap to generate confidence intervals for coefficient estimates. Cases with missing echocardiography reports were excluded from the analysis. Bootstrap results were summarized as the percentage of bootstrap iterations in which each variable’s coefficient was 1) chosen and negative, 2) shrunk to zero, or 3) chosen and positive.
We assessed different weighting schemes to generate a risk score from significant variables identified by regression modeling. These included weighting by regression coefficients rounded to the nearest integer and simple summation of the presence or absence of each variable.
Based on these results, a predictive score was developed to risk stratify patients on their probability of major, clinically significant findings on TTE. The sensitivity and specificity of a score of zero to predict findings on TTE was calculated. For confidence intervals, we used Wilson’s method for binomial confidence intervals.26 The receiver operating characteristic (ROC) curve and its associated area under the curve (AUC) were calculated, and a confidence interval for the AUC was obtained through bootstrap resampling with 2,000 iterations. As part of our sensitivity analyses, we also calculated the ROC curve and AUC after excluding the patients with a known history of CHF and significant finding on TTE. Data analyses were performed in R.27 Two sensitivity analyses were performed: 1) we used multiple imputation to impute 1,000 complete data sets and then used the same LASSO methodology as with the complete data to assess whether incorporating missing data changed the results; and 2) we simulated a conventional troponin assay by raising the positive threshold for hs-TnT to >30 pg/mL (corresponding to the limit of detection for conventional troponin).28
RESULTS
Characteristics of Study Subjects
Patient screening occurred from April 2013 to September 2016. There were 6,930 patients who met eligibility criteria, of whom 3,686 (53%) consented and enrolled in the study (See Figure 1). Of these, 995 (27%) received TTE. The mean age of patients receiving TTE was 74 years; 55% were male. Characteristics of patients obtaining and not obtaining TTE are presented in Appendix Table 2. Patients who received TTE were more likely to be older, have abnormal heart sounds, abnormal EKGs, elevated hs-TnT, elevated NT-proBNP, and have a history of CHF. Of the 995 subjects receiving TTE, 215 (21.6%) had a major, clinically significant finding.
Main Results
Univariate analysis identified 14 variables significantly associated with major findings on TTE. These included male gender, shortness of breath, abnormal heart sounds, history of renal failure, diabetes, CHF, CAD, abnormal ECG, and elevated cardiac biomarkers, among others (See Table 1). The most common major finding on TTE was regional wall motion abnormality, followed by reduced left ventricular ejection fraction (See Table 2). Of the 995 patients who received TTE, 20 (2%) were discharged directly from the ED, 444 (45%) were observed, and 531 (53%) were admitted. On average, patients who received TTE had a longer length of stay than did those that did not (3.4 days vs 1.9 days).
LASSO multivariable logistic regression produced five predictors associated with major findings on TTE: 1) history of CHF, 2) history of CAD, 3) abnormal ECG, 4) hs-TnT above 14 pg/mL, and 5) NT-proBNP above 125 pg/mL (See Table 3).
These five high-risk clinical variables retained their importance after multivariate analysis and form the ROMEO score.
The sensitivity and specificity of a ROMEO score of zero for excluding major findings on TTE was 99.5% (95% CI: 97.4%-99.9%) and 15.4% (95% CI: 13.0%-18.1%), respectively. Patients with a ROMEO score of 0 were at very low risk of having a major finding on TTE: 0.8% (95% CI: 0.02%-4.5%; Appendix Table 3). Only one out of 121 patients with none of the ROMEO criteria was found to have a major finding on TTE (regional wall motion abnormality). Patients with a score of 1 or more were at moderate-to-high risk of having a major finding (7.3% to 55.6%).
There was a linear relationship between the ROMEO score and probability of major findings on TTE (See Appendix Figure 1). The AUC was 0.77 (95% CI = 0.72-0.79) indicating good accuracy of the combination of the five high-risk clinical variables to predict major findings on TTE (See Appendix Figure 2). After excluding the 72 patients with known CHF and significant findings on TTE, the AUC was similar: 0.73 (95% CI: 0.69-0.77). There were 139 patients with at least one missing variable (14%) (See Appendix Table 4). A multiple imputation sensitivity analysis identified the same five high-risk clinical variables in 85% of imputations.
There were 253 patients with high-sensitivity troponin levels between 15 and 30 pg/mL (inclusive). Using a higher hs-TnT threshold (>30 pg/mL) to simulate a conventional troponin assay again identified the same five high-risk variables along with shortness of breath as a potential sixth variable though with an odds ratio approaching unity (See Appendix Table 5). The ROMEO score would have missed two additional patients with major findings if the troponin cutoff were raised to 30 pg/mL from 14 pg/mL, ie, it would have identified 212/215 (98.6%) of the major findings rather than 214/215 (99.5%).
DISCUSSION
Older adults with syncope often present to the ED and undergo a variety of diagnostic tests, including TTE, and a significant proportion are admitted to the hospital.2 There is currently no standardized, evidence-based approach to guide TTE ordering for these patients. Using a large, prospective dataset of syncope patients, we sought to develop a risk-stratification tool to help clinicians identify which syncope patients would be at very low risk for clinically significant findings on TTE. We found that in the absence of these five high-risk clinical variables, the rate of significant findings on TTE in our sample was less than 1%. All five high-risk variables included in the tool remained predictive in our sensitivity analyses, speaking to the robustness of our model.
Other retrospective, and smaller prospective, studies have identified a combination of low-risk criteria including: a normal ECG alone,15 a normal physical exam and normal ECG,14,17 a negative cardiac history and normal ECG.16 Han et al. performed a chart review of 241 patients presenting to the ED with syncope and identified three risk factors for abnormal TTE findings using multiple logistic regression: age, abnormal ECG, and BNP greater than 100 pg/mL.13 While these studies’ results are generally consistent with ours, the retrospective nature and small sample size of these studies limit the generalizability of these results. Thus, using a large, multicenter prospective dataset, we derived a clinical decision instrument (the ROMEO score) to determine which older adults with syncope are at very low risk for major, clinically significant findings on TTE.
Our results add to the recent American College of Cardiology/American Heart Association/Heart Rhythm Society guidelines on the management of syncope which recommend TTE in “selected patients presenting with syncope if structural heart disease is suspected.”18 Our risk-stratification tool offers a simple, standardized approach to determine specifically when to defer TTE testing.
Our findings can guide clinicians in deciding when to obtain TTE for ED syncope patients in the following way: Older adults presenting with syncope or near-syncope to the ED who have none of the ROMEO criteria are at extremely low risk for clinically significant findings on TTE and thus need not undergo such testing solely because of the syncopal event. Patients who have only one or more high-risk clinical variables are at higher risk (7.3%-56%) of significant TTE findings. In this subset, other factors, (eg, physician gestalt, recent previous echocardiography, patient preference, availability of echocardiography) can help guide TTE ordering. Patients with a greater number of high-risk variables may benefit from a more urgent echocardiographic evaluation.
Although on average, patients undergoing TTE had a longer length of stay than those that did not, this finding does not necessarily imply that ordering a TTE was the cause of the increased length of stay. It is possible that this positive association was due to greater underlying medical complexity or acuity of illness that resulted in a greater likelihood of admission/observation, and in turn, a greater length of stay.
Prior to implementation, our results should be externally validated in other clinical settings. In the interim, this risk-stratification tool may be used by clinicians, in conjunction with clinical judgement, to help guide the appropriate use of TTE in older adults presenting with syncope.
Our study has certain limitations. As we only enrolled patients 60 years and older, our findings may not necessarily be valid in younger populations of syncope patients. However, structural heart disease is less common in younger patients and is generally more of a concern for clinicians when evaluating syncope patients in the older age range.29 In our study, 47% of eligible patients declined to participate and thus sampling bias may have occurred. TTEs were ordered at the discretion of treating providers, which was likely subject to physician, institutional, and regional variation; the prevalence of major TTE findings may be lower in the overall cohort than in patients who received TTE. Prior TTE reports were not available; therefore, we were not able to determine if these major findings were previously known. Importantly, we did not perform an internal or external validation of the ROMEO score due to time and resource constraints. Thus, this study represents a derivation of the score solely and would require external validation prior to clinical implementation. Also, to calculate the ROMEO score, both an hs-TnT and NT-proBNP level must be obtained. Thus, the cost savings of any potential reduction in TTE ordering may be partially offset by the costs of increased laboratory testing. Lastly, hs-TnT assays are not currently widely available in hospitals in the United States; earlier generation cardiac troponin assays may not be a perfect substitute for hs-TnT assays. Our sensitivity analysis using an elevated threshold for hs-TnT attempted to mitigate this limitation and resulted in similar findings.
In summary, this risk-stratification tool, using five simple criteria, could help clinicians determine which older adult syncope patients can safely forgo TTE. If validated, this tool could help optimize resource utilization, and increase the value of healthcare for patients presenting with syncope.
Acknowledgments
The authors would like to thank the research assistants at all 11 sites who enrolled patients and collected data for this study.
Disclosures
Dr. Adler has received research funding from Roche. Dr. Bastani has received research funding from Radiometer and Portola and has been a consultant for Portola. Dr. Baugh has received advisory board and speaker’s fees from Roche, research funding from Janssen and Boehringer Ingelheim, and consulting and advisory board fees from Janssen. Dr. Casterino has received funding from Astra Zeneca. Dr. Clark has received research funding from Radiometer, Ortho Clinical Trials, Janssen, Pfizer, NIH, Portola, Biocryst, Glaxo Smith Klein, Hospital Quality Foundation, and Abbott. She is a consultant for Portola, Janssen, and Hospital Quality Foundation. Dr. Diercks is a consultant for Siemens, Janssen, and Roche has received institutional research support from Novartis, Ortho Scientific, and Roche. Dr. Hollander has received research funding from Alere, Siemens, Roche, Portola, and Trinity. Dr. Hollander has also received royalties from UpToDate. Dr. Nishijima has received an honorarium from Pfizer. Dr. Storrow is a consultant for Siemens and Quidel, has received speaking fees from MCM Education, and is on the Data and Safety Monitoring Board for Trevena. Dr. Sun is a consultant for Medtronic. The other authors report no relevant conflicts of interest.
Funding
This project was supported by the National Heart, Lung, And Blood Institute of the National Institutes of Health under Award Number R01 HL111033. Dr. Probst is supported by the National Heart, Lung, And Blood Institute of the National Institutes of Health under Award Number K23HL132052-02. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. Roche Diagnostics supplied the high-sensitivity troponin-T assays. The sponsoring organizations had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, or review of the manuscript.
1. Sun BC, Emond JA, Camargo CA, Jr. Characteristics and admission patterns of patients presenting with syncope to U.S. emergency departments, 1992-2000. Acad Emerg Med. 2004;11(10):1029-1034. doi: 10.1197/j.aem.2004.05.032. PubMed
2. Probst MA, Kanzaria HK, Gbedemah M, Richardson LD, Sun BC. National trends in resource utilization associated with ED visits for syncope. Am J Emerg Med. 2015;33(8):998-1001. doi: 10.1016/j.ajem.2015.04.030. PubMed
3. Kapoor WN, Karpf M, Maher Y, Miller RA, Levey GS. Syncope of unknown origin. The need for a more cost-effective approach to its diagnosis evaluation. JAMA. 1982;247(19):2687-2691. doi: 10.1001/jama.247.19.2687. PubMed
4. Pires LA, Ganji JR, Jarandila R, Steele R. Diagnostic patterns and temporal trends in the evaluation of adult patients hospitalized with syncope. Arch Intern Med. 2001;161(15):1889-1895. doi: 10.1001/archinte.161.15.1889. PubMed
5. Quinn JV, Stiell IG, McDermott DA, Sellers KL, Kohn MA, Wells GA. Derivation of the San Francisco Syncope Rule to predict patients with short-term serious outcomes. Ann Emerg Med. 2004;43(2):224-232. doi: 10.1016/S0196064403008230. PubMed
6. Linzer M, Yang EH, Estes NA, 3rd, Wang P, Vorperian VR, Kapoor WN. Diagnosing syncope. Part 1: Value of history, physical examination, and electrocardiography. Clinical Efficacy Assessment Project of the American College of Physicians. Ann Intern Med. 1997;126(12):989-996. doi: 10.7326/0003-4819-126-12-199706150-00012. PubMed
7. Linzer M, Yang EH, Estes NA, 3rd, Wang P, Vorperian VR, Kapoor WN. Diagnosing syncope. Part 2: Unexplained syncope. Clinical Efficacy Assessment Project of the American College of Physicians. Ann Intern Med. 1997;127(1):76-86. doi: 10.7326/0003-4819-127-1-199707010-00014. PubMed
8. Sun BC, Emond JA, Camargo CA, Jr. Direct medical costs of syncope-related hospitalizations in the United States. Am J Cardiol. 2005;95(5):668-671. doi: 10.1016/j.amjcard.2004.11.013. PubMed
9. American College of Cardiology Foundation. Appropriate Use Criteria Task F, American Society of Echocardiography, American Heart Association, et al. ACCF/ASE/AHA/ASNC/HFSA/HRS/SCAI/SCCM/SCCT/SCMR 2011 Appropriate Use Criteria for Echocardiography. A Report of the American College of Cardiology Foundation Appropriate Use Criteria Task Force, American Society of Echocardiography, American Heart Association, American Society of Nuclear Cardiology, Heart Failure Society of America, Heart Rhythm Society, Society for Cardiovascular Angiography and Interventions, Society of Critical Care Medicine, Society of Cardiovascular Computed Tomography, and Society for Cardiovascular Magnetic Resonance Endorsed by the American College of Chest Physicians. J Am Coll Cardiol. 2011;57(9):1126-1166. doi: 10.1016/j.echo.2010.12.008.
10. Maganti K, Rigolin VH, Sarano ME, Bonow RO. Valvular heart disease: diagnosis and management. Mayo Clin Proc. 2010;85(5):483-500. doi: 10.4065/mcp.2009.0706. PubMed
11. Mendu ML, McAvay G, Lampert R, Stoehr J, Tinetti ME. Yield of diagnostic tests in evaluating syncopal episodes in older patients. Arch Intern Med. 2009;169(14):1299-1305. doi: 10.1001/archinternmed.2009.204. PubMed
12. Madeira CL, Craig MJ, Donohoe A, Stephens JR. Things we do for no reason: echocardiogram in unselected patients with syncope. J Hosp Med. 2017;12(12):984–988. doi: http://dx.doi.org/10.12788/jhm.2864. PubMed
13. Han SK, Yeom SR, Lee SH, et al. Transthoracic echocardiogram in syncope patients with normal initial evaluation. Am J Emerg Med. 2017;35(2):281-284. doi: 10.1016/j.ajem.2016.10.078. PubMed
14. Chang NL, Shah P, Bajaj S, Virk H, Bikkina M, Shamoon F. Diagnostic yield of echocardiography in syncope patients with normal ECG. Cardiol Res Pract. 2016;2016:1251637. doi: http://dx.doi.org/10.1155/2016/1251637. PubMed
15. Anderson KL, Limkakeng A, Damuth E, Chandra A. Cardiac evaluation for structural abnormalities may not be required in patients presenting with syncope and a normal ECG result in an observation unit setting. Ann Emerg Med. 2012;60(4):478–84.e1. doi: 10.1016/j.annemergmed.2012.04.023. PubMed
16. Sarasin FP, Junod AF, Carballo D, Slama S, Unger PF, Louis-Simonet M. Role of echocardiography in the evaluation of syncope: a prospective study. Heart. 2002;88(4):363-367. doi: 10.1136/heart.88.4.363. PubMed
17. Recchia D, Barzilai B. Echocardiography in the evaluation of patients with syncope. J Gen Intern Med. 1995;10(12):649-655. doi: 10.1007/BF02602755. PubMed
18. Shen WK, Sheldon RS, Benditt DG, et al. ACC/AHA/HRS guideline for the evaluation and management of patients With syncope: executive summary: A report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines and the Heart Rhythm Society. J Am Coll Cardiol. 2017;2017(70(5)):620-663. PubMed
19. Chiu DT, Shapiro NI, Sun BC, Mottley JL, Grossman SA. Are echocardiography, telemetry, ambulatory electrocardiography monitoring, and cardiac enzymes in emergency department patients presenting with syncope useful tests? A preliminary investigation. J Emerg Med. 2014;47(1):113-118. doi: 10.1016/j.jemermed.2014.01.018. PubMed
20. Sun BC, Costantino G, Barbic F, et al. Priorities for emergency department syncope research. Ann Emerg Med. 2014;64(6):649–55.e2. doi: 10.1016/j.annemergmed.2014.04.014. PubMed
21. Sun BC, Derose SF, Liang LJ, et al. Predictors of 30-day serious events in older patients with syncope. Ann Emerg Med. 2009;54(6):769–778.e1-5. doi: 10.1016/j.annemergmed.2009.07.027. PubMed
22. Tibshirani R. Regression shrinkage and selection via the lasso. J R Stat Soc. 1996;58(1):267-288.
23. Friedman J, Hastie T, Tibshirani R. He Elements of Statistical Learning;Vol 1. New York, NY: Springer-Verlag; 2001. PubMed
24. Friedman J, Hastie T, Tibshirani R. Regularization paths for generalized linear models via coordinate descent. J Stat Softw. 2010;33(1):1-22. doi: 10.18637/jss.v033.i01. PubMed
25. James G, Witten D, Hastie T, Tibshirani R. An Introduction to Statistical Learning;Vol 112. New York, NY: Springer-Verlag; 2013.
26. Wilson EB. Probable inference, the law of succession, and statistical inference. J Am Stat Assoc. 1927 ;22(158):209-212. doi: 10.1080/01621459.1927.10502953. PubMed
27. R Core Team (2018). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL http://www.R-project.org/.
28. Chew DP, Zeitz C, Worthley M, et al. Randomized comparison of high-sensitivity troponin reporting in undifferentiated chest pain assessment. Circ Cardiovasc Qual Outcomes. 2016;9(5):542-553. doi: 10.1161/CIRCOUTCOMES.115.002488. PubMed
29. Chen RS, Bivens MJ, Grossman SA. Diagnosis and management of valvular heart disease in emergency medicine. Emerg Med Clin North Am. 2011;29(4):801–10, vii. doi: 10.1016/j.emc.2011.08.001. PubMed
Syncope, defined as a transient loss of consciousness and postural tone followed by complete, spontaneous return to neurological baseline, accounts for over 1 million (or approximately 1%) of all emergency department (ED) visits per year in the United States (US).12 Given the breadth of etiologies for syncope, including certain life-threatening conditions, extensive diagnostic evaluation and hospitalization for this complaint is common.3-7 The estimated costs of syncope-related hospitalizations are over $2.4 billion annually in the US.8
The 2011 American College of Cardiology Foundation appropriate use criteria for echocardiography state that syncope is an appropriate indication for transthoracic echocardiography (TTE) even when there are no other symptoms or signs of cardiovascular disease.9 This broad recommendation may be appropriate since a finding of severe valvular disease would generally merit consultation with a cardiothoracic surgeon to assess the potential for surgical intervention.10 However, routine use of echocardiogram in all syncope patients could result in increased healthcare costs, patient discomfort, and incidental findings of unclear significance, while rarely changing diagnosis or management.11,12
In an attempt to reduce potentially unnecessary TTE testing, several studies have tried to identify patients at very low risk of structural heart disease.13-17 These investigations suggest that TTE is not indicated in syncope patients with a normal ECG and a normal cardiac exam. However, this literature is limited by retrospective study design and/or small sample sizes. The 2017 American Heart Association/American College of Cardiology/Heart Rhythm Society syncope guidelines recommend TTE for a patient in whom structural heart disease is suspected, but they are not explicit about how to make this determination. 18 Thus, it is still unclear which syncope patients require TTE since a standardized approach to assessing risk of clinically significant findings on TTE has not yet been rigorously developed.
