Sorry, you need to enable JavaScript to visit this website.
Skip to main content

Failed Interpretation of Screening Tool: Delayed Treatment

Casey A. Cable, MD; David J. Murphy, MD, PhD; and Greg S. Martin, MD, MSc | September 1, 2017
View more articles from the same authors.

The Case

An 88-year-old man presented to the emergency department (ED) with a 2-day history of upper back pain. His medical history was significant for hypertension and longstanding mitral valve prolapse with mitral regurgitation. His initial vital signs showed a blood pressure of 130/75 mm hg, heart rate of 65 beats per minute, respiratory rate of 16 breaths per minute, temperature of 36.3°C, and an oxygen saturation of 90% on room air. Physical examination revealed a frail elderly man in no distress. He had normal heart sounds and a soft systolic murmur heard best at the apex; his lungs showed faint bilateral crackles. He had no spinal tenderness and a normal neurological examination, including mental status. The rest of his physical examination was unremarkable. Initial laboratory investigation demonstrated a normal white blood cell count, serum lactate, electrolytes, liver enzymes, and kidney function. A chest radiograph showed hilar fullness and mild bilateral infiltrates, consistent with pulmonary edema or early pneumonia.

The emergency physician assessed the patient as having either new onset congestive heart failure or an atypical pneumonia. She considered starting antibiotics but held off as the patient looked well and did not "screen positive" for sepsis. (The hospital had recently implemented a formal sepsis screening system for all ED patients.) The patient was admitted for workup of his hypoxemia. The admitting internist also deferred starting antibiotics, in part guided by the negative sepsis screening. He ordered an echocardiogram for the next morning and prescribed diuretic medications. At the time of ward admission (4 hours after presentation), the patient remained hemodynamically stable and afebrile.

Early the next morning, the patient acutely decompensated with worsened hypoxemia, hypotension, and delirium. He remained afebrile, but repeat labs now showed a markedly elevated serum lactate level and white blood cell count. A repeat chest radiograph demonstrated worsened bilateral infiltrates. He was started on broad-spectrum antibiotics and transferred to the intensive care unit. However, he rapidly experienced further respiratory decompensation requiring intubation and mechanical ventilation, as well as hemodynamic compromise requiring vasopressor support. Later that day, his blood and sputum cultures grew Escherichia coli. Over the next 4 days, he developed progressive organ dysfunction and died.

The Commentary

by Casey A. Cable, MD; David J. Murphy, MD, PhD; and Greg S. Martin, MD, MSc

This case provides an important example of diagnostic error as an unintended consequence of a decision support tool. On initial presentation of this 88-year-old man, the differential diagnosis of the emergency physician included new-onset heart failure and atypical pneumonia. Two diagnostic tests were ordered: an echocardiogram for heart failure and both blood and sputum cultures for a pulmonary infection. However, this patient was only empirically treated for one of the two potential diagnoses. He received diuretics for the potential diagnosis of heart failure, but he was not empirically treated for pneumonia with antibiotics, largely because of the negative sepsis screen and the clinician's interpretation that this indicated that an infection was not present. Unfortunately, this, and subsequent decisions, delayed antibiotic initiation until after the patient acutely decompensated. The case illustrates several important lessons, including the clinical presentation of infections and sepsis in older patients, the use and limitations of sepsis screening tools, and the genesis of diagnostic errors in general.

Sepsis disproportionately affects elderly patients, who often present with atypical or vague presentations, and who experience significantly greater mortality. The recent Third Internal Consensus Definitions Task Force defined sepsis as life-threatening organ dysfunction caused by a dysregulated host response to infection.(1) The mortality of sepsis is approximately 20% to 40%, and the incidence and prevalence of sepsis increases with age, such that patients older than 65 years (who account for about 12% of the United States population) account for more than 65% of sepsis cases.(2)

Older patients may exhibit a blunted or absent initial inflammatory response to infection, followed by a subsequent rapid progression into sepsis or septic shock. One study showed nearly half of septic elderly patients had a blunted febrile response.(3) As in this case, the patient remained afebrile even as he developed sepsis and progressed to septic shock. Furthermore, elderly patients with infection can initially present with nonspecific symptoms including fatigue, generalized weakness, falls, anorexia, and delirium.(4) Of course, noninfectious causes of these symptoms are also common, making the diagnosis of an infection and/or sepsis even more difficult in older patients.

Quickly recognizing and initiating effective treatment is critical to improving sepsis outcomes.(5) Tools to facilitate the timely identification of septic patients are important but have limitations. For example, a common electronic health record (EHR) sepsis alert is based upon meeting at least two systemic inflammatory response syndrome (SIRS) criteria (Table 1), yet SIRS criteria misses up to 54% of critically ill patients in the emergency department (6) and 1 out of 8 severely ill patients with sepsis in the intensive care unit.(7) Existing hospital EHR sepsis alerts have been shown to improve care processes, but they have a poor positive predictive value (approximately 40%) and may not improve in-hospital survival.(8) Although the accuracy of these systems has not been studied specifically in the older population, in light of the often atypical presentation of infection and sepsis in older patients, the test characteristics of sepsis alerts may well be even worse in this population.

