Anthony M. Napoli

BACKGROUND Accurate emergency department (ED) volume forecasts support staffing, capacity planning, and budgeting, but conventional time-series forecasting can require specialized analytic expertise. Consumer-facing large language model (LLM) interfaces may lower this barrier, although their forecasting accuracy, consistency, and sensitivity to data recency are uncertain. OBJECTIVE To compare forecasts produced by commonly available generative AI interfaces with a Holt-Winters exponential smoothing baseline and to determine whether progressively updating the historical data window (“proximal forecasting”) identifies a pragmatic recency point for fiscal-year planning. METHODS We conducted a two-part retrospective observational forecasting study using monthly ED visit volumes from four hospitals. In part 1, 12-month forecasts were generated from 12-, 24-, and 36-month historical windows using Holt-Winters and five LLM interfaces (ChatGPT, Claude, Copilot, Perplexity, and Gemini); LLM forecasts were repeated 10 times to characterize run-to-run variability. In part 2, ChatGPT and Gemini were evaluated using 24-month windows whose end dates were systematically reduced from 7 months before to immediately before FY2025 and were compared with an R-implemented Holt-Winters baseline. Accuracy was assessed using mean absolute percentage error (MAPE) and root mean square error (RMSE). RESULTS Across all four sites, Holt-Winters had the lowest aggregate MAPE with a 24-month historical window (3.26±0.68) compared with 12 months (4.29±0.46) and 36 months (3.43±0.73). With 24 months of history, two LLM interfaces (ChatGPT 5.5 and Gemini 3.1) produced MAPE values similar to Holt-Winters at ED A (2.17±1.22 and 2.10±1.04 vs 2.56±1.08) and ED B (2.35±1.16 and 2.21±1.21 vs 2.46±1.39), but had higher error at the pediatric ED (7.35±5.26 and 7.29±5.27 vs 5.90±2.05) and the summer-seasonal community ED (8.74±3.11 and 8.69±3.11 vs 2.12±1.35). No approach met a predefined threshold in the Ped ED and community ED. In adult EDs with typical seasonality, forecasts generated up to three months from the fiscal-year start were comparable to fiscal-year–proximal forecasts (mean MAPE difference -0.4%–1.8%). Pediatric and highly seasonal sites demonstrated higher baseline error and recency bias susceptibility. CONCLUSIONS Selected LLM interfaces can produce monthly ED volume forecasts comparable to Holt-Winters in some adult ED settings when given sufficient historical context. A forecast update up to 3 months before the fiscal year may be a pragmatic planning point for sites with typical seasonality. Pediatric and strongly seasonal EDs may require local validation and tailored models.

BACKGROUND Forecasting emergency department (ED) arrivals is foundational to operational decision-making, including clinician scheduling, bed management, throughput initiatives, and revenue forecasting. The historical baseline for many health systems has been expert consensus combined with prior-year volume—a process that is often opaque, variably calibrated, and limited in its ability to accommodate seasonality and structural disruption. OBJECTIVE To compare the accuracy of forecasts produced by commonly available generative AI interfaces with a Holt–Winters (HW) exponential smoothing baseline, and to determine whether interactive updating (“proximal forecasting”) identifies a pragmatic recency window for fiscal-year planning. METHODS We conducted a two-part observational forecasting study using monthly ED visit volumes from four hospitals. 12-month forecasts were generated using 12- and 24-month historical windows and compared across HW and five large language model (LLM) interfaces (ChatGPT, Claude, Copilot, Perplexity, Gemini) using standardized prompts and adjusting for seasonality. We then generated FY2025 forecasts using a fixed training window and systemically varying the end month of available data to simulate a budget-cycle lag. Forecast accuracy was measured using mean absolute percentage error (MAPE) and root mean square error (RMSE). RESULTS Forecasts produced by one LLM interface (ChatGPT 4.0) demonstrated accuracy comparable to HW (MAPE 3.63±1.64 vs 3.26±0.92), whereas other interfaces showed higher error and greater variability (Perplexity 5.39±1.88; Copilot 5.15±1.77; Claude 8.14±3.19; Gemini 4.27±1.31). No approach met a predefined threshold in the pediatric ED. In adult EDs with typical seasonality, forecasts generated up to four months from the fiscal-year start were comparable to fiscal-year–proximal forecasts (mean MAPE difference 0–0.4%). Pediatric and highly seasonal sites demonstrated higher baseline error and recency bias susceptibility. CONCLUSIONS With adequate historical context, an LLM interface can produce monthly ED volume forecasts with accuracy similar to Holt-Winters exponential smoothing and enable “proximal forecasting” workflows for fiscal-year planning. Strongly seasonal and pediatric settings may require tailored approaches.