The objective of this study was to develop a risk-stratification tool to identify older adults at very low risk of having a major, clinically significant finding on rest TTE after presenting to the ED with syncope or near-syncope. Using clinical, ECG, and cardiac biomarker data, we created the ROMEO (Risk Of Major Echocardiography findings in Older adults with syncope) score to help optimize resource utilization for syncope.
METHODS
Study Design and Setting
We conducted a large, multicenter, prospective, observational cohort study of older adults who presented to an ED with syncope or near-syncope (ClinicalTrials.gov identifier: NCT01802398). The study was approved by the institutional review boards at all sites and written informed consent was obtained from all participating subjects. The study was conducted at 11 academic EDs across the US (See Appendix Table 1).
Study Population
Patient inclusion criteria for eligibility were age ≥60 years with a complaint of syncope or near-syncope. Syncope was defined as transient loss of consciousness, associated with postural loss of tone, with immediate, spontaneous, and complete recovery. Near syncope was defined as the sensation of imminent loss of consciousness. Patients were excluded if their symptoms were thought to be due to intoxication, seizure, stroke, head trauma, or hypoglycemia. Additional exclusion criteria were the need for medical intervention to restore consciousness (eg, defibrillation), new or worsening confusion, and inability to obtain informed consent from the patient or a legally authorized representative.
This analysis included only patients who received a TTE during the index visit (either in the ED, observation unit, or while admitted to the hospital). This dataset was also used for other analyses addressing questions relevant to the ED management of syncope.
Measurements
All patients underwent a standardized history, physical examination, laboratory, and 12-lead ECG testing. Trained research assistants (RA) directly queried patients about symptoms associated with the syncopal episode. Data on the patient’s past medical history, medications, and physical examination findings were collected prospectively from treating providers.
Research staff obtained blood samples for testing at a core laboratory (University of Rochester, Rochester, NY). Two assays were performed using the Roche Elecsys platform: N-terminal pro B-type natriuretic peptide (NT-proBNP) and the 5th generation high-sensitivity cardiac troponin T (hs-TnT). NT-proBNP was classified as abnormal above a cutoff of 125 pg/mL. Hs-TnT was classified as abnormal above the 99th percentile for a reference population (14 pg/mL). Although hs-TnT was not approved by the U.S. Food and Drug Administration (FDA) at the time of the study, we anticipated that this assay would receive approval and be integrated into future standard of care (FDA approval was granted in January 2017). Rest TTEs were ordered at the discretion of the treating providers.
Outcome Measures
The primary outcome for this secondary analysis was a major, clinically significant finding on TTE.13,14,16,19 These included severe aortic stenosis (<1 cm2), severe mitral stenosis, severe aortic/mitral regurgitation, reduced ejection fraction (defined either quantitatively as less than 45% or qualitatively as “severe left ventricular dysfunction”), hypertrophic cardiomyopathy with outflow tract obstruction, severe pulmonary hypertension, right ventricular dysfunction/strain, large pericardial effusion, atrial myxoma, or regional wall motion abnormalities.
All echocardiogram reports were independently reviewed by two research physicians. Discrepant reviews were resolved by the research physicians and two of the study investigators (BS, CB). Of note, all the TTEs obtained were formal echocardiographic studies, not bedside ultrasonography performed by the emergency physician.
Candidate Predictors
Potential candidate predictors were identified through a prior expert panel process.20,21 Candidate predictors included age, sex, abnormal heart sounds, exertional syncope, shortness of breath, chest pain, near-syncope, family history of sudden cardiac death, high (>180 mm Hg) or low (<90 mm Hg) systolic blood pressure, abnormal ECG, elevated hs-TnT, elevated NT-proBNP, and history of the following: hypertension, cardiac dysrhythmia, renal failure, diabetes, congestive heart failure (CHF), and coronary artery disease (CAD).
The first obtained ECG was abstracted by one of five research study physicians blinded to all clinical data. Research study physicians demonstrated high interrater reliability (kappa > 0.80) in distinguishing normal from abnormal ECGs in a training set of 50 ECGs. Abnormal ECG interpretations included nonsinus rhythms (including paced rhythms), multiple premature ventricular complexes, sinus bradycardias (≤40 bpm), ventricular hypertrophies, short PR segment intervals (<100 milliseconds [ms]), axis deviations, first degree blocks (>200 ms), complete bundle branch blocks, Brugada patterns, Wolff-Parkinson-White patterns, abnormal QRS duration (>120 ms) or abnormal QTc prolongations (>450 ms), and Q/ST/T segment abnormalities suggestive of acute or chronic ischemia.
Statistical Analysis
We calculated descriptive statistics for each predictor variable, stratified by the presence or absence of TTE findings. Chi-square and t-tests were used to test associations between categorical or continuous variables and TTE findings using a significance level of 0.05 and 2-sided hypothesis testing. To identify a robust set of predictors of the primary outcome, we used multivariate logistic regression with the LASSO (Least Absolute Shrinkage and Selection Operator) to fit a parsimonious model.22 The LASSO selects variables and shrinks the associated coefficients to avoid overfitting.23-25 We then used a bootstrap to generate confidence intervals for coefficient estimates. Cases with missing echocardiography reports were excluded from the analysis. Bootstrap results were summarized as the percentage of bootstrap iterations in which each variable’s coefficient was 1) chosen and negative, 2) shrunk to zero, or 3) chosen and positive.
We assessed different weighting schemes to generate a risk score from significant variables identified by regression modeling. These included weighting by regression coefficients rounded to the nearest integer and simple summation of the presence or absence of each variable.
Based on these results, a predictive score was developed to risk stratify patients on their probability of major, clinically significant findings on TTE. The sensitivity and specificity of a score of zero to predict findings on TTE was calculated. For confidence intervals, we used Wilson’s method for binomial confidence intervals.26 The receiver operating characteristic (ROC) curve and its associated area under the curve (AUC) were calculated, and a confidence interval for the AUC was obtained through bootstrap resampling with 2,000 iterations. As part of our sensitivity analyses, we also calculated the ROC curve and AUC after excluding the patients with a known history of CHF and significant finding on TTE. Data analyses were performed in R.27 Two sensitivity analyses were performed: 1) we used multiple imputation to impute 1,000 complete data sets and then used the same LASSO methodology as with the complete data to assess whether incorporating missing data changed the results; and 2) we simulated a conventional troponin assay by raising the positive threshold for hs-TnT to >30 pg/mL (corresponding to the limit of detection for conventional troponin).28
RESULTS
Characteristics of Study Subjects
Patient screening occurred from April 2013 to September 2016. There were 6,930 patients who met eligibility criteria, of whom 3,686 (53%) consented and enrolled in the study (See Figure 1). Of these, 995 (27%) received TTE. The mean age of patients receiving TTE was 74 years; 55% were male. Characteristics of patients obtaining and not obtaining TTE are presented in Appendix Table 2. Patients who received TTE were more likely to be older, have abnormal heart sounds, abnormal EKGs, elevated hs-TnT, elevated NT-proBNP, and have a history of CHF. Of the 995 subjects receiving TTE, 215 (21.6%) had a major, clinically significant finding.
Main Results
Univariate analysis identified 14 variables significantly associated with major findings on TTE. These included male gender, shortness of breath, abnormal heart sounds, history of renal failure, diabetes, CHF, CAD, abnormal ECG, and elevated cardiac biomarkers, among others (See Table 1). The most common major finding on TTE was regional wall motion abnormality, followed by reduced left ventricular ejection fraction (See Table 2). Of the 995 patients who received TTE, 20 (2%) were discharged directly from the ED, 444 (45%) were observed, and 531 (53%) were admitted. On average, patients who received TTE had a longer length of stay than did those that did not (3.4 days vs 1.9 days).
LASSO multivariable logistic regression produced five predictors associated with major findings on TTE: 1) history of CHF, 2) history of CAD, 3) abnormal ECG, 4) hs-TnT above 14 pg/mL, and 5) NT-proBNP above 125 pg/mL (See Table 3).
These five high-risk clinical variables retained their importance after multivariate analysis and form the ROMEO score.
The sensitivity and specificity of a ROMEO score of zero for excluding major findings on TTE was 99.5% (95% CI: 97.4%-99.9%) and 15.4% (95% CI: 13.0%-18.1%), respectively. Patients with a ROMEO score of 0 were at very low risk of having a major finding on TTE: 0.8% (95% CI: 0.02%-4.5%; Appendix Table 3). Only one out of 121 patients with none of the ROMEO criteria was found to have a major finding on TTE (regional wall motion abnormality). Patients with a score of 1 or more were at moderate-to-high risk of having a major finding (7.3% to 55.6%).
There was a linear relationship between the ROMEO score and probability of major findings on TTE (See Appendix Figure 1). The AUC was 0.77 (95% CI = 0.72-0.79) indicating good accuracy of the combination of the five high-risk clinical variables to predict major findings on TTE (See Appendix Figure 2). After excluding the 72 patients with known CHF and significant findings on TTE, the AUC was similar: 0.73 (95% CI: 0.69-0.77). There were 139 patients with at least one missing variable (14%) (See Appendix Table 4). A multiple imputation sensitivity analysis identified the same five high-risk clinical variables in 85% of imputations.
There were 253 patients with high-sensitivity troponin levels between 15 and 30 pg/mL (inclusive). Using a higher hs-TnT threshold (>30 pg/mL) to simulate a conventional troponin assay again identified the same five high-risk variables along with shortness of breath as a potential sixth variable though with an odds ratio approaching unity (See Appendix Table 5). The ROMEO score would have missed two additional patients with major findings if the troponin cutoff were raised to 30 pg/mL from 14 pg/mL, ie, it would have identified 212/215 (98.6%) of the major findings rather than 214/215 (99.5%).
DISCUSSION
Older adults with syncope often present to the ED and undergo a variety of diagnostic tests, including TTE, and a significant proportion are admitted to the hospital.2 There is currently no standardized, evidence-based approach to guide TTE ordering for these patients. Using a large, prospective dataset of syncope patients, we sought to develop a risk-stratification tool to help clinicians identify which syncope patients would be at very low risk for clinically significant findings on TTE. We found that in the absence of these five high-risk clinical variables, the rate of significant findings on TTE in our sample was less than 1%. All five high-risk variables included in the tool remained predictive in our sensitivity analyses, speaking to the robustness of our model.
Other retrospective, and smaller prospective, studies have identified a combination of low-risk criteria including: a normal ECG alone,15 a normal physical exam and normal ECG,14,17 a negative cardiac history and normal ECG.16 Han et al. performed a chart review of 241 patients presenting to the ED with syncope and identified three risk factors for abnormal TTE findings using multiple logistic regression: age, abnormal ECG, and BNP greater than 100 pg/mL.13 While these studies’ results are generally consistent with ours, the retrospective nature and small sample size of these studies limit the generalizability of these results. Thus, using a large, multicenter prospective dataset, we derived a clinical decision instrument (the ROMEO score) to determine which older adults with syncope are at very low risk for major, clinically significant findings on TTE.
Our results add to the recent American College of Cardiology/American Heart Association/Heart Rhythm Society guidelines on the management of syncope which recommend TTE in “selected patients presenting with syncope if structural heart disease is suspected.”18 Our risk-stratification tool offers a simple, standardized approach to determine specifically when to defer TTE testing.
Our findings can guide clinicians in deciding when to obtain TTE for ED syncope patients in the following way: Older adults presenting with syncope or near-syncope to the ED who have none of the ROMEO criteria are at extremely low risk for clinically significant findings on TTE and thus need not undergo such testing solely because of the syncopal event. Patients who have only one or more high-risk clinical variables are at higher risk (7.3%-56%) of significant TTE findings. In this subset, other factors, (eg, physician gestalt, recent previous echocardiography, patient preference, availability of echocardiography) can help guide TTE ordering. Patients with a greater number of high-risk variables may benefit from a more urgent echocardiographic evaluation.
Although on average, patients undergoing TTE had a longer length of stay than those that did not, this finding does not necessarily imply that ordering a TTE was the cause of the increased length of stay. It is possible that this positive association was due to greater underlying medical complexity or acuity of illness that resulted in a greater likelihood of admission/observation, and in turn, a greater length of stay.
Prior to implementation, our results should be externally validated in other clinical settings. In the interim, this risk-stratification tool may be used by clinicians, in conjunction with clinical judgement, to help guide the appropriate use of TTE in older adults presenting with syncope.
Our study has certain limitations. As we only enrolled patients 60 years and older, our findings may not necessarily be valid in younger populations of syncope patients. However, structural heart disease is less common in younger patients and is generally more of a concern for clinicians when evaluating syncope patients in the older age range.29 In our study, 47% of eligible patients declined to participate and thus sampling bias may have occurred. TTEs were ordered at the discretion of treating providers, which was likely subject to physician, institutional, and regional variation; the prevalence of major TTE findings may be lower in the overall cohort than in patients who received TTE. Prior TTE reports were not available; therefore, we were not able to determine if these major findings were previously known. Importantly, we did not perform an internal or external validation of the ROMEO score due to time and resource constraints. Thus, this study represents a derivation of the score solely and would require external validation prior to clinical implementation. Also, to calculate the ROMEO score, both an hs-TnT and NT-proBNP level must be obtained. Thus, the cost savings of any potential reduction in TTE ordering may be partially offset by the costs of increased laboratory testing. Lastly, hs-TnT assays are not currently widely available in hospitals in the United States; earlier generation cardiac troponin assays may not be a perfect substitute for hs-TnT assays. Our sensitivity analysis using an elevated threshold for hs-TnT attempted to mitigate this limitation and resulted in similar findings.
In summary, this risk-stratification tool, using five simple criteria, could help clinicians determine which older adult syncope patients can safely forgo TTE. If validated, this tool could help optimize resource utilization, and increase the value of healthcare for patients presenting with syncope.
Acknowledgments
The authors would like to thank the research assistants at all 11 sites who enrolled patients and collected data for this study.
Disclosures
Dr. Adler has received research funding from Roche. Dr. Bastani has received research funding from Radiometer and Portola and has been a consultant for Portola. Dr. Baugh has received advisory board and speaker’s fees from Roche, research funding from Janssen and Boehringer Ingelheim, and consulting and advisory board fees from Janssen. Dr. Casterino has received funding from Astra Zeneca. Dr. Clark has received research funding from Radiometer, Ortho Clinical Trials, Janssen, Pfizer, NIH, Portola, Biocryst, Glaxo Smith Klein, Hospital Quality Foundation, and Abbott. She is a consultant for Portola, Janssen, and Hospital Quality Foundation. Dr. Diercks is a consultant for Siemens, Janssen, and Roche has received institutional research support from Novartis, Ortho Scientific, and Roche. Dr. Hollander has received research funding from Alere, Siemens, Roche, Portola, and Trinity. Dr. Hollander has also received royalties from UpToDate. Dr. Nishijima has received an honorarium from Pfizer. Dr. Storrow is a consultant for Siemens and Quidel, has received speaking fees from MCM Education, and is on the Data and Safety Monitoring Board for Trevena. Dr. Sun is a consultant for Medtronic. The other authors report no relevant conflicts of interest.
Funding
This project was supported by the National Heart, Lung, And Blood Institute of the National Institutes of Health under Award Number R01 HL111033. Dr. Probst is supported by the National Heart, Lung, And Blood Institute of the National Institutes of Health under Award Number K23HL132052-02. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. Roche Diagnostics supplied the high-sensitivity troponin-T assays. The sponsoring organizations had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, or review of the manuscript.
Syncope, defined as a transient loss of consciousness and postural tone followed by complete, spontaneous return to neurological baseline, accounts for over 1 million (or approximately 1%) of all emergency department (ED) visits per year in the United States (US).12 Given the breadth of etiologies for syncope, including certain life-threatening conditions, extensive diagnostic evaluation and hospitalization for this complaint is common.3-7 The estimated costs of syncope-related hospitalizations are over $2.4 billion annually in the US.8
The 2011 American College of Cardiology Foundation appropriate use criteria for echocardiography state that syncope is an appropriate indication for transthoracic echocardiography (TTE) even when there are no other symptoms or signs of cardiovascular disease.9 This broad recommendation may be appropriate since a finding of severe valvular disease would generally merit consultation with a cardiothoracic surgeon to assess the potential for surgical intervention.10 However, routine use of echocardiogram in all syncope patients could result in increased healthcare costs, patient discomfort, and incidental findings of unclear significance, while rarely changing diagnosis or management.11,12
In an attempt to reduce potentially unnecessary TTE testing, several studies have tried to identify patients at very low risk of structural heart disease.13-17 These investigations suggest that TTE is not indicated in syncope patients with a normal ECG and a normal cardiac exam. However, this literature is limited by retrospective study design and/or small sample sizes. The 2017 American Heart Association/American College of Cardiology/Heart Rhythm Society syncope guidelines recommend TTE for a patient in whom structural heart disease is suspected, but they are not explicit about how to make this determination. 18 Thus, it is still unclear which syncope patients require TTE since a standardized approach to assessing risk of clinically significant findings on TTE has not yet been rigorously developed.
The objective of this study was to develop a risk-stratification tool to identify older adults at very low risk of having a major, clinically significant finding on rest TTE after presenting to the ED with syncope or near-syncope. Using clinical, ECG, and cardiac biomarker data, we created the ROMEO (Risk Of Major Echocardiography findings in Older adults with syncope) score to help optimize resource utilization for syncope.
