Leverage humanity: How the deepest learning is when technology is a tool, not a panacea

 

Man and the machine

Upon graduating from high school, without any particular honors or distinctions, my general practitioner, Almeda Decker, DO, assured me that I did not want to make a living sweeping floors in my dad’s stores. Subsequently, she got me a job at the Chicago College of Osteopathy.
 
Albert Kelso, PhD, my new boss, was an esteemed academic, the institution’s director of research and chair of the school’s Departments of Physiology and Pharmacology. My first task was to assemble a dozen Heath Kit electronic switches, the design of which had been modified by the department’s two engineers and used with polygraph equipment in the teaching labs. My workbench was in the same room with the engineers, whose focus was building a patient scheduling file program that would both index and store lab test results for each patient. The room, however, actually belonged to three Digital Equipment Corporation towers, which required window air conditioners that ran year-round (in Chicago in the 1970s). The guys were very bright though lacking social skills and any orientation in medical nomenclature.
 
Moving on to the University of Illinois as a freshman undergraduate, I took a course in computer science, a requirement-fulfilling, pass/fail elective — a dream come true for an aspiring pre-med student who had no desire to spend time in the home of HAL 9000 from director Stanley Kubrick’s “2001: A Space Odyssey.” There, I punched enough data cards to pass each assignment, “learning” enough Fortran IV and V and whatnot to get by. Upon graduation in large part as a result of maintaining summer employment there, I was accepted to the Chicago College of Osteopathic Medicine. In a few years, I was family doctor in Columbus, Ohio.
 
In the mid-80s, I got my first PC at a cost of $5,500 and equipped with a program called Enable, which had word-processing software and a calculator program. I also purchased a telephone modem and a subscription to America Online. I was now connected to the world. I was not then aware of Jerrold S. Maxmen’s 1976 book, The Post-Physician Era: Medicine in the 21^st Century, ¹ or Philip C. Anderson, MD, who in JAMA likened the work to lacking the intellectual insight of a monthly science fiction magazine for its “cultish forecast” of healthcare in the 21^st century being provided by computers and paramedics.²
 
It would be another several years before there was any significant connection of my medical practice with computers. When Medicare created a format that allowed electronic submission of claims, the number of vendors exploded. Soon, competition for combining word processing, office notes and billing, juiced by regulatory mandates and federal financial incentives, birthed the foundation for a new industry whose products were not functionally ready for prime time. Early EHR systems were built as audit and compliance tools to satisfy billing, coding and documentation requirements or challenges to validate payment. With physicians and hospitals armed with these tools, Medicare’s mantra became, “if it’s not documented it wasn’t done — and we’re not paying for it.” Needless to say, the plaintiff lawyers loved it.
 
While most electronic systems manage to be interoperable with CMS and commercial insurance companies for submitting, processing, paying and denying claims, interoperability among those disparate systems to support the vision of improved care, communication and patient safety are yet to be realized. Health systems are now making nine- and 10-figure investments to convert existing IT and IS systems to a common and interoperable platform. The coupling of HIPAA’s intention to protect patient privacy with vendors’ profit motives, communication across health systems and platforms remains problematic.
 
For all its shortcoming, the electronification of health records has created a cesspool of data. CMS has abused its unique privilege and access to claims data to manufacture distorted, “black box-generated” performance benchmarks, rather than those that are clinical and evidence-based, to support an array of payment-reduction, incentive and alternative payment options. Third-party payers of all types have applied data in similar fashion to define outlier behaviors related primarily to resource utilization as a self-serving surrogate for quality.
 
Cynicism aside, data and analytic tools — the latter trailing the needs created by the former — are being developed and applied to facilitate evidenced-based assessments of best practices. Sepsis is an applicable example. Even armed with data and evidence, physicians who wish to assert control over the art of medicine are slow adopters of order sets. In their defense, many algorithms and analytic determinations are not, in fact, based upon evidence, but instead are founded upon pathways that achieve a desired (sometimes predefined) outcome, such as cost.
 
In a 2018 presentation, Siddhartha Mukherjee wonderfully illustrated how, without artificial intelligence (A.I.), the breakthrough of the human genome would be as functional as a room full of encyclopedias: We would have it yet not possibly gain any insight beyond the mapping. ³
 
We have all this data, and people and places of different motives trying to apply it to their own purpose — with assumed good intent. But back to the sepsis model cell: A patient can now check in to an emergency department, have an initial screen and labs drawn, and the machine can assemble those inputs and trigger an alert. The call to action can range from beeper notifications to the launch of a cascade of evidence-based, best-practice interventions (e.g., more tests, drugs, treatments). When ideally applied, the process may improve clinical outcomes, including avoiding harm and saving of lives.
 