Background: Health care organizations have started to implement artificial intelligence–powered ambient scribe technology in clinical documentation workflows. Early outpatient studies have shown mixed results. Few studies have evaluated ambient scribes in the emergency department (ED). Due to differences in setting and patient acuity between the ED and ambulatory clinics, there remains a pressing need to research this technology in the ED. Objective: This study aimed to evaluate emergency physicians’ (EPs) satisfaction, perceived documentation efficiency, after-shift documentation time, and trust in ambient scribe–generated notes compared with in-person scribes or independent documentation, and to identify ED-specific challenges. Methods: A cross-sectional survey was conducted among 16 board-certified adult and pediatric EPs who were granted access to the ambient scribe technology across 4 EDs. EPs used the ambient scribe for several months before completing a survey with multiple-choice and free-text responses. We performed a mixed methods analysis by summarizing quantitative data through descriptive statistics and performing a practical thematic analysis on free-text responses. Results: Of the 16 EPs, 14 (87.5%) completed the survey. Among respondents, 9 (64.3%; 95% CI 38.7%-83.7%) reported being satisfied or very satisfied with the ambient scribe, while 3 (21.4%; 95% CI 7.6%-47.6%) expressed dissatisfaction. When given the option, 7 (50%; 95% CI 26.8%-73.2%) respondents preferred the ambient scribe, 4 (28.5%; 95% CI 11.7%-54.6%) preferred in-person scribes, and none (95% CI 0%-20.6%) preferred independent documentation. Among previous users of in-person scribes, 50% (4/8; 95% CI 21.5%-78.5%) favored the ambient scribe. The ambient scribe was reported to improve documentation efficiency (10/14, 71.4%; 95% CI 45.4%-88.3%) and reduce after-shift documentation time (9/14, 64.3%; 95% CI 38.7%-83.7%). However, only 42.9% (6/14; 95% CI 22.3%-69.2%) of respondents trusted the accuracy of ambient scribe–generated notes, compared with 75% (6/8; 95% CI 40.9%-92.9%) who trusted in-person scribes. Few respondents found the ambient scribe helpful for physical examinations (3/13, 23.1%; 95% CI 8.2%-50.3%) or medical decision-making documentation (5/14, 35.7%; 95% CI 16.3%-61.2%). A thematic analysis identified 5 themes: challenges due to the workplace environment, challenges due to the patient population, workflow improvement, workflow harm, and narrow usefulness. Conclusions: This mixed methods pilot study is among the first to evaluate ambient scribe technology in the ED. Our results add ED-specific insights to literature focused on the outpatient setting. Our findings reveal the potential for enhancing documentation efficiency and reducing administrative burden while highlighting setting-specific challenges. While most EPs preferred artificial intelligence–assisted documentation over independent charting, confidence in documentation accuracy and functionality remains limited compared with human scribes and varies by note component. As the demand for efficiency in emergency medicine continues to grow, scalable solutions such as ambient scribes could play a pivotal role if functionality, reliability, and physician trust can be further optimized.

BACKGROUND Health care organizations have started to implement artificial intelligence–powered ambient scribe technology in clinical documentation workflows. Early outpatient studies have shown mixed results. Few studies have evaluated ambient scribes in the emergency department (ED). Due to differences in setting and patient acuity between the ED and ambulatory clinics, there remains a pressing need to research this technology in the ED. OBJECTIVE This study aimed to evaluate emergency physicians’ (EPs) satisfaction, perceived documentation efficiency, after-shift documentation time, and trust in ambient scribe–generated notes compared with in-person scribes or independent documentation, and to identify ED-specific challenges. METHODS A cross-sectional survey was conducted among 16 board-certified adult and pediatric EPs who were granted access to the ambient scribe technology across 4 EDs. EPs used the ambient scribe for several months before completing a survey with multiple-choice and free-text responses. We performed a mixed methods analysis by summarizing quantitative data through descriptive statistics and performing a practical thematic analysis on free-text responses. RESULTS Of the 16 EPs, 14 (87.5%) completed the survey. Among respondents, 9 (64.3%; 95% CI 38.7%-83.7%) reported being satisfied or very satisfied with the ambient scribe, while 3 (21.4%; 95% CI 7.6%-47.6%) expressed dissatisfaction. When given the option, 7 (50%; 95% CI 26.8%-73.2%) respondents preferred the ambient scribe, 4 (28.5%; 95% CI 11.7%-54.6%) preferred in-person scribes, and none (95% CI 0%-20.6%) preferred independent documentation. Among previous users of in-person scribes, 50% (4/8; 95% CI 21.5%-78.5%) favored the ambient scribe. The ambient scribe was reported to improve documentation efficiency (10/14, 71.4%; 95% CI 45.4%-88.3%) and reduce after-shift documentation time (9/14, 64.3%; 95% CI 38.7%-83.7%). However, only 42.9% (6/14; 95% CI 22.3%-69.2%) of respondents trusted the accuracy of ambient scribe–generated notes, compared with 75% (6/8; 95% CI 40.9%-92.9%) who trusted in-person scribes. Few respondents found the ambient scribe helpful for physical examinations (3/13, 23.1%; 95% CI 8.2%-50.3%) or medical decision-making documentation (5/14, 35.7%; 95% CI 16.3%-61.2%). A thematic analysis identified 5 themes: challenges due to the workplace environment, challenges due to the patient population, workflow improvement, workflow harm, and narrow usefulness. CONCLUSIONS This mixed methods pilot study is among the first to evaluate ambient scribe technology in the ED. Our results add ED-specific insights to literature focused on the outpatient setting. Our findings reveal the potential for enhancing documentation efficiency and reducing administrative burden while highlighting setting-specific challenges. While most EPs preferred artificial intelligence–assisted documentation over independent charting, confidence in documentation accuracy and functionality remains limited compared with human scribes and varies by note component. As the demand for efficiency in emergency medicine continues to grow, scalable solutions such as ambient scribes could play a pivotal role if functionality, reliability, and physician trust can be further optimized.

