Better Patient Forecasts and Schedule Optimization Improve Patient Care and Curb Staffing Costs
Staffing accounts for over 50 percent1 of an average hospital's costs. Too little staff can significantly impact quality of care and patient experience; too much staff can jeopardize a hospital's financial viability. More than 50 percent of administrators still schedule manually and base those schedules on number of beds available or on historical averages. The process is time-consuming, cumbersome, inflexible and inaccurate. New advances in predictive analytics can revolutionize hospitals' ability to predict patient demand, reduce staffing shortages and overages and improve the scheduling process for hospital staff and their managers.
By incorporating all relevant and available data and by applying proven optimization techniques, hospital groups can both preserve their operating margins and improve patient care. Sophisticated analytics models generate patient forecasts that are as much as 20 percent more accurate than predictions made on historical averages alone.2 Some of these new solutions can flexibly handle certification time, paid time off requests, shift swaps and more to generate patient-optimized schedules — all in far less time than manual approaches or legacy solutions require. Finally, new platforms provide comprehensive oversight into staffing-related metrics, letting directors and administrators monitor patient wait times, nurse-to-patient ratios, paid hours per visit and other hospital-specific, predetermined metrics. The most sophisticated solutions can generate iterative schedules to meet department-specific patient and financial targets.
All departments can benefit from a staffing solution that uses advanced analytics. The most natural department to start with is the emergency department, not just because it has unscheduled patient encounters, but because it's the top of the funnel for the majority of patients within a hospital. Efficiencies that occur here have positive ripple effects throughout the facility.
I will share with you the key components to consider in a highly effective scheduling solution. Let's start with the factors that differentiate solutions around patient forecasts and then review the capabilities required to generate schedules by leveraging hospital-specific data. Finally, we'll look at the reporting components that can now be leveraged prior to a schedule being published and alerts that facilitate intervention when necessary.
Accurate patient forecasts are the foundation of schedule optimization
Correct baseline forecasting is paramount in achieving the accurate staffing levels that will help minimize wait times and improve overall patient satisfaction. However, many existing staffing solutions' "forecasts" are based solely on historical averages — a highly inaccurate method of predicting the future. In a recent study we conducted in a top-20 hospital system, historical averages delivered results that were up to 20 percent less accurate than those delivered by newer methods incorporating more sophisticated predictive analytics.
A better approach is to use more advanced techniques. Here are the models, techniques and attributes to look for when searching for a more effective staffing solution:
- Models that predict the inflow of patients at a more granular level (e.g., by hour of the day, day of the week, and month of the year). The widely accepted practice of scheduling based on the number of beds in the ED and a patient-to-nurse ratio target (with the anticipation of calling off staff if the census is light) leads to higher-than-necessary staffing costs and more dissatisfied staff.
- Multiple regression models tuned to assign more weight to recent data (something scientists call "exponential decay average"). This ensures that volume shifts with long-term effects are taken into account properly. For instance, if a new facility opens nearby, that may have a lasting effect on future patient census, significantly reducing the relevance of data that was gathered before that facility opened.
- Self-learning models that adapt to changes in baseline volume, including seasonal, weekly or hourly patterns without human intervention. This makes their predictions more accurate for a longer period of time and it obviates the need for continual human tuning and rebuilding.
- Models built and trained for each individual hospital. Every hospital operates in its own market and exhibits its own demand curve for services. Baseline forecasts should be built based on both global and local factors, such as seasonality, holidays, location-based health issues and one-time occurrences, such as large conventions or major sporting events. All this data, when combined with sophisticated algorithms, allows the scheduling solution to identify the expected patient demand for a specific hospital, within very specific time windows.
- Simulation models for patient arrival and acuity or treatment needs. Because staffing requirements are highly dependent on not just overall numbers of patients but also the arrival patterns of those patients and their acuity levels or treatment needs, simulation models that run various patient arrival and acuity scenarios and estimate wait time for each scenario will provide a more accurate staffing picture. This information can be used to feed into optimization models that allow each hospital to meet its expense and wait time targets.
- Ability to generate simulations and incorporate hospital targets and metrics. Once all the forecasted patients have been "seen" by a nurse in the simulation, the solution should determine whether or not the wait times have met the targets set by the hospital. If the wait times exceed the target, it runs the simulation again and continues to do so until the target wait time is met. From there, it needs to be able to calculate the optimum number of nurses for each defined block of time.
- Ability to generate a recommended schedule. Inevitably, conflicts will arise when dealing with the hundreds of requirements, preferences and constraints that appear in a typical four- or six-week schedule. Sophisticated techniques can minimize these conflicts and help create a nearly complete schedule that requires minimal human input.
Simplifying the scheduling process benefits both staff and management
Faced with the urgent demands of an active hospital department, patient issues, certification training and company meetings, directors have little time to think strategically about staff schedules. A staffing optimization tool that takes into consideration the characteristics of each individual staff member (e.g., his or her work agreement, weekend rotation and PTO requests), as well as the number of full-time staff members on the floor and the preferences and available days for per diem resources, dramatically reduces the time required for scheduling. In many cases, schedule production time is reduced from several days to less than an hour.
Set targets — and track and report against them in real-time
When administrators set clear, specific targets, a good staffing solution ensures that those targets are met. State-of-the-art solutions alert directors to changes in patient census and wait times, excessive PTO utilization, dips in patient satisfaction and more.
The ability to aggregate data from disparate sources allows all of these factors to be incorporated into scheduling decisions. Hospital administrators need to have access to each metric, a comparison to the department target and historical trends. These metrics must also be incorporated into the model as foundational components to ensure staffing guidelines align to patient and financial targets.
With all of these tools, targets and metrics in place, management has the ability to adjust the schedule for lighter periods, level out non-productive hours across the weeks and minimize overtime.
Implementing an accurate staffing solution doesn’t have to be disruptive
Top-notch staffing solutions don't require a heavy IT investment or an overhaul of your existing system. They incorporate analytics-based forecasts with machine learning, customize each model based on an individual hospital's patient data and are delivered as a cloud-based solution. This means no changes to IT or infrastructure. They use data already in the hospital system, and the implementation process takes just a few months.
Hospital groups that choose to invest in staffing optimization that leverages advanced analytics will see benefits throughout their facilities. Implementing a solution that leverages Big Data through an accurate forecast, iterative schedule components, and the presentation of subsequent financial metrics will deliver positive outcomes across the hospital. Patients experience shorter wait times; staff are more satisfied and turnover is reduced; management has more capacity to tend to patients and staff; and hospital administrators have improved visibility into departmental metrics and performance. Analytic-driven staffing systems are reducing nurse hours by as much as 10 percent while meeting or beating financial and patient wait time targets. Overtime and nurse turnover have decreased, while patient and staff satisfaction Have increased.
Pieter Schouten is the general manager of Healthcare at Opera Solutions. Opera has developed a comprehensive set of healthcare analytics solutions that easily integrate with existing technology infrastructures and drive significant revenue, efficiency and patient outcome improvements at leading healthcare companies. Prior to Opera Solutions, Mr. Schouten held analytic and healthcare leadership roles at FICO, Medtronic, and at two analytics start-ups. Pieter received his BS in Commerce from the University of Virginia and earned his MBA from Harvard University.
1 Avalere Health analysis (2010) American Hospital Association Annual (Survey data from community hospitals).
2 Opera Solutions analysis (2013) (Independent research based on averages from a cross-section of hospitals).
© Copyright ASC COMMUNICATIONS 2017. Interested in LINKING to or REPRINTING this content? View our policies by clicking here.