Three Designed Objects: A Metaphor for Healthcare and Healthcare Data

Doctor as Designer
13 min readJul 11, 2022

--

Look at these 3 designed objects.

They are essential for having a meal, but there is something fundamentally flawed about how they are designed.

The Uncomfortable Pot

Katarina Kamprani https://www.theuncomfortable.com/#

Would you like to drain a gallon of boiling water with this pot?

The Slipper Spoon

Katarina Kamprani https://www.theuncomfortable.com/#

How effectively could you eat your soup with this spoon?

The Chain Fork

Katarina Kamprani https://www.theuncomfortable.com/#

How long would it take you to eat your salad with this fork?

You should be able to accomplish your ultimate goal of eating with these 3 objects, but the individual design of each of the objects make it extremely difficult.

Katarina Kamprani https://www.theuncomfortable.com/#

These objects represent just a few of the multitude of designed artifacts from a collection called the Uncomfortable, created by Athens-based architect Katerina Kamprani. She took a series of regular everyday objects, and then made each of them difficult and annoying to use.

I love this tweet from my colleague Jack Iwashyna, whose description of the design is spot on:

https://twitter.com/iwashyna/status/1533573827854315523

“What is especially great about those three artifacts you picked is that they are brilliantly designed to optimize potentially rational criteria — just not core uses “Easy to pull forward from back burner of stove” “Reduce risk of spoon spillage” “packs well in small space”” — Jack Iwashyna

These 3 objects are a perfect metaphor for the current design of healthcare, which remind me of these three inconveniently designed verticals in healthcare.

The silos of healthcare operations (clinical care), quality, and research.

They all share the singular goal of achieving the triple aim of healthcare (improving the health of populations, the patient experience of care, and reducing costs), yet they are each inconveniently designed to prevent success in achieving that goal, because they each prioritize their own set of motivations, behaviors, and outcomes. As a result, the overall function of the system is compromised and fails to deliver on outcomes, cost, and quality.

As someone who works in all 3 verticals, this is a core dilemma. As described by pediatric colleagues in their 2014 Health Affairs paper:

“In today’s health system, research is done by scientists, improvement is implemented by quality specialists, patient care is administered by clinicians, and management is handled by health care executives. Patients are relatively passive consumers of these services. Communication across these communities is scant, knowledge is siloed, and diffusion of evidence of best practices to achieve good outcomes into clinical practice is unacceptably slow.”

That’s why I embrace the framework of “Learning Health Systems”, which is a systems approach that asks the question: How do we intentionally and optimally design the “system”, so that all of the verticals (clinical care, improvement, and research) can work harmoniously together? How do we ensure that “learning while doing is the default?” and that “the right care is provided to the right child at the right time, every time?” Link.

“Learning Health Systems require the Effective Use of Data from the Electronic Health Record, Quality Improvement Science, Engagement in Collaborative Learning with Patients/Caregivers/Multidisciplinary Stakeholders, and a Virtuous Cycle of Learning so that research discovery can inform clinical care, and clinical care can inform research discovery.”

All of the components above are critical and essential for creating a Learning Health System, but my bias and belief, given what I do for a living as Associate Chief Medical Information Officer, is that data from the Electronic Health Record (EHR) is the linchpin of a Learning Health System.

The foundation of a learning health system starts and ends with the optimal design of data. Without it there is no capacity to measure outcomes, support and sustain improvement, or iteratively learn and make new discoveries.

The concept of leveraging the health record as a source for learning is not a novel idea, as it was promoted and back in the 1960’s by Larry Weed, MD, an academic physician at University of Vermont, who had a vision for the design of information systems that predated the dawn of the computer age. He was the creator of the Problem Oriented Medical Record (POMR), which providers still use today to share information and knowledge about patients in clinical practice. In his New England Journal of Medicine article titled “Medical Records that Guide and Teach” from 1968, he wrote:

https://pubmed.ncbi.nlm.nih.gov/5637758/

The beginning clinical clerk, the house officer and the practicing physician are all confronted with conditions that are frustrating in every phase of medical action. The purpose of this article is to identify and discuss these conditions and point out solutions. To deal effectively with these frustrations it will be necessary to develop a more organized approach to the medical record, a more rational acceptance and use of paramedical personnel and a more positive attitude about the computer in medicine.