SUMMARY: Goal: Boarding emergency department (ED) patients is associated with reductions in quality of care, patient safety and experience, and ED operational efficiency. However, ED boarding is ultimately reflective of inefficiencies in hospital capacity management. The ability of a hospital to accommodate variability in patient flow presumably affects its financial performance, but this relationship is not well studied. We investigated the relationship between ED boarding and hospital financial performance measures. Our objective was to see if there was an association between key financial measures of business performance and limitations in patient progression efficiency, as evidenced by ED boarding. Methods: Cross-sectional ED operational data were collected from the Emergency Department Benchmarking Alliance, a voluntarily self-reporting operational database that includes 54% of EDs in the United States. Freestanding EDs, pediatric EDs and EDs with missing boarding data were excluded. The key operational outcome variable was boarding time. We reviewed the financial information of these nonprofit institutions by accessing their Internal Revenue Service Form 990. We examined standard measures of financial performance, including return on equity, total margin, total asset turnover, and equity multiplier (EM). We studied these associations using quantile regressions of added ED volume, ED admission percentage, urban versus nonurban ED site location, trauma status, and percentage of the population receiving Medicare and Medicaid as covariates in the regression models. Principal Findings: Operational data were available for 892 EDs from 31 states. Of those, 127 reported a Form 990 in the year corresponding to the ED boarding measures. Median boarding time across EDs was 148 min (interquartile range [IQR]: 100–216). A significant relationship exists between boarding and the EM, along with a negative association with the hospital’s total profit margin in the highest-performing hospitals (by profit margin percentage). After adjusting for the covariates in the regression model, we found that for every 10 min above 90 min of boarding, the mean EM for the top quartile increased from 245.8% to 249.5% (p < .001). In hospitals in the top 90th percentile of total margin, every 10 min beyond the median ED boarding interval led to a decrease in total margin of 0.24%. Practical Applications: Using the largest available national registry of ED operational data and concordant nonprofit financial reports, higher boarding among the highest-profitability hospitals (i.e., top 10%) is associated with a drag on profit margin, while hospitals with the highest boarding are associated with the highest leverage (i.e., indicated by the EM). These relationships suggest an association between a key ED indicator of hospital capacity management and overall institutional financial performance.

