Patient Flow

Staffing the Shift: The Human Impact of Getting it Right

Hospitals, Staffing, Artificial Intelligence, Machine Learning, Patient Flow

“They just don’t care about us,” goes the refrain. When the shift is understaffed, nurses can feel betrayed by the managers and executives responsible for filling shifts.

As an emergency physician, I work at a wide range of Emergency Departments (“ERs”), from large referral centers with four or more providers working around the clock, to tiny ERs with eight beds and a single doc. Regardless of the size or sophistication of the ER, when nurses feel stretched beyond their limits, it’s common to hear them voice distress or to wonder who is looking out for them.

Having worked both at the front lines as a clinician as well as an administrator, I know that often enough leaders do care, and care a great deal, about filling those shifts. But sometimes it can seem impossible to find someone even for a half shift.

At the other extreme, many times I’ve been on an overnight shift at a small ER that has emptied out in the small hours of the morning. Nobody complains about those shifts. And in fact, when there is bitterness about working understaffed shifts, these overstaffed hours can seem like a well-earned respite.

The dream is, obviously, to staff every shift such that staffing precisely meets demand. But there are real-world limits to doing so. Some of these are not amenable to change. There are costs of readiness. Even if a small ER empties out for a few hours overnight, there still needs to be a doctor, a couple of nurses, and enough other staff to run a code.

However, some of these limits are imposed by a lack of information. If we are facing the prospect of staffing some nights with 10 nurses, and others with eight, because there simply aren’t enough available people, wouldn’t it be great to know which nights to staff down?

Of course some of this can be done using historical data. For example, you may know that Monday nights are always the busiest, and thus focus on using more of those staffing hours on Monday nights, leaving others possibly more short staffed.

But leveraging advanced data analytics can extend the value of historical data considerably, by finding patterns not only by day of week, but across seasons, special dates in the calendar and other events. With the right tools, shifts can be tailored to more precisely match staffing to predicted demand, hour by hour.

And still more advanced approaches can go beyond historical patterns, to forecast spikes and dips in arrivals at the ER days ahead of time. For managers, this means that if they can round up some additional hours of nursing coverage, and they know where best to put those hours to use. It also means they are not finding themselves in a position where they are making calls to ask nurses to drop their plans and come in to work that very day. Armed with several days’ advance notice, they can make the necessary staffing arrangements ahead of time, giving nurses more opportunity to make themselves available.

As hospitals adopt these powerful analytical tools, there may be fewer staff working these “slow” hours, but it can also means not getting pummeled on understaffed, “busy” shifts. I believe most of the nurses with whom I work would agree that was a good tradeoff. These cutting edge approaches will give managers the tools to look after the nurses they work with, and show them just how much they care.

This post originally appeared on the Hospital IQ website.

Patient Flow

Triggers for Surges: The Role of Advanced Data

Precision matters when hospitals are responding to surges in inpatient demand. Many hospitals learn the hard way just how expensive it can be when there is no effective process for responding to surges as they can result in lost transfers, lost direct admissions, lost admissions from ED walkouts, sometimes even canceled surgical cases, to say nothing of the negative impact to the patient experience, level of clinical care and hospital reputation. Yet hospitals that have woken up to the need for surge planning know getting the response right is just as important. Triggering surge actions when they aren’t necessary or targeting the wrong areas for action can also be expensive and can undermine staff engagement with surge planning. Advanced data systems can take hospitals to the next level of economic and operational performance by identifying triggers for surge response activation that are early, accurate, specific, and targeted.

Even without advanced data, implementing a surge plan with appropriate triggers and actions beats having no plan at all! But as we look at all the classic situations where bottlenecks cause delays and overcrowding, we see just how much can be achieved with advanced data systems that offer accurate and timely insights.

Emergency Departments

Surge triggers in the ED will typically involve how many patients are waiting to be seen, how many are currently being treated, and how many are “boarding” – admitted patients waiting for inpatient beds. These measures are good for raising the alarm when there is already a problem. However, many of the actions to address and alleviate surges require significant lead time, such as reassigning nurses or calling them in for extra shifts. Advanced data systems can forecast ED demand days in advance based on historical demand, weather patterns, seasonality and local events to predict the number of patients and mix of acuities arriving hour-by-hour. Combining this with simulation modeling produces predictions days ahead of time to warn of impending high demand situations. Additionally these predictions will be for specific bed types, highlighting whether it is critical care beds, telemetry beds, or med-surg beds that will be most in need.

Inpatient censuses

Inpatient census is another common trigger for surge responses. Once again, hospitals typically have triggers that reflect a state of bed shortage and overcrowding only once the crisis has occurred. Opening additional nursing units and beds takes time, as does calling in staff, working with case managers to prioritize discharges, and other measures to address census crunches. Advanced data approaches combine forecasting with simulation modeling of the entire hospital or health system. This allows hospitals to mobilize early and to focus upon specific levels of care. For instance, there are times when hospitals experience a shortage of critical care beds but not other bed types. Such events require targeted action rather than an overall surge response, and advanced data systems can predict these events well ahead of time.