You should watch his 1971 Grand Rounds which was just visionary. In it, Dr. Weed asks the fundamental question:

“How can the health record help guide and teach us?”

His answer was that it is in the the way you handle data:

“The practice of medicine is the way you handle data and think with it. And the way you handle it determines the way you think…the very structure of the data determines the quality of the output.” https://www.youtube.com/watch?v=qMsPXSMTpFI&t=2s

However, we have a fundamental problem in how we create, use, and take action with data because of the underlying design of the EHR. Our fundamental problem with the design of healthcare data is the metaphor/mental model that we are are using for the EHR.

I have written multiple times about the power of metaphor and how it influences the design of products and user expectations, whether it’s the design of medications like the Epi-pen for Life-Threatening Allergies or the design of algorithms to support self-driving cars. Let’s talk about the predominant metaphor/mental model of the EHR.

What do you think of when you hear the words “health record”?

You think of sheets of paper stored in a folder, such as the chart that Dr. Weed himself is holding in his hands below! A document that we type our observations into!

Our metaphor for the EHR is a Microsoft Word document, and we have designed our EHR systems accordingly.

We record critical fields needed for healthcare learning in free text.

We capture structured data for fields inconsequential for learning.

Finally, we collect patient-reported outcomes on pieces of paper that get scanned into the media tab into the EHR.

Consequently, this is how most providers feel:

“I don’t exactly know who I am taking care of, I don’t know how my patients are doing, and I can’t tell if anything I do makes a difference in their health.” -Every Provider Ever

Because of this poorly conceived EHR design, current forms of healthcare learning in the enterprise involve extensive chart review

Clinicians are writing so many notes that they don’t have time to learn, improvement experts are performing manual chart review of a subsample of charts, and research assistants are performing high-quality chart review but are gone once the grant expires, and the data is often destroyed once the project is complete.

Fascinatingly, some clinicians, in an effort to gather data for “learning” are performing double data entry, once into the EHR in their free text note, and then a second time into a “shadow EHR” manually inputting data into Microsoft Excel sheet saved on the shared drive and or a Redcap Research Database, which is a significant amount of human labor and inefficient for data collection, processing, and storage.

We need a new metaphor/mental model for design of the EHR which is NOT “note-centric”. We need to move towards a model of “Data in once/Used many times”.

Such a model has been successfully engineered by colleagues from ImproveCareNow, the international quality collaborative focused on inflammatory bowel disease. Because we now have EHRs, we have access to provider-facing tools like flowsheets, smartforms, and episodes of care, access to patient-facing tools like questionnaires and patient-entered flowsheets that can be completed inside of the patient portal, and we have the ability to utilize structured data captured from these tools to deploy clinical decision support and gain insights at a population level with analytic dashboards to improve care for our patients (which is why I love Tableau!)!

But I also want to acknowledge the ultimate conundrum with healthcare metrics, which has been best articulated by Seth Godin:

“Measurement is fabulous. Unless you’re busy measuring what’s easy to measure as opposed to what’s important.” -Seth Godin

I don’t doubt that approaches like Natural Language Processing will be critical for gleaning healthcare insights in the near future, but I also want to acknowledge that poorly designed EHR tools for documentation lead to poor data quality, and garbage in leads to garbage out. We need to capture key clinical insights relevant to patient outcomes from providers in structured data format to be able to learn. We therefore embarked on a redesign of our EHR tools and workflows at the University of Michigan to bring the vision of a Learning Health System for Type 1 Diabetes to reality.

