Using AI and behavioral science to improve hospital operations

Photo: Qventus

Behavioral science is used to get people to perform certain behaviors, including ones they may not typically embrace.

Clinicians, especially surgeons, generally do not like to be told what to do, and certainly not by technology like artificial intelligence.

Mudit Garg, CEO of Qventus, a vendor of AI-powered health IT, and a former McKinsey & Co. healthcare consultant, here shares insights with Healthcare IT News about how behavioral science can be combined with AI and machine learning technology to increase utilization and improve physician satisfaction.

Q. What is behavioral science and how is it used generally in U.S. work settings?

A. Behavioral science encompasses a broad set of fields – from economics, psychology, sociology and beyond – that study and attempt to change human behavior. Applied behavior change research has been popularized through best-selling books like Richard H. Thaler and Cass R. Sunstein’s Nudge, Charles Duhigg’s The Power of Habit, and many more.

Most well-known examples of applied behavior change focus on individuals, such as how to get them to eat healthier, save for retirement or click on more ads. Another example is how some organizations have incorporated behavioral science into their workplaces.

For example, Google ran successful experiments making healthier food more accessible without changing which unhealthy foods were available – a classic “nudge” per Thaler and Sunstein. More traditional behavior change devices are commonplace – for example, a sales team could use VIP incentive trips in order to motivate team members to hit quarterly goals.

What’s really exciting is that organizations are increasingly using behavioral science to not just motivate an individual employee’s behaviors, but to coordinate actions across stakeholders in an organization and beyond.

The same principles apply, but organizations need to be more thoughtful in designing mechanisms to drive the right behaviors in these more complex environments. This is where the careful partnership between applied behavioral science and technology comes in — and advancements in artificial intelligence and automation are well suited to helping organizations do this.

Take Uber, which tries to maximize the efficiency of every ride. Uber’s AI predicts “suggested pickup points” so that riders go to a precise location in order to get them to their destinations fastest. After booking, the AI prompts the rider to walk to a specific spot and motivates them by letting them know they’ll save a certain amount of time; the intelligence also lets the driver know that the rider has accepted the suggested pickup point.

As a result, drivers waste less time and can do more rides per hour, and riders pay less since they don’t take inefficient routes, plus they get to their destination faster. In aggregate, all of the minutes shaved off individual rides is a massive amount of value for Uber – not to mention all of the benefits to the greater good by not holding up traffic where there’s nowhere to pull over.

It’s a coordinated system of behavior change at scale, and it’s only possible with AI-enabled automation technology.

Similar opportunities exist in other complex domains that have vast numbers of interconnected actions. Healthcare, in general, and hospital operations, in particular, have enormous potential to improve outcomes while saving money and making it easier for workers.

Q. How can AI technology enable you to harness behavioral science in order to drive actions?

A. Behavioral science-based innovations from other industries are being used in healthcare to improve care operations – the set of operational activities involved in the delivery of care, such as OR scheduling and access, management of discharge planning, system-wide patient flow, and more, as well as operational processes that connect patient care beyond the hospital.

Manual operational processes in hospitals haven’t changed for decades, and leaders have reached the limits with process improvements that continue to rely on already overburdened staff. Basic analytics and visualization tools provide insight into processes, but hospitals need AI-based automation technology merged with thoughtful behavioral science to drive and sustain the necessary changes.

One of the few near-universal tenets of behavior change is to make the right thing to do the easy thing to do. This is easier said than done in the complex care operations environment. A system of action is required to do this successfully, with the first step being identifying the simplest set of actions needed to be taken to achieve the desired outcome.

Designing this set of simplest actions largely depends on new possibilities enabled by AI and machine learning. For example, AI and machine learning models that process EHR and operational data in real time can predict bottlenecks and point teams to the highest priority tasks, rather than having frontline teams spend determining what actions to take.

The second part is orchestrating and motivating those actions. In some cases, the action can be fully automated when there is high confidence, low downside risk and opportunity for a machine-led workflow. In other cases, the action needs to be confirmed or completed by a human, especially when clinical judgment is required.

The third part is accountability management, which requires giving frontline managers continuous visibility into which behaviors are or are not taking place and where to focus positive reinforcement and coaching. To be successful, an organization needs all three elements of the system, and underlying intelligence sophisticated enough to reliably power the system.

Bringing these three elements together is how hospitals are freeing up more OR time and growing surgical revenue. Most surgical departments routinely allocate dedicated block time to surgeons. Many surgeons don’t use all their block time, nor do they willingly release it soon enough for someone else to use it.

