VA Cerner EHR adoption: time to let go or innovate?
3/19/201 “After a rigorous review of our most-recent deployment … it is apparent that a strategic review is necessary. VA remains committed to the Cerner Millennium solution, and we must get this right for Veterans” - Secretary Denis McDonough
Over three years ago, the Veterans Affairs (VA), awarded Cerner with a contract to replace its current electronic health record (EHR) with Cerner Millennium. Three years and multiple articles criticizing the EHR adoption later, the opening quote by McDonough was sent out.
Leading to the statement above, press releases from the VA’s Office of EHR Modernization were generally optimistic, making this recent statement seem out of left field. Apparently, the EHR rollout is not going as planned and issues of infrastructure and staffing deficiencies need to be reviewed. No prior indication from the website hinted at this, but looking at different sources, a more critical and pessimistic narrative was being told.
What did these sources say?
“I am hearing an increasing number of complaints and pleas for help coming into my Spokane and Colville offices. They typically concern the same issues: prescriptions, the new patient portal, and training. I am getting reports of veterans not receiving their prescriptions when needed or receiving the wrong prescriptions. I have one report of a VA doctor ordering a veteran two medications, but he received 15 erroneous medications. I have multiple reports of prescriptions being delayed, which in one case caused a veteran to suffer withdrawal...These problems seemed to be caused by VA staff struggling with ordering medications in the system as well as veterans and their families being confused by the Cerner patient portal. This patient portal is not only unfamiliar but less functional than the old VA patient portal. Nurses who go to work every day to serve our veterans should not be driven to tears because software, which was intended to be an improvement, makes their jobs more difficult.”
“GAO is making two recommendations, including that VA should postpone deployment of its new EHR system at planned locations until any resulting critical and high severity test findings are appropriately addressed.
“We should have been far more transparent,” Clancy said about the initial cost estimate for the program.
“Given this month’s Inspector General reports and what we heard at today’s hearing, it’s clear we need a full accounting of all costs associated with VA’s EHRM project...I’ve heard some concerning information about the state of the EHRM project, and I’m worried that the total cost estimate was vastly underestimated by the previous administration.”
Was the project doomed from the start?
Digging deeper, it appears that Cerner was chosen out of convenience rather than a needs and compatibility assessment. The Department of Defense (DOD) currently uses Cerner Millennium, which would theoretically materialize the EHR modernization vision of the interoperability between the VA, DOD, Coast Guard, and community partners.
Unfortunately, reports from the DOD Cerner rollout display a hazardous outcome. It was reported that some providers quit due to fear of patient harm and possible death due to the flawed system. The Politico article goes into more detail worth reading to truly convey the outcome of the DOD cerner implementation--long story short, it was not good.
The DOD experiment was actually the trial phase for Cerner with plans for further deployment to the VA. Looking at these reports, a case can be made that the project was doomed from the start, but as a Tester in the Politico article stated, projects like these are “too big to fail”, and members of the team are unable to “freely speak” to express concerns.
Before I propose my two solution scenarios, let’s list issues indicated from the stories above:
Decision making not in line goals
Inefficient communication between involved parties
Incompatible workplace culture
Looking at the surrounding issues, one may say the efforts by the VA to “commit” to Cerner is a case of sunk cost fallacy--in order to avoid the outcome of the project being a failure, more effort and resources are being spent. Entertaining this idea, some validity to the argument can be made. So far, the budget increased by 6 billion and counting, as reports stated initial budgets underestimated cost and an additional 1.2 billion was requested following recent reviews for infrastructure development.
When looking at a cost benefit analysis, cost for the program is increasing while the project has not gotten any closer to meeting the proposed benefit that is:
“...the delivery of quality health care to Veterans, enhance the provider experience and promote interoperability with the Department of Defense and community care providers.”
Stopping the project prevents further misuse of resources and allows the VA to pivot to a solution that meets the proposed goals. The benefit to this, besides avoiding further cost, is resources can be averted to another solution. If so, this 16 billion dollar experiment (actually 36 billion if you add the DOD budget) creates a learning experience which could make future efforts more successful; that is if the right lessons were learned.
But, the sunken cost fallacy would not be a thing if it was not so pervasive. Like everyone else, I am also thinking, “16 billion is a lot to spend and just walk away without a product.” So, let's see if an alternative solution is viable.
Oxford defines innovate as: make changes in something established, especially by introducing new methods, ideas, or products
Based on the information available, it is difficult to say the EHR project is a failure, because there are many unanswered questions. Hence, I will propose approaches and ideas that intentionally expose the truth as to whether this is an issue of incompatibility between Cerner’s capabilities and the needs of the VA, or an implementation flaw. First, a refresher on the objective of the EHR program.
Known goals and objectives for the EHR:
What does it accomplish?
Provide a unified EHR software
9 million veterans
Qualified family members
Who will use the EHR program?
VA, 60,000 community care partners, DOD, and coast guard
Who is involved in the development?
