Applying natural language processing to identify emergency department and observation encounters for worsening heart failure

AIMS

Worsening heart failure (WHF) events occurring in non-inpatient settings are becoming increasingly recognized, with implications for prognostication. We evaluate the performance of a natural language processing (NLP)-based approach compared with traditional diagnostic coding for non-inpatient clinical encounters and left ventricular ejection fraction (LVEF).

METHODS AND RESULTS

We compared characteristics for encounters that did vs. did not meet WHF criteria, stratified by care setting [i.e. emergency department (ED) and observation stay]. Overall, 8407 (22%) encounters met NLP-based criteria for WHF (3909 ED visits and 4498 observation stays). The use of an NLP-derived definition adjudicated 3983 (12%) of non-primary HF diagnoses as meeting consensus definitions for WHF. The most common diagnosis indicated in these encounters was dyspnoea. Results were primarily driven by observation stays, in which 2205 (23%) encounters with a secondary HF diagnosis met the WHF definition by NLP.

CONCLUSIONS

The use of standard claims-based adjudication for primary diagnosis in the non-inpatient setting may lead to misclassification of WHF events in the ED and overestimate observation stays. Primary diagnoses alone may underestimate the burden of WHF in non-hospitalized settings.

The Brain Injury Guidelines (BIG) and emergency department observation and admission rates: A retrospective cohort study

BACKGROUND

Emergency Department (ED) Observation Units (OU) can provide safe, effective care for low risk patients with intracranial hemorrhages. We compared current ED OU use for patients with subdural hematomas (SDH) to the validated Brain Injury Guidelines (BIG) to evaluate the potential impact of implementing this risk stratification tool.

METHODS

Retrospective cohort of patients ≥18 years old with SDH of any cause from 2014 to 2020 to evaluate for potential missed OU cases. Missed OU cases were defined as patients with an initial Glasgow Coma Score (GCS) of 15 with hospital length of stays (LOS) <2 days, who did not meet the composite outcome and were not cared for in the OU or discharged from the ED. Composite outcome included in-hospital death or transition to hospice care, neurosurgical intervention, GCS decline, and worsening SDH size. Secondary outcomes were whether application of BIG would increase ED OU use or reduce CT use.

RESULTS

264 patients met inclusion criteria over 5.3 year study timeframe. Mean age was 61 years (range 19-93) and 61.4% were male. SDH were traumatic in 76.9% and 60.2% of the cohort had additional injuries. The admission rate was 81.4% (n = 215). Fourteen (6.5%) missed OU cases were identified (2.6/year). Retrospective application of BIG resulted in 82.6% (n = 217) at BIG 3, 10.2% (n = 27) at BIG 2 and 7.6% (n = 20) at BIG 1. Application of BIG would not have decreased admission rates (82.6% BIG 3) and BIG 1 and 2 admissions were often for medical co-morbidities. The composite outcome was met in 50% of BIG 3, 22% of BIG 2, and no BIG 1 patients.

CONCLUSION

In a level 1 trauma center with an established observation unit, current clinical care processes missed very few patients who could be discharged or placed in ED OU for SDH. Hospital admissions in BIG 1/2 were driven by co-morbidities and/or injuries, limiting applicability of BIG to this population.

The feasibility of emergency department observation units in the management of mild to moderate hyponatremia

OBJECTIVE

To describe the feasibility of managing hyponatremia patients under outpatient observation status in an academic medical center, and compare outcomes based on the use of an emergency department observation unit (EDOU).

METHODS

This is a retrospective cohort study of emergency department hyponatremic patients managed in four hospitals within a large urban academic medical center over 27 months. All patients had an admit-to-observation order, ICD-10 codes for hyponatremia, and mild (130-135 mmol/L) to moderate (121-129 mmol/L) hyponatremia. Observation settings were divided into two groups: EDOU and Non-Observation Unit (NOU) inpatient beds. Severe hyponatremia (≤120 mmol/L) was excluded. Primary clinical outcomes were inpatient admit rate, length of stay (LOS), total direct cost, the rate of adverse events and 30-day recidivism.