Advanced data approaches

Advanced data approaches can combine several key elements, such as data from disparate hospital IT systems (e.g., health records, physician order entry systems, bed management and finance). They integrate these disparate data sources into a single data warehouse. They can run thousands of simulations with real data, to quantify not only expected metrics, but variability. They allow “what if” scenario testing to establish whether specific actions will work to improve operations, avoiding putting staff through trial and error initiatives. Complex forecasting can bring in arrivals through all sources, whether it’s emergencies, direct admits, or planned surgeries. Holistic modeling across an entire system can allow decisions to fully realize synergies across sites. Dashboards break down walls between disciplines and departments, allowing everyone to share the same data and the same forecasting in real time. And these systems allow for radical transparency, drilling down to the level of individual encounters behind the metrics to achieve trust and buy-in.

All of these features are important for improving healthcare operations generally. When it comes to census surges specifically, these capabilities will powerfully augment traditional triggers to ensure surge responses are early, accurate, specific, and targeted.

This article originally appeared on the Hospital IQ website

Patient Flow

Only a Holistic Approach Can Fix Patient Flow

Patient Flow, Hospital Administration, Operations Management, Overcrowding, Discrete Event Simulation
Modeling the Entire Hospital

When working to improve patient flow, it is vital to recognize that the different parts of a hospital are deeply interconnected. The nursing units where overcrowding is most evident may not be the units where the mismatch between demand and capacity is causing the problem. Further, patients have complex journeys through their hospital stays, and do not simply move “downstream” from ED to ICU to the floor. In order to improve patient flow in hospitals, sophisticated data analytics are needed to guide decisions on beds and staffing, and to target efforts on length of stay. And to deliver meaningful, actionable information, such analytics must be “holistic”. By holistic, I mean that analytics must encompass the entire interconnected picture of a hospital or even a system, as well as the full complexity of patient movements.

Those of us who work in patient flow have long realized that we cannot address any one part of the hospital in isolation. When tasked with “fixing the ED” to reduce crowding, walkouts, and ambulance diversion, and to improve door-to-doctor time, we know that process changes in the emergency department alone won’t get us to where we need to be. In most cases, emergency departments with significant crowding and delays are that way because beds in the emergency department during peak hours are filled with patients who are admitted and who are waiting to be moved to inpatient beds. Thus, freeing up inpatient capacity is key to unblocking the emergency department.

The same things hold true for other parts of the hospital. We cannot address problems in intensive care units without addressing the floor. We cannot address the PACU without addressing the intensive care units.

We also know that to fix problems in one part of the hospital we have to look both “upstream” and “downstream” to ensure we are actually addressing the real problem and not a symptom. The ICUs may be full because there aren’t enough telemetry beds for patients to move out to, not because there are too few ICU beds.

However, the picture is far more complicated than words like “upstream” and “downstream” can convey. Patients may move from the floor up to the ICU or go from the ICU to the OR, and sometimes they are directly discharged home from the ICU.

The true complexity of all the different patient journeys between different parts of the hospital has to be adequately represented if we are to target our true goals: namely, how often do we want to be able to place patients in the right level of care and the right nursing unit and how much of a delay are we willing to tolerate.

Nineteenth century approaches cannot yield the quantitative answers needed to fix flow. Walking around and looking at units that are full or at beds that are empty does not adequately explain what actions are required. Midnight and noon census, basic arithmetic and spreadsheets all fall short, as does queuing theory alone, which simplifies patients’ journeys into flow between compartments based on formulas and probability distributions.

We need computationally intensive advanced data modeling, such as discrete event simulation and what-if scenario testing, if we are to make the best use of our beds. Truly holistic models represent all the different parts of the hospital, including the ED, the OR, PACU, ICUs, floor; they take into account multiple sites across health systems; they factor in real-life constraints from hospital policies; and they represent the full complexity of patient journeys which crisscross throughout institutions. Only such holistic approaches can accurately represent the complexity we wish to manage. And only such approaches can provide us the operational clarity we need to improve patient access, and to better align staffing structure for overall financial performance.

In the broader culture, when the word holistic comes up, it can come across as a throw away adjective, used to sell everything from skin care products to dog food. But in hospital administration, a 21st century data analytics platform that offers a holistic view for operational planning and management is far from being a consumer luxury. It is an essential requirement for delivering efficient, effective patient care.

Paris Lovett, MD, MBA, FACEP, FACHE
Chief Medical Officer, Hospital IQ
Post originally appeared on the Hospital IQ Website