OLDER ADULTS COMPRISE the fastest growing emergency department demographic. Undertriage among this population may increase delays in care, emergency department (ED) revisits, and poor treatment outcomes. This article critiques a recent study assessing the accuracy of the current emergency severity index (ESI) triage system in predicting mortality among older adults and examines whether additional triage components reduce the risk of undertriage and poor treatment outcomes.
Mrs P is a 75-year-old divorced woman, with a medical history of well-controlled hypertension, hypothyroidism, and hyperlipidemia who presents to the ED complaining of generalized weakness, gradually worsening over the past 3 months. She denied fever, chills, headaches, dizziness, nausea, vomiting, diarrhea, abdominal or chest pain, shortness of breath, dysuria, or hematuria. She described her weakness as not feeling well and stated that she just doesn't “want to get out of the bed in the morning.” Mrs P drove herself to the ED, was able to ambulate from the valet parking to the registration desk, and was triaged within 15 min of arrival. Mrs P clearly did not require immediate lifesaving intervention, and due to the chronic nature of her compliant, she did not appear to be at high risk. Her vital signs were within normal limits, so she received an ESI rating of 3 (patient with vital signs that do not meet the danger zone criteria and requiring more than one resource). Because of this ESI level, Mrs P waited 7.5 hr to see a provider without any laboratory work or imaging ordered. During her wait time her shortness of breath increased, but she decided not to “bother” the busy triage nurse with her complaint.
When she was seen by a nurse practitioner (NP), a more detailed history was obtained. The NP questioned Mrs P about her activities of daily living (ADL). Mrs P stated that she had been eating a normal volume of food but has not been motivated to prepare a meal and has been mostly eating snack foods, which is typical for her. She showers every day and dresses herself, even if she does not have anywhere to go that day. She denied bowel or bladder incontinence and stated that she never has problems making it to the toilet in time. She denied exercising and endorsed driving everywhere including to her neighbor's house down the street. Further discussion revealed that she drives to the neighbor's because she cannot “get a good breath” if she walks. She further admitted that she plans her day's activities to limit the number of trips she makes upstairs in her home because she becomes “so winded” when she climbs the stairs.
Mrs P endorsed changing primary care providers (PCPs) 10 years ago when she became Medicare eligible because her previous PCP did not accept it. She stated that upon changing providers she received a “Welcome to Medicare” physical examination (Centers for Medicare & Medicaid Services, 2020) and has been diligent about having annual Medicare wellness examinations since. Mrs P is currently taking losartan 20 mg once a day, aspirin 81 mg once a day, and levothyroxine 75 mg once a day and is adherent with her medication regimen. She added that her PCP had discussed starting her on a statin medication but she declined because she does not like taking medications. She endorsed having occasional headaches, over the past 30 years, usually resolved with over-the-counter ibuprofen. She stated that 6 years ago she was worked up for chest pain and received a negative stress test. She did not follow up with a cardiologist and has not had any further symptoms since. She has no medication allergies.
Mrs P stated that she lives alone since her divorce 14 years ago and has no children. She endorsed drinking a glass of wine three to four times a week with dinner and having an occasional cocktail with a friend. She denied ever smoking or using recreational drugs. She is a retired flight attendant and still enjoys traveling internationally, but has not planned any trips recently because she has not been feeling well.
Upon physical examination, her vital signs were: blood pressure 145/87 mmHg, heart rate 98 bpm, respiratory rate 20, SPO2 94%, temperature of 36.7°C, and body mass index of 37 kg/m2 (morbidly obese). She was well nourished, well hydrated, in no apparent distress, and was alert and oriented to person, place, time, and situation. Her skin was warm and dry, with moist mucosal membranes. Her lungs were clear to auscultation with normal work of breathing. Her heart had a regular rate and rhythm with a grade 2 diastolic murmur. Peripheral pulses were present and equal bilaterally with 2+ pitting edema in her lower extremities bilaterally. Her abdomen was soft, nontender, and nondistended with normal bowel sounds. There were no pulsatile masses, bruits, or hepatosplenomegaly. Her neurological examination was normal with no focal deficits and with 3+ reflexes in upper and lower extremities. Her musculoskeletal examination revealed a full range of motion with 5/5 strength in all extremities.
