With the shift to a value-based reimbursement model, hospitals and clinicians are looking for ways to increase efficiencies and improve patient outcomes. Artificial intelligence can help to streamline diagnoses and treatments by culling through volumes of data and pinpointing specific disease types or other patient data. Patients who need to be seen are seen quicker because a doctor or nurse wasn’t spending time looking through reams of reports, which in turn increases satisfaction all around.
The goal of cognitive computing is to make knowledge workers more effective, not to replace them, says Hal Andrews, president of healthcare at software company Digital Reasoning. “Any workflow that requires humans to read or skim or scan vast amounts of data, technology can make that more efficient," Andrews tells Healthcare Dive. "And there are better outcomes for the patients, there are lower costs for the system as a whole, and there’s improved job satisfaction for the knowledge workers.”
Use of AI is growing in healthcare, with the market poised to reach $6 billion by 2021, up from $600 million in 2014, according to Frost & Sullivan. Cognitive solutions such as IBM’s Watson system are capable of sifting through huge volumes of patient data and providing guidance and decision support to improve workflow.
Also, Vice President Biden cancer moonshot has plans to harness big data to provide precision medicine solutions for veterans and to expand the use of mobile and wearable technologies for cancer diagnosis and treatment. The initiative also is developing a tool that converts narrative into standardized data, making clinical data more available for research.
Enhancing outcomes
When it comes to diagnosing cancer or a heart ailment, time is of the essence. Clinicians can spend a lot of time pouring through X-rays or pathology reports to find the 10% or 20% of information that needs to be followed up. That can mean delays in diagnosis and treatment. With AI, “we’re looking for the signal in the data and we’re trying to eliminate the noise," Andrews says. "And when it comes to clinical care, the noise is all the people that you don’t need to follow up with and the signal is the people that you do.”
Today, most healthcare decisions are based on structured data. Cognitive computing can improve patient care by looking for patterns over time in the nursing and physician documentation and triggering an alert or other action based on a longitudinal trend, according to Andrews.
Improving efficiencies
One way Digital Reasoning is working to improve workflow is by using technology to automate reporting to state and federal agencies — in this case, radiology reports that indicate breast cancer. “If a radiology report suggests a patient has breast cancer, from that report I need to extract certain data about the report, about the patient, about the site of service, about the tumor that’s indicated, and I have to report that,” Andrews explains.
All of that work is typically done manually. But technology can analyze those reports, pull the ones that require follow-up, extract the data elements that need to be reported, and auto-report them, he adds.
Tackling heart disease
Cognitive computing is also being used to improve outcomes in patients with congestive heart failure. Earlier this month, New York’s Mount Sinai Hospital partnered with CloudMedx to leverage intelligent predictive insights from the company's technology to identify patients who are at higher risk of CHF, and use evidence-based care interventions to reduce hospital readmissions and improve overall wellbeing.
“As an industry, we do not have a sufficiently sophisticated tool to predict certain things such as disease progression and resulting readmissions in hospitals,” Ashish Atreja, chief technology innovation and engagement officer in Mount Sinai’s Icahn School of Medicine, said in a statement. “CloudMedx has a fast, scalable platform that can allow us to do just that.”
Using predictive algorithms, CloudMedx identifies clinical risks during patient visits, gaps in care, and potential outcomes. Those risks can then be addressed with customizing treatment guidelines, the company said.
Clinical trials and adverse drug events
On the treatment front, Digital Reasoning is working with clinical research organizations to apply technology to match patients with clinical studies. The idea is to analyze the massive amounts of unstructured data from patient questionnaires, documents about the efficacy of a drug, indications and contraindications, social media websites, and so forth, and recruit patients who fit the profile of a particular drug.
The company is developing ways to help pharmaceutical companies analyze postmarketing reports of adverse drug events to understand if patients in places around the globe are experiencing the same problem.
“When you start to introduce multilingual reporting, you have to have models that can detect changes in how people say things in Portuguese versus how they say them in Japanese,” Andrews says. “Using our technology, we’re able to match up reports in Japanese and reports in Portuguese and reports in English and say these three patients in three countries speaking three languages actually were experiencing the same symptom.”
Adopting AI technology
So is AI technology within reach of most hospitals? Andrews says it is, noting that the push to electronic medical records and computers has meant that more and more clinical data is electronic. Moreover, the HL7 standard, at least in theory, provides a common format for moving data around electronically around in healthcare. “Anyone who can get healthcare data to an HL7 format, we can analyze it,” he says.
And because patient data is processed in a streaming fashion as opposed to a batch fashion, it requires less computing power and hence less infrastructure investment, he adds.
The real obstacle to the adoption of cognitive computing in healthcare is the regulatory environment around privacy laws, Andrews argues. “The [government’s] policy of these extreme fines for inadvertent breaches of HIPAA have the hospitals scared,” he says. “And so the hospitals will not leverage cloud technology, which could make cognitive computing really affordable because they’re scared they’ll get fined.”
A new world order?
Despite such concerns, clinicians will need to work with machines to enhance their own abilities and stay competitive. “If you just look at cancer drugs, there are so many indications and contraindications and research day after day after day. Humans just can’t keep up with it,” Andrews says.
Clinicians can try to grapple with the data and wind up “flying blind.” Instead, they can apply technology to filter the noise and find the needle in a haystack, he adds. “Helping humans make better decisions in a more timely fashion is the real value of cognitive computing for healthcare.”