For Josh Laberee of Advanced Imaging in Orange City, Florida, administering an independent medical imaging center means constantly evaluating its performance, from the conventional metrics around cost controls to those harder-to-pin-down details around patient satisfaction.
Advanced Imaging is a privately operated, standalone facility that competes on a throughput advantage, marketing a 15-minute turnaround time for its walk-in patients. Delivering on that experience absolutely turns on analysis of mechanical performance data and in-house benchmarking, which Laberee sees as ultimately in service to those cost controls. Yet despite that information, some of the most valuable data is qualitative; an itch that’s far more difficult to scratch.
“People are so concerned with making everything fit within benchmark; if you dissect that another layer, what are they doing within that category?” he asked. “You have all your C-suite people who are focused on the numbers, but how is that being executed?“
“We’re all about making the number, but I don’t think people are focused on how we get there,” Laberee said. “People on the ground can most tell you what’s working and what’s not. At the end of the day, we can’t lose sight on the biggest competitive advantage: how do you still meet your numbers but not take away the personal human experience? I think there’s going to be a ton of value if you do meet your numbers without losing touch.”
Part of the reason for Laberee’s conclusion is that many of the imaging workflow benchmarks that are published in studies of corporate America are based around patient experiences and departmental performance in multisite or hospital settings, and the data won’t necessarily capture the experience of a facility like Advanced Imaging. To generate maximum efficiency from his operation – and guarantee positive word-of-mouth, which leads to repeat business – Laberee’s workflow processes are anchored in the performance of staff with clearly defined roles and an ear to the ground on patient feedback.
“What creates our biggest competitive advantage is we’re not corporate,” he said. “We have a processing system in place, but as we’ve been able to grow the business by advertising that we can turn around a walk-in X-ray in 15 minutes, I’ve been able to staff accordingly. You can’t become efficient unless you know where you’re deficient. That’s hard to determine with benchmarks.”
To get at some of the most insightful patient experience data requires independent benchmarking, Laberee said, which in his case, means “secret-shopping” competitors “very tightly.” When staff, family and friends need to get imaging studies done, Laberee sends them to a competitor’s facility and covers the cost in exchange for detailed feedback on their experiences there.
“I want to know everything,” he said. “How long did you sit in the lobby? Did you like the taste of their coffee? What was the bathroom like? Did they support bilingual patients?”
In addition, Laberee said he pays close attention to the customer reviews he receives in direct response to patient feedback surveys as well as those reviews that show up on social media sites. Sometimes the lessons learned are painful, but informative; and in at least one case, expensive to correct. The ultimate example he described came from a place he least expected it: the front door.
“We have these beautiful automated doors, one of a kind, custom, that open at the front,” Laberee said. “But guess what? Patients are creatures of habit. Your typical door slides from right to left as you walk through. Ours came at you. People were pulling at door handles. Fast-forward, and we spend thousands to put in a conventional door.”
“Sometimes you try to do something unique, and sometimes you do have to conform with the masses and what’s expected,” he said.
Although workflow performance metrics are difficult to define, institutions nonetheless continue to seek artful ways to capture them, said Dr. Christopher Austin, chief medical officer at the U.K.-based Kheiron Medical Technologies. Independent of modality, all imaging workflow begins with scheduling, and, according to Austin, the fundamental way to conceive of scheduling efficiency is by measuring system performance in terms of time elapsed from the next available appointment until that appointment is completed. Some of that comes down to the type of study being performed: chest X-rays and mammograms are among the highest-volume studies that go through an imaging department, Austin said, while MR and CT studies “have their own complexities” when it comes to scheduling.
“Schedulers are looking to see if there’s any more information that would prioritize that patient over another,” he said. “They’ll have some ability to say, ‘This is a high priority,’ and they do some triage in that function.”
From when an appointment is scheduled, some organizations will measure elapsed time from check-in to the start of the procedure, which can include time spent preparing the patient for the exam, performance of the study itself, and any time spent preparing the room for the next patient. After that, the next question is how long does it take for that image to be put on a work list, where it can be read and reported by a radiologist? And finally, when does that information make it into the hands of the referring physician and, eventually, the patient?
“All these little steps are timestamps that organizations are trying to optimize for clinical and economic efficiency, and ultimately, for the patient,” Austin said. “The challenge of all of that is standardizing how they measure those times and metrics and how to remove human error from those measurements.”
Some of the variability in taking these measurements is related to the manual entry of these timestamps by staff as the process rolls itself out; because of it, these data points may not always be reflective of actual turnaround times, which can muddy the results as they’re analyzed. Moreover, Austin said some PACS and RIS vendors have not been able to automate timestamping through their software. Instead, he believes that analytical workflow processes aided by machine learning or artificial intelligence (AI) may be used to flag automatically those moments when radiologists review a study, taking away the guesswork.
“You do have a timestamp for when the study gets pulled, but I think where AI is being used, or will be used, alongside a radiologist, is a great way to timestamp what’s happening,” he said. “I think we’re going to see a lot more automation there, and a lot more accurate reporting of turnaround time.”
“In the radiologist’s workflow, it’s getting patients into the room and the reporting turnaround that are your biggest bottlenecks,” Austin said. “All these things take time, and AI will be able to help us understand those trends better.”
Austin believes that AI, which can support radiologists in managing complex imaging studies, including those of cancerous growths, can also improve workflow and patient throughput overall by helping prioritize those studies that may take more time to complete, or which may need to be conducted because of the severity of the potential findings.
