Severe SARS-CoV-2 infection is linked to the presence of autoantibodies against multiple targets, including phospholipids and type-I interferons. We recently identified activation of an autoimmune-prone B cell response pathway as correlate of severe COVID-19, raising the possibility of de novo autoreactive antibody production during the antiviral response. Here, we identify autoreactive antibodies as a common feature of severe COVID-19, identifying biomarkers of tolerance breaks that may indicate aggressive immunomodulation.

Objective: To determine whether clinicians will use machine learned clinical order recommender systems for electronic order entry for simulated inpatient cases, and whether such recommendations impact the clinical appropriateness of the orders being placed. Materials and Methods: 43 physicians used a clinical order entry interface for five simulated medical cases, with each physician-case randomized whether to have access to a previously-developed clinical order recommendation system. A panel of clinicians scored determined whether orders placed were clinically appropriate. Outcomes assessed included The primary outcome was the differences in clinical appropriateness scores, of orders for cases randomized to the recommender system. Secondary outcomes included usage metrics, and physician opinions. Results: Clinical appropriateness scores for orders were comparable for cases randomized to the recommender system (mean difference -0.1 order per score, 95% CI:[-0.4, 0.2]). Physicians using the recommender placed more orders (mean 17.3 vs. 15.7 orders; incidence ratio 1.09, 95% CI:[1.01-1.17]). Case times were comparable with the recommender system. Order suggestions generated from the recommender system were more likely to match physician needs than standard manual search options. Approximately 95% of participants agreed the system would be useful for their workflows. Discussion: Machine-learned clinical order options can meet physician needs better than standard manual search systems. This may increase the number of clinical orders placed per case, while still resulting in similar overall clinically appropriate choices. Conclusions: Clinicians can use and accept machine learned clinical order recommendations integrated into an electronic order entry interface. The clinical appropriateness of orders entered was comparable even when supported by automated recommendations.

Amid growing rates of burnout, physicians report increasing electronic health record (EHR) usage alongside decreasing clinical facetime with patients. There exists a pressing need to improve physician-computer-patient interactions by streamlining EHR workflow.To identify interventions to improve EHR design and usage, we systematically characterize EHR activity among internal medicine residents at a tertiary academic hospital across various inpatient rotations and roles from June 2013 to November 2016.Logged EHR timestamps were extracted from Stanford Hospital's EHR system (Epic) and crossreferenced against resident rotation schedules. We tracked the quantity of EHR logs across 24-hour cycles to reveal daily usage patterns. In addition, we decomposed daily EHR time into time spent on specific EHR actions (e.g. chart review, note entry and review, results review).In examining 24-hour usage cycles from general medicine day and night team rotations, we identified a prominent trend in which night team activity promptly ceased at the shift's end, while day team activity tended to linger post-shift. Across all rotations and roles, residents spent on average 5.38 hours (standard deviation=2.07) using the EHR. PGY1 (post-graduate year one) interns and PGY2+ residents spent on average 2.4 and 4.1 times the number of EHR hours on information review (chart, note, and results review) as information entry (note and order entry).Analysis of EHR event log data can enable medical educators and programs to develop more targeted interventions to improve physician-computer-patient interactions, centered on specific EHR actions.

BackgroundGreater US local public health department (LPHD) spending has been associated with decreases in population-wide mortality. We examined the association between changes in LPHD spending between 2008 and 2016 and county-level sociodemographic indicators of public health need.MethodsMultivariable linear regression was used to estimate the association between changes in county-level per-capita LPHD spending and 2008 sociodemographic indicators of interest: percent of population that was over 65 years old, Black, Hispanic, in poverty, unemployed, and uninsured. A second model assessed the relationship between changes in LPHD spending and sociodemographic shifts between 2008 and 2016.ResultsLPHD spending increases were associated with higher percentage points of 2008 adults over 65 years of age (+$0.53 per higher percentage point; 95% CI: +$0.01 to +$1.06) and unemployment (+$1.31; 95% CI: -$2.34 to -$0.27). Spending did not increase for communities with a higher proportion of people who identified as Black or Hispanic, or those with a greater proportion of people in poverty or uninsured, using either baseline or sociodemographic shifts between 2008 and 2016.ConclusionFuture LPHD funding decisions should consider increasing investments in counties serving disadvantaged communities to counteract the social, political, and structural barriers which have historically prevented these communities from achieving better health.