Since I took my first “data processing” class in college, the rule has been that only “data” in a defined box could be used in computing. All our various notes and information that falls outside the structure was valuable on an individual basis but could not be aggregated.
This has been the challenge for medical records. Those piles of hand-written notes recorded in a consumer file, the many faxes laying on the floor of the emergency room, and the promise of information technology seemed to be out of reach for many health care situations. But, a recent announcement from Amazon raises the possibility that technology in general, and artificial or augmented intelligence (AI) specifically, may offer some solutions sooner than we think.
The announcement I’m referring to came from Amazon late last year—describing the launch of its Comprehend Medical machine learning service to extract medical information from unstructured text in electronic health data, such as clinical notes, prescriptions, audio interview transcripts, and pathology and radiology reports (see Amazon Launches Comprehend Medical AI Service To Analyze Unstructured Electronic Health Data). Amazon Comprehend Medical uses natural language processing (a form of machine learning) to identify the key common types of medical information automatically. Natural language processing (NLP) is a branch of artificial intelligence “that helps computers understand, interpret and manipulate human language” (see Natural Language Processing—What It Is And Why It Matters).
AI and NLP are making their way to other applications in health and human services. We reported on Woebot—a chatbot resembling an instant messaging service that “asks” about user moods and thoughts and uses the answers to “learn” about the consumer. It then offers consumers the use of cognitive behavior therapy (CBT) tools, all while emulating a face-to-face meeting (see Woebot Raises $8 Million For Its AI Therapist). And there are several consumer monitoring applications. In December, the AI health monitoring and consumer wearable organization Myia Labs (Myia) announced it had raised $6.75 in funding for wearables that are attachable to devices such as Apple Watch and Fitbit to monitor individuals with chronic illness or a post-operative condition (see Myia Labs Raises $6.75 Million To Build AI Consumer Monitoring Into Wearables). And there is Mindstrong, which collects data on consumer smartphone activities such as scrolling, typing, and clicking, and then uses that data to identify consumers whose mental health is changing and intervene via telehealth and messaging. Clinical professionals can also view the data 24/7 and are alerted when individuals may need intervention (see Five California Counties Approved To Launch Virtual Therapies Under Mental Health Service Act and Optum Backs Mindstrong Health To Develop Smartphone Data Mining Technology).
I think the question for most executives of health and human service organizations is where to start. These applications seem far removed from most day-to-day service system operations. For more on adapting these important technologies to health and human services, I reached out to my colleague, OPEN MINDS Senior Associate David Young, who wrote:
The one big concern in the AI field is that “how” artificial intelligence and machine learning works is often very difficult to explain and somewhat of a black box. I am reminded of Arthur C. Clark’s famous quote “Any sufficiently advanced technology is indistinguishable from magic.” AI falls into that category for most folks right now.
To understand where AI and NLP fit, he said it is important to answer, “what is the question,” noting:
The reason AI is coming to the forefront now (the concepts that are found in the math are quite old) in health care is threefold. The first is we now have huge databases that can be accessed in some form. The second is that computing horsepower is now at the level that it can process so much data in reasonable time frames. And the third people are beginning to shape relevant question about consumer care, consumer needs, and consumer costs. The “questions asked” are critical since machine learning will only answer the question. If my question is “how can I treat someone with the absolute minimum cost, the answer will be “do nothing.” However, if I ask how to treat at the minimum cost and obtain a low PHQ9 score in 16 weeks with a high patient satisfaction rate, AI will do that and can also give comparisons of the different tracks.
An example that is popping up right now is using NLP to replace the practice of writing everything down. In the very near future clinical professionals will merely talk out loud in an exam room and the AI will use its NLP to electronically document, and extract relevant clinical data to complete treatment plans, prescribe, etc. Then using models like the AWS Comprehend model, it will detect and sort that information to do things like predict costs, consumer compliance, early consumer returns. You name it and the system will provide that information. Despite the unfounded fears, this won’t be expensive. Certainly not as expensive as humans writing everything and getting care wrong. And can be done fairly quickly. One of the biggest barriers of course is the current payment system, but already VBR demands a sea change in thinking about health care delivery.
Mr. Young did have some advice about how to approach applying emerging technologies to improving health and human service organization performance. And, as usual, it’s all in the planning.
To get this done requires that organizations think about data governance early and often. Our currently multi-siloed system of information collection and storage make it so difficult to get that “rich” data that make AI and machine learning work so well. So, for a provider organization executive to start down this road, they would need an organized approach to technology-facilitated performance improvement. The approach has four key elements:
Incorporate AI into strategic planning—What are the overall objectives and performance requirements of the organization? What predictive and prescriptive information would assist in meeting the goals of the organization? How could the use of AI provide this information better, cheaper, and/or faster?
Set clear goals for predictive and prescriptive information—Are there clear questions for AI to solve? Does the executive team know what they would do with this information to improve strategic positioning?
Organize the data around the questions that provide the predictive and prescription information—How is the AI/ML tool part of the larger data system?
Assign responsibility—Who is going to make this happen? (The answer is not the IT department.)
I think this process illustrates a principle that our team at OPEN MINDS uses as the basis for strategy implementation—leveraging technology doesn’t happen in a vacuum. And, technology investments and management cannot be separate from the overall organizational operations. Mr. Young said it very succinctly in our discussion:
AI and machine learning don’t “stand alone.” One does not go out and “buy some AI.” It is one of the tools of a comprehensive data/information system and its purpose is to use all the information you can feed it to answer those questions about the future. That said, AI isn’t that far removed from what we already have in health care. We are always peering at our spreadsheets try to discern the current state and future of our operations. Now we have the data, the computer power and the algorithms to find the future in the data at incredible speeds.”
So, to move your organization “beyond spreadsheets” in performance management and operations optimization, check out these resources in The OPEN MINDS Industry Library:
- Digital Transformations Demand Digital Dexterity
- Your Digital Tech Integration Checklist
- ‘Productizing’ Services For Competitive Success
- Thinking About Partnering With A Tech Start-Up?
- Failure To Launch
- Yes, There Are Organizations Using Augmented Intelligence
- Ready Or Not, Cognitive Computing Will Change Your Organization
- Crossing The Digital Health Chasm
- Myia Labs Raises $6.75 Million To Build AI Consumer Monitoring Into Wearables
- New Tool Predicts Life Expectancy For Long-Term Care Users; Goal Is To Facilitate End-Of-Life Planning
And for even more, join OPEN MINDS Senior Associate Joseph P. Naughton-Travers, Ed.M. for his Executive Seminar, Making The Right Tech Investments For Your Organization: An Executive Seminar On Technology Budgeting & Planning, on February 13 in Clearwater, Florida (for a preview of the seminar, see How To Make The Right Tech Investments For Your Organization: An OPEN MINDS Executive Seminar On Technology Budgeting & Planning). And don’t forget to mark your calendars for The 2019 OPEN MINDS Technology & Informatics Institute on October 28-30, 2019, at the Loews Philadelphia Hotel, in Philadelphia, Pennsylvania.