An electrocardiogram (EKG), complete metabolic panel (CMP), complete blood count (CBC), troponin, b-type natriuretic peptide (BNP), thyroid-stimulating hormone level (TSH), and chest radiographs were ordered to rule out a metabolic, pulmonary, or cardiac cause for the weakness.
REVIEW OF THE ARTICLE
Malinovska, A., Pitasch, L., Geigy, N., Nickel, C. H., & Bingisser, R. (2019). Modification of the emergency severity index improves mortality prediction in older patients. The Western Journal of Emergency Medicine, 20(4), 633–640. doi:10.5811/westjem.2019.4.40031
Malinovska, Pitasch, Geigy, Nickel, and Bingisser's (2019) study was designed to address the concern that older adults presenting to the ED with nonspecific complaints (NSC) are often undertriaged using the ESI method, leading to increased adverse outcomes including permanent functional decline. Malinovska et al. (2019) examined whether adding additional clinical factors, including additional vital sign assessments, inquiring about recent changes in ADL, incoherence in the patient's history, and incorporating provider “first impression” to the ESI triage method, improved triage accuracy and prediction of 30-day mortality.
The authors' (Malinovska et al., 2019) defined NSCs as symptoms, such as weakness, that are not suggestive of a working diagnosis that can be managed by evidence-based protocols. They defined undertriage as a misclassification of a patient into a lower urgency group, occurring most often between ESI level 2 (patients who should not wait due to the high-risk nature of their presentation) and ESI level 3 (patients with vital signs that meet certain criteria for urgency or a need for additional resources for diagnosis and treatment) (Christ, Grossmann, Winter, Bingisser, & Platz, 2010). The authors recognized that older adults often present to the ED with NSC and atypical presentations. Because of this, they considered adding other factors to the traditional ESI triage classification system as a means of improving triage accuracy, improving clinical outcomes, and reducing the undertriaging this vulnerable population.
Study Design and Methods
The study used a prospective, observational design including a 30-day follow-up. Data were collected over a 4-year period at three different EDs ranging in size from a 700-bed tertiary care hospital to a 400-bed secondary care hospital.
Participant inclusion criteria included patients 18 years and older presenting to the ED with NSCs and initially triaged as ESI 2 or 3. Exclusion criteria included patients with traumatic injuries, unstable vital signs, having specific complaints such as chest pain or demonstrating EKG changes, or those with known abnormal laboratory values or incomplete data. Patients meeting inclusion criteria were enrolled after a review of their medical history, completion of a focused physical examination, and prior to the return of laboratory results. Of 1401 possible candidates, 948 met inclusion criteria and were enrolled in the study.
Data Analysis and Results
Malinovska et al. (2019) examined the unique and combined effects of 12 additional clinical variables, combined with the usual ESI triage system, in predicting 30-day mortality among adults. The additional variables included age, sex, heart rate, respiratory rate, oxygen saturation, systolic blood pressure, low temperature, EKG changes, mode of arrival—ambulance versus walk-in, incoherence of history, ADL deterioration, and the provider's first impression of the patient's condition. ADL deterioration was assessed by having the provider ask the patient if they were experiencing any difficulty in their ability to bathe, dress, move, feed themselves, or maintain toilet hygiene. Incoherence of history was rated as a yes or no based on the presence of inconsistencies in the history provided by the patient. The first impression of the provider was a subjective 0–10 score, ranging from “looks healthy” to “looks critically ill.” Univariate logistic regression was used to test which variables were significantly associated with 30-day mortality at a 95% confidence interval (CI).
Descriptive statistics, including χ2, t tests and Kruskal–Wallis tests, were used to compare effects of each of the 12 clinical variables. To compare the predictive accuracy of the additional variables, Malinovskaya et al. (2019) used the area under the curve of the receiver operating characteristic curve with a 95% CI.