RESULTS

188 patients were managed as an observation patient, with 64 managed in an EDOU setting (age 74.0 yr, 70.3% female) and 124 managed in a NOU setting (age 71.5 yr, 64.5% female). Patient subgroups were similar in terms of presenting complaints, comorbidities, and medication histories. Initial and final sodium levels were similar between settings: EDOU (125.1 to 132.6 mmol/L) vs NOU (123.5 to 132.0 mmol/L). However, outcomes differed by setting for observation to inpatient admit rate (EDOU 28.1% vs NOU 37.9%, adjusted effect 0.70), overall length of stay (EDOU 19.2 h vs NOU 31.9 h; adjusted effect -10.5 h and total direct cost ($1230 vs $1531; adjusted effect -$167). EDOU sodium correction rates were faster (EDOU 0.44 mmol/L/h vs 0.24 mmol/L/h; adjusted effect 0.15 mmol/L/h) and 30-day recidivism rate was similar (EDOU 13% vs NOU 15%). There were no index visit deaths or intensive care unit admissions.

CONCLUSION

Management of selected hyponatremia patients under observation status is feasible, with the EDOU setting demonstrating lower admit rates, shorter length of stay, and lower total direct costs with similar clinical outcomes.

Implementing a multidisciplinary rapid geriatric observation unit for non-critical older patients referred to hospital: observational study on real-world data

BACKGROUND

Geriatric rapid observation units may represent an alternative to hospitalization in older patients with non-critical acute illness.

AIMS

To describe the characteristics and outcomes of patients admitted to a geriatric observation unit called URGe (Unità Geriatrica Rapida), implemented in an Italian hospital and characterized by multidisciplinary medical staff with geriatric expertise, fast-track access to diagnostic resources, regular use of point-of-care ultrasound and predicted length of stay (LOS) < 72 h.

METHODS

The medical records of patients admitted to URGe during a 3-month period (452 subjects, 247 F and 205 M, median age 82 years, IQR 77-87) were retrospectively examined. The primary study endpoint was transferral from URGe to regular wards. Baseline covariates included demographics, comprehensive geriatric assessment, acute illnesses, comorbidities, vital signs and routine laboratory tests.

RESULTS

Despite elevated burden of multimorbidity (median number of chronic diseases 4, IQR 2-5) and frailty (median Rockwood Clinical Frailty Scale score 4, IQR 3-6), only 137 patients (30.3%) required transferral from URGe to regular wards. The main factors positively associated with this outcome were Rockwood score, fever, cancer and red cell distribution width (P < 0.05 on multivariate logistic regression model). The rate of complications (mortality, delirium, and falls) during URGe stay was low (0.5%, 7% and 2%, respectively). Overall duration of hospital stay was lower than that of a group of historical controls matched by age, sex, main diagnosis, multimorbidity and frailty.

CONCLUSIONS

The URGe model of acute geriatric care is feasible, safe and has the potential of reducing unnecessary hospitalizations of older patients.

Strategies to measure and improve emergency department performance: a scoping review

Background

Over the last two decades, Emergency Department (ED) crowding has become an increasingly common occurrence worldwide. Crowding is a complex and challenging issue that affects EDs’ capacity to provide safe, timely and quality care. This review aims to map the research evidence provided by reviews to improve ED performance.

Methods and findings

We performed a scoping review, searching Cochrane Database of Systematic Reviews, Scopus, EMBASE, CINAHL and PubMed (from inception to July 9, 2019; prospectively registered in Open Science Framework https://osf.io/gkq4t/). Eligibility criteria were: (1) review of primary research studies, published in English; (2) discusses a) how performance is measured in the ED, b) interventions used to improve ED performance and their characteristics, c) the role(s) of patients in improving ED performance, and d) the outcomes attributed to interventions used to improve ED performance; (3) focuses on a hospital ED context in any country or healthcare system. Pairs of reviewers independently screened studies’ titles, abstracts, and full-texts for inclusion according to pre-established criteria. Discrepancies were resolved via discussion. Independent reviewers extracted data using a tool specifically designed for the review. Pairs of independent reviewers explored the quality of included reviews using the Risk of Bias in Systematic Reviews tool. Narrative synthesis was performed on the 77 included reviews. Three reviews identified 202 individual indicators of ED performance. Seventy-four reviews reported 38 different interventions to improve ED performance: 27 interventions describing changes to practice and process (e.g., triage, care transitions, technology), and a further nine interventions describing changes to team composition (e.g., advanced nursing roles, scribes, pharmacy). Two reviews reported on two interventions addressing the role of patients in ED performance, supporting patients’ decisions and providing education. The outcomes attributed to interventions used to improve ED performance were categorised into five key domains: time, proportion, process, cost, and clinical outcomes. Few interventions reported outcomes across all five outcome domains.

Conclusions

ED performance measurement is complex, involving automated information technology mechanisms and manual data collection, reflecting the multifaceted nature of ED care. Interventions to improve ED performance address a broad range of ED processes and disciplines.