METHODS
Study Design and Setting
We conducted a large, multicenter, prospective, observational cohort study of older adults who presented to an ED with syncope or near-syncope (ClinicalTrials.gov identifier: NCT01802398). The study was approved by the institutional review boards at all sites and written informed consent was obtained from all participating subjects. The study was conducted at 11 academic EDs across the US (See Appendix Table 1).
Study Population
Patient inclusion criteria for eligibility were age ≥60 years with a complaint of syncope or near-syncope. Syncope was defined as transient loss of consciousness, associated with postural loss of tone, with immediate, spontaneous, and complete recovery. Near syncope was defined as the sensation of imminent loss of consciousness. Patients were excluded if their symptoms were thought to be due to intoxication, seizure, stroke, head trauma, or hypoglycemia. Additional exclusion criteria were the need for medical intervention to restore consciousness (eg, defibrillation), new or worsening confusion, and inability to obtain informed consent from the patient or a legally authorized representative.
This analysis included only patients who received a TTE during the index visit (either in the ED, observation unit, or while admitted to the hospital). This dataset was also used for other analyses addressing questions relevant to the ED management of syncope.
Measurements
All patients underwent a standardized history, physical examination, laboratory, and 12-lead ECG testing. Trained research assistants (RA) directly queried patients about symptoms associated with the syncopal episode. Data on the patient’s past medical history, medications, and physical examination findings were collected prospectively from treating providers.
Research staff obtained blood samples for testing at a core laboratory (University of Rochester, Rochester, NY). Two assays were performed using the Roche Elecsys platform: N-terminal pro B-type natriuretic peptide (NT-proBNP) and the 5th generation high-sensitivity cardiac troponin T (hs-TnT). NT-proBNP was classified as abnormal above a cutoff of 125 pg/mL. Hs-TnT was classified as abnormal above the 99th percentile for a reference population (14 pg/mL). Although hs-TnT was not approved by the U.S. Food and Drug Administration (FDA) at the time of the study, we anticipated that this assay would receive approval and be integrated into future standard of care (FDA approval was granted in January 2017). Rest TTEs were ordered at the discretion of the treating providers.
Outcome Measures
The primary outcome for this secondary analysis was a major, clinically significant finding on TTE.13,14,16,19 These included severe aortic stenosis (<1 cm2), severe mitral stenosis, severe aortic/mitral regurgitation, reduced ejection fraction (defined either quantitatively as less than 45% or qualitatively as “severe left ventricular dysfunction”), hypertrophic cardiomyopathy with outflow tract obstruction, severe pulmonary hypertension, right ventricular dysfunction/strain, large pericardial effusion, atrial myxoma, or regional wall motion abnormalities.
All echocardiogram reports were independently reviewed by two research physicians. Discrepant reviews were resolved by the research physicians and two of the study investigators (BS, CB). Of note, all the TTEs obtained were formal echocardiographic studies, not bedside ultrasonography performed by the emergency physician.
Candidate Predictors
Potential candidate predictors were identified through a prior expert panel process.20,21 Candidate predictors included age, sex, abnormal heart sounds, exertional syncope, shortness of breath, chest pain, near-syncope, family history of sudden cardiac death, high (>180 mm Hg) or low (<90 mm Hg) systolic blood pressure, abnormal ECG, elevated hs-TnT, elevated NT-proBNP, and history of the following: hypertension, cardiac dysrhythmia, renal failure, diabetes, congestive heart failure (CHF), and coronary artery disease (CAD).
The first obtained ECG was abstracted by one of five research study physicians blinded to all clinical data. Research study physicians demonstrated high interrater reliability (kappa > 0.80) in distinguishing normal from abnormal ECGs in a training set of 50 ECGs. Abnormal ECG interpretations included nonsinus rhythms (including paced rhythms), multiple premature ventricular complexes, sinus bradycardias (≤40 bpm), ventricular hypertrophies, short PR segment intervals (<100 milliseconds [ms]), axis deviations, first degree blocks (>200 ms), complete bundle branch blocks, Brugada patterns, Wolff-Parkinson-White patterns, abnormal QRS duration (>120 ms) or abnormal QTc prolongations (>450 ms), and Q/ST/T segment abnormalities suggestive of acute or chronic ischemia.
Statistical Analysis
We calculated descriptive statistics for each predictor variable, stratified by the presence or absence of TTE findings. Chi-square and t-tests were used to test associations between categorical or continuous variables and TTE findings using a significance level of 0.05 and 2-sided hypothesis testing. To identify a robust set of predictors of the primary outcome, we used multivariate logistic regression with the LASSO (Least Absolute Shrinkage and Selection Operator) to fit a parsimonious model.22 The LASSO selects variables and shrinks the associated coefficients to avoid overfitting.23-25 We then used a bootstrap to generate confidence intervals for coefficient estimates. Cases with missing echocardiography reports were excluded from the analysis. Bootstrap results were summarized as the percentage of bootstrap iterations in which each variable’s coefficient was 1) chosen and negative, 2) shrunk to zero, or 3) chosen and positive.
We assessed different weighting schemes to generate a risk score from significant variables identified by regression modeling. These included weighting by regression coefficients rounded to the nearest integer and simple summation of the presence or absence of each variable.
Based on these results, a predictive score was developed to risk stratify patients on their probability of major, clinically significant findings on TTE. The sensitivity and specificity of a score of zero to predict findings on TTE was calculated. For confidence intervals, we used Wilson’s method for binomial confidence intervals.26 The receiver operating characteristic (ROC) curve and its associated area under the curve (AUC) were calculated, and a confidence interval for the AUC was obtained through bootstrap resampling with 2,000 iterations. As part of our sensitivity analyses, we also calculated the ROC curve and AUC after excluding the patients with a known history of CHF and significant finding on TTE. Data analyses were performed in R.27 Two sensitivity analyses were performed: 1) we used multiple imputation to impute 1,000 complete data sets and then used the same LASSO methodology as with the complete data to assess whether incorporating missing data changed the results; and 2) we simulated a conventional troponin assay by raising the positive threshold for hs-TnT to >30 pg/mL (corresponding to the limit of detection for conventional troponin).28
RESULTS
Characteristics of Study Subjects
Patient screening occurred from April 2013 to September 2016. There were 6,930 patients who met eligibility criteria, of whom 3,686 (53%) consented and enrolled in the study (See Figure 1). Of these, 995 (27%) received TTE. The mean age of patients receiving TTE was 74 years; 55% were male. Characteristics of patients obtaining and not obtaining TTE are presented in Appendix Table 2. Patients who received TTE were more likely to be older, have abnormal heart sounds, abnormal EKGs, elevated hs-TnT, elevated NT-proBNP, and have a history of CHF. Of the 995 subjects receiving TTE, 215 (21.6%) had a major, clinically significant finding.
Main Results
Univariate analysis identified 14 variables significantly associated with major findings on TTE. These included male gender, shortness of breath, abnormal heart sounds, history of renal failure, diabetes, CHF, CAD, abnormal ECG, and elevated cardiac biomarkers, among others (See Table 1). The most common major finding on TTE was regional wall motion abnormality, followed by reduced left ventricular ejection fraction (See Table 2). Of the 995 patients who received TTE, 20 (2%) were discharged directly from the ED, 444 (45%) were observed, and 531 (53%) were admitted. On average, patients who received TTE had a longer length of stay than did those that did not (3.4 days vs 1.9 days).
LASSO multivariable logistic regression produced five predictors associated with major findings on TTE: 1) history of CHF, 2) history of CAD, 3) abnormal ECG, 4) hs-TnT above 14 pg/mL, and 5) NT-proBNP above 125 pg/mL (See Table 3).
These five high-risk clinical variables retained their importance after multivariate analysis and form the ROMEO score.
The sensitivity and specificity of a ROMEO score of zero for excluding major findings on TTE was 99.5% (95% CI: 97.4%-99.9%) and 15.4% (95% CI: 13.0%-18.1%), respectively. Patients with a ROMEO score of 0 were at very low risk of having a major finding on TTE: 0.8% (95% CI: 0.02%-4.5%; Appendix Table 3). Only one out of 121 patients with none of the ROMEO criteria was found to have a major finding on TTE (regional wall motion abnormality). Patients with a score of 1 or more were at moderate-to-high risk of having a major finding (7.3% to 55.6%).
There was a linear relationship between the ROMEO score and probability of major findings on TTE (See Appendix Figure 1). The AUC was 0.77 (95% CI = 0.72-0.79) indicating good accuracy of the combination of the five high-risk clinical variables to predict major findings on TTE (See Appendix Figure 2). After excluding the 72 patients with known CHF and significant findings on TTE, the AUC was similar: 0.73 (95% CI: 0.69-0.77). There were 139 patients with at least one missing variable (14%) (See Appendix Table 4). A multiple imputation sensitivity analysis identified the same five high-risk clinical variables in 85% of imputations.
There were 253 patients with high-sensitivity troponin levels between 15 and 30 pg/mL (inclusive). Using a higher hs-TnT threshold (>30 pg/mL) to simulate a conventional troponin assay again identified the same five high-risk variables along with shortness of breath as a potential sixth variable though with an odds ratio approaching unity (See Appendix Table 5). The ROMEO score would have missed two additional patients with major findings if the troponin cutoff were raised to 30 pg/mL from 14 pg/mL, ie, it would have identified 212/215 (98.6%) of the major findings rather than 214/215 (99.5%).
DISCUSSION
Older adults with syncope often present to the ED and undergo a variety of diagnostic tests, including TTE, and a significant proportion are admitted to the hospital.2 There is currently no standardized, evidence-based approach to guide TTE ordering for these patients. Using a large, prospective dataset of syncope patients, we sought to develop a risk-stratification tool to help clinicians identify which syncope patients would be at very low risk for clinically significant findings on TTE. We found that in the absence of these five high-risk clinical variables, the rate of significant findings on TTE in our sample was less than 1%. All five high-risk variables included in the tool remained predictive in our sensitivity analyses, speaking to the robustness of our model.
Other retrospective, and smaller prospective, studies have identified a combination of low-risk criteria including: a normal ECG alone,15 a normal physical exam and normal ECG,14,17 a negative cardiac history and normal ECG.16 Han et al. performed a chart review of 241 patients presenting to the ED with syncope and identified three risk factors for abnormal TTE findings using multiple logistic regression: age, abnormal ECG, and BNP greater than 100 pg/mL.13 While these studies’ results are generally consistent with ours, the retrospective nature and small sample size of these studies limit the generalizability of these results. Thus, using a large, multicenter prospective dataset, we derived a clinical decision instrument (the ROMEO score) to determine which older adults with syncope are at very low risk for major, clinically significant findings on TTE.
Our results add to the recent American College of Cardiology/American Heart Association/Heart Rhythm Society guidelines on the management of syncope which recommend TTE in “selected patients presenting with syncope if structural heart disease is suspected.”18 Our risk-stratification tool offers a simple, standardized approach to determine specifically when to defer TTE testing.
Our findings can guide clinicians in deciding when to obtain TTE for ED syncope patients in the following way: Older adults presenting with syncope or near-syncope to the ED who have none of the ROMEO criteria are at extremely low risk for clinically significant findings on TTE and thus need not undergo such testing solely because of the syncopal event. Patients who have only one or more high-risk clinical variables are at higher risk (7.3%-56%) of significant TTE findings. In this subset, other factors, (eg, physician gestalt, recent previous echocardiography, patient preference, availability of echocardiography) can help guide TTE ordering. Patients with a greater number of high-risk variables may benefit from a more urgent echocardiographic evaluation.
Although on average, patients undergoing TTE had a longer length of stay than those that did not, this finding does not necessarily imply that ordering a TTE was the cause of the increased length of stay. It is possible that this positive association was due to greater underlying medical complexity or acuity of illness that resulted in a greater likelihood of admission/observation, and in turn, a greater length of stay.
Prior to implementation, our results should be externally validated in other clinical settings. In the interim, this risk-stratification tool may be used by clinicians, in conjunction with clinical judgement, to help guide the appropriate use of TTE in older adults presenting with syncope.
Our study has certain limitations. As we only enrolled patients 60 years and older, our findings may not necessarily be valid in younger populations of syncope patients. However, structural heart disease is less common in younger patients and is generally more of a concern for clinicians when evaluating syncope patients in the older age range.29 In our study, 47% of eligible patients declined to participate and thus sampling bias may have occurred. TTEs were ordered at the discretion of treating providers, which was likely subject to physician, institutional, and regional variation; the prevalence of major TTE findings may be lower in the overall cohort than in patients who received TTE. Prior TTE reports were not available; therefore, we were not able to determine if these major findings were previously known. Importantly, we did not perform an internal or external validation of the ROMEO score due to time and resource constraints. Thus, this study represents a derivation of the score solely and would require external validation prior to clinical implementation. Also, to calculate the ROMEO score, both an hs-TnT and NT-proBNP level must be obtained. Thus, the cost savings of any potential reduction in TTE ordering may be partially offset by the costs of increased laboratory testing. Lastly, hs-TnT assays are not currently widely available in hospitals in the United States; earlier generation cardiac troponin assays may not be a perfect substitute for hs-TnT assays. Our sensitivity analysis using an elevated threshold for hs-TnT attempted to mitigate this limitation and resulted in similar findings.
In summary, this risk-stratification tool, using five simple criteria, could help clinicians determine which older adult syncope patients can safely forgo TTE. If validated, this tool could help optimize resource utilization, and increase the value of healthcare for patients presenting with syncope.
Acknowledgments
The authors would like to thank the research assistants at all 11 sites who enrolled patients and collected data for this study.
Disclosures
Dr. Adler has received research funding from Roche. Dr. Bastani has received research funding from Radiometer and Portola and has been a consultant for Portola. Dr. Baugh has received advisory board and speaker’s fees from Roche, research funding from Janssen and Boehringer Ingelheim, and consulting and advisory board fees from Janssen. Dr. Casterino has received funding from Astra Zeneca. Dr. Clark has received research funding from Radiometer, Ortho Clinical Trials, Janssen, Pfizer, NIH, Portola, Biocryst, Glaxo Smith Klein, Hospital Quality Foundation, and Abbott. She is a consultant for Portola, Janssen, and Hospital Quality Foundation. Dr. Diercks is a consultant for Siemens, Janssen, and Roche has received institutional research support from Novartis, Ortho Scientific, and Roche. Dr. Hollander has received research funding from Alere, Siemens, Roche, Portola, and Trinity. Dr. Hollander has also received royalties from UpToDate. Dr. Nishijima has received an honorarium from Pfizer. Dr. Storrow is a consultant for Siemens and Quidel, has received speaking fees from MCM Education, and is on the Data and Safety Monitoring Board for Trevena. Dr. Sun is a consultant for Medtronic. The other authors report no relevant conflicts of interest.
Funding
This project was supported by the National Heart, Lung, And Blood Institute of the National Institutes of Health under Award Number R01 HL111033. Dr. Probst is supported by the National Heart, Lung, And Blood Institute of the National Institutes of Health under Award Number K23HL132052-02. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. Roche Diagnostics supplied the high-sensitivity troponin-T assays. The sponsoring organizations had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, or review of the manuscript.
1. Sun BC, Emond JA, Camargo CA, Jr. Characteristics and admission patterns of patients presenting with syncope to U.S. emergency departments, 1992-2000. Acad Emerg Med. 2004;11(10):1029-1034. doi: 10.1197/j.aem.2004.05.032. PubMed
2. Probst MA, Kanzaria HK, Gbedemah M, Richardson LD, Sun BC. National trends in resource utilization associated with ED visits for syncope. Am J Emerg Med. 2015;33(8):998-1001. doi: 10.1016/j.ajem.2015.04.030. PubMed
3. Kapoor WN, Karpf M, Maher Y, Miller RA, Levey GS. Syncope of unknown origin. The need for a more cost-effective approach to its diagnosis evaluation. JAMA. 1982;247(19):2687-2691. doi: 10.1001/jama.247.19.2687. PubMed
4. Pires LA, Ganji JR, Jarandila R, Steele R. Diagnostic patterns and temporal trends in the evaluation of adult patients hospitalized with syncope. Arch Intern Med. 2001;161(15):1889-1895. doi: 10.1001/archinte.161.15.1889. PubMed
5. Quinn JV, Stiell IG, McDermott DA, Sellers KL, Kohn MA, Wells GA. Derivation of the San Francisco Syncope Rule to predict patients with short-term serious outcomes. Ann Emerg Med. 2004;43(2):224-232. doi: 10.1016/S0196064403008230. PubMed
6. Linzer M, Yang EH, Estes NA, 3rd, Wang P, Vorperian VR, Kapoor WN. Diagnosing syncope. Part 1: Value of history, physical examination, and electrocardiography. Clinical Efficacy Assessment Project of the American College of Physicians. Ann Intern Med. 1997;126(12):989-996. doi: 10.7326/0003-4819-126-12-199706150-00012. PubMed
7. Linzer M, Yang EH, Estes NA, 3rd, Wang P, Vorperian VR, Kapoor WN. Diagnosing syncope. Part 2: Unexplained syncope. Clinical Efficacy Assessment Project of the American College of Physicians. Ann Intern Med. 1997;127(1):76-86. doi: 10.7326/0003-4819-127-1-199707010-00014. PubMed
8. Sun BC, Emond JA, Camargo CA, Jr. Direct medical costs of syncope-related hospitalizations in the United States. Am J Cardiol. 2005;95(5):668-671. doi: 10.1016/j.amjcard.2004.11.013. PubMed
9. American College of Cardiology Foundation. Appropriate Use Criteria Task F, American Society of Echocardiography, American Heart Association, et al. ACCF/ASE/AHA/ASNC/HFSA/HRS/SCAI/SCCM/SCCT/SCMR 2011 Appropriate Use Criteria for Echocardiography. A Report of the American College of Cardiology Foundation Appropriate Use Criteria Task Force, American Society of Echocardiography, American Heart Association, American Society of Nuclear Cardiology, Heart Failure Society of America, Heart Rhythm Society, Society for Cardiovascular Angiography and Interventions, Society of Critical Care Medicine, Society of Cardiovascular Computed Tomography, and Society for Cardiovascular Magnetic Resonance Endorsed by the American College of Chest Physicians. J Am Coll Cardiol. 2011;57(9):1126-1166. doi: 10.1016/j.echo.2010.12.008.