The Opportunity in Type 1 Diabetes

To give you some background, across the US, the healthcare enterprise has not been successful with helping our patients achieve the lofty American Diabetes Association goal of having a Hemoglobin A1c<7.0%. Currently only ~10% of pediatric patients with T1D are achieving that goal, and there are significant disparities in these outcomes by race/ethnicity and socioeconomic status, as documented by colleagues in the T1D Exchange.

Because of the landmark Diabetes Control and Complications Trial (DCCT) published in 1993, we have known for almost 30 years that intensive therapy (frequent blood glucose monitoring and three or more insulin injections daily) delays the onset and slows the progression of diabetes complications (retinopathy, nephropathy, and neuropathy) among individuals with type 1 diabetes. However, few diabetes centers were recording these and other key metrics for diabetes self-management associated with improved glycemic outcomes.

Use of a Quality Improvement Framework

We used a quality improvement framework to inform our EHR redesign, as we have been collaborating with colleagues from the @T1DExchange Quality Improvement Collaborative since its inception.

https://t1dexchange.org/quality-improvement/

Creation of Provider Tools to Capture Evidence-based Self-Management Metrics and Other Key Data

We formulated a parsimonious set of evidence-based self-management metrics associated with optimal glycemic outcomes in type 1 diabetes, based on the literature, what we call the “six habits”, which could be captured as part of a low-burden EHR workflow at every visit.

(1) Checks glucose at least 4 times/day or uses continuous glucose monitor (CGM);

(2) Gives at least 3 rapid-acting insulin boluses per day;

(3) Uses an insulin pump;

(4) Delivers boluses before meals;

(5) Reviewed glucose data at least once since the last clinic visit;

(6) Changed insulin doses at least once since the last clinic visit.

We created provider “flowsheets” to capture the “six habits” at each clinical visit in structured format, and then included additional elements important to all stakeholders to provide key information that would support operations, quality, and research, such as additional quality metrics for the T1D Exchange, metrics for the US News and World Report Survey, and Nursing Magnet and Billing Requirements.

Creation of Patient Tools to collect Patient-Reported Outcomes

We converted our patient questionnaires to electronic questionnaires in the portal and tablets in clinic which would automatically be administered at specified intervals (every visit vs yearly). We did a Marie Kondo, removing a significant number of patient-reported questions that were not being used by the team or the institution, and we added patient-reported outcomes that are important for capturing all aspects of well-being in diabetes care including additional psychosocial domains such as Self-Efficacy, Parent-Child Conflict, Problem Solving, Fear of Hypoglycemia, Peer Support, and Diabetes Distress, and conducted user testing with stakeholders.

Integration of Provider and Patient Tools into the EHR Workflow

As everyone knows in the Health IT world, it’s all about workflow, workflow, workflow. We made it a team based workflow (CDE/MD/Psych/SW) and we placed the six habits at the top of the flowsheet as they were the most important data elements for measurement. We asked providers to focus their workflow on the flowsheet, rather than the free-text notes, to support structured data collection of these key evidence-based metrics.

With the team-based EHR workflow, notes become a byproduct of structured data collection

We created a simple smartphrase so that with one typed phrase, providers could bring in all the relevant data elements recorded in the flowsheet and questionnaires into notes for MDs, nurses, and dieticians, which reduced the burden and time of documentation, and allowed us to start learning from the data.

What did we learn from the data captured from this EHR redesign?

We published the findings in JAMA Network Open - Feasibility of Electronic Health Record Assessment of 6 Pediatric Type 1 Diabetes Self-management Habits and Their Association With Glycemic Outcomes. (Blogpost Link)

We found that individuals who perform habits 1–6 have significantly lower HbA1c (Blue Bars) than those who do not perform the habit (Grey Bars).

6 key diabetes self-management habits: (1) checks glucose at least 4 times/day or uses continuous glucose monitor (CGM); (2) gives at least 3 rapid-acting insulin boluses per day; (3) uses insulin pump; (4) delivers boluses before meals; (5) reviewed glucose data since last clinic visit, and (6) has changed insulin doses since the last clinic visit.