To examine the predictive effectiveness of vital signs, Malinovska et al. (2019) combined respiratory rate, heart rate, and oxygen saturation to calculate an ESI “vital score” and found that this combined vital score had greater accuracy in predicting 30-day mortality than the traditional ESI method. After analyzing the accuracy of the ESI “vital score,” Malinovska et al. (2019) added in each of the above individual additional variables to the ESI “vital score” and found that each additional variable resulted in a statistically significant improvement (66%–71%) in predicting 30-day mortality but that no one variable was statistically better than the other.
Next, they compared the combined effects of different variables with the ESI vital score in a pair-wise fashion, including ADL deterioration with incoherency of history, ADL deterioration with first impression by provider, and incoherency of history with first impression of provider. Their results showed that although each pair improved the predictive power of 30-day mortality when combined with the ESI “vital score,” the combination of the “vital score” with provider first impression had the greatest predictive power.
Study Strengths, Limitations, and Conclusions
Perhaps the greatest strength of the study was inclusion of a broad range of novel variables to the usual ESI triage system in assessing older adults presenting with NSC complaints. In using multiple variables both in isolation and in combinations, Malinovska et al. (2019) were able to discern which additional factors improve triage accuracy to better expedite care for patients who are the most vulnerable when presenting with NSCs.
Another strength was that the study was conducted at three different emergency departments at hospitals of various sizes and populations lending to its generalizability of results.
One limitation is that the investigators excluded patients presenting with abnormal vital signs, such as tachycardia and hypoxia, which may have had strong predictive accuracy if included in the analysis. Another study limitation was the use of 30-day mortality as an outcome measure. Because the study included a relatively stable population of adult patients 18 years and older who were not at an immediate risk of death based on their presenting complaints, measuring 30-day mortality may underestimate the accuracy of adding in the individual triage components. Many older adults who present to the ED and are assigned an ESI rating of 2 or 3 are experiencing new onset or acute exacerbations of conditions that require ongoing, long-term management. A more sensitive outcome measure than mortality may be rates of hospital admission versus ED discharge, or the use of the PARIS score (Brabrand, Knudsen, Hallas, Graham, & Kellett, 2018), which includes loss of independence as an outcome measure, along with 30-day readmission rates. Further research evaluating these outcomes in concert with acute morbidity and loss of function may be more beneficial when assessing potential adverse outcomes and undertriage in the older adult population.
Malinovska et al. (2019) recognized that age-related vital sign changes in older adults can result in undertriage because vital signs are less sensitive indicators of serious illness in patients 75 years and older. Therefore, an age-adjusted vital sign cutoff in triage may reduce undertriaging these patients as earlier research has suggested (Platts-Mills et al., 2010). Malinovska et al.'s (2019) findings that the best predictors of 30-day mortality were ADL deterioration and first impression by the provider are consistent with previous investigations (LaMantia et al., 2013) highlighting the importance of including these components when assessing older adults.
From 2014 to 2017, patients 60 years and older comprised 20%, or approximately 29 million, ED visits annually in the United States (Ashman, Schappert, & Santo, 2020). Therefore, undertriage of this older adult population is a serious and growing concern that can have devastating outcomes.
Grossmann et al. (2012) examined ESI triage accuracy among patients 65 years and older, by reviewing acute resuscitation procedures performed within 1 hr for patients triaged as ESI level 1. For level 2 and 3 patients, Grossmann et al. (2012) reviewed the resources that were used during the ED stay, including blood draws, EKG, imaging, interventions (e.g., fluid resuscitation), procedures (e.g., laceration repair), and specialty consultations. They also assessed interrater reliability of the ESI algorithm by comparing the ESI rating assigned by the triage nurse to one assigned by an emergency physician and an advanced practice nurse, who were blinded to the triage nurse's initial assigned ESI level. Grossmann et al. (2012) found that interrater reliability was highest for patients triaged at ESI level 1 (those requiring immediate care), and was lowest with more undertriage among patients assigned an initial triage ESI level of 3 when their actual presentation met ESI level 2 (requiring emergent evaluation) criteria. Of concern is that if a patient is undertriaged as an ESI level of 3, when presenting with an unrecognized emergent condition (ESI 2), they may quickly decompensate and become unstable during a prolonged waiting room stay when necessary treatment is delayed.
