Validation of parsimonious prognostic models for patients infected with COVID-19.

OBJECTIVES: Predictive studies play important roles in the development of models informing care for patients with COVID-19. Our concern is that studies producing ill-performing models may lead to inappropriate clinical decision-making. Thus, our objective is to summarise and characterise performance of prognostic models for COVID-19 on external data. METHODS: We performed a validation of parsimonious prognostic models for patients with COVID-19 from a literature search for published and preprint articles. Ten models meeting inclusion criteria were either (a) externally validated with our data against the model variables and weights or (b) rebuilt using original features if no weights were provided. Nine studies had internally or externally validated models on cohorts of between 18 and 320 inpatients with COVID-19. One model used cross-validation. Our external validation cohort consisted of 4444 patients with COVID-19 hospitalised between 1 March and 27 May 2020. RESULTS: Most models failed validation when applied to our institution's data. Included studies reported an average validation area under the receiver-operator curve (AUROC) of 0.828. Models applied with reported features averaged an AUROC of 0.66 when validated on our data. Models rebuilt with the same features averaged an AUROC of 0.755 when validated on our data. In both cases, models did not validate against their studies' reported AUROC values. DISCUSSION: Published and preprint prognostic models for patients infected with COVID-19 performed substantially worse when applied to external data. Further inquiry is required to elucidate mechanisms underlying performance deviations. CONCLUSIONS: Clinicians should employ caution when applying models for clinical prediction without careful validation on local data.

Megakaryocytes and platelet-fibrin thrombi characterize multi-organ thrombosis at autopsy in COVID-19: A case series.

BACKGROUND: There is increasing recognition of a prothrombotic state in COVID-19. Post-mortem examination can provide important mechanistic insights. METHODS: We present a COVID-19 autopsy series including findings in lungs, heart, kidneys, liver, and bone, from a New York academic medical center. FINDINGS: In seven patients (four female), regardless of anticoagulation status, all autopsies demonstrated platelet-rich thrombi in the pulmonary, hepatic, renal, and cardiac microvasculature. Megakaryocytes were seen in higher than usual numbers in the lungs and heart. Two cases had thrombi in the large pulmonary arteries, where casts conformed to the anatomic location. Thrombi in the IVC were not found, but the deep leg veins were not dissected. Two cases had cardiac venous thrombosis with one case exhibiting septal myocardial infarction associated with intramyocardial venous thrombosis, without atherosclerosis. One case had focal acute lymphocyte-predominant inflammation in the myocardium with no virions found in cardiomyocytes. Otherwise, cardiac histopathological changes were limited to minimal epicardial inflammation (n = 1), early ischemic injury (n = 3), and mural fibrin thrombi (n = 2). Platelet-rich peri­tubular fibrin microthrombi were a prominent renal feature. Acute tubular necrosis, and red blood cell and granular casts were seen in multiple cases. Significant glomerular pathology was notably absent. Numerous platelet-fibrin microthrombi were identified in hepatic sinusoids. All lungs exhibited diffuse alveolar damage (DAD) with a spectrum of exudative and proliferative phases including hyaline membranes, and pneumocyte hyperplasia, with viral inclusions in epithelial cells and macrophages. Three cases had superimposed acute bronchopneumonia, focally necrotizing. INTERPRETATION: In this series of seven COVID-19 autopsies, thrombosis was a prominent feature in multiple organs, in some cases despite full anticoagulation and regardless of timing of the disease course, suggesting that thrombosis plays a role very early in the disease process. The finding of megakaryocytes and platelet-rich thrombi in the lungs, heart and kidneys suggests a role in thrombosis. FUNDING: None.

Bending the cost curve: time series analysis of a value transformation programme at an academic medical centre.

BACKGROUND: Reducing costs while increasing or maintaining quality is crucial to delivering high value care. OBJECTIVE: To assess the impact of a hospital value-based management programme on cost and quality. DESIGN: Time series analysis of non-psychiatric, non-rehabilitation, non-newborn patients discharged between 1 September 2011 and 31 December 2017 from a US urban, academic medical centre. INTERVENTION: NYU Langone Health instituted an institution-wide programme in April 2014 to increase value of healthcare, defined as health outcomes achieved per dollar spent. Key features included joint clinical and operational leadership; granular and transparent cost accounting; dedicated project support staff; information technology support; and a departmental shared savings programme. MEASUREMENTS: Change in variable direct costs; secondary outcomes included changes in length of stay, readmission and in-hospital mortality. RESULTS: The programme chartered 74 projects targeting opportunities in supply chain management (eg, surgical trays), operational efficiency (eg, discharge optimisation), care of outlier patients (eg, those at end of life) and resource utilisation (eg, blood management). The study cohort included 160 434 hospitalisations. Adjusted variable costs decreased 7.7% over the study period. Admissions with medical diagnosis related groups (DRG) declined an average 0.20% per month relative to baseline. Admissions with surgical DRGs had an early increase in costs of 2.7% followed by 0.37% decrease in costs per month. Mean expense per hospitalisation improved from 13% above median for teaching hospitals to 2% above median. Length of stay decreased by 0.25% per month relative to prior trends (95% CI -0.34 to 0.17): approximately half a day by the end of the study period. There were no significant changes in 30-day same-hospital readmission or in-hospital mortality. Estimated institutional savings after intervention costs were approximately $53.9 million. LIMITATIONS: Observational analysis. CONCLUSION: A systematic programme to increase healthcare value by lowering the cost of care without compromising quality is achievable and sustainable over several years.