With the new Sepsis-3 definitions, a new screening tool, qSOFA (quick Sequential Organ Failure Assessment score), is now in the spotlight. When screening infected patients outside the ICU, qSOFA has slightly better predictive value for in-hospital mortality than SIRS.(9) In the case presented above, none of the most widely used sepsis screening tools—whether based on SIRS or qSOFA—would have triggered an alert upon the patient's initial presentation (Table 1). In addition to the potential for false negative results, sepsis screens are also susceptible to false positive results that can lead to inappropriate treatment. Thus, clinicians should understand the appropriate use of sepsis screening when an infection is suspected, while recognizing that other etiologies of end organ dysfunction should be clinically evaluated and appropriately treated. Clearly, additional work is needed to evaluate and improve our ability to identify sepsis particularly in older patients.

Given the imperfection of screening tools, clinicians should consider the results of these tools in light of the overall clinical presentation. In this case, antibiotics were withheld despite the clinician's assessment of potential infection because the patient did not "screen positive" for sepsis. Having an infection and being septic are not synonymous. Sepsis screening tools are not intended to screen for an infection, but rather to screen for the life-threatening organ dysfunction that may result from an infection. In this case, the negative sepsis screen was treated as a negative infection screen, leading to an important diagnostic error and the withholding of antibiotics. Organizations implementing clinical decision support tools such as sepsis screening alerts should present the results of these tools to minimize the risk of diagnostic error due to result misinterpretation. For example, emphasis and education on the role of sepsis screening and highlighting that they are not intended as infection screens.

Defined as a mistake or failure in the process leading to a misdiagnosis, a missed diagnosis, or delay in diagnosis (10), diagnostic errors are difficult to quantify and frequently underappreciated. The National Academy of Medicine (formerly the Institute of Medicine) reports that approximately 1 in 10 diagnoses are incorrect, and a systematic review of autopsies found 9% of patients experienced a major diagnostic error.(11,12) Diagnostic errors are commonly attributable to cognitive biases (e.g., availability bias, anchoring bias, blind obedience) that can adversely influence the diagnostic process (Table 2). In this case, the clinicians placed undue weight on the negative sepsis screen in their decision to withhold antibiotics for this patient with an atypical presentation of pneumonia, an example of the cognitive bias (or heuristic) known as blind obedience. Given the expanding number of available decision support tools, clinicians should be aware of the potential benefits and pitfalls of applying these tools at the bedside.

In summary, early identification of infection and sepsis is critical to providing the proper and timely life-sustaining therapies. Sepsis may present atypically in elderly patients, who have a higher risk of dying. Therefore, clinicians should have a low threshold to suspect infection in older patients with atypical presentations, as they can deteriorate quickly, and to initiate empiric antibiotics. Additional work is needed to better characterize the infection and sepsis phenotypes in elderly patients as well as to improve diagnostic support tools across all relevant populations. Sepsis screening tools are important in improving early sepsis detection, but they can provide a false sense of security that can lead to diagnostic errors. Employing these tools requires an understanding of their specific role and clinical judgment should always be utilized.

Take-Home Points

  • Infection is common in elderly patients, but they may initially exhibit an absent or blunted clinical response, followed by a rapid deterioration.
  • There is a need to develop better tools to identify sepsis in different populations.
  • Sepsis alerts and screening tools are designed to identify patients who have organ dysfunction due to an infection and should not be equated to an "infection screen." Clinicians should be aware that false negative results of sepsis screening tools may provide a false sense of security.
  • Sepsis screens and other decision support tools can be helpful in guiding appropriate decision making. However, overreliance on decision tools at the expense of clinical expertise is dangerous.

Casey A. Cable, MD Pulmonary and Critical Care Fellow Division of Pulmonary, Allergy, Sleep, and Critical Care Medicine Emory University School of Medicine

David J. Murphy, MD, PhD Patient Safety Officer, Emory Healthcare Director of Quality, Emory Critical Care Center Assistant Professor, Division of Pulmonary, Allergy, Sleep, and Critical Care Medicine Emory University School of Medicine

Greg S. Martin, MD, MSc Professor of Medicine Section Chief and Associate Division Director for Critical Care Division of Pulmonary, Allergy, Sleep, and Critical Care Medicine Emory University School of Medicine


1. Singer M, Deutschman CS, Seymour CW, et al. The Third International Consensus Definitions for Sepsis and Septic Shock (Sepsis-3). JAMA. 2016;315:801-810. [go to PubMed]

2. Martin GS, Mannino DM, Moss M. The effect of age on the development and outcome of adult sepsis. Crit Care Med. 2006;34:15-21. [go to PubMed]

3. Castle SC, Norman DC, Yeh M, Miller D, Yoshikawa TT. Fever response in elderly nursing home residents: are the older truly colder? J Am Geriatr Soc. 1991;39:853-857. [go to PubMed]