Despite this finding, Grossmann et al. (2012) concluded that the current ESI methodology was sufficient for the older adult population.
In a similar study, Platts-Mills et al. (2010) assessed triage accuracy in a retrospective review of 782 cases of adults 65 years and older requiring resuscitative interventions in the ED and found that less than half of these unstable patients were appropriately triaged using the ESI method. Among this sample, 34% of patients were undertriaged, and no cases were overtriaged (assigned a higher triage level than the presentation warranted). And like Grossmann et al. (2012),Platts-Mills et al. (2010) did not propose changing the ESI system to prevent undertriage among older adults or suggest alternative measures to improve accuracy. The findings from both of these studies demonstrate that triage accuracy increases when it is performed by expert nurses and experienced providers.
Malinovska et al.'s (2019) finding that the most significant predictor of 30-day mortality in adults presenting with NSCs was clinician first impression suggests there is an art in “recognizing” a vulnerable patient that transcends objective findings alone and can lead to improved clinical outcomes. This has been concluded in other investigations. For example, Herzog, Jenny, Nickel, Nieves-Ortega, and Bingisser. (2020) found that a physician's “impression” of acute morbidity and mortality in patients presenting to the ED with generalized weakness and fatigue accurately predicted mortality, hospitalization, and transfer to the intensive care unit. Herzog et al.'s finding supports that clinical acumen is important when practicing in the ED where rapid decision-making can prevent adverse outcomes by expediting care.
In an ED process investigation, Spencer, Stephens, Swanson-Biearman, and Whiteman. (2019) found that time to provider, length of stay, and the number of patients who left without being seen (LWBS) were all significantly reduced by assigning a physician or advanced practice provider (APP) to triage. Although their study did not specifically review the financial implications of such a model, it did reveal that the presence of a physician or APP in triage prevented a loss of revenue. For example, the investigators found that every LWBS patient represented an average potential loss of $480 in revenue equating to approximately $1.3 million in lost revenue over a 6-month period (Spencer et al., 2019). By assigning APPs in the triage area, they generated an additional gross profit (increase in revenue minus the APP salary cost) of over $806,000 (Spencer et al., 2019). These findings further support that having experienced clinicians in triage expedites care and improves triage accuracy that is likely associated with an experienced provider's clinical reasoning and first “impression.”
Strategies that strengthen medical and NP student clinical reasoning and decision-making skills are an increasing focus within interprofessional education (Graber et al., 2018). Suggested strategies for improving analytical reasoning include use of evidence-based clinical guidelines and algorithms, use of rationale decision-making strategies, and reflective methods to improve awareness of personal biases that can negatively affect the diagnostic process and patient care (Graber et al., 2018). These recommendations seem to contradict the finding that a provider's first impression, which results from rapid pattern recognition and an intuitive mode of reasoning, improves medical decision-making and patient management. However, Croskerry (2020) discusses that using intuition in decision-making in the ED is not wrong. ED clinical decisions typically involve a dual process where analytical reasoning is coupled with intuition, which can save time when decisions are straightforward or routine (Croskerry, 2020). However, it is imperative that learners and ED providers practice techniques to improve awareness of when they are using intuition to avoid diagnostic pitfalls such as premature closure and anchoring that can adversely affect patient care. Teaching these skills to NPs preparing to work in emergency care settings is essential.
Malinovska et al.'s (2019) finding that asking older adult patients about changes in ADL improves triage accuracy is significant because it is easy to add to a rapid triage assessment. Park, Lee, and Kim. (2017) also found, in a study of older adult hospice patients, that low ADL scores of functional status were associated with a greater risk of death. This suggests that ADL deterioration as a clinical indicator of health outcomes among older adults may improve their care and disposition and represents a gap in current research. Calculating an objective ADL score would also be helpful in assessing patients over time or at repeat ED visits.
