Background: Malnutrition has been linked to adverse health economic outcomes. There is a paucity of data on malnutrition in patients admitted with COVID-19. Methods: This is a retrospective cohort study consisting of 4311 COVID-19 adult (18 years and older) inpatients at 5 Johns Hopkins-affiliated hospitals between 1 March and 3 December 2020. Malnourishment was identified using the malnutrition universal screening tool (MUST), then confirmed by registered dietitians. Statistics were conducted with SAS v9.4 (Cary, NC, USA) software to examine the effect of malnutrition on mortality and hospital length of stay among COVID-19 inpatient encounters, while accounting for possible covariates in regression analysis predicting mortality or the log-transformed length of stay. Results: COVID-19 patients who were older, male, or had lower BMIs had a higher likelihood of mortality. Patients with malnutrition were 76% more likely to have mortality (p < 0.001) and to have a 105% longer hospital length of stay (p < 0.001). Overall, 12.9% (555/4311) of adult COVID-19 patients were diagnosed with malnutrition and were associated with an 87.9% increase in hospital length of stay (p < 0.001). Conclusions: In a cohort of COVID-19 adult inpatients, malnutrition was associated with a higher likelihood of mortality and increased hospital length of stay.
Surgical resection for localized hepatocellular carcinoma (HCC) is typically reserved for a minority of patients with favorable tumor features and anatomy. Neoadjuvant immunotherapy can expand the number of patients who are candidates for surgical resection and potentially reduce the chance for recurrence, but its role in HCC not defined. We retrospectively examined the outcomes of patients who underwent surgical resection for HCC at the Johns Hopkins Hospital and compared the clinical outcomes of patients who received neoadjuvant immunotherapy with those who underwent upfront resection. The clinical cohort included a total of 92 patients, 36 of whom received neoadjuvant immune checkpoint inhibitor (ICI)-based treatment. A majority of patients (61.1%) who received neoadjuvant ICI–based therapy were outside of standard resectability criteria and were more likely to have features known to confer risk of disease recurrence, including α-fetoprotein ≥ 400 ng/mL (P = 0.02), tumor diameter ≥ 5 cm (P = 0.001), portal vein invasion (P < 0.001), and multifocality (P < 0.001). Patients who received neoadjuvant immunotherapy had similar rates of margin-negative resection (P = 0.47) and recurrence-free survival (RFS) as those who underwent upfront surgical resection (median RFS 44.8 months compared with 49.3 months, respectively, log-rank P = 0.66). There was a nonsignificant trend toward superior RFS in the subset of patients with a pathologic response (tumor necrosis ≥ 70%) with neoadjuvant immunotherapy. Neoadjuvant ICI-based therapy may allow high-risk patients, including those who are outside traditional resectability criteria, to achieve comparable clinical outcomes with those who undergo upfront resection. Significance: Surgical resection for localized HCC is typically only reserved for those with solitary tumors without vascular invasion. In this retrospective analysis, we show that neoadjuvant immunotherapy may allow high-risk patients, including those who are outside of standard resection criteria, to undergo successful margin-negative resection and achieve comparable long-term clinical outcomes compared with upfront resection. These findings highlight need for prospective studies on neoadjuvant immunotherapy in HCC.
Hepatic steatosis is a precursor to more severe liver disease, increasing morbidity and mortality risks. In the Emergency Department, routine abdominal imaging often reveals incidental hepatic steatosis that goes undiagnosed due to the acute nature of encounters. Imaging reports in the electronic health record contain valuable information not easily accessible as discrete data elements. We hypothesized that large language models could reliably detect hepatic steatosis from reports without extensive natural language processing training. We identified 200 adults who had CT abdominal imaging in the Emergency Department between August 1, 2016, and December 31, 2023. Using text from imaging reports and structured prompts, 3 Azure OpenAI models (ChatGPT 3.5, 4, 4o) identified patients with hepatic steatosis. We evaluated model performance regarding accuracy, inter-rater reliability, sensitivity, and specificity compared to physician reviews. The accuracy for the models was 96.2% for v3.5, 98.3% for v4, and 98.8% for v4o. Inter-rater reliability ranged from 0.99 to 1.00 across 10 iterations. Mean model confidence scores were 2.9 (SD 0.8) for v3.5, 3.9 (SD 0.3) for v4, and 4.0 (SD 0.07) for v4o. Incorrect evaluations were 76 (3.8%) for v3.5, 34 (1.7%) for v4, and 25 (1.3%) for v4o. All models showed sensitivity and specificity above 0.9. Large language models can assist in identifying incidental conditions from imaging reports that otherwise may be missed opportunities for early disease intervention. Large language models are a democratization of natural language processing by allowing for a user-friendly, expansive analyses of electronic medical records without requiring the development of complex natural language processing models.