They might hesitate to release it because they think they might fill it at the last minute, or they might simply have no incentive to be good stewards of the extremely valuable resource of a staffed OR room.

This is where the magic of AI and behavioral science come into play. AI can predict a specific time within a block that is likely to be unused – up to one month in advance. When the AI reaches a confidence threshold, it automatically emails the surgeon to encourage them to release the time (prompt).

The email makes it easy to do, presenting the surgeon or scheduler with a simple button to accept or reject the nudge (action). Unlike a generic release reminder, an AI-based nudge gets sent as soon as the machine learning prediction has sufficient confidence and tells the surgeon why they should release the time.

For example, the nudge may let the surgeon know that there is only a 12% chance that they’ll use the time based on their historical booking patterns, they can increase their quarterly OR utilization by 8% if they release the time now, and they will be given priority on future open time (reward). The surgeon also knows that if they participate in this system and encourage others to do so, more OR time will become available for them (investment).

There are three important distinctions here that set this AI and behavioral science-based approach apart from basic analytic tools, which often give delayed insights at high effort and leave a lot of opportunity on the table.

One is the use of AI to identify the specific time likely to go unused. Generic email release reminders just create noise and are easily ignored. But because AI is targeted, the system can send fewer nudges with clear rationale to explain the recommended action, and an easy way to act.

The second is the use of AI to generate individualized rationale to encourage release, as I already described. If there’s no clear and compelling reason for the surgeon to release the time, they’ll be inclined to hold on to it.

The third is the use of automation to make the right action incredibly easy to perform. The surgeon scheduler can automatically release the time at the click of a button in the email.

What’s more, performance analytics can then assist OR leaders in closing the loop by helping them understand how often specific surgeon schedulers respond to the nudges and accrue benefits, or which surgeon schedulers may not be taking full advantage of the system. This gives them the visibility to become better managers, supported by a sophisticated and largely automated system of action.

An AI-based platform enables hospitals to integrate behavioral science throughout their care operations in ways never before possible, and there are countless use cases.

Q. What outcomes can you achieve by successfully using behavioral science?

A. Each year, the U.S. healthcare system loses more than $1 trillion in waste, according to The Journal of the American Medical Association. One major cause is inefficient operations. As a result, care teams and staff are experiencing epidemic levels of burnout, patients are unable to access the care they need in a timely manner, and hospital margins are at record lows.

By using AI-based technology that incorporates behavioral science, hospitals can make it easy for individuals across the hospital – and even outside stakeholders, like independent surgeons in the community – to take the right actions in order to reduce or eliminate inefficiencies. Significant benefits accrue to staff, patients and the hospital’s bottom line.

In the example I just described, behavioral science can encourage release of OR block, which unlocks more OR time for the hospital without having to add resources. With early release, the hospital has greater lead time to fill the slot with a high-value case.

This increases OR utilization and enables a single facility to grow revenue by over $10 million. The individual surgeon releasing the time improves their utilization, and all surgeons operating at the facility have more OR access because less block time goes unused. Most importantly, patients get more timely access to surgical care, which has a meaningful impact on experience and outcomes.

Another example in the OR is using behavioral science to optimize utilization of assets such as surgical robots. Unlike the OR block release example that uses behavioral science to prompt a binary decision (to release or not release the block time), here AI and behavioral science prompt the OR scheduler with an action that they otherwise may not know to take.

AI can suggest specific alternative rooms or times that won’t impact the case, making it easier for a scheduler to act. It further motivates the scheduler to make the change by letting them know the number of minutes that will be freed up for a robotic case. This enables hospitals to increase robotic cases by 33% and decrease the number of non-robotic cases in robotic rooms by 11%.

Similarly, behavioral science can be used to create inpatient bed capacity by improving discharge planning processes. During multidisciplinary discharge planning rounds (or MDRs), AI can predict an earlier estimated date of discharge (EDD) compared to what the care team had decided.

That’s the prompt for the care team to discuss whether that earlier discharge is feasible and what would need to be to achieve it. Such discussions are further facilitated by AI through the proactive identification of barriers to discharge so that they can be resolved before they become the binding constraint to length of stay.

Discharge planning is a lot of work, and it’s hard for care teams to feel that impact day to day. So, if the care team is shown the impact of their discharge planning efforts – for example, letting them know that their proactive actions have resulted in patients in the past week going home to families a total of 9 days earlier – the care team is more motivated to continue conducting high quality MDRs so that patients are discharged in a safe, timely manner.