In reviewing Mrs P's diagnostic studies, the NP noted nonspecific T-wave abnormalities on the EKG with a normal axis and sinus rate of 97. Her CBC, CMP, and TSH, and serial troponins were within normal limits. However, her BNP was 1,200 ng/L, and her chest radiograph was remarkable for pulmonary edema that was subsequently treated in the ED with furosemide 100 mg intravenously. The NP ordered an echocardiogram and admitted her to the hospital medicine service for new-onset heart failure.
Mrs P's initial triage level of 3 resulted in nearly an 8-hr delay in care in the waiting room. When asked about her ADL, Mrs P complained that she was unable to walk to her neighbor's house a block away, and added that in the past month or so she had started planning her day around limiting the number of trips upstairs or trips for errands, because of shortness of breath. Since her symptoms had developed gradually, she ignored them until they became the focus of her attention. If during triage, an experienced NP had queried Mrs P about any recent changes in her ADL, and noticed her 2+ pitting edema, the NP's initial impression of possible heart failure would have resulted in assigning a triage level of 2, emergent, though not critical. A triage level of 2 would have expedited her transfer to an ED for further evaluation reducing her wait time from 7.5 hours to 30 min, and by initiating her workup at triage ED visit time and total length of stay would have been significantly less, improving ED flow and efficiency. Mrs P would also have received treatment for her low oxygen saturation and dyspnea sooner reducing her symptoms and improving the quality of her care. As a result of being triaged as level 3, Mrs P's respiratory distress was unnecessarily prolonged. However, her ED management and inpatient hospitalization had a good outcome. After spending 4 days in the hospital her condition significantly improved, and she was discharged home with a cardiology follow-up appointment and for cardiac rehabilitation to improve her physical conditioning. Her 30-day follow-up found her much improved and able to perform her ADL without difficulty.
An alternative scenario leading to a more serious outcome is that Mrs P waited in the lobby for approximately an hour noticing that patients arriving after her were being taken from the waiting room to the ED treatment area. She approached the registration desk and was informed that patients are seen on the basis of how sick they are, not on a first-come-first-serve basis. Mrs P acknowledged that she understood that and told the front desk attendant that she was not feeling well, was weak and needed to lie down. The front desk attendant informed her that she would be seen as soon as they could get to her. Mrs P returned to her seat. After another hour passed, Mrs P asked the front desk attendant how much longer would she have to wait and the attendant informed her that, based on her acuity level, several people were ahead of her, so it would be a couple of more hours at least. Mrs P decided to leave without being seen.
Approximately 3 weeks later, Mrs P started up her staircase when she became dizzy, lost consciousness, and fell down four stairs to the ground level. When she regained consciousness, she attempted to get herself up from the floor, but was unable to stand because of severe pain in her left hip. She was able to crawl to the telephone where she dialed 911 for help. The dispatch operator instructed her to crawl to the door and unlock it so the paramedics would not have to break it down. She did as she was instructed, though very slowly and with great pain. The paramedics soon arrived and transported her to the same hospital from where she left 3 weeks earlier. This time, she was taken straight to a room where she received a stat EKG. She also had a CBC, CMP, troponin, BNP, and TSH drawn. Chest and left hip radiographs were obtained and a computerized tomography (CT) scan of her head was performed. Her syncope workup also included a stat echocardiogram. Although the CT showed no evidence of intracranial bleeding, the BNP result was at 1,950 ng/L and the echocardiogram revealed an ejection fraction of 15%. Diuresis was initiated in the ED. The chest radiograph was remarkable for significant pulmonary effusion, interstitial edema, and cardiomegaly. The radiograph of the left hip revealed a fracture to the femoral head. Mrs P was admitted to the hospital for new-onset heart failure and left hip fracture. She was released 2 weeks later to rehabilitation while awaiting placement into an assisted living facility. She would not be able to return to her home of 40 years, and her prognosis remained uncertain.
Had Mrs P been appropriately triaged upon her first visit to the ED, she would have been assigned ESI level 2, waited approximately 30 min, and been rapidly diagnosed with heart failure. She would have spent 3–5 days in the hospital, and received appropriate ongoing care with a good outcome as described. In both scenarios, inaccurate triage led to an extended ED stay and persistent distress and in scenario two a devastating outcome and subsequent loss of independence.