The U.S. healthcare system is in a bad place. One area of consensus is that it’s too expensive, supported by an annual spend of $3.5 trillion, or $11,000 per person. Since 2009, the value of technology has been overstated, with the delusion that technology would improve care, cost and outcomes. After all, the practice of medicine is an intrinsically human endeavor, with patient engagement and activation critical to outcomes. Automation and electronics — be they EHRs or wearable devices — don’t make anybody better; they’re simply tools that, hopefully, help to produce a desired result.
 
Our processes and measures are out of synch with aspiration and mandates to reduce cost and improve the health of our population. Physician compensation and provider revenues continue to rely upon volume of services rendered. RVUs are a CMS-defined, payer-neutral measure of physician productivity and remain, in most cases, the basis for their compensation. There are no RVUs for managing care without seeing, touching or doing something to a person. Bond ratings important to fund and finance “innovation” and operations of health systems continue to emphasize revenue and market growth, which are consumption- and volume-driven. As Adam Davidson wrote of the early days of GM, “[t]he secret of the corporation’s success, however, was that it generally did not focus on truly transformative innovations. Most firms found that the surest way to grow was to perfect the manufacturing of the same products year after year,”⁴ — so is the case in healthcare.
 
Physicians are working more hours and providing care to fewer patients as they spend time with their computers, completing fields and boxes on checklists and clicking through “best practice” advisories and alerts. Physician burnout is a national epidemic because physicians are increasingly dissatisfied with their work. The literature informs us that physician engagement has a direct correlation to the patient experience — and that the patient experience has a direct correlation to clinical outcomes.
 
In a recent issue of the New England Journal of Medicine, Drs. Rajkomar, Dean and Kohane state wrote: “EHRs have improved the availability of data. However, these systems have also frustrated clinicians with a panoply of checkboxes for billing or administrative documentation, clunky user interfaces, increased time spent entering data, and new opportunities for medical errors.”⁵
 
Physicians’ primary task has become completing “the note,” rather than communicating with the person for whom they’re providing care... the person behind the computer screen. Physicians are enslaved by EHRs and charged with translating patient stories to binary form. Lost are the intellectual and intuitive functions of the physician and care team. The patients’ symptoms and stories are nuanced and colored — translation to binary input of EHR formats loses context, meaning, relevance and value: Garbage in, garbage out. Time on a computer clicking and entering data reduces time talking with the patients and negatively impacts interaction and understanding. The result? Less time and opportunity for insight and increased opportunity for more errors; misses as we convert to the unstructured verbiage, body language, nuance and color of conversation to structured data and entry fields which, in turn, feed our flowsheets and trigger myopic BPAs. Lack of integrative thinking. The larger consequence is the disconnect that occurs among the patient, care team, and physician. In fact, the contagion only serves to deteriorate the relationship among care team members.
 
While I am not going to attribute all woes to the EHR, it is a major contributor. Physicians are killing themselves at the highest rate of any profession in the United State — physician suicides are almost three times as prevalent as the general U.S. population. Trends indicate that physicians are retiring earlier and those in practice are becoming increasing disengaged, as measured by standardized surveys.
 
It’s a sobering fact that appears to be especially true among primary care physicians. All the while, evidence tells us that people with established primary care relationships have higher value care outcomes. Primary care panel sizes — the number of people whose care a physician manages — have gone from 3,500 in my day to studies showing that today’s figure ranges closer to about half that. It’s even lower in the VA system and still lower among concierge and direct primary care practice settings. Among the early promises of the Patient-Centered Medical Home (PCMH) concept was that team-based coordinated care was going to deliver higher value to a larger population of people. That model had the dream smashed from it and life sucked out of it by the EHR. The negative impact on patient access and physician productivity is staggering.
 
All of this occurs in the background as patients, now consumers, are expecting and demanding access and accessibility comparable to banking, airlines and other industries. It adds burden to both the EHR and tool platform, which is not conducive to workflow efficiency or the interpersonal connectivity that healthcare requires.
 
A recent study reported in JAMA Network Open demonstrated that targeted usability enhancement to the EHR system appeared to be associated with better physician cognitive workload and performance. The finding suggests that next-generation systems should strip away non-value-added EHR interaction, which may help physicians eliminate the need to develop their own suboptimal workflows.⁶ We must insist that it happen now.
 