“The decisions that are being made throughout the course of their staging and their follow-up is a precise art, and those are the studies that take the most time for radiologists,” Austin said. “If you allow AI to provide a lot more accuracy, you remove some of the subjectivity that’s in the hands of a radiologist. That’s where a lot of new efficiencies are going to happen.”
“As we create more and better diagnostic technology, it has an impact on the radiologist in terms of the time it takes,” he said. “That’s the tension: high volume, new technologies, and legacy imaging protocols. That’s the stuff that’s putting radiologists on the hamster wheels. Can AI actually help remove some of that fatigue? Definitely it can. We’re not there yet because the humans are doing a lot of the work to check what the machines are doing.”
There are, of course, bottlenecks to any multi-stage process, including those far less complicated than medical imaging. One such issue in organizations that offer multiple imaging modalities may simply be the challenge of additional variables. Maximizing usage of one imaging system may be complicated enough; factor in multiple suites with various modalities at multiple locations, and the problem becomes far more complex, even under normal conditions, to say nothing of times of stress. During the novel coronavirus (COVID-19) pandemic, for example, imaging throughput stalled for a number of reasons. Even discounting the government-mandated shutdown of non-essential operations, once facilities reopened, enhanced sanitization procedures were required to clean work environments between patients. Yet Austin argues that AI-based workflow solutions can help here, too, in prioritizing which patients should be seen first based on information analyzed from prior studies.
“You’re talking about as many as 10 million patients who missed their cancer screening appointments” during the pandemic, he said, citing an April 2021 investigation from the Journal of the American Medical Association. “It’s a scheduling as well as a workflow nightmare. How do we bring these patients back? We’re using AI to look at the previous exams of these patients, and generate a scheduling list based on features the AI saw in previous images that might suggest that someone is at greater risk of evolving into a more progressive cancer. I think that’s a really exciting development that really is putting a different tool into the hands of the schedulers.”
Imaging workflow efficiency can also be derailed by technical recalls; those circumstances in which imaging studies must be repeated because of staff errors in their initial execution. Although such circumstances account for maybe just 2 or 3 percent of all studies, Austin argues that AI-powered algorithms can aid throughput here again by tagging images at the time of the study and certifying that they’re of sufficient diagnostic quality.
“You’re not going to get paid for that repeat exam in this model, and you’re refining lower-quality images and causing consternation for the radiologist,” he said. “When it does happen, it’s an added burden, and an added cost for the patient. In a large organization, that can have a lot of financial consequences and certainly interrupt your scheduling.”
Payers also have an interest in improving radiology workflow efficiencies because processes that deliver better results – fewer technical recalls, optimized scheduling, faster turnaround times – can lead to better health outcomes for patients (and therefore, less expensive interventions are necessary to treat them). Payer-providers especially also rely on the kinds of data that can help optimize the patient experience for positioning their practices in the markets they serve. But like with all metrics, the pursuit of excellence can expose facilities to different pitfalls.
“Radiologists are looking for tools that would help them be more efficient and safer,” Austin said. “The challenge is that they don’t want to make mistakes, or push the productivity to a place where they’re sacrificing clinical quality. I think they really just want to be able to work smarter.”
Working smarter is especially critical when imaging stroke patients, said Rachel Witalec, vice president of product at the Menlo Park, California-based Rapid AI. The company was founded by Stanford University physicians who were looking to improve clinical decision support in order to better identify and diagnose cerebrovascular disorders that require the fastest possible treatment responses. The workflow problem to solve used to be finding ways to improve patient selection for treatment, Witalec said; then it shifted to finding the best places for them to receive necessary care as quickly as possible.
“In the world of stroke, you’ve got 6,000 hospitals, and only a subset of them have the resources to treat stroke,” she said. “Patients show up, but they have to be transferred. Every second you have that clot in your brain, your brain is withering away; if you have a clot in your brain for three hours, you’ve aged up 10 to 20 years. The problem becomes workflow.”
The RapidAI stroke mobile app tries to coordinate the various elements of imaging workflow by providing notifications for different stages of the patient journey, from their arrival in the facility to when imaging results are available to be viewed. Timer functionality shows where patients are throughout the process, counting up from the last known condition. Relevant personnel can communicate within the app, all with the aim of getting the best possible response within the shortest amount of time.
“Our sweet spot is fusing AI and workflow to help teams understand how they help patients,” Witalec said. “Hospitals get more patients to treat who need that treatment, which means reimbursement, and the team centralizes that tool, which minimizes burnout.”
To Witalec, stroke imaging workflow is less about optimizing throughput performance from a revenue perspective and more about aligning the discrete segments that comprise a rapid response to patient needs, from transport to treatment. That means understanding more about the health network itself, its transfer protocols, and the technologies available to practitioners there. In her experience, the best processes are derived in environments with institutional champions for workflow efficiency.
“For us it’s important to make these hospitals feel like they’re supported,” she said. “These are complex things; these are people’s lives, so it’s important to do that hand-holding, support the team and help them understand what changes with the new process.”
“As I’ve talked to more and more physicians and hospital teams, the pain points are very similar,” Witalec said. “They say, ‘I don’t know that an EMS team is going to arrive with a patient. Maybe I’m on-call, and I’m not even at the hospital. Maybe EMS didn’t take that patient to the most appropriate point of care. Maybe the ED, EMS and specialist teams are all using different tools.’ You then have very disconnected teams, and then the specialist team is often late to the party, and they have to mobilize to take care of the patient.”
“I think people are logjams naturally,” she said. “I think we’re really in the business of change management. That’s a psychological thing because people don’t like change, and technology is scary, and we’ve used the same process flow diagram for 10 years. How do we help train people so that they feel comfortable?”