Further research to examine the most accurate and efficient parameters for triaging older adults, a growing ED population, is needed to improve health outcomes, costs, and clinical resources. Development and validation of a standardized tool for triaging patients older than 65 years may improve clinical outcomes and reduce ED and hospital readmissions, and health care costs. Malinovska et al.'s (2019) study suggests that some simple changes may be made immediately to lower the rate of undertriage. Accurately recording respiratory rates and asking about ADL deterioration over the past 2 weeks would be parameters easily captured without additional costs, time, or resources. The addition of a provider's first impression would also be a parameter that could be practically implemented by careful assignment of providers to triage experienced in assessment of older adults. These simple changes may ultimately save lives. Emergency NP education and competency evaluation must incorporate this critical content, along with preparation to improve diagnostic reasoning, as the ED population of older adults who present with NSC and complex conditions continues to grow.
Ashman J. J., Schappert S. M., Santo L. (2020, June). Emergency department visits among adults aged 60 and over: United States, 2014–2017. National Center for Health Statistics Data Brief, (367), 1–8.
Brabrand M., Knudsen T., Hallas J., Graham C. A., Kellett J. (2018). The PARIS score can reliably predict 7-day all-cause mortality for both acute medical and surgical patients: An international validation study. QJM: An International Journal of Medicine, 111(10), 721–25. doi:10.1093/qjmed/hcy174
Croskerry P. (2020). Becoming less wrong (and more rational) in clinical decision making. Annals of Emergency Medicine, 75(2), 218–220. doi:10.1016/j.annemergmed.2019.09.002
Christ M., Grossmann F., Winter D., Bingisser R., Platz E. (2010). Modern triage in the emergency department. Deutsches Ärzteblatt International, 107(50), 892–898. doi:10.3238/arztebl.2010.089
Graber M. L., Rencic J., Rusz D., Papa F., Croskerry P., Zierler B., Olson A. P. J. Improving diagnosis by improving education: A policy brief on education in healthcare professions. Diagnosis, 5(3), 107–118. doi: 10.1515/dx-2018-0033
Grossmann F. F., Zumbrunn T., Frauchiger A., Delport K., Bingisser R., Nickel C. H. (2012). At risk of under-triage? Testing the performance and accuracy of the emergency severity index in older emergency department patients. Annals of Emergency Medicine, 60(3), 317–325.e3. doi:10.1016/j.annemergmed.2011.12.013
Herzog S. M., Jenny M. A., Nickel C. H., Nieves-Ortega R., Bingisser R. (2020). Emergency department patients with weakness or fatigue: Can physicians predict their outcomes at the front door? A prospective observational study. PLoS One, 15(11), e0239902. doi:10.1371/journal.pone.0239902
Lamantia M. A., Stewart P. W., Platts-Mills T. F., Biese K. J., Forbach C., Zamora E., Kizer J. S. (2013). Predictive value of initial triage vital signs for critically ill older adults. The Western Journal of Emergency Medicine, 14(5), 453–460.
Malinovska A., Pitasch L., Geigy N., Nickel C. H., Bingisser R. (2019). Modification of the emergency severity index improves mortality prediction in older patients. The Western Journal of Emergency Medicine, 20(4), 633–640. doi:10.5811/westjem.2019.4.40031
Park H. S., Lee K. T., Kim T. W. (2017). Role of physical activity in mortality prediction in elderly hospice patients. Journal of Exercise Rehabilitation, 13(2), 250–254. doi:10.12965/jer.1734950.475
Platts-Mills T. F., Travers D., Biese K., McCall B., Kizer S., LaMantia M., Cairns C. B. (2010). Accuracy of the emergency severity index triage instrument for identifying elder emergency department patients receiving an immediate life-saving Intervention. Academic Emergency Medicine, 17(3), 238–243.
Spencer S., Stephens K., Swanson-Biearman B., Whiteman K. (2019). Health care provider in triage to improve outcomes. Journal of Emergency Nursing, 45(5), 561–566. doi:10.1016/j.jen.2019.01.008