Prospective, multi-site study of patient outcomes after implementation of the TREWS machine learning-based early warning system for sepsis.

Early recognition and treatment of sepsis are linked to improved patient outcomes. Machine learning-based early warning systems may reduce the time to recognition, but few systems have undergone clinical evaluation. In this prospective, multi-site cohort study, we examined the association between patient outcomes and provider interaction with a deployed sepsis alert system called the Targeted Real-time Early Warning System (TREWS). During the study, 590,736 patients were monitored by TREWS across five hospitals. We focused our analysis on 6,877 patients with sepsis who were identified by the alert before initiation of antibiotic therapy. Adjusting for patient presentation and severity, patients in this group whose alert was confirmed by a provider within 3 h of the alert had a reduced in-hospital mortality rate (3.3%, confidence interval (CI) 1.7, 5.1%, adjusted absolute reduction, and 18.7%, CI 9.4, 27.0%, adjusted relative reduction), organ failure and length of stay compared with patients whose alert was not confirmed by a provider within 3 h. Improvements in mortality rate (4.5%, CI 0.8, 8.3%, adjusted absolute reduction) and organ failure were larger among those patients who were additionally flagged as high risk. Our findings indicate that early warning systems have the potential to identify sepsis patients early and improve patient outcomes and that sepsis patients who would benefit the most from early treatment can be identified and prioritized at the time of the alert.

Factors driving provider adoption of the TREWS machine learning-based early warning system and its effects on sepsis treatment timing.

Machine learning-based clinical decision support tools for sepsis create opportunities to identify at-risk patients and initiate treatments at early time points, which is critical for improving sepsis outcomes. In view of the increasing use of such systems, better understanding of how they are adopted and used by healthcare providers is needed. Here, we analyzed provider interactions with a sepsis early detection tool (Targeted Real-time Early Warning System), which was deployed at five hospitals over a 2-year period. Among 9,805 retrospectively identified sepsis cases, the early detection tool achieved high sensitivity (82% of sepsis cases were identified) and a high rate of adoption: 89% of all alerts by the system were evaluated by a physician or advanced practice provider and 38% of evaluated alerts were confirmed by a provider. Adjusting for patient presentation and severity, patients with sepsis whose alert was confirmed by a provider within 3 h had a 1.85-h (95% CI 1.66-2.00) reduction in median time to first antibiotic order compared to patients with sepsis whose alert was either dismissed, confirmed more than 3 h after the alert or never addressed in the system. Finally, we found that emergency department providers and providers who had previous interactions with an alert were more likely to interact with alerts, as well as to confirm alerts on retrospectively identified patients with sepsis. Beyond efforts to improve the performance of early warning systems, efforts to improve adoption are essential to their clinical impact and should focus on understanding providers' knowledge of, experience with and attitudes toward such systems.

A randomized controlled trial of online symptom searching to inform patient generated differential diagnoses.

Patient online health searching is now commonplace, however, the accuracy of patient generated differentials for new symptoms and potential for patient anxiety are concerns. We aimed primarily to determine the accuracy of patient generated differentials for new symptoms with and without online searching, and secondarily, to evaluate the impact of searching on anxiety levels. In the waiting room prior to seeing a clinician, 300 patients with new symptoms were randomly assigned 1:1:1 to Google searching with health related features including a symptom search tool vs Google searching with health related features disabled vs no searching. Participants were 18 years or older and presenting to the emergency department of an urban academic medical center with new low-acuity symptoms that were not due to exacerbation of a chronic condition. Search groups received access on a tablet/smartphone to Google searching with or without health related features. Both search groups could access any websites; health related features led the patient to common diagnoses and physician-validated information. The primary outcome was accuracy of the patient generated differential assessed by matching at least two of the top three diagnoses on the clinician's differential. A secondary outcome was anxiety by a visual analogue scale. Patients were a median of 33.1 (IQI 26.2-45.9) years old, 60% women, 63% black, 82% had a high school education or less, and 45.7% reported having performed an online search prior to presentation. Search group patients spent a median of 3.82 (2.53-5.72) minutes searching online. Similar proportions of patients in each group matched at least two of three clinician diagnoses: 27.0% and 28.3% for Google searching with and without health related features vs 23.8% in the no search group. Patients in the search groups had a similar odds of matching ≥2/3 diagnoses as the no search group [OR (95% CI): 1.23 (0.70-2.13), p = 0.47]. Anxiety was unchanged with online searching. In conclusion, brief online searching in the waiting room did not improve accuracy of patient generated differential diagnoses for new symptoms. The absence of an increase in patient anxiety provides reassurance for subsequent work to refine and investigate online symptom search tools.

Correction to: Do Words Matter? Stigmatizing Language and the Transmission of Bias in the Medical Record.

In the HTML version of this paper, Ms. Goddu was incorrectly listed. The correct citation should be AP Goddu. Exposure to the stigmatizing language note was associated with more negative attitudes towards the patient (20.3 stigmatizing vs. 25.1 neutral, p < 0.001). Reading the stigmatizing language note was associated with less aggressive management of the patient's pain (4.7 stigmatizing vs. 5.3 neutral, p < 0.001).

The Effects of a Systemwide Diagnostic Stewardship Change on West Nile Virus Disease Ordering Practices.

We report that removing the clinically insensitive West Nile virus CSF nucleic acid amplification test (NAAT) from the electronic health record (EHR) test. This diagnostic stewardship intervention decreased costs and may have improved diagnostic yield.

Do Words Matter? Stigmatizing Language and the Transmission of Bias in the Medical Record.