Introduction: Big data and improved analytic techniques, such as triple exponential smoothing (TES), allow for prediction of emergency department (ED) volume. We sought to determine 1) which method of TES was most accurate in predicting pre-coronavirus 2019 (COVID-19), during COVID-19, and post-COVID-19 ED volume; 2) how the pandemic would affect TES prediction accuracy; and 3) whether TES would regain its pre-COVID-19 accuracy in the early post-pandemic period.Methods: We studied monthly volumes of four EDs with a combined annual census of approximately 250,000 visits in the two years prior to, during the 25-month COVID-19 pandemic, and the 14 months following. We compared the accuracy of four models of TES forecasting by measuring the mean absolute percentage error (MAPE), mean square errors (MSE) and mean absolute deviation (MAD), comparing actual to predicted monthly volume.Results: In the 23 months prior to COVID-19, the overall average MAPE across four forecasting methods was 3.88% ± 1.88% (range 2.41–6.42% across the four ED sites), rising to 15.21% ± 6.67%during the 25-month COVID-19 period (range 9.97–25.18% across the four sites), and falling to 6.45% ± 3.92% in the 14 months after (range 3.86–12.34% across the four sites). The 12-month Holt-Winter method had the greatest accuracy prior to COVID-19 (3.18% ± 1.65%) and during the pandemic (11.31% ± 4.81%), while the 24-month Holt-Winter offered the best performance following the pandemic (5.91% ± 3.82%). The pediatric ED had an average MAPE more than twice that of the average MAPE of the three adult EDs (6.42% ± 1.54% prior to COVID-19, 25.18% ± 9.42% during the pandemic, and 12.34% ± 0.55% after COVID-19). After the onset of the pandemic, there was no immediate improvement in forecasting model accuracy until two years later; however, these still had not returned to baseline accuracy levels.Conclusion: We were able to identify a TES model that was the most accurate. Most of the models saw an approximate four-fold increase in MAPE after onset of the pandemic. In the months following the most severe waves of COVID-19, we saw improvements in the accuracy of forecasting models, but they were not back to pre-COVID-19 accuracies.

Background: Individuals experiencing homelessness residing in congregate shelters have increased risk for SARS-CoV-2 infection. Prevalence of asymptomatic SARS-CoV-2 in congregate shelters is high, but shelter characteristics associated with SARS-CoV-2 transmission are currently unknown. Methods: We conducted a cross-sectional, multicenter cohort study across five congregate shelters in Rhode Island. We tested people 18 years of age and older staying in Rhode Island congregate shelters in April 2020 during the COVID-19 pandemic. A survey instrument was designed and implemented based on an a priori sample size. All consented participants reported basic demographics, recent travel, duration of time at the shelter, any symptomatology, and had their temperature and pulse oximetry measured. Each participant was tested for COVID-19 using nasopharyngeal swabbing. Shelter characteristics about location, occupancy, resident length of stay, and COVID-19 mitigation strategies were collected through structured phone questionnaire with shelter staff. Results: A total of 302 individuals were screened and 299 participated across five homeless shelters. The median age was 47.9 (range 18-85) and 20% were female. Of the 299 participants, 35 (11.7%) were positive for SARS-CoV-2; rates varied among shelters, ranging from 0% to 35%. Among the participants in the study, 5% had a new cough, 4% shortness of breath, and 3% reported loss of taste or smell. Symptom prevalence did not vary significantly between positive and negative SARS-CoV-2 groups. Regular symptom screening was not associated with lower infection rates. Shelters with higher rates of positivity were in more densely populated areas, cared for a more transient populations, and instituted fewer social distancing practices for sleeping arrangements or mealtimes. Conclusions and Relevance: Residents of congregate shelters are at increased risk for asymptomatic transmission of SARS-CoV-2. To reduce transmission and enable continuation of low-threshold shelter services, there is a need for universal testing, implementation of infection control and physical distancing measures within congregate shelters, and expansion of non-congregate supportive housing.

Introduction: Boarding of patients in the emergency department (ED) is associated with decreased ED efficiency. The provider-in-triage (PIT) model has been shown to improve ED throughput, but it is unclear how these improvements are affected by boarding. We sought to assess the effects of boarding on ED throughput and whether implementation of a PIT model mitigated those effects.Methods: We performed a multi-site retrospective review of 955 days of ED operations data at a tertiary care academic ED (AED) and a high-volume community ED (CED) before and after implementation of PIT. Key outcome variables were door to provider time (D2P), total length of stay of discharged patients (LOSD), and boarding time (admit request to ED departure [A2D]).Results: Implementation of PIT was associated with a decrease in median D2P by 22 minutes or 43% at the AED (p < 0.01), and 18 minutes (31%) at the CED (p < 0.01). LOSD also decreased by 19 minutes (5.9%) at the AED and 8 minutes (3.3%) at the CED (p<0.01). After adjusting for variations in daily census, the effect of boarding (A2D) on D2P and LOSD was unchanged, despite the implementation of PIT. At the AED, 7.7 minutes of boarding increased median D2P by one additional minute (p < 0.01), and every four minutes of boarding increased median LOSD by one minute (p < 0.01). At the CED, 7.1 minutes of boarding added one additional minute to D2P (p < 0.01), and 4.8 minutes of boarding added one minute to median LOSD (p < 0.01).Conclusion: In this retrospective, observational multicenter study, ED operational efficiency was improved with the implementation of a PIT model but worsened with boarding. The PIT model was unable to mitigate any of the effects of boarding. This suggests that PIT is associated with increased efficiency of ED intake and throughput, but boarding continues to have the same effect on ED efficiency regardless of upstream efficiency measures that may be designed to minimize its impact.