Every time a patient doesn't show up or reschedules an appointment, a health system loses money.
Health systems save money using digital tools for scheduling appointments, administrative work
That may sound like a minor problem in an industry dogged by astronomically high costs that show little sign of falling, but like an untreated ailment, minor problems often evolve into bigger ones.
Minor problems also tend to be great matches for new digital tools, like machine learning and natural language processing. While much of the hype focuses on artificial intelligence for clinical uses—diagnosing patients and predicting outcomes, for instance—AI-based tools are actually great for more mundane tasks, like self-service scheduling.
Remarkably, healthcare leads the pack of industries turning to AI. About 15% of healthcare companies are using AI, tied for the top spot with mining and minerals, according to a 2017 Accenture survey. The most common use, according to the survey, is managing information technology, followed by managing financial resources.
Experts have found that when patients schedule their own appointments, they're more likely to show up. So more health systems are offering self-scheduling options and using algorithms to predict which patients still might not come in.
“Things that are core to operations that are less sexy—the things that aren't going to make the front page of the newspaper—are really driving efficiency,” said Dr. Joe Kvedar, vice president of connected health for Partners HealthCare, which is rolling out self-scheduling for its patients.
Scheduling and other administrative tasks are ripe for cost-cutting through digital tools. Health systems are turning to those kinds of tools because of “the tremendous cost pressures they're facing,” said Brian Kalis, managing director of digital health and innovation for Accenture's health business. “They want to increase efficiency to maintain margins and serve as large of a population as they can,” he said. “Automation becomes a way to reimagine processes to get those cost efficiencies.”
That, in turn, can help address the huge labor costs health systems incur. “The healthcare business model is so broken, so wrong in that we keep hiring more and more people,” said Dr. Eric Topol, director of the Scripps Translational Science Institute. “It's really imperative to start to look at why we have such out-of-control growth in the healthcare workforce and what we can do about it.”
The use of digital tools begins as soon as a patient considers visiting a health system, during scheduling. Patients might call a clinic and make an appointment by talking to a digital assistant, they might type out their request to a chatbot online, or they might self-schedule through a patient portal. Each of these methods saves money by freeing up administrative staff time.
At Providence St. Joseph Health, of the more than 100,000 appointments last year for its Express Care services—on-demand virtual, clinic and at-home visits—more than half were scheduled online. “Consumers love it, and it lets us save money by not having someone who has to do it manually,” said Aaron Martin, chief digital officer at Providence St. Joseph. Each appointment booked online saves the system $3 to $4, and doing the math is fairly simple: That's about $300,000 or more saved.
When patients show up for their appointments, there are potential areas for cost-saving too. By using registration kiosks, there's less need for front-desk personnel. That doesn't necessarily mean layoffs would follow, Kvedar said. “Our volumes of new patients continue to grow, and we haven't hired as many people as a result of having these kiosks,” he said.
Labor accounts for roughly 60% of a hospital's costs that aren't capital spending. “Especially with the downward pressure on revenue and reimbursement, health systems are putting labor costs in the crosshairs,” said David Gregory, a principal at Baker Tilly and leader of the firm's healthcare consulting practice. “Many senior executives are looking for any way possible to reduce the administrative burden on their clinical team.”
Just as they're doing with nonclinical tasks, executives are also seeking fixes for clinical tasks that aren't necessarily earth-shattering but are, nevertheless, effective at driving efficiency and improving outcomes.
Electronic consults are one way to reduce waste. A primary-care physician, for instance, might electronically consult with a specialist, eliminating the need for a specialist visit entirely. At Partners, 60% of patients whose primary-care physicians virtually consulted with dermatologists did not need specialty visits. “That's efficiency in a market where there's more demand than supply,” Kvedar said.
But e-consults don't always work properly. Primary-care physicians might send hastily written notes, or they might send blurry photos. In these cases, the virtual consult might be for naught, with the patient needing a specialty visit anyway. Such care is redundant, not efficient, Kvedar said.
The trouble is the data are not always available in a usable form. But as data improve, coming not only from clinicians but also directly from patients, health systems can become even more efficient.
In certain areas, they already are. Some health systems, for instance, are tracking inpatient data to predict adverse events, allowing clinicians to intervene before anything goes wrong.
At Geisinger Health, physicians and radiologists can use voice recognition for notes and reports. Then, using natural language processing, the health system can pull out discrete data elements. It can also tag radiologists' readings results to comparatively analyze them.
Other health systems are using digital tools to predict patients' discharge dates to make planning more accurate. “It synchronizes the system,” said Manu Tandon, chief information officer at Beth Israel Deaconess Medical Center, where clinicians predict discharge dates both manually and with machine learning. “Otherwise, everybody is doing localized optimization.”
Without predicted dates, for instance, a custodial worker cleaning one room on a floor might as well clean all the rooms on that floor just because they're there. But with predicted dates, that worker can prioritize which ones should be cleaned, working in order of importance, not necessarily in order of proximity.
Opportunities exist for more administrative efficiency even after patients leave the system. “Over time, technology can really reduce the labor required for revenue-cycle activities,” Baker Tilly's Gregory said.
Because UPMC is both a payer and provider, it has an incentive to make the claims process more efficient and accurate. “The payer wants to lower the costs, and the provider wants to give better-quality care,” said Tal Heppenstall, executive vice president and treasurer of UPMC and president of UPMC Enterprises. “These digital tools in general, and AI in particular, can do both,” he said.
The health system has partnered with and invested in Health Fidelity, for instance, in the risk-adjustment cycle. Health Fidelity's software uses natural language processing to analyze clinical notes, optimizing coding for Medicare Advantage patients. “It generates significant benefits for our health plan,” Heppenstall said, “making sure the health plan and clinician are getting paid for the services the patient consumed.”
Because health systems are getting access to more and more claims data, they're finding new opportunities for lowering costs, said David Wildebrandt, a member of the Berkeley Research Group's healthcare performance improvement practice. “We're seeing health systems really looking at the cost drivers in their systems and inefficiencies,” he said.
Claims and prior authorization are prime opportunities for automation, OptumLabs CEO Dr. Paul Bleicher said. In those areas, AI-based tools might be more accurate than people. And even if they're not, they don't put patient safety at risk the way inaccurate AI tools would in clinical settings. Because AI tools for claims and other administrative tasks are, in that sense, safer, “it's an opportunity to really be able to be aggressive with augmented intelligence,” Bleicher said.
But accuracy is still important, Scripps' Topol said. That's where humans come in—though maybe fewer of them, as they're needed less for the actual tasks and more for validation. “We're not talking about elimination of any of these jobs, just reduction,” he said, as health systems increasingly turn to automation for tasks that machines are better at than humans.
“This is the perfect synergy of using machines with humans so you increase productivity and efficiency by basically outsourcing some of this work to machines that can do it better,” Topol said. “This is the sweet spot for machine learning.”
Send us a letter
Have an opinion about this story? Click here to submit a Letter to the Editor, and we may publish it in print.