A National Dataset Analysis of older adults in emergency department observation units

BACKGROUND

Emergency Department (ED) Observation Units (Obs Units) are prevalent in the US, but little is known regarding older adults in observation. Our objective was to describe the Obs Units nationally and observation patients with specific attention to differences in care with increasing age.

DESIGN

This is an analysis of 2010-2013 data from the National Hospital Ambulatory Medical Care Survey (NHAMCS), a national observational cohort study including ED patients. Weighted means are presented for continuous data and weighted percent for categorical data. Multivariable logistic regression was used to identify variables associated with placement in and admission from observation.

RESULTS

The number of adult ED visits varied from 100 million to 107 million per year and 2.3% of patients were placed in observation. Adults ≥65 years old made up a disproportionate number of Obs Unit patients, 30.6%, compared to only 19.7% of total ED visits (odds ratio 1.5 (95% CI 1.5-1.6), adjusting for sex, race, month, day of week, payer source, and hospital region). The overall admission rate from observation was 35.6%, ranging from 31.3% for ages 18-64 years to 47.5% for adults ≥85 years old (p < 0.001). General symptoms (e.g., nausea, dizziness) and hypertensive disease were the most common diagnoses overall. Older adults varied from younger adults in that they were frequently observed for diseases of the urinary system (ICD-9 590-599) and metabolic disorders (ICD-9 270-279).

CONCLUSIONS

Older adults are more likely to be cared for in Obs Units. Older adults are treated for different medical conditions than younger adults.

Role of emergency department observation units in the management of patients with unexplained syncope: a critical review and meta-analysis

This meta-analysis aimed to establish the role of standardized emergency department (ED) observation protocols in the management of syncopal patients as an alternative to ordinary admission. A systematic electronic literature search was performed to identify randomized controlled trials or observational studies evaluating syncopal patients managed in ED observation units. Data regarding mean length of stay, rate of etiological diagnosis, admission rate, and incidence of short-term serious outcomes were extracted. Six mostly single-center, small sized studies characterized by high heterogeneity, were included. A total of 458 patients were included with a balanced sex distribution (male 50.2%), a mean age of 60.1 years, and a considerable prevalence of heart disease (32.4%). Pooled analysis of the outcomes showed a mean stay of 28.2 hours, an etiological diagnosis rate of 67.3%, an admission rate of 18.5%, and a very low incidence of short-term serious outcomes (2.8%). Due to elevated diagnostic yield and low incidence of short-term adverse events, ED observation units-based management strategy seems ideal for patients with syncope. Nevertheless, further research is needed to identify criteria for selecting patients to be managed with this approach, define evaluation protocols, and confirm the safety of this strategy.

Utilization of Observation Units for the Care of Poisoned Patients: Trends from the Toxicology Investigators Consortium Case Registry

Many poisoned patients may only require a period of observation after their exposure. There are limited data describing the use of observation units for managing poisoned adult and pediatric patients. We performed a retrospective review of all patients reported to the ToxIC Case Registry between January 1, 2012 and December 31, 2013. Eligible patients included those who received a bedside consultation by a medical toxicologist and whose care was provided in an observation unit, or those who were admitted under the care of a medical toxicologist in an observation unit. A total of 15,562 poisonings were reported to the registry during the study period, of which 340 (2.2 %) involved patients who were cared for in an observation unit. Of these patients, 22.1 % were 18 years of age or younger, and the remaining 77.9 % were greater than 18 years of age. The most common reason for exposure was the intentional ingestion of a pharmaceutical agent in both adult (30.2 %) and pediatric patients (36.0 %). Alcohols (ethanol) (24.9 %), opioids (20.0 %), and sedative-hypnotics (17.7 %) were the most common agent classes involved in adult patient exposures. The most common agent classes involved in pediatric exposures were antidepressants (12.0 %), anticonvulsants (10.7 %), and envenomations (10.7 %). In adult patients, the most common signs and symptoms involved the nervous system (52.0 %), a toxidrome (17.0 %), or a major vital sign abnormality (14.7 %). In pediatric patients, the most common signs and symptoms involved the nervous system (53.3 %), a toxidrome (21.3 %), or a major vital sign abnormality (17.3 %). The results of this study demonstrate that a wide variety of poisoned patients have been cared for in an observation unit in consultation with a board-certified medical toxicologist. Patterns for the reasons for exposure, agents responsible for the exposure, and toxicological treatments will continue to evolve. Further study is needed to identify better those poisoned patients who can be appropriately managed in an observation unit.