Advances in hospital data that track current patient demand, staffing levels, and available resources allow administrators to anticipate and plan for extreme events like COVID-19 outbreaks. This type of data-driven leadership can reduce burdens on staff and prevent hospitals from becoming overwhelmed.
Hospitalization data, despite reporting mandates from the Centers for Medicare and Medicaid Services (CMS), has been insufficient to truly gauge the complete on-the-ground situation in hospitals. This was a major topic of discussion in our previous Expert Forum: COVID Patient Data. Both accessing and analyzing data on hospital resources and demand can give a better picture of institutional resilience and even predict future needs. Dr. Kimia Ghobadi of the Whiting School for Engineering’s Center for Systems Science and Engineering has been using hospital operations data to demonstrate how hospitals can preemptively reduce burden on staff, share resources, transfer patients, and prevent any one hospital from becoming overwhelmed during a crisis. These data should not just reflect the past, but help hospitals prepare for the future.
Hospitalization needs are one part of it — the demand side. The other side are the available resources. That could be the number of beds or the personnel — a whole array of people needed from doctors and nurses to janitorial staff. When we think about hospital data, we sometimes forget how dynamic it is. It's one thing to have a physical bed, but another thing to have all the required devices, staff, medications, and services to make the bed functional at the appropriate level of care.
This is not specific to the pandemic. Before COVID-19 there were also considerations about beds in semi-private or shared rooms when the other occupants may have had a communicable disease. The counts would show two beds, but one of them has to be closed for patient safety. Another example is if there is a disruptive patient in one bed or there are personnel shortages. These nuances eventually impact the actual hospital capacity and function, and are therefore critical to record and analyze.
We had the idea of pooling resources together among hospitals, but we could not really investigate that effectively without data. We started to think about it back in March 2020, and the mandate went into effect in the summer of 2020 with data releases beginning in late August. That's when we actually got to start the project. The mandate has been enabling us, but not to the extent that we were hoping. Part of the problem is that the data is being collected on a daily basis, but it's only being publicly released weekly.
The published data is aggregated in a lot of different manners, providing less detail than the data that is actually collected. For example, the admission and occupancy data specific to COVID-19 are released weekly, but bed capacity data is only provided as an aggregate with all other admissions. We know how many COVID-19 patients are admitted, but we don't know how many COVID-19 beds there are. Subsequently we don’t know how close to capacity hospitals are. Hospitals also collect data on other required resources like personnel shortages, but this is not included in the public dataset. Emergency Department data about patient overflow is not included either. All of these data points would have been helpful for operational analysis and for assessing future needs. They exist, but we don't have access to them. Data uncertainty and inaccuracy add another layer of complexity.
Overwhelmed hospitals may need to pick up the phone and call other hospitals to find beds for patients they cannot provide care to anymore. If the data existed publicly, hospital managers and clinicians could see the status of other hospitals and expedite the transfer process by calling those hospitals directly. That additional, stressful work to find space for a patient would lessen, and patients would get the care they need more quickly, just by making data freely available. Similarly, if data on hospital occupancy was available and accurate, other clinical facilities, emergency services, ambulances, and even patients themselves could use it to go to hospitals that have more available capacity. That would create a natural load distribution by better matching the number of patients arriving at hospitals with their available capacity. For decision makers on a city, county, or health system level, there can be substantial benefits to having these data points freely accessible without needing to jump through hoops. One of such benefit includes better resource allocation — determining if a hospital will need extra beds or a city will need an emergency field hospital in the near future.
These data are really useful, although they focus on the past and the present. The data don’t tell us much about the future in regards to resources or demand, which, in a way, is the more interesting question. Outside of a crisis like COVID-19, future demand could be extrapolated based on historical data. Then we also need insights into future resources. While the number of physical beds is pretty constant, available staffed beds for each service is a value that constantly changes. For individual hospitals, we can combine the data on current and past resources and demands to predict the future and advise on how to adjust the resources. In the case of COVID-19, hospitals could convert some units from normal operations to COVID-19. And once the COVID-19 cases decrease, they could bring them back to standard operations as soon as possible, which usually takes a week or two.
With predictive data and accurate estimation of available resources, we can tell a hospital that they will need a certain number of COVID-19 beds available. That will make the decision of converting beds to COVID-19 beds easier. These are tough decisions because these beds have to be taken away from another service to be added to COVID-19 units. Data can help individual hospitals with their capacity management, but that also means that they get to think about their personnel and equipment needs. If there is a component of their resources that they don't have, e.g. ventilators, they can address it with predictive capacity data.
Apart from being able to pool various resources together, one of the major benefits of this preemptive strategizing is reducing the burden on staff. If 100% of beds are full at a hospital, that hospital is already over capacity. The next person to arrive will not have a bed, so we need to keep operational utilization under 100%, which can be done by large-scale strategizing between hospitals. Currently, we often keep admitting patients until the hospital is almost overwhelmed. We then start to take actions to reduce the burden, such as sending patients to other locations. That adds a significant burden to clinicians, patients, and everyone involved. It's also not ideal for patient care because their care usually has to restart again with new providers in a new system. We recommended the concept of an optimal transfer, where we anticipate how many patients need to be transferred to avoid immediate or near-future overcrowding, and then transfer that many (appropriate) patients at the beginning of their care, e.g. from the Emergency Department, rather than delaying these decisions until capacity alarms are raised in the hospital. Preemptive but consistent transfers, as small as one patient every few days, can avoid large downstream overcrowding. If we have a sense of the demand and resources in the future, and we have data on operations today and in the recent past, it's possible to predict upcoming hospital utilization. We can know how many patients to admit and how many to transfer to keep all hospitals operational and know when more beds are required.