1,200 subject matter experts, 69% from practicing veterans health administration providers and other staff
Members of the VA central office for organizational alignment
We start here, because if a team is not unified in their mindset, confusion and waste occurs. Seeing that this is largely a user experience effort, key questions regarding the customer experience and provider experience must be central throughout the development process. Whether it is a team meeting, deciding on user interface, or data security, the question of “how does this improve the user experience?” needs to be constantly referenced.
A problem I see with the current approach is the emphasis on the technical aspect of the program. Yes, this is a highly complex and technically challenging process, but what good is solving that problem if it does not address the central goal that is improving user experience? I would argue it is wasteful due to the large unit of effort expended without resulting in a meaningful change to the customer experience.
This is an approach backed by the government emphasizing the need for government programs to meet customer satisfaction.
Next, the workplace culture must encourage radical progress and truth seeking. While the general consensus is that government programs are slow and ineffective, efforts have been made to address the issue with some success (see the work of Todd Park with HHS).
I’ll take this process further using Ray Dalio’s “idea meritocracy.” The definition of an idea meritocracy is “a decision-making system where the best ideas win out.” A departure from hierarchy-based decision making.
For an idea meritocracy to work, two practices are needed:
Radical Truthfulness: “...not filtering one’s thoughts and one’s questions, especially about problems and weaknesses.”
Radical Transparency: “...giving mostly everyone the ability to see mostly everything. To give people anything less would deny them what they need to form their own opinions about what’s happening around them … Radical transparency reduces both harmful office politics and bad behavior because these things tend to take place behind closed doors rather than out in the open.”
For three years to pass, and a review to essentially state, “it is apparent we need to go back to the drawing board“, is representative of a workplace culture that lacks adequate transparency and truthfulness, which statements from the above mentioned. Having a culture where everyone has the ability to voice their ideas and concerns and having the “best” ideas win improves the likelihood of efforts reaching the truth.
With the established perspective and culture in place, a lean approach could prove effective in meeting goals due to the novelty of the task.
Before going into the lean principles, let's start with key questions that will help organize and prioritize efforts. Remember, the goal is to come up with productive answers that meet the goal of user satisfaction:
How is quality measured?
What is the minimum standard for quality?
What level of quality is adequate for regional rollout?
What does optimal quality look like?
What is the time frame to reach optimal quality?
What level of quality is adequate for national rollout?
When will issues in quality be addressed?
How will these issues be addressed? Individual feedback? Departmental feedback? Real-time vs. delayed feedback?
With these questions in mind, let’s look at some lean principles and possible applications to the problem at hand.
Cross functional teams
Creating teams from different departments that are interoperable (like the goal for this EHR product) allows for free flow of communication and collaboration. Two benefits to this approach are increased responsiveness as requests no longer have to travel between silos, and concurrent feedback during product design which reduces back and forth iterations of the product.
Validated learning is a process of systematically answering questions through intentional experiments. These experiments are designed to answer questions on quality improvement. This process must also occur in real time as post-hoc analyses, like the beginning statement, are haphazard and lead to the situation at hand: moving forward with a project only to realize you are not getting the answers you were looking for.
Minimum viable product (MVP)
The problem with seeking perfection before execution is that the work is perfected mostly using internal feedback, putting the product at risk for irrelevance when it is exposed to external use, and increasing cost. A MVP prevents this by putting out a working version of the product to be tested. Restructuring the rollout to use a MVP allows concurrent external feedback during the development process and introduces shortcomings early in the development stage.
Concierge MVP and early adopters
Not all customers are created equally, and for a product rollout to be efficient, the team must cater to specific customers instead of trying to meet the needs of ALL customers. In this case, the user base/customers are even more complex as the product must meet the needs of providers, patients, administration, billing, and much more. A concierge MVP is an approach to developing a MVP that uses a more hands-on approach with a smaller pool of participants, which we will call early adopters. (Technically, early adopters are customers who are ready for a solution and will be the first to adopt a product that solves that solution. Here, it is referring to clinical scenarios that are more likely to succeed with the transition).
A criteria for early adopters should have been made where the risk is low but validated learning can occur. An example of a sample cohort could be patients with less than three active diagnoses that are well managed, no hospitalization in the past 2 years, employed, age 55 and lower. These are likely to be patients with low utilization, less medical data to sort through, and adhere to medication. Also, due to relative stability of the patient, adverse events would be less hazardous, as noted from the Spokane rollout. Unfortunately, with a task this large, every player involved needs to play a part. Having increased participation by the patient could prevent some of the issues that have risen with improper medication refills and patient portal access.
This model also allows for adequate time prior to and after any visits to have debriefing sessions to test various hypotheses and promote validated learning. These concierge efforts over the span of a year could have exposed the team to the technical and user operability concerns that arose, while addressing them rapidly. Dealt with in this low risk situation, a MVP arises that is better suited to test the management of higher complexity patients (outpatient), inpatient cases, larger networks, and larger and more rigorous tasks.
What I am proposing is actually not innovative, as I mentioned earlier. Todd Park and many others in the government have and continue to push for these styles of product development and management. With the recent pause on the rollout, the question remains whether the information so far has provided validated learning for the team to move forward with or will they drudge forward in the name of avoiding failure.