We also created the total habit score by summing the number of habits performed for each individual. We found that each 1-unit increase in total habit score was associated with a 0.6% decrease in HbA1c.

Total Habit Score=Sum of all 6 key diabetes self-management habits per person.

We now had clear, actionable metrics associated with glycemic outcomes that we could continuously measure and tackle in QI interventions to improve the health of our patients. Furthermore, we identified greater opportunities to achieve health equity in diabetes outcomes, as we found that that the association of the six habits with the glycemic outcomes was similar across race and insurance subgroups.

The design of the EHR to capture structured data and reduce documentation burden now provides the opportunity to standardize clinical care and focus on the most important self-management habits, to design and develop interventions to improve adoption of the habits in the population, and to use what we learn to support observational and interventional research!

I am excited to report that these metrics are on the roadmap for the @T1DExchange quality collaborative so that every center can measure them in a systematic way, to inform improvement at a national level.

As I have tweeted before:

Here are my Design Recommendations for the EHR

  • We need to identify evidence-based, clinically meaningful, and parsimonious structured data elements that are critical for operations, quality and research; we need to
  • We need to deemphasize the note as the center of the workflow; notes should be the byproduct of structured data workflows
  • We need specialty-specific standardization of data definitions and elements across multiple institutions that can be validated against clinical outcomes
  • Finally, we need EHR Workflow Tools that can be automatically uploaded and implemented with low burden once we as a provider enterprise decide on the key metrics.

I want to acknowledge that there is a lot of overarching redesign that needs to happen with the EHR; a number of professional organizations are trying to address this as a systemic issue to reduce the burden of EHR clicks and documentation on providers.

We can’t ask doctors to become data entry clerks for the EHR but there are key clinical and evidence-based metrics that require clinical judgment and can’t necessarily be inferred from the notes. We should take advantage to reduce the note writing burden of the providers and to increase opportunities for learning. Otherwise we are doomed to do chart review and stymie our efforts to do continuous learning for the rest of eternity!

Dr. Weed was very prescient in predicting the ideal future of the EHR. As he stated in his 1971 Grand Rounds:

“This record cannot be separated from the caring of that patient. This is not something…the practice of medicine here and the record over here. This is the practice of medicine. It’s intertwined with it. It determines what you do in the long run. You’re a victim of it or you’re a triumph because of it. The human mind simply cannot carry all the information about all the patients in the practice without error and so the record becomes part of your practice.”

-Dr. Larry Weed

Let’s move away from the metaphors of the 3 awkwardly and uncoordinated designed objects and “Microsoft Word for the EHR” in healthcare, and let’s move towards more coordinated design of information systems to support and improve outcomes for our patients across the operational, quality, and research divide!

I tweet and blog about design, healthcare, and innovation as “Doctor as Designer”. Follow me on Twitter and sign up for my newsletter.

Click here for information about creative commons licensing. Disclosures: Medical Advisory Board of GoodRx, Consultant: Tandem Diabetes Care.

See previous posts about EHR and data design and the role of metaphor in healthcare design:

I want to acknowledge the many individuals whose support and efforts are critical to the work I do as Associate Chief Medical Information Officer for Pediatric Research, Associate Chair, Health Metrics and Learning Health Systems, and Associate Director for Informatics and Clinical Research Innovation, Caswell Diabetes Institute (CDI) (Ranjit Aiyagari, Donna Martin, Martin Myers, Rich Medlin, Dan Stanish, Emily Dhadphale, Ashley Garrity, and the Pediatric Diabetes Team). We receive funding support from the Department of Pediatrics, NIH (P30GRANT12959224) for the Michigan Nutrition Obesity Research Center, the Elizabeth Weiser Caswell Diabetes Institute, and the JDRF Center of Excellence Grant at Michigan Medicine.

--

--

Doctor as Designer

Joyce Lee, MD, MPH, Physician, Designer, ACMIO, #EHR, #learninghealthsystems, #design, #makehealth http://www.doctorasdesigner.com/