Duality of Purpose
While data collection and information-driven modeling were promised from EHRs, it is folly to assume that they are drivers of the technology. Realistically, EHRs were designed and developed for one purpose: to improve the claims billing process6. Consider the way traditional outpatient data entry unfolds: a set purpose for the visit; vital sign capture and triggered alerts for out-of-range values (note the absence of therapeutic discourse and suggestion related to these inputs); order sets based on evidence-based treatment protocols, focused on billable endpoints (diagnoses and NDCs, not goals, recommended interventions, and resource-rich patient deliverables); and designated follow-up on endpoints and billable timetables, without regard to progress or reassessment.
It is nearly impossible to argue – improved claims billing is the primary outcome driving EHR design and improvement.
How then do we take this technology and help it evolve into a diagnostician’s toolkit instead of an expensive billing construct?
It’s All About the Data
On its own data is not a dirty word, but it’s starting to feel like one. Increasingly accessible data, in copious amounts, has led to the outcry that providers have access to more and should be doing better1,2. The reality, though, is that data, on its own, is utterly meaningless. A standalone snapshot in time, or a set of related but unconnected points offer little insight into how to effectively manage and treat a patient. Not to mention the fact that data, offered in slices across points in time, is truly no different than treating in the age of paper-charting; physicians see numbers, mentally apply gold-standard treatment algorithms, and play connect-the-dots in following established treatment patterns without any respect for individual circumstances and complications. Clearly, there is more to this picture than just the data.
Drowning in Data
The growing number of studies published each year, tides of data generated from EHRs and the Internet of Things (IOT) (wearables like smartwatches and fitness trackers, as well as medical devices, like blood glucose monitors, insulin pumps, and blood pressure cuffs), and limited, finite hours in a day have cumulatively led to data overload5,7. Physicians are surrounded by data, being pushed with more data, and are expected to act on data alone.
But data lacks meaning. Data lacks context. Data is not information. This is the part that almost everyone gets wrong: access to data without contextual meaning and application is pointless. And, without that needed meaning, providers are left to revert to the one thing they can trust: history. So, instead of relying on data, even real-time data, providers fall back on established treatment algorithms and gold-standard therapies; which, ironically, is exactly what those aforementioned billing entities want. While being held accountable for disease management and symptom improvement, innumerable boxes also need to be checked indicating appropriate testing and pharmacologic intervention without regard to patient engagement or circumstances.
In the end, the result is a vicious cycle. Providers aren’t given any discernible insights from EHR data, so they defer to what they know and are comfortable with. This, in turn, falls short of systemic and third-party goals, whose unequivocal answer is: they need more data. More recently, this cycle has devolved into a spiral; namely, the spiral that leads to burnout6. Rest assured, there is a better way.
Data and information are not synonymous. Data, as discussed, is a series of raw, unfiltered and often unrelated points. It can come in the form of words, codes, measured values, and much more. But it is fairly useless without context or a story, a concept missed by most EHR developers.
Providers don’t need more data. They need more and better information. They need improved, informed decision-making tools. And they need all of this designed on the premise of improving outcomes.
Knowing a patient’s HgbA1c today – good, but isolated. Knowing that same HgbA1c in reference to the past 10 values for the patient – helpful, as it creates a story, but still little more than a pretty graph. However, seeing that HgbA1c trend in relation to other patients with similar demographics, comorbid conditions, and disease progression – Gold. Add in the specific details of what treatment plans worked and didn’t work for those comparable patients and now you’re operating in the 21st century of medicine. But digitized insights like this are rare, and information-based, progressive treatment algorithms are even more scarce. So how do we make them the norm instead of the exception?
Predictive Analytics and a Question of Trust
To generate information and insight like that highlighted above, data must be processed. Through the use of advanced data analytics, machine learning, artificial intelligence (AI) applications, and natural-language processing (NLP) tools (to name a few), raw, overwhelming data can be transformed into tangible, actionable insight and truly personalized care. Taken one step further, predictive analytics can model historical data and establish likelihoods of success, failure, prevalence, incidence, and severity of any number of qualitative and quantitative health measures8. Put simply, effective analysis of historical data can help predict the efficacy and success of future applications and endeavors8,9. Why, then, are predictive analytics not more prevalent in medicine?