It’s a win for the care team, the patient and the hospital, and these actions can reduce excess hospital days by 30-50%, creating the equivalent capacity of up to 30 additional inpatient beds using existing resources.

Q. How can you find the right balance between provider and staff autonomy and automation?

A. Coming back to one of the core tenets of behavior change, we must make the right thing to do the easy thing to do – and this requires capabilities that can intelligently and automatically trigger different actions based on context and other factors.

Using the discharge planning example, an advanced machine learning model can predict EDDs based on thousands of patient characteristics, past performance of the care team, and national benchmarks. Evaluating these recommendations is easier and more accurate than the care team generating them manually.

The AI can automatically populate EDDs for patients who do not have one entered by the care team. This eliminates work for the care team and helps automate the ideal outcome.

But if the care team has already entered an EDD, the AI can let them know they are on track or ahead of target, or it can prompt them to consider an earlier date if the care team has chosen one that is later than what the AI predicts. In this case, AI and behavioral science guide the care team’s decision making toward the ideal outcome.

One way to make the right thing to do the easy thing to do is to reduce friction wherever possible. In a manual system, staff may need to flip between multiple screens to view information or transcribe data from one system to another. These unnecessary steps make it harder to act and therefore make the right actions less frequent.

Integrating behavioral science and automations at the point of action help overcome this. For example, directly writing the AI-predicted EDD into the EHR eliminates the need to view multiple screens during discharge planning, resulting in a higher likelihood that the care team member will act based on the prompt.

In this context with AI-guided decisions and human-led actions, it’s important that a system reinforce desired behaviors as well as identify improvement opportunities. In a manual environment, this requires a hospital departmental leader to pull a report, analyze the data and then figure out what to do with it.

By the time they can have a conversation with their team, the data might be months old, and they are not equipped to provide specific feedback. This is another opportunity to use behavioral science with technology.

We know that teams are more likely to improve if they get granular feedback on behaviors that they know they can directly control. Performance analytics tools can provide managers with concrete measures their teams can work on, rather than inactionable data that managers need to analyze in order to figure out what’s going on.

Ultimately, AI and behavioral science-based automation technology reduces manual work and guides decision making, enabling top-of-licensure work for providers and staff.

Q. What are ways automation can be used to reduce staff burnout and address staffing challenges?

A. Hospitals are stuck in a vicious cycle due to staffing shortages, turnover and rising costs.

Given these challenges, hospitals need new ways to do more with available resources – and reduce the burden on those resources to retain them.

If hospitals continue to rely on people to manually manage operational processes, they’ll only perpetuate provider team and staff burnout, and won’t achieve the efficiency needed to actually bend the cost curve.

Current stopgap measures like agency staffing, scheduling flexibility and compensation adjustments aren’t sustainable. Hospitals need a sustainable approach that addresses burnout and solves operational and financial challenges.

One major root cause of staff burnout is the burden of manual tasks and complexity of the clinical environment that hold teams back from top-of-licensure work. Whether you’re a physician, nurse, case manager, scheduler or other staff member, time spent making back-and-forth phone calls, digging through the EHR to find data, or transcribing information takes time away from higher order tasks like being at the patient bedside.

This is where AI and behavioral science-based automation software enables hospitals to transform the work environment. By using AI and behavioral science to guide decisions, it reduces cognitive load. And by automating manual tasks, it reduces workload. This enables more top-of-licensure work, increasing provider and staff satisfaction as well as retention.

At the same time, automation lets hospitals increase resource productivity and asset utilization, which strengthens the bottom line. This is particularly important in the OR department, which provides a significant amount of revenue to support other hospital services that are not financially viable on their own.

Across the country, many hospitals are closing ORs due to staffing shortages, losing an extraordinary amount of revenue. But hospitals using automation have actually been able to do more surgical cases using existing staff.

For example, by using automation, Saint Luke’s Hospital of Kansas City was able to increase the number of surgical cases by 7%, despite having to temporarily close a fifth of its ORs due to staffing shortages. This strengthened the hospital’s bottom line and let them continue providing surgical care for patients.

What’s incredibly exciting is that AI-based automation software addresses burnout while simultaneously solving staffing challenges. It enables a work environment where providers and staff can thrive. It also enables the hospital to increase revenue, generating more ROI from investment in assets like ORs and surgical robots.

Twitter: @SiwickiHealthIT
Email the writer: [email protected]
Healthcare IT News is a HIMSS Media publication.

Source: Read Full Article