4. Gavazzi G, Krause KH. Ageing and infection. Lancet Infect Dis. 2002;2:659-666. [go to PubMed]

5. De Backer D, Dorman T. Surviving sepsis guidelines: a continuous move toward better care of patients with sepsis. JAMA. 2017;317:807-808. [go to PubMed]

6. Liao MM, Lezotte D, Lowenstein SR, et al. Sensitivity of systemic inflammatory response syndrome for critical illness among ED patients. Am J Emerg Med. 2014;32:1319-1325. [go to PubMed]

7. Kaukonen KM, Bailey M, Pilcher D, Cooper DJ, Bellomo R. Systemic inflammatory response syndrome criteria in defining severe sepsis. N Engl J Med. 2015;372:1629-1638. [go to PubMed]

8. Makam AN, Nguyen OK, Auerbach AD. Diagnostic accuracy and effectiveness of automated electronic sepsis alert systems: a systematic review. J Hosp Med. 2015;10:396-402. [go to PubMed]

9. Seymour CW, Liu VX, Iwashyna TJ, et al. Assessment of clinical criteria for sepsis: for the Third International Consensus Definitions for Sepsis and Septic Shock (Sepsis-3). JAMA. 2016;315:762-774. [go to PubMed]

10. Schiff GD, Hasan O, Kim S, et al. Diagnostic error in medicine: analysis of 583 physician-reported errors. Arch Intern Med. 2009;169:1881-1887. [go to PubMed]

11. Improving Diagnosis in Health Care. Committee on Diagnostic Error in Health Care, National Academies of Science, Engineering, and Medicine. Washington, DC: National Academies Press; 2015. [go to PubMed]

12. Shojania KG, Burton EC, McDonald KM, Goldman L. Changes in rates of autopsy-detected diagnostic errors over time: a systematic review. JAMA. 2003;289:2849-2856. [go to PubMed]


Table 1. Sepsis Screening Tools.
Sepsis Screening Tool Criteria Evaluated Scale Range (min–max) Positive Screen
SIRS (Systemic Inflammatory Response Syndrome) Temperature >38°C or <36°C Heart rate >90 beats per min Respiratory rate >20 breaths per min or PaCO2 <32 mm Hg WBC <4,000/μL or >12,000/μL or >10% bands 0–4 ≥ 2
qSOFA (quick Sequential Organ Failure Assessment score) Systolic blood pressure <90 mm Hg Respiratory rate $gt;22 breaths per min Altered mental status 0–3 ≥ 2
SOFA (Sequential Organ Failure Assessment score) 6 organ systems:   Cardiovascular (level of hypotension)   Hepatic (bilirubin)   Coagulation (platelet count)   Renal (creatinine, urine output)   Respiratory (PaO2/FiO2 ratio)   Neurologic (GCS) 0–24 ≥ 2
MEWS (Modified Early Warning Score) Systolic blood pressure Heart rate Respiratory rate Temperature AVPU score (alert, voice, pain, unresponsive) 0–15 ≥ 4
LODS (Logistic Organ Dysfunction System score) 6 organ systems:   Neurologic (GCS)   Cardiologic (heart rate, systolic blood pressure)   Renal (BUN, creatinine, urine output)   Pulmonary (PaO2/FiO2 ratio)   Hematologic (WBC, platelet count)   Hepatic (bilirubin, PT) 0–22 ≥ 2

WBC (white blood cell count); GCS (Glasgow Coma Scale), BUN (blood urea nitrogen), PT (prothrombin time)

Table 2. Example Cognitive Biases.
Cognitive Bias Definition Potential Examples
Availability bias Diagnosis of current patient biased by experience with past cases Having recently treated numerous elderly patients with heart failure, incorrectly treating this patient for heart failure and fluid overload even though there were no physical exam findings evident of fluid overload or prior history of heart failure. There is no mention of prior cases influencing decision making for this patient in this case, so this does not appear to be a case of availability bias.
Anchoring bias (premature closure) Relying on initial diagnostic impression, despite subsequent information to the contrary Normal initial laboratory blood work, vital signs, crackles on lung exam, and mention of pulmonary edema on the chest radiograph report prompted clinical focus on new-onset heart failure and not consider infection. In this case, the clinicians did consider infection in their differential diagnosis.
Blind obedience Placing undue reliance on test results or "expert" opinion The ED false negative sepsis screen gives false sense of security that the patient does not have sepsis or an infection. Thus, this is a good example of blind obedience.
This project was funded under contract number 75Q80119C00004 from the Agency for Healthcare Research and Quality (AHRQ), U.S. Department of Health and Human Services. The authors are solely responsible for this report’s contents, findings, and conclusions, which do not necessarily represent the views of AHRQ. Readers should not interpret any statement in this report as an official position of AHRQ or of the U.S. Department of Health and Human Services. None of the authors has any affiliation or financial involvement that conflicts with the material presented in this report. View AHRQ Disclaimers
Related Resources From the Same Author(s)
Related Resources