Healthcare data is messy and incomplete
If not for the healthcare and real estate industries, fax machines would be about as common as rotary phones. Rarely do you work with research institutes and technology companies where a detailed technical discussion ends with “will you fax me that information?”. This is largely due to doctors’ handwritten notes, prescriptions and other printed documents which are not easily shared. This information can be digitized, but it is not easily accessible as organized and structured computer data. In order to leverage computer systems to analyze and disseminate information efficiently, the information must be correct, well-defined, and complete.
Although tools exist to convert printed and faxed documents into structured data, they are not 100% accurate – especially if the document is not a well-structured form. The field of Natural Language Processing (NLP) is advancing quickly and will help, but like Apple’s Siri or Amazon’s Echo which use NLP, the understanding and conversion of spoken or written text is far from perfect. While the information is held in this manner, understanding a patient’s health and medical history feels more like researching a historical figure through artifacts rather than reading a well-documented biography.
Patient Should Own Their Data
Another issue is that the primary purpose of Electronic Health Record systems (EHRs) is not to create a complete picture of the patient’s health. EHRs focus on organizing and orchestrating the workflow of the healthcare provider that purchases them. Along with documenting the patient’s health status, diagnosis, and treatment for the provider’s use, providers are making notes for themselves, documenting details for reimbursement from insurance companies, and coordinating tasks and additional care. Progress has been made in standardizing the health information and supporting EHR interoperability to create a more complete picture of the patient’s health and medical history across providers. Unfortunately, acquiring patient health information is never effortless and rarely free, leaving the patient without a comprehensive record of their health or healthcare.
Patient’s need to be able to own – not just access – their health data. Data sharing must improve between healthcare providers and patients: it cannot only exist between healthcare providers. A data ownership will allow patients to maintain the complete picture of their health and healthcare.
You are unique, just like everyone else
As patient health information becomes more complete and is augmented by genomic data, a quote attributed to Margaret Mead has never been more relevant: “Always remember that you are unique. Just like everyone else.” This quote strikes me as both humorous and absolutely true. The study of medicine is moving forward toward Precision Medicine, which embraces the specifics of an individual’s physiology and genetic makeup. But this level of analysis and insight into our health and treatment has the potential to become mind-bogglingly complex.
While researchers and healthcare organizations are working to garner more insight from larger populations of patients, the amount of data being generated for and about each patient is exploding. For one, the quality and detail of imaging and diagnostic information is increasing rapidly. At the same time, available genetic information for each individual is enormous and not yet well understood. Achieving the dual goals of a) improving treatment accuracy and b) including more patients to improve insights about the efficacy and side effects of treatments is very complex, and the complexity increases as the volume of data grows.
In addition, the data from patient monitoring systems, clinically accurate wearables, and in-home diagnostic solutions adds another piece to the puzzle. These devices generate data multiple times a day, an hour, or even a minute. Through analysis of this data, physicians can become increasingly predictive, proactive and preventive in a patient’s care. But again, the risk of redundant or less valuable data grows as the number of measurements increase. There is value in this deluge of data, but it is still a deluge. As Nate Silver coined, we have to be able to find the signal in the noise!
Problems only a machine can solve
Now we have arrived at the crux of the problem. The amount of data that can be leveraged to a) clarify diagnosis, b) improve treatment plans, c) increase the precision of treatment, and d) monitor health to predict and prevent illness is immense. Health professionals are very data-driven and try to use the most up-to-date evidence-based medicine, but the speed of advancement in medicine and the volume of data needed to make the most accurate decisions can be beyond the ability of any individual to address in the short time allotted to make healthcare decisions.
Healthcare professionals need insight, not just data! Insight is seeing the blood pressure trend along with the significant outliers – not every data point. Insight is seeing the genomic data which is known to affect the patient’s current medical issues – not every piece of the patient’s genomic data. Insight is visibility to the occurrences of illness in patients with similar symptoms in the last three months in the region that the patient lives – not a listing of number of occurrences of every illness in the world. Insight is visibility to effectiveness of treatments for the diagnosed illness in the last 12 months – not a listing of the clinical trials for potential medications. We must use the data and systems at our disposal to determine the signals that are predictive and dampen the overabundant noise.
The promise of machine learning and specialized artificial intelligence
With new technologies and systems, there is always a learning curve – a time when the effort to realize the benefit is greater than the benefit. But these efforts create learnings that improve the benefit and reduce the effort. Over time, many new technologies move from complex and cumbersome to intuitive and integrated parts of our lives. Websites and smartphones have gone from toys for technologists to ubiquitous parts of our society. Although we may be a long way from the initial efforts in machine learning and artificial intelligence in the 1960s, we are only at the beginning of the transition from complex and cumbersome to intuitive and integral.
The childish misinterpretations of our requests to Apple’s Siri or Amazon’s Alexa obscure the pace of improvements in Natural Language Processing, machine learning, and specialized artificial intelligence. I am not referring to human level, general artificial intelligence like Terminator movies, but am instead referring to the acceleration of systems’ ability to process data, find patterns, and make predictions. Some people may not be excited that a machine is now better at identifying a cat in a picture than a person can. But real benefits are being driven when a system determines from real-time data that a part will fail in the coming days, especially if that part is on a car, train, or plane.
This ability to a) process enormous volumes of data, b) identify trends and patterns, c) find specific data upon request, and d) make predictions about outcomes based on known situations or data is remarkable and potentially lifesaving. The question is not if these technologies will be valuable for healthcare, but when will the technologies be at the level of capability to incorporate them into the healthcare workflow. The question is how to leverage the insight that can be gained from these technologies. The speed of advancement in medicine is too great to allow physicians or patients to absorb or even scan all the research. The personal health data for many older patients is already too vast and dispersed to be found in a single medical record for use in diagnosis and avoid complications from medication interactions. Clinicians need tools that enhance their ability to track, assess, and assist patients. The current systems slow them down when they need to be assisting them.
Augmenting the abilities of the physician and the patient
The best systems and user interfaces become part of our world and integrate seamlessly into our lifestyle. Many people no longer think of a phone, text, or email as technology, but just a tool. Google’s navigation apps are so capable that few of us own, much less carry, maps. When physicians have the probabilities of diagnosis, treatment outcomes, and medication side effects ready to review as soon as the patient’s symptoms and biometrics are updated in their medical record, these health systems will have transitioned from complex and cumbersome to intuitive tools that assist and inform. When patients can self-triage simple medical issues using their own medical records, the technology will become a tool that is a part of life. We are not there yet. But isn’t this the goal…insight, not data?
Converging to Consumer Centric Care – next article
Our next article, Converging to Consumer Centric Care, concludes the discussion of how the emerging medical, societal, and technology trends will extend our knowledge, improve our health, and reduce the cost of healthcare.
Join Our Newsletter
If you would like to get updates on our published articles and get our quarterly newsletter, please sign up at Join Our Newsletter.
by Matt Larsen, Principal, Healthscient
Published on Healthscient.com: October 24, 2017