Three words: empirical scientific data
The premise of evidence-based medicine relies on reproducibility. Almost all facets of Western medicine are predicated in the fact that results are validated through repeated experimentation with consistent, reproducible results. This is almost in direct opposition to the idea that value-based care, in its most effective form, is entirely personalized. In a system that values historical validation, untested and non-generalizable predictions are difficult to trust10.
Ironically, all of the variables that predictive models take into account are not new; genetic variability, biologic response, adverse effects, allergies, and others have always existed. There just wasn’t necessarily a way to consider them in the treatment paradigm. Now that there is, it is understandable that there might be a period of adjustment. However, utilization of predictive analytics needs to continue to drive that period of adjustment until clinicians become more comfortable with their use.
Modeling to Move Forward
Clearly, data itself is not a tool. It is raw material used with tools like deductive and predictive analytics to improve quality of care and health outcomes. So, to stop the overwhelming data cycle providers face, and improve the underwhelming outcomes that ultimately result, systems need to provide clinicians with organized, actionable information.
Clinical insight and information need to be presented in EHRs in a way that serves clinical decision making and not just claims billing processes. Organizations must give clinicians the time, support, and resources needed to connect with patients on a human level, instead of seeing them as sources of data.
Above all, software systems and EHRs must be designed to serve clinicians, and not the reverse. Through AI, predictive analytics, intuitive and user-oriented interfaces, and patient- and provider-centric systems processes, it is possible to turn data into a key pillar of the foundations and delivery of value-based care.
- Dash S, Shakyawar SK, Sharma M, Kaushik S. Big data in healthcare: management, analysis and future prospects. Journal of Big Data. 2019/06/19 2019;6(1):54. doi:10.1186/s40537-019-0217-0
- Shilo S, Rossman H, Segal E. Axes of a revolution: challenges and promises of big data in healthcare. Nature Medicine. 2020/01/01 2020;26(1):29-38. doi:10.1038/s41591-019-0727-5
- Adnan K, Akbar R, Khor SW, Ali ABA. Role and Challenges of Unstructured Big Data in Healthcare. In: Sharma N, Chakrabarti A, Balas V, eds. Data Management, Analytics and Innovation. Advances in Intelligent Systems and Computing, 2020, vol 1042. Springer, Singapore.
- Ambigavathi M, Sridharan D. A Survey on Big Data in Healthcare Applications. In: Choudhury S, Mishra R, Mishra R, Kumar A, eds. Intelligent Communication, Control and Devices. Advances in Intelligent Systems and Computing, 2020, vol 989. Springer, Singapore.
- Akhtar U, Lee JW, Bilal HSM, Ali T, Khan WA, Lee S. The Impact of Big Data In Healthcare Analytics. 2020:61-63.
- Collier R. Electronic health records contributing to physician burnout. Canadian Medical Association Journal. 2017;189(45):E1405-E1406. doi:10.1503/cmaj.109-5522
- Rashid M, Singh H, Goyal V, Parah SA, Wani AR. Chapter 3 – Big data based hybrid machine learning model for improving performance of medical Internet of Things data in healthcare systems. In: Balas VE, Pal S, eds. Healthcare Paradigms in the Internet of Things Ecosystem. Academic Press; 2021:47-62.
- Venkatesh R, Balasubramanian C, Kaliappan M. Development of Big Data Predictive Analytics Model for Disease Prediction using Machine learning Technique. Journal of Medical Systems. 2019/07/05 2019;43(8):272. doi:10.1007/s10916-019-1398-y
- Muniasamy A, Tabassam S, Hussain MA, Sultana H, Muniasamy V, Bhatnagar R. Deep Learning for Predictive Analytics in Healthcare. In: Hassanien A, Azar A, Gaber T., Bhatnagar R, F Tolba M, eds. The International Conference on Advanced Machine Learning Technologies and Applications (AMLTA2019). AMLTA 2019. Advances in Intelligent Systems and Computing, 2020, vol 921. Springer, Cham. https://doi.org/10.1007/978-3-030-14118-9_4
- Van Calster B, Wynants L, Timmerman D, Steyerberg EW, Collins GS. Predictive analytics in health care: how can we know it works? Journal of the American Medical Informatics Association. 2019;26(12):1651-1654. doi:10.1093/jamia/ocz130