The path forward is complex and must be addressed in multiple dimensions. Technology needs to improve. Will future technology be replete with the ability to transcribe and translate the patient/physician interaction and, in lieu of Alexa or Google developing targeted ads and applying advanced intuitive logic, add the requisite color to the interactive input to optimize care algorithms that augment the physician? A recent Syneos Health Communications’ publication7 states: “A.I. will help put back the humanity in healthcare by allowing physicians to focus on the patient versus drowning in data... A.I. supports and complements a human physician.”⁷
 
However, it doesn’t end there. Technology alone neither created the problem nor represents the singular solution. Depression, disengagement and isolation represent an epidemic among care providers. We, in turn, need to restore civility and humanity — not just between patient and physician, but among all care team members to understand that their role in care delivery contributes to a civil, collegial world. We all should strive to be happier, more satisfied and able to support each other, in sickness and for health.

Arrested development

EHR and big data have not realized error or workload reduction — nor has it cured cancer.

A.I. and machine learning (ML) are tools that provide the opportunity to improve health care. The EHR, as a component of the workflow, is a barrier. Ideally, an EHR provides a platform to identify care gaps, as well as the opportunity to interact directly with patients as people. Broader use of digital tools supports activation and communication and education, but they’re only intended as tools to augment human interaction — and let’s not confuse “interact” with “interface.” People interact, computers interface. In an ideal world, humanity and technology integrate.
 
To borrow a point from Swiss designer Yves Béhar, that integration of technology should serve human needs and be a tool, rather than it being an end in itself.

Value of intuition, compassion is not dead

Consistent with Steve Jobs’ “bicycle for the mind” metaphor, in which he compared the efficiency of A.I. to that of a human being powering a bicycle, human-driven technology leverages the human intellect —the point being that computers will only take us as far as our own intellect allows.
 
No industry holds greater potential for the application of A.I. than healthcare. It generates massive of amounts of data that, if properly analyzed, could have enormous implications for costs, health outcomes and human longevity.
 
Healthcare A.I. may turn out to be the ultimate example of the late researcher, scientist and futurist Roy Amara’s law: We tend to overstate the effects of technology in the short run and underestimate the effects in the long run. Regardless of those results, IBM’s Dr. John E. Kelly III expressed the view of many that A.I. isn’t going to replace the judgment of medical professionals: “It is always going to be human plus machines making the decisions.”⁸

While data scientists and clinical research plan for Big Data to formulate clinically significant and actionable applications that support health outcomes, operations teams need to passionately combine common sense and investment to develop an intuitive interface and functionality of various inputs to include those mediated between physician and patient. Front-end enhancement will support the aspiration of the Triple Aim, as technology effectively augments the physician and care team in translating and applying human elements to assure patient-centered outcomes.

Dr. Joseph Smith said it best:
Health isn’t digital. Not at all. But digital healthcare makes perfect sense. Healthcare is a human construction made up of a series of decisions, interventions, and outcomes, based on insights, values, and options. The options are finite, the choices are discrete, and the outcomes are often binary.⁹
 
Personal computing came bound with the (vacant) promise to improve care and enhance safety, yet the patient — the very reason for our profession and the epicenter of the healthcare quarrel — is effectively removed from the equation. It’s no wonder that patients often find that their appendix is often easier to remove than communication barriers with the care team. While EHRs were the elixir for audits, coding, and billing, it was far from the panacea for care or safety. Old-school processes and procedures continue to win the day — gather data, study best practices, invoke A.I. and augmented reality, and neural networks that learn from observational data. In other words, deep learning by and for humanity.

Notes

1. Maxmen J. The Post-Physician Era: Medicine in the Twenty-First Century. New York: Wiley- Interscience Publication, 1976.
2. Anderson P. “The Post-Physician Era: Medicine in the 21^st Century.” Journal American Medical Association 1977;237(21):2336-2337.
3. Kauffman Hall Leadership Conference; October 18, 2018; Chicago, IL.
4. Davidson A. “Welcome to the Failure Age!” New York Times Magazine. Nov. 16, 2014. Available from: nyti.ms/2pj6II0.
5. Rajkomar A, Dean J, Kohane I. “Frontiers in medicine: Machine learning in medicine.” The New England Journal of Medicine 2019;380(14):1347-1358.
6. Mazur LM, Mosaly PR, Moore C, Marks L. “Association of the usability of electronic health records with cognitive workload and performance levels among physicians.” JAMA Network Open. 2019;2(4): e191709.
7. “Artificial Intelligence for Authentic Engagement – Patient perspectives on healthcare’s evolving AI conversation.” Syneos Health Communications 2018
8. Fan S. “Forget humans vs. machines: It’s a humans + machines future.” SingularityHub. Oct. 14, 2015. Available from: bit.ly/2poummD.
9. Smith J. “Why health isn’t digital.” Forbes. April 27, 2018. Available from: bit.ly/2Wsqr4o.