When the COVID-19 pandemic arrived in early 2020, Northwell Health was at the epicenter of the epicenter. In addition to providing care for more COVID patients than any healthcare system in the region, Northwell was presented with the opportunity to conduct groundbreaking research on COVID utilizing a vast, unique, and diverse data set.
During this time, a data science team under Northwell’s Feinstein Institutes for Medical Research set out to derive a model that could predict survivability for hospitalized COVID-positive patients. The team analyzed labs and patient data from Northwell’s database of over 14,000 COVID-positive patient encounters and hospitalizations, and derived the Northwell COVID-19 Survival (NOCOS) predictive model. (The NOCOS derivation manuscript can be found in pre-print form at https://www.medrxiv.org/content/10.1101/2020.04.22.20075416v3.)
Methods and Results
The NOCOS model features six data points that can be ascertained for patients who have been admitted to the ED: patient age (years); oxygen saturation (%); absolute neutrophil count (K/uL); red cell distribution width (%); serum sodium (mmol/l); and serum blood urea nitrogen (mg/dL). Based on these six data points, the NOCOS formula is able to provide a percent probability that the patient will survive hospitalization. Performance in the internal validation, prospective validation, and external validation for NOCOS were marked by AUCs of 0.86, 0.82, and 0.82, respectively.
Due to the time sensitive nature of the COVID-19 crisis, the data science team looked to find a way to operationalize their work as quickly as possible, and the Usability Lab research team provided a solution in the form of the EvidencePoint clinical decision support software platform. In a manner of weeks, the NOCOS prediction model was implemented as a website running on top of the EvidencePoint platform. In April 2020, the NOCOS website was released publicly at https://cbmi.northwell.edu/nocos/.
Following the initial release, focus was turned to integrating the NOCOS website with Northwell’s Allscripts Sunrise Clinical Manager (SCM) EHR. The EHR-integrated version of NOCOS would be designed to pre-populate its six input fields with data from the EHR, dramatically simplifying the workflow and automating much of the tool’s use.
In August of 2020, the Usability Lab released the NOCOS 2.0 iCPR web app, which was fully integrated with SCM. This release pre-populated the six data fields, and wrote the resulting NOCOS survival percentage back into the SCM database as a Health Issue for reference over the course of a patient’s hospitalization.
Northwell’s large population of COVID patients presented an opportunity for researchers to perform essential research during the early days of the pandemic. The Northwell COVID-19 Survival (NOCOS) calculator was one of the results of this early research, and provided clinicians with a valuable tool to help treat patients with COVID-19. The Usability Lab’s EvidencePoint clinical decision support software platform was used to rapidly create a publicly facing website for the NOCOS tool, followed by a version that integrated with Northwell’s EHR systems to enhance usability.