BACKGROUND: Clinician bias contributes to healthcare disparities, and the language used to describe a patient may reflect that bias. Although medical records are an integral method of communicating about patients, no studies have evaluated patient records as a means of transmitting bias from one clinician to another. OBJECTIVE: To assess whether stigmatizing language written in a patient medical record is associated with a subsequent physician-in-training's attitudes towards the patient and clinical decision-making. DESIGN: Randomized vignette study of two chart notes employing stigmatizing versus neutral language to describe the same hypothetical patient, a 28-year-old man with sickle cell disease. PARTICIPANTS: A total of 413 physicians-in-training: medical students and residents in internal and emergency medicine programs at an urban academic medical center (54% response rate). MAIN MEASURES: Attitudes towards the hypothetical patient using the previously validated Positive Attitudes towards Sickle Cell Patients Scale (range 7-35) and pain management decisions (residents only) using two multiple-choice questions (composite range 2-7 representing intensity of pain treatment). KEY RESULTS: Exposure to the stigmatizing language note was associated with more negative attitudes towards the patient (20.6 stigmatizing vs. 25.6 neutral, p < 0.001). Furthermore, reading the stigmatizing language note was associated with less aggressive management of the patient's pain (5.56 stigmatizing vs. 6.22 neutral, p = 0.003). CONCLUSIONS: Stigmatizing language used in medical records to describe patients can influence subsequent physicians-in-training in terms of their attitudes towards the patient and their medication prescribing behavior. This is an important and overlooked pathway by which bias can be propagated from one clinician to another. Attention to the language used in medical records may help to promote patient-centered care and to reduce healthcare disparities for stigmatized populations.

Low high-sensitivity troponin I and zero coronary artery calcium score identifies coronary CT angiography candidates in whom further testing could be avoided.

RATIONALE AND OBJECTIVES: Pilot study to determine whether among subjects receiving coronary computed tomography angiography (CTA), the combination of high-sensitivity troponin I (hsTnI) and coronary artery calcium score (CACS) identifies a low-risk population in whom CTA might be avoided. MATERIALS AND METHODS: A cross-sectional study of 314 symptomatic patients receiving CTA as part of their acute coronary syndrome evaluation was conducted. hsTnI was measured with Abbott Laboratories' hsTnI assay. CACSs were calculated via the Agatston method. Patients were followed for at least 30 days after discharge for the occurrence of major adverse cardiac events (MACEs; all-cause mortality, acute coronary syndrome, and revascularization). RESULTS: Of 314 subjects studied, 213 (67.8%) had no coronary artery stenosis, and 67 (21.3%), 28 (8.9%), and 6 (1.9%) had maximal coronary artery stenosis of 1%-49%, 50%-69%, and 70% or greater, respectively. All MACEs occurred during index hospitalization and include one myocardial infarction and four revascularizations. Sixty-two percent (189/307) of subjects had zero CACS, and 24% (76/314) of subjects had undetected hsTnI. No subjects with undetectable hsTnI or zero CACS had an MACE. A strategy of avoiding further testing in subjects with undetectable initial hsTnI, performing CACS on subjects with detectable initial hsTnI but nonincreased hsTnI (less than 99th percentile), and obtaining CTA in subjects with Agatston greater than 0 will have a negative predictive value of 100.0% (95% confidence interval, 98.2%-100.0%). This strategy will avoid CTA in 63% (198/314) of subjects. CONCLUSIONS: In this pilot study, the addition of CACS to hsTnI improves the identification of low-risk subjects in whom CTA might be avoided.

Troponin elevations only detected with a high-sensitivity assay: clinical correlations and prognostic significance.

OBJECTIVES: With clinical use of high-sensitivity troponin I (hsTnI), more frequent troponin elevations will occur. However, the burden and implications of these elevations are not well understood. The authors quantified the prevalence of elevated hsTnI in patients presenting with possible acute coronary syndrome (ACS) who do not have elevated troponin with a current generation assay (cardiac troponin I [cTnI]) and determined the association of these newly detected elevations with a composite of all-cause mortality and subsequent cardiac hospitalization. METHODS: This was a prospective observational study of 808 subjects evaluated for possible ACS and followed for up to 1 year. Troponin values were measured with hsTnI (Abbott Laboratories) and cTnI (Abbott and Beckman Coulter). Cardiac hospitalization was defined as hospitalization for ACS, revascularization, acute heart failure (AHF), or tachy/brady arrhythmia that occurred after the index emergency department (ED) visit or hospital discharge. RESULTS: Forty subjects (5%) were diagnosed with ACS (26 myocardial infarction and 14 unstable angina). On the initial sample, the prevalence of elevated hsTnI among subjects with nonelevated cTnI was 9.2% using a gender-neutral cutoff (95% confidence interval [CI] = 7.1% to 11.4%) and 11.1% using a gender-specific cutoff (95% CI = 8.8% to 13.4%). Adjudicated diagnoses for subjects whose initial samples had elevated hsTnI but nonelevated cTnI (gender-neutral cutoff) were as follows: three (4.6%) ACS, 15 (23.1%) AHF, three (4.6%) volume overload etiology unclear/noncardiac, three (4.6%) cardiac (non-ACS), and 41 (63.1%) other. Of the 65 patients whose initial samples had hsTnI but nonelevated cTnI, eight developed cTnI elevation on subsequent serial sampling. After traditional cardiovascular risk factors and renal function were adjusted for, subjects with elevated initial hsTnI but nonelevated cTnI (initial and serial sampling) had a higher risk of all-cause mortality and subsequent cardiac hospitalization than subjects with both nonelevated hsTnI and nonelevated cTnI (hazard ratio [HR] = 1.91, 95% CI = 1.14 to 3.19). CONCLUSIONS: On the initial sample, 9% to 11% of subjects without cTnI elevation had hsTnI elevation. Although the majority of the patients with these newly detected hsTnI elevations did not have ACS, they had a higher risk for all-cause mortality and subsequent cardiac hospitalization.