10. Maganti K, Rigolin VH, Sarano ME, Bonow RO. Valvular heart disease: diagnosis and management. Mayo Clin Proc. 2010;85(5):483-500. doi: 10.4065/mcp.2009.0706. PubMed
11. Mendu ML, McAvay G, Lampert R, Stoehr J, Tinetti ME. Yield of diagnostic tests in evaluating syncopal episodes in older patients. Arch Intern Med. 2009;169(14):1299-1305. doi: 10.1001/archinternmed.2009.204. PubMed
12. Madeira CL, Craig MJ, Donohoe A, Stephens JR. Things we do for no reason: echocardiogram in unselected patients with syncope. J Hosp Med. 2017;12(12):984–988. doi: http://dx.doi.org/10.12788/jhm.2864. PubMed
13. Han SK, Yeom SR, Lee SH, et al. Transthoracic echocardiogram in syncope patients with normal initial evaluation. Am J Emerg Med. 2017;35(2):281-284. doi: 10.1016/j.ajem.2016.10.078. PubMed
14. Chang NL, Shah P, Bajaj S, Virk H, Bikkina M, Shamoon F. Diagnostic yield of echocardiography in syncope patients with normal ECG. Cardiol Res Pract. 2016;2016:1251637. doi: http://dx.doi.org/10.1155/2016/1251637. PubMed
15. Anderson KL, Limkakeng A, Damuth E, Chandra A. Cardiac evaluation for structural abnormalities may not be required in patients presenting with syncope and a normal ECG result in an observation unit setting. Ann Emerg Med. 2012;60(4):478–84.e1. doi: 10.1016/j.annemergmed.2012.04.023. PubMed
16. Sarasin FP, Junod AF, Carballo D, Slama S, Unger PF, Louis-Simonet M. Role of echocardiography in the evaluation of syncope: a prospective study. Heart. 2002;88(4):363-367. doi: 10.1136/heart.88.4.363. PubMed
17. Recchia D, Barzilai B. Echocardiography in the evaluation of patients with syncope. J Gen Intern Med. 1995;10(12):649-655. doi: 10.1007/BF02602755. PubMed
18. Shen WK, Sheldon RS, Benditt DG, et al. ACC/AHA/HRS guideline for the evaluation and management of patients With syncope: executive summary: A report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines and the Heart Rhythm Society. J Am Coll Cardiol. 2017;2017(70(5)):620-663. PubMed
19. Chiu DT, Shapiro NI, Sun BC, Mottley JL, Grossman SA. Are echocardiography, telemetry, ambulatory electrocardiography monitoring, and cardiac enzymes in emergency department patients presenting with syncope useful tests? A preliminary investigation. J Emerg Med. 2014;47(1):113-118. doi: 10.1016/j.jemermed.2014.01.018. PubMed
20. Sun BC, Costantino G, Barbic F, et al. Priorities for emergency department syncope research. Ann Emerg Med. 2014;64(6):649–55.e2. doi: 10.1016/j.annemergmed.2014.04.014. PubMed
21. Sun BC, Derose SF, Liang LJ, et al. Predictors of 30-day serious events in older patients with syncope. Ann Emerg Med. 2009;54(6):769–778.e1-5. doi: 10.1016/j.annemergmed.2009.07.027. PubMed
22. Tibshirani R. Regression shrinkage and selection via the lasso. J R Stat Soc. 1996;58(1):267-288.
23. Friedman J, Hastie T, Tibshirani R. He Elements of Statistical Learning;Vol 1. New York, NY: Springer-Verlag; 2001. PubMed
24. Friedman J, Hastie T, Tibshirani R. Regularization paths for generalized linear models via coordinate descent. J Stat Softw. 2010;33(1):1-22. doi: 10.18637/jss.v033.i01. PubMed
25. James G, Witten D, Hastie T, Tibshirani R. An Introduction to Statistical Learning;Vol 112. New York, NY: Springer-Verlag; 2013.
26. Wilson EB. Probable inference, the law of succession, and statistical inference. J Am Stat Assoc. 1927 ;22(158):209-212. doi: 10.1080/01621459.1927.10502953. PubMed
27. R Core Team (2018). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL http://www.R-project.org/.
28. Chew DP, Zeitz C, Worthley M, et al. Randomized comparison of high-sensitivity troponin reporting in undifferentiated chest pain assessment. Circ Cardiovasc Qual Outcomes. 2016;9(5):542-553. doi: 10.1161/CIRCOUTCOMES.115.002488. PubMed
29. Chen RS, Bivens MJ, Grossman SA. Diagnosis and management of valvular heart disease in emergency medicine. Emerg Med Clin North Am. 2011;29(4):801–10, vii. doi: 10.1016/j.emc.2011.08.001. PubMed
1. Sun BC, Emond JA, Camargo CA, Jr. Characteristics and admission patterns of patients presenting with syncope to U.S. emergency departments, 1992-2000. Acad Emerg Med. 2004;11(10):1029-1034. doi: 10.1197/j.aem.2004.05.032. PubMed
2. Probst MA, Kanzaria HK, Gbedemah M, Richardson LD, Sun BC. National trends in resource utilization associated with ED visits for syncope. Am J Emerg Med. 2015;33(8):998-1001. doi: 10.1016/j.ajem.2015.04.030. PubMed
3. Kapoor WN, Karpf M, Maher Y, Miller RA, Levey GS. Syncope of unknown origin. The need for a more cost-effective approach to its diagnosis evaluation. JAMA. 1982;247(19):2687-2691. doi: 10.1001/jama.247.19.2687. PubMed
4. Pires LA, Ganji JR, Jarandila R, Steele R. Diagnostic patterns and temporal trends in the evaluation of adult patients hospitalized with syncope. Arch Intern Med. 2001;161(15):1889-1895. doi: 10.1001/archinte.161.15.1889. PubMed
5. Quinn JV, Stiell IG, McDermott DA, Sellers KL, Kohn MA, Wells GA. Derivation of the San Francisco Syncope Rule to predict patients with short-term serious outcomes. Ann Emerg Med. 2004;43(2):224-232. doi: 10.1016/S0196064403008230. PubMed
6. Linzer M, Yang EH, Estes NA, 3rd, Wang P, Vorperian VR, Kapoor WN. Diagnosing syncope. Part 1: Value of history, physical examination, and electrocardiography. Clinical Efficacy Assessment Project of the American College of Physicians. Ann Intern Med. 1997;126(12):989-996. doi: 10.7326/0003-4819-126-12-199706150-00012. PubMed
7. Linzer M, Yang EH, Estes NA, 3rd, Wang P, Vorperian VR, Kapoor WN. Diagnosing syncope. Part 2: Unexplained syncope. Clinical Efficacy Assessment Project of the American College of Physicians. Ann Intern Med. 1997;127(1):76-86. doi: 10.7326/0003-4819-127-1-199707010-00014. PubMed
8. Sun BC, Emond JA, Camargo CA, Jr. Direct medical costs of syncope-related hospitalizations in the United States. Am J Cardiol. 2005;95(5):668-671. doi: 10.1016/j.amjcard.2004.11.013. PubMed
9. American College of Cardiology Foundation. Appropriate Use Criteria Task F, American Society of Echocardiography, American Heart Association, et al. ACCF/ASE/AHA/ASNC/HFSA/HRS/SCAI/SCCM/SCCT/SCMR 2011 Appropriate Use Criteria for Echocardiography. A Report of the American College of Cardiology Foundation Appropriate Use Criteria Task Force, American Society of Echocardiography, American Heart Association, American Society of Nuclear Cardiology, Heart Failure Society of America, Heart Rhythm Society, Society for Cardiovascular Angiography and Interventions, Society of Critical Care Medicine, Society of Cardiovascular Computed Tomography, and Society for Cardiovascular Magnetic Resonance Endorsed by the American College of Chest Physicians. J Am Coll Cardiol. 2011;57(9):1126-1166. doi: 10.1016/j.echo.2010.12.008.
10. Maganti K, Rigolin VH, Sarano ME, Bonow RO. Valvular heart disease: diagnosis and management. Mayo Clin Proc. 2010;85(5):483-500. doi: 10.4065/mcp.2009.0706. PubMed
11. Mendu ML, McAvay G, Lampert R, Stoehr J, Tinetti ME. Yield of diagnostic tests in evaluating syncopal episodes in older patients. Arch Intern Med. 2009;169(14):1299-1305. doi: 10.1001/archinternmed.2009.204. PubMed
12. Madeira CL, Craig MJ, Donohoe A, Stephens JR. Things we do for no reason: echocardiogram in unselected patients with syncope. J Hosp Med. 2017;12(12):984–988. doi: http://dx.doi.org/10.12788/jhm.2864. PubMed
13. Han SK, Yeom SR, Lee SH, et al. Transthoracic echocardiogram in syncope patients with normal initial evaluation. Am J Emerg Med. 2017;35(2):281-284. doi: 10.1016/j.ajem.2016.10.078. PubMed
14. Chang NL, Shah P, Bajaj S, Virk H, Bikkina M, Shamoon F. Diagnostic yield of echocardiography in syncope patients with normal ECG. Cardiol Res Pract. 2016;2016:1251637. doi: http://dx.doi.org/10.1155/2016/1251637. PubMed
15. Anderson KL, Limkakeng A, Damuth E, Chandra A. Cardiac evaluation for structural abnormalities may not be required in patients presenting with syncope and a normal ECG result in an observation unit setting. Ann Emerg Med. 2012;60(4):478–84.e1. doi: 10.1016/j.annemergmed.2012.04.023. PubMed
16. Sarasin FP, Junod AF, Carballo D, Slama S, Unger PF, Louis-Simonet M. Role of echocardiography in the evaluation of syncope: a prospective study. Heart. 2002;88(4):363-367. doi: 10.1136/heart.88.4.363. PubMed
17. Recchia D, Barzilai B. Echocardiography in the evaluation of patients with syncope. J Gen Intern Med. 1995;10(12):649-655. doi: 10.1007/BF02602755. PubMed
18. Shen WK, Sheldon RS, Benditt DG, et al. ACC/AHA/HRS guideline for the evaluation and management of patients With syncope: executive summary: A report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines and the Heart Rhythm Society. J Am Coll Cardiol. 2017;2017(70(5)):620-663. PubMed
19. Chiu DT, Shapiro NI, Sun BC, Mottley JL, Grossman SA. Are echocardiography, telemetry, ambulatory electrocardiography monitoring, and cardiac enzymes in emergency department patients presenting with syncope useful tests? A preliminary investigation. J Emerg Med. 2014;47(1):113-118. doi: 10.1016/j.jemermed.2014.01.018. PubMed
20. Sun BC, Costantino G, Barbic F, et al. Priorities for emergency department syncope research. Ann Emerg Med. 2014;64(6):649–55.e2. doi: 10.1016/j.annemergmed.2014.04.014. PubMed
21. Sun BC, Derose SF, Liang LJ, et al. Predictors of 30-day serious events in older patients with syncope. Ann Emerg Med. 2009;54(6):769–778.e1-5. doi: 10.1016/j.annemergmed.2009.07.027. PubMed
22. Tibshirani R. Regression shrinkage and selection via the lasso. J R Stat Soc. 1996;58(1):267-288.
23. Friedman J, Hastie T, Tibshirani R. He Elements of Statistical Learning;Vol 1. New York, NY: Springer-Verlag; 2001. PubMed
24. Friedman J, Hastie T, Tibshirani R. Regularization paths for generalized linear models via coordinate descent. J Stat Softw. 2010;33(1):1-22. doi: 10.18637/jss.v033.i01. PubMed
25. James G, Witten D, Hastie T, Tibshirani R. An Introduction to Statistical Learning;Vol 112. New York, NY: Springer-Verlag; 2013.
26. Wilson EB. Probable inference, the law of succession, and statistical inference. J Am Stat Assoc. 1927 ;22(158):209-212. doi: 10.1080/01621459.1927.10502953. PubMed
27. R Core Team (2018). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL http://www.R-project.org/.
28. Chew DP, Zeitz C, Worthley M, et al. Randomized comparison of high-sensitivity troponin reporting in undifferentiated chest pain assessment. Circ Cardiovasc Qual Outcomes. 2016;9(5):542-553. doi: 10.1161/CIRCOUTCOMES.115.002488. PubMed
29. Chen RS, Bivens MJ, Grossman SA. Diagnosis and management of valvular heart disease in emergency medicine. Emerg Med Clin North Am. 2011;29(4):801–10, vii. doi: 10.1016/j.emc.2011.08.001. PubMed
© 2018 Society of Hospital Medicine
Impaired Arousal and Mortality
Arousal is defined as the patient's overall level of responsiveness to the environment. Its assessment is standard of care in most intensive care units (ICUs) to monitor depth of sedation and underlying brain dysfunction. There has been recent interest in expanding the role of arousal assessment beyond the ICU. Specifically, the Veterans Affairs Delirium Working Group proposed that simple arousal assessment be a vital sign to quantify underlying brain dysfunction.[1] The rationale is that impaired arousal is closely linked with delirium,[2] and is an integral component of multiple delirium assessments.[3, 4, 5] Chester et al. observed that the presence of impaired arousal was 64% sensitive and 93% specific for delirium diagnosed by a psychiatrist.[2] Delirium is an under‐recognized public health problem that affects up to 25% of older hospitalized patients,[6, 7] is associated with a multitude of adverse outcomes such as death and accelerated cognitive decline,[8] and costs the US healthcare system an excess of $152 billion dollars.[9]
Most delirium assessments require the patient to undergo additional cognitive testing. The assessment of arousal, however, requires the rater to merely observe the patient during routine clinical care and can be easily integrated into the clinical workflow.[10] Because of its simplicity and brevity, assessing arousal alone using validated scales such as the Richmond Agitation‐Sedation Scale (RASS) may be a more appealing alternative to traditional, more complex delirium screening in the acute care setting. Its clinical utility would be further strengthened if impaired arousal was also associated with mortality, and conferred risk even in the absence of delirium. As a result, we sought to determine if impaired arousal at initial presentation in older acutely ill patients predicted 6‐month mortality and whether this relationship was present in the absence of delirium.
METHODS
Design Overview
We performed a planned secondary analysis of 2 prospective cohorts that enrolled patients from May 2007 to August 2008 between 8 am and 10 pm during the weekdays, and July 2009 to February 2012 between 8 am and 4 pm during the weekdays. The first cohort was designed to evaluate the relationship between delirium and patient outcomes.[11, 12] The second cohort was used to validate brief delirium assessments using a psychiatrist's assessment as the reference standard.[5, 13] The local institutional review board approved these studies.
Setting and Participants
These studies were conducted in an urban emergency department located within an academic, tertiary care hospital with over 57,000 visits annually. Patients were included if they were 65 years or older and in the emergency department for <12 hours at the time of enrollment. The 12‐hour cutoff was used to include patients who presented to the emergency department in the evening and early morning hours. Patients were excluded if they were previously enrolled, non‐English speaking, comatose, or were nonverbal and unable to follow simple commands prior to the acute illness. Because the July 2009 to February 2012 cohort was designed to validate delirium assessments with auditory and visual components, patients were also excluded if they were deaf or blind.
Measurement of Arousal
RASS is an arousal scale commonly used in ICUs to assess depth of sedation and ranges from 5 (unarousable) to +4 (combative); 0 represents normal arousal.[10, 14] The RASS simply requires the rater to observe the patient during their routine interactions and does not require any additional cognitive testing. The RASS terms sedation was modified to drowsy (Table 1), because we wanted to capture impaired arousal regardless of sedation administration. We did not use the modified RASS (mRASS) proposed by the Veteran's Affairs Delirium Working Group, because it was published after data collection began.[1] The mRASS is very similar to the RASS, except it also incorporates a very informal inattention assessment. The RASS was ascertained by research assistants who were college students and graduates, and emergency medical technician basics and paramedics. The principal investigator gave them a 5‐minute didactic lecture about the RASS and observed them perform the RASS in at least 5 patients prior to the start of the study. Inter‐rater reliability between trained research assistants and a physician was assessed for 456 (42.0%) patients of the study sample. The weighted kappa of the RASS was 0.61, indicating very good inter‐rater reliability. Because the 81.7% of patients with impaired arousal had a RASS of 1, the RASS dichotomized as normal (RASS=0) or impaired (RASS other than 0).
| Score | Term | Description |
|---|---|---|
| ||
| +4 | Combative | Overtly combative, violent, immediate danger to staff |
| +3 | Very agitated | Pulls or removes tube(s) or catheter(s), aggressive |
| +2 | Agitated | Frequent nonpurposeful movement |
| +1 | Restless | Anxious but movements not aggressive or vigorous |
| 0 | Alert and calm | |
| 1 | Slight drowsy | Not fully alert, but has sustained awakening (eye opening/eye contact) to voice (>10 seconds) |
| 2 | Moderately drowsy | Briefly awakens with eye contact to voice (<10 seconds) |
| 3 | Very drowsy | Movement or eye opening to voice (but no eye contact) |
| 4 | Awakens to pain only | No response to voice, but movement or eye opening to physical stimulation |
| 5 | Unarousable | No response to voice or physical stimulation |
Death Ascertainment
Death within 6 months was ascertained using the following algorithm: (1) The electronic medical record was searched to determine the patient's death status. (2) Patients who had a documented emergency department visit, outpatient clinic visit, or hospitalization after 6 months were considered to be alive at 6 months. (3) For the remaining patients, date of death was searched in the Social Security Death Index (SSDI). (4) Patients without a death recorded in the SSDI 1 year after the index visit was considered to be alive at 6 months. Nine hundred thirty‐one (85.9%) out of 1084 patients had a recorded death in the medical record or SSDI, or had an emergency department or hospital visit documented in their record 6 months after the index visit.
Additional Variables Collected
Patients were considered to have dementia if they had: (1) documented dementia in the medical record, (2) a short form Informant Questionnaire on Cognitive Decline in the Elderly score (IQCODE) greater than 3.38,[15] or (3) prescribed cholinesterase inhibitors prior to admission. The short form IQCODE is an informant questionnaire with 16 items; a cutoff of 3.38 out of 5.00 is 79% sensitive and 82% specific for dementia.[16] Premorbid functional status was determined by the Katz Activities of Daily Living (Katz ADL) and ranges from 0 (completely dependent) to 6 (completely independent).[17] Patients with a score <5 were considered to be functionally dependent. Both the IQCODE and Katz ADL were prospectively collected in the emergency department at the time of enrollment.
The Charlson Comorbidity Index was used to measure comorbid burden.[18] The Acute Physiology Score (APS) of the Acute Physiology and Chronic Health Evaluation II score was used to quantify severity of illness.[19] The Glasgow Coma Scale was not included in the APS because it was not collected. Intravenous, intramuscular, and oral benzodiazepine and opioids given in the prehospital and emergency department were also recorded. The Charlson Comorbidity Index, APS, and benzodiazepine and opioid administration were collected after patient enrollment using the electronic medical record.
