Why we need an overall vision for health IT

Healthcare is a vast, intensely complex ecosystem. In order to make sense off all the variety found in the sector, we can categorize health IT efforts into two segments: organizing the world's medical information, and making the data universally useful and available.

healthcare cloud
Thinkstock

Healthcare is a vast, intensely complex ecosystem. Perhaps it is one of the most complex arenas of endeavor of any. Not surprisingly, the business sectors that create tools needed for this ecosystem are also extraordinarily varied – a quick trip to one of the many health IT and digital health conferences will attest to the overwhelming variety of solutions being pursued.

Is there any way of pulling all this together into a coherent direction? An overall health IT mission statement? Clearly, without a vision for what we are trying to accomplish in the healthcare and health IT space, we will fall prey to inadequate, limited, and partial solutions that fizzle out, waste resources, and do not accomplish what we, as a society, need getting done.

Perhaps an overarching goal for the general direction of health IT, and healthcare generally, may be summed up as follows:

Our goal is to organize all the world’s medical information and make it immediately and universally useful and available.

This overall vision helps segment health IT activity into a variety of efforts.

Organizing the world’s medical information

Central to making the world’s medical information useful is the creation of a Medical Knowledge Graph. This is the result of advanced machine-learning algorithms which can discover patterns in the data that may have been hidden previously. This Medical Knowledge Graph can enable a whole ecosystem of uses for the insights and knowledge. It can be the ‘Intel Inside’ of health IT.

The key challenges with building artificial intelligence (AI) in healthcare lie less in the realm of machine-learning and data science, and more in the limitations of data access. Health data is sensitive, protected (under HIPAA and other laws), and fragmented into institution-centered silos. There are several kinds of health data which need to be brought into a coordinated framework:

  1. Research data, from all medical studies conducted to date. Though there are limitations to the applicability of these findings to individual cases, the insights derived are the basis of “evidence based medicine” that has guided treatment up until now. The weakness of published medical data (e.g., in journal articles collected via PubMed), however, is that the findings are aggregated, not individualized and quantitative. The underlying raw, objective data that fuels the published studies can contribute greatly to the knowledge graph.
  2. Clinical data, from all the different Electronic Health Record (EHR) systems used by clinicians and institutions, also need to be centralized. Historically, this data has  been fragmented, and numerous initiatives have sought to improve secure sharing  of data between institutions. Health Information Exchanges (the more successful ones are vendor-associated rather than regional or institutional ones), the Direct project, and even using blockchain (the technology behind Bitcoin) have been methods of exchanging sensitive clinical information securely between institutions. But none of these activities solve the problem of data fragmentation into silos – they just allow those silos to exchange copies of data better.
  3. Genomics data is increasingly becoming available. DNA sequencing has become faster, cheaper, and more readily available in recent years. It has become commoditized. Pulling all that data into a coherent knowledge graph allows for insights to be built integrating clinical observations and genomics. The use of genetic information will become an increasingly important element in the individualization of medical recommendations, and achieving the goals of Precision Medicine.
  4. Billing transactional data can also supplement the clinical data records. Clinical data, suffering from segmentation, may often be only locally-informed. Billing data from payers, though limited in terms of the kinds of data it contains, is a helpful layer of data that crosses providers, and includes pharmacy-fill data, hospital data, and all other activities for which bills have been sent. The Centers for Medicare and Medicaid Services (CMS) holds a vast trove of data for Medicare recipients; health plans hold similar (though more segmented) data for privately insured enrollees.
  5. Consumer-generated (patient-generated) data is a huge reservoir of information currently existing in vendor-centered data stores. This includes website and smartphone application data where patients/consumers have entered their own information, and also includes device-generated data that alert consumers (via phones, smart watches, custom wearable devices, etc.). This data, when organized in ways that can interact with personal, private medical information (HIPAA-compliant EHR data) can help generate some powerful insights of unprecedented benefit.

Making data immediately and universally useful and available

With all the activity in health IT focused on accumulating and sharing medical information, the next step is to put it all to good use. In healthcare, there are thousands of data workflows addressing allsorts of data and use cases, some byzantine, others quite efficient. But understanding the workflows is necessary in order to leverage the Medical Knowledge Graph in ways that improve efficiency and accuracy, improve patient and provider satisfaction, and lower cost.

There are general research needs for the data, there are population-specific needs, and there are individual patient needs:

  1. Research needs: Using de-identified medical data can propel research in an unprecedented fashion. Discovering patterns in clinical observations, especially when linked to variants in genomic data, can identify risks, help us understand disease better, and point to solutions – some individual, some environmental, some social, and some economic. Health policy at all levels can be informed by data like never before.
  2. Population based needs: As the healthcare system continues to embrace value-based compensation, understanding populations is increasingly important to all healthcare organizations responsible for delivering or managing care at the population level – insurers, accountable care organizations, medical groups, hospitals, and so forth. It involves understanding the current status of the population of patients currently being cared for, and suggesting ways to improve health status. It involves streamlining access to indicated resources, and removing the current “blunt instruments” of authorization that vex all stakeholders in the healthcare system.
  3. Individual patient needs: Clinicians need access to data about their patients, regardless of where that care has been delivered, with specific recommendations for care, based on broadest possible knowledge base. The questions of “what do I need to do for this specific person in front of me to prevent disease or treat symptoms already present” are central to the art and science of care. The most appropriate diagnostic or therapeutic options need to be delivered at the point of care. Clinicians need to be able to communicate with each other about the patient-at-hand in ways that are secure, and informed by all known data. This is a vast area of endeavor, as the workflows of clinicians vary tremendously, based on specialty, preference, location of care, and best practices identified and promoted by professional societies, provider organizations and payors.
  4. Individual consumer needs: In addition to individual clinicians giving advice, patients/consumers also need to use this universal medical knowledge in all sorts of ways. Understanding about disease in a fact-based fashion is important; as we know, there is all sorts of advocacy out there in the world which has no basis in fact. People want trusted facts, interpreted in ways that are useful and understandable. In addition, people want to know more specifically about what decisions they should make about their own care. “Is this the best medication for me?” “What if I don’t do what is recommended – does it really matter?” “This set of symptoms I have, given that I have these genetic factors – what do I make of it? Should I be worried? Should I see a doctor? Who, exactly?”
  5. Device-associated needs: Consumer devices, from smartphones to smart watches, to specific wearables, can all be made much more useful when they are powered by AI that can deliver individualized feedback (as well as add to the knowledge library which the AI uses). Development of medical AI-powered devices is an emerging field which will result in a future that we can only glimpse currently.

***

The world of healthcare and health IT is enormously complex, and becoming more so with each passing year. Many efforts by many companies have visions that are focused on specific niches of need – useful, but perhaps lacking the overall view of where we are going as a sector and as a society. We need a global “mission statement” for health IT, from which we can coordinate our efforts, innovate effectively, and achieve the vision that we have laid out.

I invite you to join me on this journey.

This article is published as part of the IDG Contributor Network. Want to Join?

Time is running our to share your experience. Take the 2019 State of the CIO survey today!