2 Rightsizing Response: The Optimization of Critical Care Resources during COVID-19

Study Objectives: As the number of COVID-19 patients increased across the US, health care systems required a variety of approaches to meet the demand for critical care resources We sought to determine the ability of the existing health care system to meet these demands and explored the intersection of critical care bed (CCB) capacity and staffing availability in U S counties using two-week-ahead projections for April 13th, 2020 Methods: A linear optimization model was developed and solved using the revised simplex method The model aimed to minimize unmet demand for COVID-19 critical care through an optimal combination of (i) redistribution of nurses and physicians within each state (within 250 miles) and (ii) provision of additional CCB capacity and staff Staffing ratios of 2 CCBs/nurse and 10 CCBs/physician were applied Advanced practice practitioners (APPs) were used to “extend” physician coverage with each APP equal to 0 5 physicians Staffing counts were estimated using American Hospital Association and Health Resources and Services Administration Data To account for critical care training, 15% of RNs, 12% of NPs, 1 4% of PAs, and 50% of CRNAs were considered as available critical care trained staff Intensivists (100%) and Medical and Surgical specialists (30%) were included with 45% of these available for hospital staffing Case count projections were taken from the Columbia University models (Shaman, 2020) and 70% of CCBs in each county were assumed to be occupied by non-COVID-19 patients For each county, three potential constraints on increasing capacity were estimated: the number of nurses, the number of physicians (including APPs), and the number of CCBs One or more constraints could be active at any time Results: Prior to optimization, 91% of counties were able to meet the demand for projected case counts In contrast, 8 4% were limited by nursing resources, 0 09% by physicians, and 0 8% by the number of CCBs After optimization, 16 9% of counties sent nurses to a different county(s) (median 6 nurses sent, IQR 13 75) compared with 5 5% counties receiving them (median 23, IQR 43 5) Fewer physicians were relocated (0 09% sent, median 1, IQR 1;0 06% received, median 2 5, IQR 1 5) (Figure) Using baseline staffing ratios and availability, these redistributions led to a reduction in total unmet demand from 24,155 to 19,976 In order to fully meet demand across the US under these conditions, an additional 1,225 physicians, 41,939 nurses and 13,905 CCBs would have been needed Conclusion: This work shows that with the redeployment of resources even within state boundaries may provide relief to areas of need without causing strain in other locations While validation with actual redeployment during the pandemic can improve estimates, these models can provide decision support to stakeholders by suggesting optimal reallocation or the ability of existing resources to support additional capacity [Formula presented]