Within 3 hours of the RASS, a subset of 406 patients was evaluated by a consultation‐liaison psychiatrist who determined the patient's delirium status using Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, Text Revision (DSM‐IV‐TR) criteria.[20] Details of their comprehensive assessments have been described in a previous report.[5]
Statistical Analysis
Measures of central tendency and dispersion for continuous variables were reported as medians and interquartile ranges. Categorical variables were reported as proportions. For simple comparisons, Wilcoxon rank sum tests were performed for continuous data, and 2 analyses or Fisher exact test were performed for categorical data. To evaluate the predictive validity of impaired arousal on 6‐month mortality, the cumulative probability of survival was estimated within 6 months from the study enrollment date using the Kaplan‐Meier method. Cox proportional hazards regression was performed to assess if impaired arousal was independently associated with 6‐month mortality after adjusting for age, gender, nonwhite race, comorbidity burden (Charlson Comorbidity Index), severity of illness (APS), dementia, functional dependence (Katz ADL <5), nursing home residence, admission status, and benzodiazepine or opioid medication administration. Patients were censored at the end of 6 months. The selection of covariates was based upon expert opinion and literature review. The number of covariates used for the model was limited by the number of events to minimize overfitting; 1 df was allowed for every 10 to 15 events.[21] Because severity of illness, psychoactive medication administration, and admission status might modify the relationship between 6‐month mortality and impaired arousal, 2‐way interaction terms were incorporated. To maintain parsimony and minimize overfitting and collinearity, nonsignificant interaction terms (P>0.20) were removed in the final model.[22] Hazard ratios (HR) with their 95% confidence interval (95% CI) were reported.
To determine if arousal was associated with 6‐month mortality in the absence of delirium, we performed another Cox proportional hazard regression in a subset of 406 patients who received a psychiatrist assessment. Six‐month mortality was the dependent variable, and the independent variable was a 3‐level categorical variable of different arousal/delirium combinations: (1) impaired arousal/delirium positive, (2) impaired arousal/delirium negative, and (3) normal arousal (with or without delirium). Because there were only 8 patients who had normal arousal with delirium, this group was collapsed into the normal arousal without delirium group. Because there were 55 deaths, the number of covariates that could be entered into the Cox proportional hazard regression model was limited. We used the inverse weighted propensity score method to help minimize residual confounding.[23] Traditional propensity score adjustment could not be performed because there were 3 arousal/delirium categories. Similar to propensity score adjustment, inverse weighted propensity score method was used to help balance the distribution of patient characteristics among the exposure groups and also allow adjustment for multiple confounders while minimizing the degrees of freedom expended. A propensity score was the probability of having a particular arousal/delirium category based upon baseline patient characteristics. Multinomial logistic regression was performed to calculate the propensity score, and the baseline covariates used were age, gender, nonwhite race, comorbidity burden, severity of illness, dementia, functional dependence, and nursing home residence. For the Cox proportional hazard regression model, each observation was weighted by the inverse of the propensity score for their given arousal/delirium category; propensity scores exceeding the 95th percentile were trimmed to avoid overly influential weighting. Benzodiazepine and opioid medications given in the emergency department and admission status were adjusted as covariates in the weighted Cox proportional hazard regression model.
Nineteen patients (1.8%) had missing Katz ADL; these missing values were imputed using multiple imputation. The reliability of the final regression models were internally validated using the bootstrap method.[21] Two thousand sets of bootstrap samples were generated by resampling the original data, and the optimism was estimated to determine the degree of overfitting.[21] An optimism value >0.85 indicated no evidence of substantial overfitting.[21] Variance inflation factors were used to check multicollinearity. Schoenfeld residuals were also analyzed to determine goodness‐of‐fit and assess for outliers. P values <0.05 were considered statistically significant. All statistical analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC) and open source R statistical software version 3.0.1 (
RESULTS
A total of 1903 patients were screened, and 1084 patients met enrollment criteria (Figure 1). Of these, 1051 (97.0%) were non‐ICU patients. Patient characteristics of this cohort can be seen in Table 2. Enrolled patients and potentially eligible patients who presented to the emergency department during the enrollment window were similar in age, gender, and severity of illness, but enrolled patients were slightly more likely to have a chief complaint of chest pain and syncope (unpublished data).
| Variables | Normal Arousal, n=835 | Impaired Arousal, n=249 | P Value |
|---|---|---|---|
| |||
| Median age, y (IQR) | 74 (6980) | 75 (7083) | 0.005 |
| Female gender | 459 (55.0%) | 132 (53.0%) | 0.586 |
| Nonwhite race | 122 (14.6%) | 51 (20.5%) | 0.027 |
| Residence | <0.001 | ||
| Home | 752 (90.1%) | 204 (81.9%) | |
| Assisted living | 29 (3.5%) | 13 (5.2%) | |
| Rehabilitation | 8 (1.0%) | 5 (2.0%) | |
| Nursing home | 42 (5.0%) | 27 (10.8%) | |
| Dementia* | 175 (21.0%) | 119 (47.8%) | <0.001 |
| Dependent | 120 (14.4%) | 99 (39.8%) | <0.001 |
| Median Charlson (IQR) | 2 (1, 4) | 3 (2, 5) | <0.001 |
| Median APS (IQR) | 2 (1, 4) | 2 (1, 5) | <0.001 |
| Primary complaint | <0.001 | ||
| Abdominal pain | 45 (5.4%) | 13 (5.2%) | |
| Altered mental status | 12 (1.4%) | 36 (14.5%) | |
| Chest pain | 128 (15.3%) | 31 (12.5%) | |
| Disturbances of sensation | 17 (2.0%) | 2 (0.8%) | |
| Dizziness | 16 (1.9%) | 2 (0.8%) | |
| Fever | 11 (1.3%) | 7 (2.8%) | |
| General illness, malaise | 26 (3.1%) | 5 (2.0%) | |
| General weakness | 68 (8.1%) | 29 (11.7%) | |
| Nausea/vomiting | 29 (3.5%) | 4 (1.6%) | |
| Shortness of breath | 85 (10.2%) | 21 (8.4%) | |
| Syncope | 46 (5.5%) | 10 (4.0%) | |
| Trauma, multiple organs | 19 (2.3%) | 8 (3.2%) | |
| Other | 333 (39.9%) | 81 (32.5%) | |
| Benzodiazepines or opioid medications administration | 188 (22.5%) | 67 (26.9%) | 0.152 |
| Admitted to the hospital | 478 (57.3%) | 191 (76.7%) | 0.002 |
| Internal medicine | 411 (86.0%) | 153 (80.1%) | |
| Surgery | 38 (8.0%) | 21 (11.0%) | |
| Neurology | 19 (4.0%) | 13 (6.8%) | |
| Psychiatry | 1 (0.2%) | 2 (1.1%) | |
| Unknown/missing | 9 (1.9%) | 2 (1.1%) | |
| Death within 6 months | 81 (9.7%) | 59 (23.7%) | <0.001 |
Of those enrolled, 249 (23.0%) had an abnormal RASS at initial presentation, and their distribution can be seen in Figure 2. Within 6 months, patients with an abnormal RASS were more likely to die compared with patients with a RASS of 0 (23.7% vs 9.7%, P<0.001). The Kaplan‐Meier survival curves for all enrolled patients with impaired and normal RASS can be seen in Figure 3; the survival curve declined more slowly in patients with a normal RASS compared with those with an abnormal RASS.
Using Cox proportional hazards regression, the relationship between an abnormal RASS at initial presentation and 6‐month mortality persisted (HR: 1.73, 95% CI: 1.21‐2.49) after adjusting for age, sex, nonwhite race, comorbidity burden, severity of illness, dementia, functional dependence, nursing home residence, psychoactive medications given, and admission status. The interaction between an abnormal RASS and APS (severity of illness) had a P value of 0.52. The interaction between an abnormal RASS and benzodiazepine or opioid medication administration had a P value of 0.38. The interaction between an abnormal RASS and admission status had a P value of 0.57. This indicated that severity of illness, psychoactive medication administration, and admission status did not modify the relationship between an abnormal RASS and 6‐month mortality.
We analyzed a subset of 406 patients who received a psychiatrist's assessment to determine if an abnormal RASS was associated with 6‐month mortality regardless of delirium status using Cox proportional hazard regression weighted by the inverse of the propensity score. Patients with an abnormal RASS and no delirium were significantly associated with higher mortality compared to those with a normal RASS (HR: 2.20, 95% CI: 1.10‐4.41). Patients with an abnormal RASS with delirium also had an increased risk for 6‐month mortality (HR: 2.86, 95% CI: 1.29‐6.34).
All regression models were internally validated. There was no evidence of substantial overfitting or collinearity. The Schoenfeld residuals for each model were examined graphically and there was good model fit overall, and no significant outliers were observed.
DISCUSSION
Vital sign measurements are a fundamental component of patient care, and abnormalities can serve as an early warning signal of the patient's clinical deterioration. However, traditional vital signs do not include an assessment of the patient's brain function. Our chief finding is that impaired arousal at initial presentation, as determined by the nonphysician research staff, increased the risk of 6‐month mortality by 73% after adjusting for confounders in a diverse group of acutely ill older patients. This relationship existed regardless of severity of illness, administration of psychoactive medications, and admission status. Though impaired arousal is closely linked with delirium,[2, 24] which is another well‐known predictor of mortality,[11, 25, 26] the prognostic significance of impaired arousal appeared to extend beyond delirium. We observed that the relationship between 6‐month mortality and impaired arousal in the absence of delirium was remarkably similar to that observed with impaired arousal with delirium. Arousal can be assessed for by simply observing the patient during routine clinical care and can be performed by nonphysician and physician healthcare providers. Its assessment should be performed and communicated in conjunction with traditional vital sign measurements in the emergency department and inpatient settings.[1]
Most of the data linking impaired arousal to death have been collected in the ICU. Coma, which represents the most severe form of depressed arousal, has been shown to increase the likelihood of death regardless of underlying etiology.[27, 28, 29, 30, 31] This includes patients who have impaired arousal because they received sedative medications during mechanical ventilation.[32] Few studies have investigated the effect of impaired arousal in a non‐ICU patient population. Zuliani et al. observed that impaired arousal was associated with 30‐day mortality, but their study was conducted in 469 older stroke patients, limiting the study's external validity to a more general patient population.[33] Our data advance the current stage of knowledge; we observed a similar relationship between impaired arousal and 6‐month mortality in a much broader clinical population who were predominantly not critically ill regardless of delirium status. Additionally, most of our impaired arousal cohort had a RASS of 1, indicating that even subtle abnormalities portended adverse outcomes.
In addition to long‐term prognosis, the presence of impaired arousal has immediate clinical implications. Using arousal scales like the RASS can serve as a way for healthcare providers to succinctly communicate the patient's mental status in a standardized manner during transitions of care (eg, emergency physician to inpatient team). Regardless of which clinical setting they are in, older acutely ill patients with an impaired arousal may also require close monitoring, especially if the impairment is acute. Because of its close relationship with delirium, these patients likely have an underlying acute medical illness that precipitated their impaired arousal.
Understanding the true clinical significance of impaired arousal in the absence of delirium requires further study. Because of the fluctuating nature of delirium, it is possible that these patients may have initially been delirious and then became nondelirious during the psychiatrist's evaluation. Conversely, it is also possible that these patients may have eventually transitioned into delirium at later point in time; the presence of impaired arousal alone may be a precursor to delirium. Last, these patients may have had subsyndromal delirium, which is defined as having 1 or more delirium symptoms without ever meeting full DSM‐IV‐TR criteria for delirium.[34] Patients with subsyndromal delirium have poorer outcomes, such as prolonged hospitalizations, and higher mortality than patients without delirium symptoms.[34]
Additional studies are also needed to further clarify the impact of impaired arousal on nonmortality outcomes such as functional and cognitive decline. The prognostic significance of serial arousal measurements also requires further study. It is possible that patients whose impaired arousal rapidly resolves after an intervention may have better prognoses than those who have persistent impairment. The measurement of arousal may have additional clinical applications in disease prognosis models. The presence of altered mental status is incorporated in various disease‐specific risk scores such as the CURB‐65 or Pneumonia Severity Index for pneumonia,[35, 36] and the Pulmonary Embolism Severity Index for pulmonary embolism.[37] However, the definition of altered mental status is highly variable; it ranges from subjective impressions that can be unreliable to formal cognitive testing, which can be time consuming. Arousal scales such as the RASS may allow for more feasible, reliable, and standardized assessment of mental status. Future studies should investigate if incorporating the RASS would improve the discrimination of these disease‐severity indices.
This study has several notable limitations. We excluded patients with a RASS of 4 and 5, which represented comatose patients. This exclusion, however, likely biased our findings toward the null. We enrolled a convenience sample that may have introduced selection bias. However, our enrolled cohort was similar to all potentially eligible patients who presented to the emergency department during the study period. We also attempted to mitigate this selection bias by using multivariable regression and adjusting for factors that may have confounded the relationship between RASS and 6‐month mortality. This study was performed at a single, urban, academic hospital and enrolled patients who were aged 65 years and older. Our findings may not be generalizable to other settings and to those who are under 65 years of age. Because 406 patients received a psychiatric evaluation, this limited the number of covariates that could be incorporated into the multivariable model to evaluate if impaired arousal in the absence of delirium is associated with 6‐month mortality. To minimize residual confounding, we used the inverse weighted propensity score, but we acknowledge that this bias may still exist. Larger studies are needed to clarify the relationships between arousal, delirium, and mortality.
CONCLUSION
In conclusion, impaired arousal at initial presentation is an independent predictor for 6‐month mortality in a diverse group of acutely ill older patients, and this risk appears to be present even in the absence of delirium. Because of its ease of use and prognostic significance, it may be a useful vital sign for underlying brain dysfunction. Routine standardized assessment and communication of arousal during routine clinical care may be warranted.
Disclosures: Research reported in this publication was supported by the Vanderbilt Physician Scientist Development Award, Emergency Medicine Foundation, and National Institute on Aging of the National Institutes of Health under award number K23AG032355. This study was also supported by the National Center for Research Resources, grant UL1 RR024975‐01, and is now at the National Center for Advancing Translational Sciences, grant 2 UL1 TR000445‐06. Dr. Vasilevskis was supported in part by the National Institute on Aging of the National Institutes of Health under award number K23AG040157. Dr. Powers was supported by Health Resources and Services Administration Geriatric Education Centers, grant 1D31HP08823‐01‐00. Dr. Storrow was supported by the National Heart, Lung, and Blood Institute of the National Institutes of Health under award number K12HL1090 and the National Center for Advancing Translational Sciences under award number UL1TR000445. Dr. Ely was supported in part by the National Institute on Aging of the National Institutes of Health under award numbers R01AG027472 and R01AG035117, and a Veteran Affairs MERIT award. Drs. Vasilevskis, Schnelle, Dittus, Powers, and Ely were supported by the Veteran Affairs Geriatric Research, Education, and Clinical Center. The content is solely the responsibility of the authors and does not necessarily represent the official views of Vanderbilt University, Emergency Medicine Foundation, National Institutes of Health, and Veterans Affairs. The funding agencies did not have any role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, review, or approval of the manuscript.
J.H.H., E.W.E., J.F.S., A.B.S., and R.D.S. conceived the trial. J.H.H., E.W.E., A.B.S., J.F.S., R.D.S., A.S., and A.W. participated in the study design. J.H.H. and A.W. recruited patients and collected the data. J.H.H., A.J.G., and A.S. analyzed the data. All authors participated in the interpretation of results. J.H.H. drafted the manuscript, and all authors contributed to the critical review and revision of the manuscript.
The authors report no conflicts of interest.
- , , , et al. The development of a mental status vital sign for use across the spectrum of care. J Am Med Dir Assoc. 2009;10:379–380.
- , , , . Serial administration of a modified Richmond Agitation and Sedation Scale for delirium screening. J Hosp Med. 2012;7:450–453.
- , , , , , . Clarifying confusion: the confusion assessment method. A new method for detection of delirium. Ann Intern Med. 1990;113:941–948.
- , , , et al. Delirium in mechanically ventilated patients: validity and reliability of the confusion assessment method for the intensive care unit (CAM‐ICU). JAMA. 2001;286:2703–2710.
- , , , et al. Diagnosing delirium in older emergency department patients: validity and reliability of the Delirium Triage Screen And The Brief Confusion Assessment Method. Ann Emerg Med. 2013;62:457–465.
- , , , , . Does delirium contribute to poor hospital outcomes? A three‐site epidemiologic study. J Gen Intern Med. 1998;13:234–242.
- , , , . Prognostic significance of delirium in frail older people. Dement Geriatr Cogn Disord. 2005;19:158–163.
- , , , , , . Delirium in elderly patients and the risk of postdischarge mortality, institutionalization, and dementia: a meta‐analysis. JAMA. 2010;304:443–451.
- , , , , . One‐year health care costs associated with delirium in the elderly population. Arch Intern Med. 2008;168:27–32.
- , , , et al. The Richmond Agitation‐Sedation Scale: validity and reliability in adult intensive care unit patients. Am J Respir Crit Care Med. 2002;166:1338–1344.
- , , , et al. Delirium in the emergency department: an independent predictor of death within 6 months. Ann Emerg Med. 2010;56:244–252.
- , , , et al. Delirium in older emergency department patients is an independent predictor of hospital length of stay. Acad Emerg Med. 2011;18:451–457.
- , , , et al. Validation of the Confusion Assessment Method For The Intensive Care Unit in older emergency department patients. Acad Emerg Med. 2014;21:180–187.
- , , , et al. Monitoring sedation status over time in ICU patients: reliability and validity of the Richmond Agitation‐Sedation Scale (RASS). JAMA. 2003;289:2983–2991.
- , , , . Does this patient have dementia? JAMA. 2007;297:2391–2404.
- . A short form of the Informant Questionnaire on Cognitive Decline in the Elderly (IQCODE): development and cross‐validation. Psychol Med. 1994;24:145–153.
- . Assessing self‐maintenance: activities of daily living, mobility, and instrumental activities of daily living. J Am Geriatr Soc. 1983;31:721–727.
- , , , , . Charlson Index is associated with one‐year mortality in emergency department patients with suspected infection. Acad Emerg Med. 2006;13:530–536.
- , , , . APACHE II: a severity of disease classification system. Crit Care Med. 1985;13:818–829.
- American Psychiatric Association. Task Force on DSM‐IV. Diagnostic and Statistical Manual of Mental Disorders: DSM‐IV‐TR. 4th ed. Washington, DC: American Psychiatric Association; 2000.
- . Regression Modeling Strategies: With Applications to Linear Models, Logistic Regression, and Survival Analysis. New York, NY: Springer; 2001.
- . Power for tests of interaction: effect of raising the Type I error rate. Epidemiol Perspect Innov. 2007;4:4.
- . An Introduction to Propensity Score Methods for Reducing the Effects of Confounding in Observational Studies. Multivariate Behav Res. 2011;46:399–424.
