Importance When hospitals are at capacity, accurate deterioration indices could help identify low-risk patients as potential candidates for home care programs and alleviate hospital strain. To date, many existing deterioration indices are based entirely on structured data from the electronic health record (EHR) and ignore potentially useful information from other sources. Objective To improve the accuracy of existing deterioration indices by incorporating unstructured imaging data from chest radiographs. Design, setting, and participants Machine learning models were trained to predict deterioration of patients hospitalized with acute dyspnea using existing deterioration index scores and chest radiographs. Models were trained on hospitalized patients without coronavirus disease 2019 (COVID-19) and then subsequently tested on patients with COVID-19 between January 2020 and December 2020 at a single tertiary care center who had at least one radiograph taken within 48 hours of hospital admission. Main outcomes and measures Patient deterioration was defined as the need for invasive or non-invasive mechanical ventilation, heated high flow nasal cannula, IV vasopressor administration or in-hospital mortality at any time following admission. The EPIC deterioration index was augmented with unstructured data from chest radiographs to predict risk of deterioration. We compared discriminative performance of the models with and without incorporating chest radiographs using area under the receiver operating curve (AUROC), focusing on comparing the fraction and total patients identified as low risk at different negative predictive values (NPV). Results Data from 6278 hospitalizations were analyzed, including 5562 hospitalizations without COVID-19 (training cohort) and 716 with COVID-19 (216 in validation, 500 in held-out test cohort). At a NPV of 0.95, the best-performing image-augmented deterioration index identified 49 more (9.8%) individuals as low-risk compared to the deterioration index based on clinical data alone in the first 48 hours of admission. At a NPV of 0.9, the EPIC image-augmented deterioration index identified 26 more individuals (5.2%) as low-risk compared to the deterioration index based on clinical data alone in the first 48 hours of admission. Conclusion and relevance Augmenting existing deterioration indices with chest radiographs results in better identification of low-risk patients. The model augmentation strategy could be used in the future to incorporate other forms of unstructured data into existing disease models.

No abstract

BackgroundHealthcare-associated Clostridioides difficile infection (C diff infection, or CDI) imposes a substantial burden on the healthcare system. The impact of an individual C diff infection on onward transmission is not well understood. We developed a model of incident infections using self-exciting stochastic processes, known as Hawkes processes. These models can be used to improve our understanding of the factors that affect the likelihood of new infections to result in additional infections.MethodsAll patients admitted to a large urban hospital between January 2013 and June 2014 were included. We used Hawkes processes to model the influence of each new CDI case (index infection) on transmission to other patients resulting in additional CDI. We developed separate Hawkes processes for each unit in the hospital to understand the differential impact of a C diff case across units. Units included both semi- and private-room wards, intensive care units, an emergency department, and specialty units such as oncology.ResultsThe magnitude of influence of an index infection on additional infections in the 2 days prior to a C diff test being sent varied across units. Results for an oncology unit, the emergency department, and an all private-room unit are provided (Table 1). An index infection in the emergency department demonstrated the greatest influence, leading to the largest number of additional infections, and increasing in the days leading up to the C diff test being sent. The impact 2 days prior to sample collection was similar across all unit types, and remained constant for oncology unit patients.ConclusionWe used Hawkes processes to model the impact of an index C diff infection on onward transmission. We identified differential impacts associated with the unit where the index patient was located in the days leading up to diagnosis. These differences, which could relate to unit-specific factors such as cleaning practices, patient turnover rates, use of portable medical equipment, antibiotic use, and other factors that vary across units, suggest that interventions aimed at controlling CDI may need to consider unit-specific approaches. Disclosures All authors: No reported disclosures.