Search for: All records

Abstract ObjectiveTo develop predictive models of coronavirus disease 2019 (COVID-19) outcomes, elucidate the influence of socioeconomic factors, and assess algorithmic racial fairness using a racially diverse patient population with high social needs. Materials and MethodsData included 7,102 patients with positive (RT-PCR) severe acute respiratory syndrome coronavirus 2 test at a safety-net system in Massachusetts. Linear and nonlinear classification methods were applied. A score based on a recurrent neural network and a transformer architecture was developed to capture the dynamic evolution of vital signs. Combined with patient characteristics, clinical variables, and hospital occupancy measures, this dynamic vital score was used to train predictive models. ResultsHospitalizations can be predicted with an area under the receiver-operating characteristic curve (AUC) of 92% using symptoms, hospital occupancy, and patient characteristics, including social determinants of health. Parsimonious models to predict intensive care, mechanical ventilation, and mortality that used the most recent labs and vitals exhibited AUCs of 92.7%, 91.2%, and 94%, respectively. Early predictive models, using labs and vital signs closer to admission had AUCs of 81.1%, 84.9%, and 92%, respectively. DiscussionThe most accurate models exhibit racial bias, being more likely to falsely predict that Black patients will be hospitalized. Models that are only based on the dynamic vital score exhibited accuracies close to the best parsimonious models, although the latter also used laboratories. ConclusionsThis large study demonstrates that COVID-19 severity may accurately be predicted using a score that accounts for the dynamic evolution of vital signs. Further, race, social determinants of health, and hospital occupancy play an important role. 

Background Effective implementation of technologies into clinical workflow is hampered by lack of integration into daily activities. Normalisation process theory (NPT) can be used to describe the kinds of ‘work’ necessary to implement and embed complex new practices. We determined the suitability of NPT to assess the facilitators, barriers and ‘work’ of implementation of two clinical decision support (CDS) tools across diverse care settings. Methods We conducted baseline and 6-month follow-up quantitative surveys of clinic leadership at two academic institutions’ primary care clinics randomised to the intervention arm of a larger study. The survey was adapted from the NPT toolkit, analysing four implementation domains: sense-making, participation, action, monitoring. Domains were summarised among completed responses (n=60) and examined by role, institution, and time. Results The median score for each NPT domain was the same across roles and institutions at baseline, and decreased at 6 months. At 6 months, clinic managers’ participation domain (p=0.003), and all domains for medical directors (p<0.003) declined. At 6 months, the action domain decreased among Utah respondents (p=0.03), and all domains decreased among Wisconsin respondents (p≤0.008). Conclusions This study employed NPT to longitudinally assess the implementation barriers of new CDS. The consistency of results across participant roles suggests similarities in the work each role took on during implementation. The decline in engagement over time suggests the need for more frequent contact to maintain momentum. Using NPT to evaluate this implementation provides insight into domains which can be addressed with participants to improve success of new electronic health record technologies. Trial registration number NCT02534987 .