- , , . Defining delirium for the International Classification of Diseases, 11th Revision. J Psychosom Res. 2008;65:207–214.
- , , , , . Delirium predicts 12‐month mortality. Arch Intern Med. 2002;162:457–463.
- , , , et al. Delirium as a predictor of mortality in mechanically ventilated patients in the intensive care unit. JAMA. 2004;291:1753–1762.
- , , . Predicting mortality of intensive care unit patients. The importance of coma. Crit Care Med. 1982;10:86–95.
- , . Assessment of outcome after severe brain damage. Lancet. 1975;1:480–484.
- , , , , , . Predicting outcome from hypoxic‐ischemic coma. JAMA. 1985;253:1420–1426.
- , , , et al. Prediction of intracerebral hemorrhage survival. Ann Neurol. 1988;24:258–263.
- , , , . Is this patient dead, vegetative, or severely neurologically impaired? Assessing outcome for comatose survivors of cardiac arrest. JAMA. 2004;291:870–879.
- , , , et al. Early intensive care sedation predicts long‐term mortality in ventilated critically ill patients. Am J Respir Crit Care Med. 2012;186:724–731.
- , , , , , . Risk factors for short‐term mortality in older subjects with acute ischemic stroke. Gerontology. 2006;52:231–236.
- , , , . The prognostic significance of subsyndromal delirium in elderly medical inpatients. J Am Geriatr Soc. 2003;51:754–760.
- , , , et al. Defining community acquired pneumonia severity on presentation to hospital: an international derivation and validation study. Thorax. 2003;58:377–382.
- , , , et al. A prediction rule to identify low‐risk patients with community‐acquired pneumonia. N Engl J Med. 1997;336:243–250.
- , , , et al. Derivation and validation of a prognostic model for pulmonary embolism. Am J Respir Crit Care Med. 2005;172:1041–1046.
Arousal is defined as the patient's overall level of responsiveness to the environment. Its assessment is standard of care in most intensive care units (ICUs) to monitor depth of sedation and underlying brain dysfunction. There has been recent interest in expanding the role of arousal assessment beyond the ICU. Specifically, the Veterans Affairs Delirium Working Group proposed that simple arousal assessment be a vital sign to quantify underlying brain dysfunction.[1] The rationale is that impaired arousal is closely linked with delirium,[2] and is an integral component of multiple delirium assessments.[3, 4, 5] Chester et al. observed that the presence of impaired arousal was 64% sensitive and 93% specific for delirium diagnosed by a psychiatrist.[2] Delirium is an under‐recognized public health problem that affects up to 25% of older hospitalized patients,[6, 7] is associated with a multitude of adverse outcomes such as death and accelerated cognitive decline,[8] and costs the US healthcare system an excess of $152 billion dollars.[9]
Most delirium assessments require the patient to undergo additional cognitive testing. The assessment of arousal, however, requires the rater to merely observe the patient during routine clinical care and can be easily integrated into the clinical workflow.[10] Because of its simplicity and brevity, assessing arousal alone using validated scales such as the Richmond Agitation‐Sedation Scale (RASS) may be a more appealing alternative to traditional, more complex delirium screening in the acute care setting. Its clinical utility would be further strengthened if impaired arousal was also associated with mortality, and conferred risk even in the absence of delirium. As a result, we sought to determine if impaired arousal at initial presentation in older acutely ill patients predicted 6‐month mortality and whether this relationship was present in the absence of delirium.
METHODS
Design Overview
We performed a planned secondary analysis of 2 prospective cohorts that enrolled patients from May 2007 to August 2008 between 8 am and 10 pm during the weekdays, and July 2009 to February 2012 between 8 am and 4 pm during the weekdays. The first cohort was designed to evaluate the relationship between delirium and patient outcomes.[11, 12] The second cohort was used to validate brief delirium assessments using a psychiatrist's assessment as the reference standard.[5, 13] The local institutional review board approved these studies.
Setting and Participants
These studies were conducted in an urban emergency department located within an academic, tertiary care hospital with over 57,000 visits annually. Patients were included if they were 65 years or older and in the emergency department for <12 hours at the time of enrollment. The 12‐hour cutoff was used to include patients who presented to the emergency department in the evening and early morning hours. Patients were excluded if they were previously enrolled, non‐English speaking, comatose, or were nonverbal and unable to follow simple commands prior to the acute illness. Because the July 2009 to February 2012 cohort was designed to validate delirium assessments with auditory and visual components, patients were also excluded if they were deaf or blind.
Measurement of Arousal
RASS is an arousal scale commonly used in ICUs to assess depth of sedation and ranges from 5 (unarousable) to +4 (combative); 0 represents normal arousal.[10, 14] The RASS simply requires the rater to observe the patient during their routine interactions and does not require any additional cognitive testing. The RASS terms sedation was modified to drowsy (Table 1), because we wanted to capture impaired arousal regardless of sedation administration. We did not use the modified RASS (mRASS) proposed by the Veteran's Affairs Delirium Working Group, because it was published after data collection began.[1] The mRASS is very similar to the RASS, except it also incorporates a very informal inattention assessment. The RASS was ascertained by research assistants who were college students and graduates, and emergency medical technician basics and paramedics. The principal investigator gave them a 5‐minute didactic lecture about the RASS and observed them perform the RASS in at least 5 patients prior to the start of the study. Inter‐rater reliability between trained research assistants and a physician was assessed for 456 (42.0%) patients of the study sample. The weighted kappa of the RASS was 0.61, indicating very good inter‐rater reliability. Because the 81.7% of patients with impaired arousal had a RASS of 1, the RASS dichotomized as normal (RASS=0) or impaired (RASS other than 0).
| Score | Term | Description |
|---|---|---|
| ||
| +4 | Combative | Overtly combative, violent, immediate danger to staff |
| +3 | Very agitated | Pulls or removes tube(s) or catheter(s), aggressive |
| +2 | Agitated | Frequent nonpurposeful movement |
| +1 | Restless | Anxious but movements not aggressive or vigorous |
| 0 | Alert and calm | |
| 1 | Slight drowsy | Not fully alert, but has sustained awakening (eye opening/eye contact) to voice (>10 seconds) |
| 2 | Moderately drowsy | Briefly awakens with eye contact to voice (<10 seconds) |
| 3 | Very drowsy | Movement or eye opening to voice (but no eye contact) |
| 4 | Awakens to pain only | No response to voice, but movement or eye opening to physical stimulation |
| 5 | Unarousable | No response to voice or physical stimulation |
Death Ascertainment
Death within 6 months was ascertained using the following algorithm: (1) The electronic medical record was searched to determine the patient's death status. (2) Patients who had a documented emergency department visit, outpatient clinic visit, or hospitalization after 6 months were considered to be alive at 6 months. (3) For the remaining patients, date of death was searched in the Social Security Death Index (SSDI). (4) Patients without a death recorded in the SSDI 1 year after the index visit was considered to be alive at 6 months. Nine hundred thirty‐one (85.9%) out of 1084 patients had a recorded death in the medical record or SSDI, or had an emergency department or hospital visit documented in their record 6 months after the index visit.
Additional Variables Collected
Patients were considered to have dementia if they had: (1) documented dementia in the medical record, (2) a short form Informant Questionnaire on Cognitive Decline in the Elderly score (IQCODE) greater than 3.38,[15] or (3) prescribed cholinesterase inhibitors prior to admission. The short form IQCODE is an informant questionnaire with 16 items; a cutoff of 3.38 out of 5.00 is 79% sensitive and 82% specific for dementia.[16] Premorbid functional status was determined by the Katz Activities of Daily Living (Katz ADL) and ranges from 0 (completely dependent) to 6 (completely independent).[17] Patients with a score <5 were considered to be functionally dependent. Both the IQCODE and Katz ADL were prospectively collected in the emergency department at the time of enrollment.
The Charlson Comorbidity Index was used to measure comorbid burden.[18] The Acute Physiology Score (APS) of the Acute Physiology and Chronic Health Evaluation II score was used to quantify severity of illness.[19] The Glasgow Coma Scale was not included in the APS because it was not collected. Intravenous, intramuscular, and oral benzodiazepine and opioids given in the prehospital and emergency department were also recorded. The Charlson Comorbidity Index, APS, and benzodiazepine and opioid administration were collected after patient enrollment using the electronic medical record.
Within 3 hours of the RASS, a subset of 406 patients was evaluated by a consultation‐liaison psychiatrist who determined the patient's delirium status using Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, Text Revision (DSM‐IV‐TR) criteria.[20] Details of their comprehensive assessments have been described in a previous report.[5]
Statistical Analysis
Measures of central tendency and dispersion for continuous variables were reported as medians and interquartile ranges. Categorical variables were reported as proportions. For simple comparisons, Wilcoxon rank sum tests were performed for continuous data, and 2 analyses or Fisher exact test were performed for categorical data. To evaluate the predictive validity of impaired arousal on 6‐month mortality, the cumulative probability of survival was estimated within 6 months from the study enrollment date using the Kaplan‐Meier method. Cox proportional hazards regression was performed to assess if impaired arousal was independently associated with 6‐month mortality after adjusting for age, gender, nonwhite race, comorbidity burden (Charlson Comorbidity Index), severity of illness (APS), dementia, functional dependence (Katz ADL <5), nursing home residence, admission status, and benzodiazepine or opioid medication administration. Patients were censored at the end of 6 months. The selection of covariates was based upon expert opinion and literature review. The number of covariates used for the model was limited by the number of events to minimize overfitting; 1 df was allowed for every 10 to 15 events.[21] Because severity of illness, psychoactive medication administration, and admission status might modify the relationship between 6‐month mortality and impaired arousal, 2‐way interaction terms were incorporated. To maintain parsimony and minimize overfitting and collinearity, nonsignificant interaction terms (P>0.20) were removed in the final model.[22] Hazard ratios (HR) with their 95% confidence interval (95% CI) were reported.
To determine if arousal was associated with 6‐month mortality in the absence of delirium, we performed another Cox proportional hazard regression in a subset of 406 patients who received a psychiatrist assessment. Six‐month mortality was the dependent variable, and the independent variable was a 3‐level categorical variable of different arousal/delirium combinations: (1) impaired arousal/delirium positive, (2) impaired arousal/delirium negative, and (3) normal arousal (with or without delirium). Because there were only 8 patients who had normal arousal with delirium, this group was collapsed into the normal arousal without delirium group. Because there were 55 deaths, the number of covariates that could be entered into the Cox proportional hazard regression model was limited. We used the inverse weighted propensity score method to help minimize residual confounding.[23] Traditional propensity score adjustment could not be performed because there were 3 arousal/delirium categories. Similar to propensity score adjustment, inverse weighted propensity score method was used to help balance the distribution of patient characteristics among the exposure groups and also allow adjustment for multiple confounders while minimizing the degrees of freedom expended. A propensity score was the probability of having a particular arousal/delirium category based upon baseline patient characteristics. Multinomial logistic regression was performed to calculate the propensity score, and the baseline covariates used were age, gender, nonwhite race, comorbidity burden, severity of illness, dementia, functional dependence, and nursing home residence. For the Cox proportional hazard regression model, each observation was weighted by the inverse of the propensity score for their given arousal/delirium category; propensity scores exceeding the 95th percentile were trimmed to avoid overly influential weighting. Benzodiazepine and opioid medications given in the emergency department and admission status were adjusted as covariates in the weighted Cox proportional hazard regression model.
Nineteen patients (1.8%) had missing Katz ADL; these missing values were imputed using multiple imputation. The reliability of the final regression models were internally validated using the bootstrap method.[21] Two thousand sets of bootstrap samples were generated by resampling the original data, and the optimism was estimated to determine the degree of overfitting.[21] An optimism value >0.85 indicated no evidence of substantial overfitting.[21] Variance inflation factors were used to check multicollinearity. Schoenfeld residuals were also analyzed to determine goodness‐of‐fit and assess for outliers. P values <0.05 were considered statistically significant. All statistical analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC) and open source R statistical software version 3.0.1 (
RESULTS
A total of 1903 patients were screened, and 1084 patients met enrollment criteria (Figure 1). Of these, 1051 (97.0%) were non‐ICU patients. Patient characteristics of this cohort can be seen in Table 2. Enrolled patients and potentially eligible patients who presented to the emergency department during the enrollment window were similar in age, gender, and severity of illness, but enrolled patients were slightly more likely to have a chief complaint of chest pain and syncope (unpublished data).
| Variables | Normal Arousal, n=835 | Impaired Arousal, n=249 | P Value |
|---|---|---|---|
| |||
| Median age, y (IQR) | 74 (6980) | 75 (7083) | 0.005 |
| Female gender | 459 (55.0%) | 132 (53.0%) | 0.586 |
| Nonwhite race | 122 (14.6%) | 51 (20.5%) | 0.027 |
| Residence | <0.001 | ||
| Home | 752 (90.1%) | 204 (81.9%) | |
| Assisted living | 29 (3.5%) | 13 (5.2%) | |
| Rehabilitation | 8 (1.0%) | 5 (2.0%) | |
| Nursing home | 42 (5.0%) | 27 (10.8%) | |
| Dementia* | 175 (21.0%) | 119 (47.8%) | <0.001 |
| Dependent | 120 (14.4%) | 99 (39.8%) | <0.001 |
| Median Charlson (IQR) | 2 (1, 4) | 3 (2, 5) | <0.001 |
| Median APS (IQR) | 2 (1, 4) | 2 (1, 5) | <0.001 |
| Primary complaint | <0.001 | ||
| Abdominal pain | 45 (5.4%) | 13 (5.2%) | |
| Altered mental status | 12 (1.4%) | 36 (14.5%) | |
| Chest pain | 128 (15.3%) | 31 (12.5%) | |
| Disturbances of sensation | 17 (2.0%) | 2 (0.8%) | |
| Dizziness | 16 (1.9%) | 2 (0.8%) | |
| Fever | 11 (1.3%) | 7 (2.8%) | |
| General illness, malaise | 26 (3.1%) | 5 (2.0%) | |
| General weakness | 68 (8.1%) | 29 (11.7%) | |
| Nausea/vomiting | 29 (3.5%) | 4 (1.6%) | |
| Shortness of breath | 85 (10.2%) | 21 (8.4%) | |
| Syncope | 46 (5.5%) | 10 (4.0%) | |
| Trauma, multiple organs | 19 (2.3%) | 8 (3.2%) | |
| Other | 333 (39.9%) | 81 (32.5%) | |
| Benzodiazepines or opioid medications administration | 188 (22.5%) | 67 (26.9%) | 0.152 |
| Admitted to the hospital | 478 (57.3%) | 191 (76.7%) | 0.002 |
| Internal medicine | 411 (86.0%) | 153 (80.1%) | |
| Surgery | 38 (8.0%) | 21 (11.0%) | |
| Neurology | 19 (4.0%) | 13 (6.8%) | |
| Psychiatry | 1 (0.2%) | 2 (1.1%) | |
| Unknown/missing | 9 (1.9%) | 2 (1.1%) | |
| Death within 6 months | 81 (9.7%) | 59 (23.7%) | <0.001 |
Of those enrolled, 249 (23.0%) had an abnormal RASS at initial presentation, and their distribution can be seen in Figure 2. Within 6 months, patients with an abnormal RASS were more likely to die compared with patients with a RASS of 0 (23.7% vs 9.7%, P<0.001). The Kaplan‐Meier survival curves for all enrolled patients with impaired and normal RASS can be seen in Figure 3; the survival curve declined more slowly in patients with a normal RASS compared with those with an abnormal RASS.
Using Cox proportional hazards regression, the relationship between an abnormal RASS at initial presentation and 6‐month mortality persisted (HR: 1.73, 95% CI: 1.21‐2.49) after adjusting for age, sex, nonwhite race, comorbidity burden, severity of illness, dementia, functional dependence, nursing home residence, psychoactive medications given, and admission status. The interaction between an abnormal RASS and APS (severity of illness) had a P value of 0.52. The interaction between an abnormal RASS and benzodiazepine or opioid medication administration had a P value of 0.38. The interaction between an abnormal RASS and admission status had a P value of 0.57. This indicated that severity of illness, psychoactive medication administration, and admission status did not modify the relationship between an abnormal RASS and 6‐month mortality.
We analyzed a subset of 406 patients who received a psychiatrist's assessment to determine if an abnormal RASS was associated with 6‐month mortality regardless of delirium status using Cox proportional hazard regression weighted by the inverse of the propensity score. Patients with an abnormal RASS and no delirium were significantly associated with higher mortality compared to those with a normal RASS (HR: 2.20, 95% CI: 1.10‐4.41). Patients with an abnormal RASS with delirium also had an increased risk for 6‐month mortality (HR: 2.86, 95% CI: 1.29‐6.34).
All regression models were internally validated. There was no evidence of substantial overfitting or collinearity. The Schoenfeld residuals for each model were examined graphically and there was good model fit overall, and no significant outliers were observed.
DISCUSSION
Vital sign measurements are a fundamental component of patient care, and abnormalities can serve as an early warning signal of the patient's clinical deterioration. However, traditional vital signs do not include an assessment of the patient's brain function. Our chief finding is that impaired arousal at initial presentation, as determined by the nonphysician research staff, increased the risk of 6‐month mortality by 73% after adjusting for confounders in a diverse group of acutely ill older patients. This relationship existed regardless of severity of illness, administration of psychoactive medications, and admission status. Though impaired arousal is closely linked with delirium,[2, 24] which is another well‐known predictor of mortality,[11, 25, 26] the prognostic significance of impaired arousal appeared to extend beyond delirium. We observed that the relationship between 6‐month mortality and impaired arousal in the absence of delirium was remarkably similar to that observed with impaired arousal with delirium. Arousal can be assessed for by simply observing the patient during routine clinical care and can be performed by nonphysician and physician healthcare providers. Its assessment should be performed and communicated in conjunction with traditional vital sign measurements in the emergency department and inpatient settings.[1]
Most of the data linking impaired arousal to death have been collected in the ICU. Coma, which represents the most severe form of depressed arousal, has been shown to increase the likelihood of death regardless of underlying etiology.[27, 28, 29, 30, 31] This includes patients who have impaired arousal because they received sedative medications during mechanical ventilation.[32] Few studies have investigated the effect of impaired arousal in a non‐ICU patient population. Zuliani et al. observed that impaired arousal was associated with 30‐day mortality, but their study was conducted in 469 older stroke patients, limiting the study's external validity to a more general patient population.[33] Our data advance the current stage of knowledge; we observed a similar relationship between impaired arousal and 6‐month mortality in a much broader clinical population who were predominantly not critically ill regardless of delirium status. Additionally, most of our impaired arousal cohort had a RASS of 1, indicating that even subtle abnormalities portended adverse outcomes.
In addition to long‐term prognosis, the presence of impaired arousal has immediate clinical implications. Using arousal scales like the RASS can serve as a way for healthcare providers to succinctly communicate the patient's mental status in a standardized manner during transitions of care (eg, emergency physician to inpatient team). Regardless of which clinical setting they are in, older acutely ill patients with an impaired arousal may also require close monitoring, especially if the impairment is acute. Because of its close relationship with delirium, these patients likely have an underlying acute medical illness that precipitated their impaired arousal.
Understanding the true clinical significance of impaired arousal in the absence of delirium requires further study. Because of the fluctuating nature of delirium, it is possible that these patients may have initially been delirious and then became nondelirious during the psychiatrist's evaluation. Conversely, it is also possible that these patients may have eventually transitioned into delirium at later point in time; the presence of impaired arousal alone may be a precursor to delirium. Last, these patients may have had subsyndromal delirium, which is defined as having 1 or more delirium symptoms without ever meeting full DSM‐IV‐TR criteria for delirium.[34] Patients with subsyndromal delirium have poorer outcomes, such as prolonged hospitalizations, and higher mortality than patients without delirium symptoms.[34]
Additional studies are also needed to further clarify the impact of impaired arousal on nonmortality outcomes such as functional and cognitive decline. The prognostic significance of serial arousal measurements also requires further study. It is possible that patients whose impaired arousal rapidly resolves after an intervention may have better prognoses than those who have persistent impairment. The measurement of arousal may have additional clinical applications in disease prognosis models. The presence of altered mental status is incorporated in various disease‐specific risk scores such as the CURB‐65 or Pneumonia Severity Index for pneumonia,[35, 36] and the Pulmonary Embolism Severity Index for pulmonary embolism.[37] However, the definition of altered mental status is highly variable; it ranges from subjective impressions that can be unreliable to formal cognitive testing, which can be time consuming. Arousal scales such as the RASS may allow for more feasible, reliable, and standardized assessment of mental status. Future studies should investigate if incorporating the RASS would improve the discrimination of these disease‐severity indices.
This study has several notable limitations. We excluded patients with a RASS of 4 and 5, which represented comatose patients. This exclusion, however, likely biased our findings toward the null. We enrolled a convenience sample that may have introduced selection bias. However, our enrolled cohort was similar to all potentially eligible patients who presented to the emergency department during the study period. We also attempted to mitigate this selection bias by using multivariable regression and adjusting for factors that may have confounded the relationship between RASS and 6‐month mortality. This study was performed at a single, urban, academic hospital and enrolled patients who were aged 65 years and older. Our findings may not be generalizable to other settings and to those who are under 65 years of age. Because 406 patients received a psychiatric evaluation, this limited the number of covariates that could be incorporated into the multivariable model to evaluate if impaired arousal in the absence of delirium is associated with 6‐month mortality. To minimize residual confounding, we used the inverse weighted propensity score, but we acknowledge that this bias may still exist. Larger studies are needed to clarify the relationships between arousal, delirium, and mortality.
CONCLUSION
In conclusion, impaired arousal at initial presentation is an independent predictor for 6‐month mortality in a diverse group of acutely ill older patients, and this risk appears to be present even in the absence of delirium. Because of its ease of use and prognostic significance, it may be a useful vital sign for underlying brain dysfunction. Routine standardized assessment and communication of arousal during routine clinical care may be warranted.
Disclosures: Research reported in this publication was supported by the Vanderbilt Physician Scientist Development Award, Emergency Medicine Foundation, and National Institute on Aging of the National Institutes of Health under award number K23AG032355. This study was also supported by the National Center for Research Resources, grant UL1 RR024975‐01, and is now at the National Center for Advancing Translational Sciences, grant 2 UL1 TR000445‐06. Dr. Vasilevskis was supported in part by the National Institute on Aging of the National Institutes of Health under award number K23AG040157. Dr. Powers was supported by Health Resources and Services Administration Geriatric Education Centers, grant 1D31HP08823‐01‐00. Dr. Storrow was supported by the National Heart, Lung, and Blood Institute of the National Institutes of Health under award number K12HL1090 and the National Center for Advancing Translational Sciences under award number UL1TR000445. Dr. Ely was supported in part by the National Institute on Aging of the National Institutes of Health under award numbers R01AG027472 and R01AG035117, and a Veteran Affairs MERIT award. Drs. Vasilevskis, Schnelle, Dittus, Powers, and Ely were supported by the Veteran Affairs Geriatric Research, Education, and Clinical Center. The content is solely the responsibility of the authors and does not necessarily represent the official views of Vanderbilt University, Emergency Medicine Foundation, National Institutes of Health, and Veterans Affairs. The funding agencies did not have any role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, review, or approval of the manuscript.
J.H.H., E.W.E., J.F.S., A.B.S., and R.D.S. conceived the trial. J.H.H., E.W.E., A.B.S., J.F.S., R.D.S., A.S., and A.W. participated in the study design. J.H.H. and A.W. recruited patients and collected the data. J.H.H., A.J.G., and A.S. analyzed the data. All authors participated in the interpretation of results. J.H.H. drafted the manuscript, and all authors contributed to the critical review and revision of the manuscript.
The authors report no conflicts of interest.
Arousal is defined as the patient's overall level of responsiveness to the environment. Its assessment is standard of care in most intensive care units (ICUs) to monitor depth of sedation and underlying brain dysfunction. There has been recent interest in expanding the role of arousal assessment beyond the ICU. Specifically, the Veterans Affairs Delirium Working Group proposed that simple arousal assessment be a vital sign to quantify underlying brain dysfunction.[1] The rationale is that impaired arousal is closely linked with delirium,[2] and is an integral component of multiple delirium assessments.[3, 4, 5] Chester et al. observed that the presence of impaired arousal was 64% sensitive and 93% specific for delirium diagnosed by a psychiatrist.[2] Delirium is an under‐recognized public health problem that affects up to 25% of older hospitalized patients,[6, 7] is associated with a multitude of adverse outcomes such as death and accelerated cognitive decline,[8] and costs the US healthcare system an excess of $152 billion dollars.[9]
Most delirium assessments require the patient to undergo additional cognitive testing. The assessment of arousal, however, requires the rater to merely observe the patient during routine clinical care and can be easily integrated into the clinical workflow.[10] Because of its simplicity and brevity, assessing arousal alone using validated scales such as the Richmond Agitation‐Sedation Scale (RASS) may be a more appealing alternative to traditional, more complex delirium screening in the acute care setting. Its clinical utility would be further strengthened if impaired arousal was also associated with mortality, and conferred risk even in the absence of delirium. As a result, we sought to determine if impaired arousal at initial presentation in older acutely ill patients predicted 6‐month mortality and whether this relationship was present in the absence of delirium.
METHODS
Design Overview
We performed a planned secondary analysis of 2 prospective cohorts that enrolled patients from May 2007 to August 2008 between 8 am and 10 pm during the weekdays, and July 2009 to February 2012 between 8 am and 4 pm during the weekdays. The first cohort was designed to evaluate the relationship between delirium and patient outcomes.[11, 12] The second cohort was used to validate brief delirium assessments using a psychiatrist's assessment as the reference standard.[5, 13] The local institutional review board approved these studies.
Setting and Participants
These studies were conducted in an urban emergency department located within an academic, tertiary care hospital with over 57,000 visits annually. Patients were included if they were 65 years or older and in the emergency department for <12 hours at the time of enrollment. The 12‐hour cutoff was used to include patients who presented to the emergency department in the evening and early morning hours. Patients were excluded if they were previously enrolled, non‐English speaking, comatose, or were nonverbal and unable to follow simple commands prior to the acute illness. Because the July 2009 to February 2012 cohort was designed to validate delirium assessments with auditory and visual components, patients were also excluded if they were deaf or blind.
Measurement of Arousal
RASS is an arousal scale commonly used in ICUs to assess depth of sedation and ranges from 5 (unarousable) to +4 (combative); 0 represents normal arousal.[10, 14] The RASS simply requires the rater to observe the patient during their routine interactions and does not require any additional cognitive testing. The RASS terms sedation was modified to drowsy (Table 1), because we wanted to capture impaired arousal regardless of sedation administration. We did not use the modified RASS (mRASS) proposed by the Veteran's Affairs Delirium Working Group, because it was published after data collection began.[1] The mRASS is very similar to the RASS, except it also incorporates a very informal inattention assessment. The RASS was ascertained by research assistants who were college students and graduates, and emergency medical technician basics and paramedics. The principal investigator gave them a 5‐minute didactic lecture about the RASS and observed them perform the RASS in at least 5 patients prior to the start of the study. Inter‐rater reliability between trained research assistants and a physician was assessed for 456 (42.0%) patients of the study sample. The weighted kappa of the RASS was 0.61, indicating very good inter‐rater reliability. Because the 81.7% of patients with impaired arousal had a RASS of 1, the RASS dichotomized as normal (RASS=0) or impaired (RASS other than 0).
| Score | Term | Description |
|---|---|---|
| ||
| +4 | Combative | Overtly combative, violent, immediate danger to staff |
| +3 | Very agitated | Pulls or removes tube(s) or catheter(s), aggressive |
| +2 | Agitated | Frequent nonpurposeful movement |
| +1 | Restless | Anxious but movements not aggressive or vigorous |
| 0 | Alert and calm | |
| 1 | Slight drowsy | Not fully alert, but has sustained awakening (eye opening/eye contact) to voice (>10 seconds) |
| 2 | Moderately drowsy | Briefly awakens with eye contact to voice (<10 seconds) |
| 3 | Very drowsy | Movement or eye opening to voice (but no eye contact) |
| 4 | Awakens to pain only | No response to voice, but movement or eye opening to physical stimulation |
| 5 | Unarousable | No response to voice or physical stimulation |
Death Ascertainment
Death within 6 months was ascertained using the following algorithm: (1) The electronic medical record was searched to determine the patient's death status. (2) Patients who had a documented emergency department visit, outpatient clinic visit, or hospitalization after 6 months were considered to be alive at 6 months. (3) For the remaining patients, date of death was searched in the Social Security Death Index (SSDI). (4) Patients without a death recorded in the SSDI 1 year after the index visit was considered to be alive at 6 months. Nine hundred thirty‐one (85.9%) out of 1084 patients had a recorded death in the medical record or SSDI, or had an emergency department or hospital visit documented in their record 6 months after the index visit.
Additional Variables Collected
Patients were considered to have dementia if they had: (1) documented dementia in the medical record, (2) a short form Informant Questionnaire on Cognitive Decline in the Elderly score (IQCODE) greater than 3.38,[15] or (3) prescribed cholinesterase inhibitors prior to admission. The short form IQCODE is an informant questionnaire with 16 items; a cutoff of 3.38 out of 5.00 is 79% sensitive and 82% specific for dementia.[16] Premorbid functional status was determined by the Katz Activities of Daily Living (Katz ADL) and ranges from 0 (completely dependent) to 6 (completely independent).[17] Patients with a score <5 were considered to be functionally dependent. Both the IQCODE and Katz ADL were prospectively collected in the emergency department at the time of enrollment.
The Charlson Comorbidity Index was used to measure comorbid burden.[18] The Acute Physiology Score (APS) of the Acute Physiology and Chronic Health Evaluation II score was used to quantify severity of illness.[19] The Glasgow Coma Scale was not included in the APS because it was not collected. Intravenous, intramuscular, and oral benzodiazepine and opioids given in the prehospital and emergency department were also recorded. The Charlson Comorbidity Index, APS, and benzodiazepine and opioid administration were collected after patient enrollment using the electronic medical record.
Within 3 hours of the RASS, a subset of 406 patients was evaluated by a consultation‐liaison psychiatrist who determined the patient's delirium status using Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, Text Revision (DSM‐IV‐TR) criteria.[20] Details of their comprehensive assessments have been described in a previous report.[5]
Statistical Analysis
Measures of central tendency and dispersion for continuous variables were reported as medians and interquartile ranges. Categorical variables were reported as proportions. For simple comparisons, Wilcoxon rank sum tests were performed for continuous data, and 2 analyses or Fisher exact test were performed for categorical data. To evaluate the predictive validity of impaired arousal on 6‐month mortality, the cumulative probability of survival was estimated within 6 months from the study enrollment date using the Kaplan‐Meier method. Cox proportional hazards regression was performed to assess if impaired arousal was independently associated with 6‐month mortality after adjusting for age, gender, nonwhite race, comorbidity burden (Charlson Comorbidity Index), severity of illness (APS), dementia, functional dependence (Katz ADL <5), nursing home residence, admission status, and benzodiazepine or opioid medication administration. Patients were censored at the end of 6 months. The selection of covariates was based upon expert opinion and literature review. The number of covariates used for the model was limited by the number of events to minimize overfitting; 1 df was allowed for every 10 to 15 events.[21] Because severity of illness, psychoactive medication administration, and admission status might modify the relationship between 6‐month mortality and impaired arousal, 2‐way interaction terms were incorporated. To maintain parsimony and minimize overfitting and collinearity, nonsignificant interaction terms (P>0.20) were removed in the final model.[22] Hazard ratios (HR) with their 95% confidence interval (95% CI) were reported.
To determine if arousal was associated with 6‐month mortality in the absence of delirium, we performed another Cox proportional hazard regression in a subset of 406 patients who received a psychiatrist assessment. Six‐month mortality was the dependent variable, and the independent variable was a 3‐level categorical variable of different arousal/delirium combinations: (1) impaired arousal/delirium positive, (2) impaired arousal/delirium negative, and (3) normal arousal (with or without delirium). Because there were only 8 patients who had normal arousal with delirium, this group was collapsed into the normal arousal without delirium group. Because there were 55 deaths, the number of covariates that could be entered into the Cox proportional hazard regression model was limited. We used the inverse weighted propensity score method to help minimize residual confounding.[23] Traditional propensity score adjustment could not be performed because there were 3 arousal/delirium categories. Similar to propensity score adjustment, inverse weighted propensity score method was used to help balance the distribution of patient characteristics among the exposure groups and also allow adjustment for multiple confounders while minimizing the degrees of freedom expended. A propensity score was the probability of having a particular arousal/delirium category based upon baseline patient characteristics. Multinomial logistic regression was performed to calculate the propensity score, and the baseline covariates used were age, gender, nonwhite race, comorbidity burden, severity of illness, dementia, functional dependence, and nursing home residence. For the Cox proportional hazard regression model, each observation was weighted by the inverse of the propensity score for their given arousal/delirium category; propensity scores exceeding the 95th percentile were trimmed to avoid overly influential weighting. Benzodiazepine and opioid medications given in the emergency department and admission status were adjusted as covariates in the weighted Cox proportional hazard regression model.
Nineteen patients (1.8%) had missing Katz ADL; these missing values were imputed using multiple imputation. The reliability of the final regression models were internally validated using the bootstrap method.[21] Two thousand sets of bootstrap samples were generated by resampling the original data, and the optimism was estimated to determine the degree of overfitting.[21] An optimism value >0.85 indicated no evidence of substantial overfitting.[21] Variance inflation factors were used to check multicollinearity. Schoenfeld residuals were also analyzed to determine goodness‐of‐fit and assess for outliers. P values <0.05 were considered statistically significant. All statistical analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC) and open source R statistical software version 3.0.1 (
RESULTS
A total of 1903 patients were screened, and 1084 patients met enrollment criteria (Figure 1). Of these, 1051 (97.0%) were non‐ICU patients. Patient characteristics of this cohort can be seen in Table 2. Enrolled patients and potentially eligible patients who presented to the emergency department during the enrollment window were similar in age, gender, and severity of illness, but enrolled patients were slightly more likely to have a chief complaint of chest pain and syncope (unpublished data).
| Variables | Normal Arousal, n=835 | Impaired Arousal, n=249 | P Value |
|---|---|---|---|
| |||
| Median age, y (IQR) | 74 (6980) | 75 (7083) | 0.005 |
| Female gender | 459 (55.0%) | 132 (53.0%) | 0.586 |
| Nonwhite race | 122 (14.6%) | 51 (20.5%) | 0.027 |
| Residence | <0.001 | ||
| Home | 752 (90.1%) | 204 (81.9%) | |
| Assisted living | 29 (3.5%) | 13 (5.2%) | |
| Rehabilitation | 8 (1.0%) | 5 (2.0%) | |
| Nursing home | 42 (5.0%) | 27 (10.8%) | |
| Dementia* | 175 (21.0%) | 119 (47.8%) | <0.001 |
| Dependent | 120 (14.4%) | 99 (39.8%) | <0.001 |
| Median Charlson (IQR) | 2 (1, 4) | 3 (2, 5) | <0.001 |
| Median APS (IQR) | 2 (1, 4) | 2 (1, 5) | <0.001 |
| Primary complaint | <0.001 | ||
| Abdominal pain | 45 (5.4%) | 13 (5.2%) | |
| Altered mental status | 12 (1.4%) | 36 (14.5%) | |
| Chest pain | 128 (15.3%) | 31 (12.5%) | |
| Disturbances of sensation | 17 (2.0%) | 2 (0.8%) | |
| Dizziness | 16 (1.9%) | 2 (0.8%) | |
| Fever | 11 (1.3%) | 7 (2.8%) | |
| General illness, malaise | 26 (3.1%) | 5 (2.0%) | |
| General weakness | 68 (8.1%) | 29 (11.7%) | |
| Nausea/vomiting | 29 (3.5%) | 4 (1.6%) | |
| Shortness of breath | 85 (10.2%) | 21 (8.4%) | |
| Syncope | 46 (5.5%) | 10 (4.0%) | |
| Trauma, multiple organs | 19 (2.3%) | 8 (3.2%) | |
| Other | 333 (39.9%) | 81 (32.5%) | |
| Benzodiazepines or opioid medications administration | 188 (22.5%) | 67 (26.9%) | 0.152 |
| Admitted to the hospital | 478 (57.3%) | 191 (76.7%) | 0.002 |
| Internal medicine | 411 (86.0%) | 153 (80.1%) | |
| Surgery | 38 (8.0%) | 21 (11.0%) | |
| Neurology | 19 (4.0%) | 13 (6.8%) | |
| Psychiatry | 1 (0.2%) | 2 (1.1%) | |
| Unknown/missing | 9 (1.9%) | 2 (1.1%) | |
| Death within 6 months | 81 (9.7%) | 59 (23.7%) | <0.001 |
Of those enrolled, 249 (23.0%) had an abnormal RASS at initial presentation, and their distribution can be seen in Figure 2. Within 6 months, patients with an abnormal RASS were more likely to die compared with patients with a RASS of 0 (23.7% vs 9.7%, P<0.001). The Kaplan‐Meier survival curves for all enrolled patients with impaired and normal RASS can be seen in Figure 3; the survival curve declined more slowly in patients with a normal RASS compared with those with an abnormal RASS.
Using Cox proportional hazards regression, the relationship between an abnormal RASS at initial presentation and 6‐month mortality persisted (HR: 1.73, 95% CI: 1.21‐2.49) after adjusting for age, sex, nonwhite race, comorbidity burden, severity of illness, dementia, functional dependence, nursing home residence, psychoactive medications given, and admission status. The interaction between an abnormal RASS and APS (severity of illness) had a P value of 0.52. The interaction between an abnormal RASS and benzodiazepine or opioid medication administration had a P value of 0.38. The interaction between an abnormal RASS and admission status had a P value of 0.57. This indicated that severity of illness, psychoactive medication administration, and admission status did not modify the relationship between an abnormal RASS and 6‐month mortality.
We analyzed a subset of 406 patients who received a psychiatrist's assessment to determine if an abnormal RASS was associated with 6‐month mortality regardless of delirium status using Cox proportional hazard regression weighted by the inverse of the propensity score. Patients with an abnormal RASS and no delirium were significantly associated with higher mortality compared to those with a normal RASS (HR: 2.20, 95% CI: 1.10‐4.41). Patients with an abnormal RASS with delirium also had an increased risk for 6‐month mortality (HR: 2.86, 95% CI: 1.29‐6.34).
All regression models were internally validated. There was no evidence of substantial overfitting or collinearity. The Schoenfeld residuals for each model were examined graphically and there was good model fit overall, and no significant outliers were observed.
DISCUSSION
Vital sign measurements are a fundamental component of patient care, and abnormalities can serve as an early warning signal of the patient's clinical deterioration. However, traditional vital signs do not include an assessment of the patient's brain function. Our chief finding is that impaired arousal at initial presentation, as determined by the nonphysician research staff, increased the risk of 6‐month mortality by 73% after adjusting for confounders in a diverse group of acutely ill older patients. This relationship existed regardless of severity of illness, administration of psychoactive medications, and admission status. Though impaired arousal is closely linked with delirium,[2, 24] which is another well‐known predictor of mortality,[11, 25, 26] the prognostic significance of impaired arousal appeared to extend beyond delirium. We observed that the relationship between 6‐month mortality and impaired arousal in the absence of delirium was remarkably similar to that observed with impaired arousal with delirium. Arousal can be assessed for by simply observing the patient during routine clinical care and can be performed by nonphysician and physician healthcare providers. Its assessment should be performed and communicated in conjunction with traditional vital sign measurements in the emergency department and inpatient settings.[1]
Most of the data linking impaired arousal to death have been collected in the ICU. Coma, which represents the most severe form of depressed arousal, has been shown to increase the likelihood of death regardless of underlying etiology.[27, 28, 29, 30, 31] This includes patients who have impaired arousal because they received sedative medications during mechanical ventilation.[32] Few studies have investigated the effect of impaired arousal in a non‐ICU patient population. Zuliani et al. observed that impaired arousal was associated with 30‐day mortality, but their study was conducted in 469 older stroke patients, limiting the study's external validity to a more general patient population.[33] Our data advance the current stage of knowledge; we observed a similar relationship between impaired arousal and 6‐month mortality in a much broader clinical population who were predominantly not critically ill regardless of delirium status. Additionally, most of our impaired arousal cohort had a RASS of 1, indicating that even subtle abnormalities portended adverse outcomes.
In addition to long‐term prognosis, the presence of impaired arousal has immediate clinical implications. Using arousal scales like the RASS can serve as a way for healthcare providers to succinctly communicate the patient's mental status in a standardized manner during transitions of care (eg, emergency physician to inpatient team). Regardless of which clinical setting they are in, older acutely ill patients with an impaired arousal may also require close monitoring, especially if the impairment is acute. Because of its close relationship with delirium, these patients likely have an underlying acute medical illness that precipitated their impaired arousal.
Understanding the true clinical significance of impaired arousal in the absence of delirium requires further study. Because of the fluctuating nature of delirium, it is possible that these patients may have initially been delirious and then became nondelirious during the psychiatrist's evaluation. Conversely, it is also possible that these patients may have eventually transitioned into delirium at later point in time; the presence of impaired arousal alone may be a precursor to delirium. Last, these patients may have had subsyndromal delirium, which is defined as having 1 or more delirium symptoms without ever meeting full DSM‐IV‐TR criteria for delirium.[34] Patients with subsyndromal delirium have poorer outcomes, such as prolonged hospitalizations, and higher mortality than patients without delirium symptoms.[34]
Additional studies are also needed to further clarify the impact of impaired arousal on nonmortality outcomes such as functional and cognitive decline. The prognostic significance of serial arousal measurements also requires further study. It is possible that patients whose impaired arousal rapidly resolves after an intervention may have better prognoses than those who have persistent impairment. The measurement of arousal may have additional clinical applications in disease prognosis models. The presence of altered mental status is incorporated in various disease‐specific risk scores such as the CURB‐65 or Pneumonia Severity Index for pneumonia,[35, 36] and the Pulmonary Embolism Severity Index for pulmonary embolism.[37] However, the definition of altered mental status is highly variable; it ranges from subjective impressions that can be unreliable to formal cognitive testing, which can be time consuming. Arousal scales such as the RASS may allow for more feasible, reliable, and standardized assessment of mental status. Future studies should investigate if incorporating the RASS would improve the discrimination of these disease‐severity indices.
This study has several notable limitations. We excluded patients with a RASS of 4 and 5, which represented comatose patients. This exclusion, however, likely biased our findings toward the null. We enrolled a convenience sample that may have introduced selection bias. However, our enrolled cohort was similar to all potentially eligible patients who presented to the emergency department during the study period. We also attempted to mitigate this selection bias by using multivariable regression and adjusting for factors that may have confounded the relationship between RASS and 6‐month mortality. This study was performed at a single, urban, academic hospital and enrolled patients who were aged 65 years and older. Our findings may not be generalizable to other settings and to those who are under 65 years of age. Because 406 patients received a psychiatric evaluation, this limited the number of covariates that could be incorporated into the multivariable model to evaluate if impaired arousal in the absence of delirium is associated with 6‐month mortality. To minimize residual confounding, we used the inverse weighted propensity score, but we acknowledge that this bias may still exist. Larger studies are needed to clarify the relationships between arousal, delirium, and mortality.
CONCLUSION
In conclusion, impaired arousal at initial presentation is an independent predictor for 6‐month mortality in a diverse group of acutely ill older patients, and this risk appears to be present even in the absence of delirium. Because of its ease of use and prognostic significance, it may be a useful vital sign for underlying brain dysfunction. Routine standardized assessment and communication of arousal during routine clinical care may be warranted.
Disclosures: Research reported in this publication was supported by the Vanderbilt Physician Scientist Development Award, Emergency Medicine Foundation, and National Institute on Aging of the National Institutes of Health under award number K23AG032355. This study was also supported by the National Center for Research Resources, grant UL1 RR024975‐01, and is now at the National Center for Advancing Translational Sciences, grant 2 UL1 TR000445‐06. Dr. Vasilevskis was supported in part by the National Institute on Aging of the National Institutes of Health under award number K23AG040157. Dr. Powers was supported by Health Resources and Services Administration Geriatric Education Centers, grant 1D31HP08823‐01‐00. Dr. Storrow was supported by the National Heart, Lung, and Blood Institute of the National Institutes of Health under award number K12HL1090 and the National Center for Advancing Translational Sciences under award number UL1TR000445. Dr. Ely was supported in part by the National Institute on Aging of the National Institutes of Health under award numbers R01AG027472 and R01AG035117, and a Veteran Affairs MERIT award. Drs. Vasilevskis, Schnelle, Dittus, Powers, and Ely were supported by the Veteran Affairs Geriatric Research, Education, and Clinical Center. The content is solely the responsibility of the authors and does not necessarily represent the official views of Vanderbilt University, Emergency Medicine Foundation, National Institutes of Health, and Veterans Affairs. The funding agencies did not have any role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, review, or approval of the manuscript.
J.H.H., E.W.E., J.F.S., A.B.S., and R.D.S. conceived the trial. J.H.H., E.W.E., A.B.S., J.F.S., R.D.S., A.S., and A.W. participated in the study design. J.H.H. and A.W. recruited patients and collected the data. J.H.H., A.J.G., and A.S. analyzed the data. All authors participated in the interpretation of results. J.H.H. drafted the manuscript, and all authors contributed to the critical review and revision of the manuscript.
The authors report no conflicts of interest.
- , , , et al. The development of a mental status vital sign for use across the spectrum of care. J Am Med Dir Assoc. 2009;10:379–380.
- , , , . Serial administration of a modified Richmond Agitation and Sedation Scale for delirium screening. J Hosp Med. 2012;7:450–453.
- , , , , , . Clarifying confusion: the confusion assessment method. A new method for detection of delirium. Ann Intern Med. 1990;113:941–948.
- , , , et al. Delirium in mechanically ventilated patients: validity and reliability of the confusion assessment method for the intensive care unit (CAM‐ICU). JAMA. 2001;286:2703–2710.
- , , , et al. Diagnosing delirium in older emergency department patients: validity and reliability of the Delirium Triage Screen And The Brief Confusion Assessment Method. Ann Emerg Med. 2013;62:457–465.
- , , , , . Does delirium contribute to poor hospital outcomes? A three‐site epidemiologic study. J Gen Intern Med. 1998;13:234–242.
- , , , . Prognostic significance of delirium in frail older people. Dement Geriatr Cogn Disord. 2005;19:158–163.
- , , , , , . Delirium in elderly patients and the risk of postdischarge mortality, institutionalization, and dementia: a meta‐analysis. JAMA. 2010;304:443–451.
- , , , , . One‐year health care costs associated with delirium in the elderly population. Arch Intern Med. 2008;168:27–32.
- , , , et al. The Richmond Agitation‐Sedation Scale: validity and reliability in adult intensive care unit patients. Am J Respir Crit Care Med. 2002;166:1338–1344.
- , , , et al. Delirium in the emergency department: an independent predictor of death within 6 months. Ann Emerg Med. 2010;56:244–252.
- , , , et al. Delirium in older emergency department patients is an independent predictor of hospital length of stay. Acad Emerg Med. 2011;18:451–457.
- , , , et al. Validation of the Confusion Assessment Method For The Intensive Care Unit in older emergency department patients. Acad Emerg Med. 2014;21:180–187.
- , , , et al. Monitoring sedation status over time in ICU patients: reliability and validity of the Richmond Agitation‐Sedation Scale (RASS). JAMA. 2003;289:2983–2991.
- , , , . Does this patient have dementia? JAMA. 2007;297:2391–2404.
- . A short form of the Informant Questionnaire on Cognitive Decline in the Elderly (IQCODE): development and cross‐validation. Psychol Med. 1994;24:145–153.
- . Assessing self‐maintenance: activities of daily living, mobility, and instrumental activities of daily living. J Am Geriatr Soc. 1983;31:721–727.
- , , , , . Charlson Index is associated with one‐year mortality in emergency department patients with suspected infection. Acad Emerg Med. 2006;13:530–536.
- , , , . APACHE II: a severity of disease classification system. Crit Care Med. 1985;13:818–829.
- American Psychiatric Association. Task Force on DSM‐IV. Diagnostic and Statistical Manual of Mental Disorders: DSM‐IV‐TR. 4th ed. Washington, DC: American Psychiatric Association; 2000.
- . Regression Modeling Strategies: With Applications to Linear Models, Logistic Regression, and Survival Analysis. New York, NY: Springer; 2001.
- . Power for tests of interaction: effect of raising the Type I error rate. Epidemiol Perspect Innov. 2007;4:4.
- . An Introduction to Propensity Score Methods for Reducing the Effects of Confounding in Observational Studies. Multivariate Behav Res. 2011;46:399–424.
- , , . Defining delirium for the International Classification of Diseases, 11th Revision. J Psychosom Res. 2008;65:207–214.
- , , , , . Delirium predicts 12‐month mortality. Arch Intern Med. 2002;162:457–463.
- , , , et al. Delirium as a predictor of mortality in mechanically ventilated patients in the intensive care unit. JAMA. 2004;291:1753–1762.
- , , . Predicting mortality of intensive care unit patients. The importance of coma. Crit Care Med. 1982;10:86–95.
- , . Assessment of outcome after severe brain damage. Lancet. 1975;1:480–484.
- , , , , , . Predicting outcome from hypoxic‐ischemic coma. JAMA. 1985;253:1420–1426.
- , , , et al. Prediction of intracerebral hemorrhage survival. Ann Neurol. 1988;24:258–263.
- , , , . Is this patient dead, vegetative, or severely neurologically impaired? Assessing outcome for comatose survivors of cardiac arrest. JAMA. 2004;291:870–879.
- , , , et al. Early intensive care sedation predicts long‐term mortality in ventilated critically ill patients. Am J Respir Crit Care Med. 2012;186:724–731.
- , , , , , . Risk factors for short‐term mortality in older subjects with acute ischemic stroke. Gerontology. 2006;52:231–236.
- , , , . The prognostic significance of subsyndromal delirium in elderly medical inpatients. J Am Geriatr Soc. 2003;51:754–760.
- , , , et al. Defining community acquired pneumonia severity on presentation to hospital: an international derivation and validation study. Thorax. 2003;58:377–382.
- , , , et al. A prediction rule to identify low‐risk patients with community‐acquired pneumonia. N Engl J Med. 1997;336:243–250.
- , , , et al. Derivation and validation of a prognostic model for pulmonary embolism. Am J Respir Crit Care Med. 2005;172:1041–1046.
- , , , et al. The development of a mental status vital sign for use across the spectrum of care. J Am Med Dir Assoc. 2009;10:379–380.
- , , , . Serial administration of a modified Richmond Agitation and Sedation Scale for delirium screening. J Hosp Med. 2012;7:450–453.
- , , , , , . Clarifying confusion: the confusion assessment method. A new method for detection of delirium. Ann Intern Med. 1990;113:941–948.
- , , , et al. Delirium in mechanically ventilated patients: validity and reliability of the confusion assessment method for the intensive care unit (CAM‐ICU). JAMA. 2001;286:2703–2710.
- , , , et al. Diagnosing delirium in older emergency department patients: validity and reliability of the Delirium Triage Screen And The Brief Confusion Assessment Method. Ann Emerg Med. 2013;62:457–465.
- , , , , . Does delirium contribute to poor hospital outcomes? A three‐site epidemiologic study. J Gen Intern Med. 1998;13:234–242.
- , , , . Prognostic significance of delirium in frail older people. Dement Geriatr Cogn Disord. 2005;19:158–163.
- , , , , , . Delirium in elderly patients and the risk of postdischarge mortality, institutionalization, and dementia: a meta‐analysis. JAMA. 2010;304:443–451.
- , , , , . One‐year health care costs associated with delirium in the elderly population. Arch Intern Med. 2008;168:27–32.
- , , , et al. The Richmond Agitation‐Sedation Scale: validity and reliability in adult intensive care unit patients. Am J Respir Crit Care Med. 2002;166:1338–1344.
- , , , et al. Delirium in the emergency department: an independent predictor of death within 6 months. Ann Emerg Med. 2010;56:244–252.
- , , , et al. Delirium in older emergency department patients is an independent predictor of hospital length of stay. Acad Emerg Med. 2011;18:451–457.
- , , , et al. Validation of the Confusion Assessment Method For The Intensive Care Unit in older emergency department patients. Acad Emerg Med. 2014;21:180–187.
- , , , et al. Monitoring sedation status over time in ICU patients: reliability and validity of the Richmond Agitation‐Sedation Scale (RASS). JAMA. 2003;289:2983–2991.
- , , , . Does this patient have dementia? JAMA. 2007;297:2391–2404.
- . A short form of the Informant Questionnaire on Cognitive Decline in the Elderly (IQCODE): development and cross‐validation. Psychol Med. 1994;24:145–153.
- . Assessing self‐maintenance: activities of daily living, mobility, and instrumental activities of daily living. J Am Geriatr Soc. 1983;31:721–727.
- , , , , . Charlson Index is associated with one‐year mortality in emergency department patients with suspected infection. Acad Emerg Med. 2006;13:530–536.
- , , , . APACHE II: a severity of disease classification system. Crit Care Med. 1985;13:818–829.
- American Psychiatric Association. Task Force on DSM‐IV. Diagnostic and Statistical Manual of Mental Disorders: DSM‐IV‐TR. 4th ed. Washington, DC: American Psychiatric Association; 2000.
- . Regression Modeling Strategies: With Applications to Linear Models, Logistic Regression, and Survival Analysis. New York, NY: Springer; 2001.
- . Power for tests of interaction: effect of raising the Type I error rate. Epidemiol Perspect Innov. 2007;4:4.
- . An Introduction to Propensity Score Methods for Reducing the Effects of Confounding in Observational Studies. Multivariate Behav Res. 2011;46:399–424.
- , , . Defining delirium for the International Classification of Diseases, 11th Revision. J Psychosom Res. 2008;65:207–214.
- , , , , . Delirium predicts 12‐month mortality. Arch Intern Med. 2002;162:457–463.
- , , , et al. Delirium as a predictor of mortality in mechanically ventilated patients in the intensive care unit. JAMA. 2004;291:1753–1762.
- , , . Predicting mortality of intensive care unit patients. The importance of coma. Crit Care Med. 1982;10:86–95.
- , . Assessment of outcome after severe brain damage. Lancet. 1975;1:480–484.
- , , , , , . Predicting outcome from hypoxic‐ischemic coma. JAMA. 1985;253:1420–1426.
- , , , et al. Prediction of intracerebral hemorrhage survival. Ann Neurol. 1988;24:258–263.
- , , , . Is this patient dead, vegetative, or severely neurologically impaired? Assessing outcome for comatose survivors of cardiac arrest. JAMA. 2004;291:870–879.
- , , , et al. Early intensive care sedation predicts long‐term mortality in ventilated critically ill patients. Am J Respir Crit Care Med. 2012;186:724–731.
- , , , , , . Risk factors for short‐term mortality in older subjects with acute ischemic stroke. Gerontology. 2006;52:231–236.
- , , , . The prognostic significance of subsyndromal delirium in elderly medical inpatients. J Am Geriatr Soc. 2003;51:754–760.
- , , , et al. Defining community acquired pneumonia severity on presentation to hospital: an international derivation and validation study. Thorax. 2003;58:377–382.
- , , , et al. A prediction rule to identify low‐risk patients with community‐acquired pneumonia. N Engl J Med. 1997;336:243–250.
- , , , et al. Derivation and validation of a prognostic model for pulmonary embolism. Am J Respir Crit Care Med. 2005;172:1041–1046.
© 2014 Society of Hospital Medicine
