Celebrating nine years in business, we have talked to hundreds of physician executives, and we are well-versed in both curiosity and skepticism around clinical data. While physicians pour immeasurable terabytes into EHRs, very rarely are those systems playing back to them both where they are succeeding and how they might practice better medicine. So, when we proudly say we have the most physician-friendly way to deliver helpful, insightful data on clinical variation to physicians, leaders naturally have some challenging questions.
No worries — we have answers!
Let’s take a look at a series of questions we had from one recent diligent physician leader.
Why would a hospitalist buy into this?
This is probably the most important question we receive. This is the promise of value. If we had to give the one sentence answer to that question, it would be “because it will help them become a better physician in less than a few minutes per week.”
We know that nothing is more important to physicians than delivering the best care possible.
But for a physician to buy in, a solution must not only deliver the goods, it must be extremely easy to use, accurate, and almost irresistible. The characteristics below help explain why over 80% of physicians access our tool monthly, with 70% of the process measures showing statistically significant improvement. No other physician engagement and change tool can boast these statistics. Here is why our solution works:
- Physician-friendly design: Physicians receive one text a week at a time of their choosing. There is no need to login, and if they do need to login to review case examples, we leverage SSO to make it easy. The mobile UI/UX is easy to use, taking only a couple minutes to process the information. All in all, the solution is simple, easy, and respectful of physician time.
- Precise, fair, adaptable attribution logic: This is arguably the key and most differentiating part of our data. In short, we use notes- and orders-based attribution logic that pinpoints responsibility and agency, commensurately varies for each metric, and is adaptable to local facilities and goals. Learn more about our attribution approach in the video below.
- Curiosity, “collabetition,” data-driven clinical learning: By using unblinded peer comparison on a wide range of process metrics, physicians are naturally curious as to where they fall. This creates a sort of friendly, motivating competition, but it also sparks conversations where physicians collaborate to learn how others handle particular processes and clinical decisions. The app contains easy links to external evidence/guidelines, internal protocols, managerial suggestions, and the comparisons alone spark self-reflection like no other solution.
How do you address the attribution questions and the data accuracy questions that are foundational to hospitalist and physician acceptance?
We began to address this above, yet at a more tactical level, nightly clinical data feeds gives us access to all provider actions on and associations to given patients on given days. Each metric in our library has a default attribution model (e.g., ordering, attending as determined by progress notes, authorizing, discharging, etc.) has years of feedback from hospitalists using our product and is easily customized for special or local cases. Prior to a go-live, we have a formal QA process with medical directors for all metrics of interest (typically about a dozen per year) where we check recent patient examples to confirm attribution accuracy and identify changes (e.g., exclusions, provider type adjustments, etc.). Most importantly, we project humility and collaboration, asking hospitalists during go-live to leave comments in the app if they find a patient misattributed to them (and missed in the QA process), and then we fix it.
In the minds of the hospitalists/doctors, how is this project positioned so that it doesn’t become “one more thing the administration is pushing” today?
Agathos projects are positioned as non-punitive, physician-centric clinical learning and improvement initiatives sponsored by clinical leaders and led by site medical directors. Onboarding is onsite and co-led with medical directors. As mentioned above, they all have an invitational and curious tone, and lead with Agathos’ different philosophical approach that anticipates concerns from physicians on data initiatives (e.g., bad attribution, time-consuming, not actionable, financially-oriented). The onboarding materials and in-app messaging emphasize that Agathos is not telling physicians how to practice, but rather giving transparency and tools for self-reflection and improvement while creating team-based forums (e.g., data discussions) where physicians can share amongst (and only amongst) each other how they are doing this or that, and exchanging tips and ideas amongst colleagues.
Our organization has laborists, surgicalists, orthopedic hospitalists, neurointensivists, CVS intensivists, medical intensivists and a bunch of other “ists.” How could we use Agathos for them?
We have options here. For specialized hospitalists, we could either
- Keep them in the same cohort and mark their names/data in some conspicuous way,
- Not include them, while allowing them to view data on different “teams” such as orthopedic hospitalists, or
- Put them in a separate peer cohort altogether. The intensivists would be a separate project and would have the same set of options.
- Heavily preview and caveat any apples-and-oranges dynamics during onboarding for a united cohort (which we do at all times any way, and particularly advised if the above three options are insufficient yet there is demand for peer reference data all the same).
What is the main differentiator/competitive advantage over the data we self-produce?
To our knowledge, Agathos is the only third-party vendor doing mobile-based nudges and direct-to-hospitalist engagement on practice patterns (i.e., outside of the EHR and dashboards, at the level of provider activities, properly and flexibly attributed). Every now and then we run across homegrown tools and systems that do part of our model (for instance, a distributed attribution model for LOS, perhaps a text notification system for off-target stays), yet we have never come across even close the scope of all the process metrics, proactively generated, logically cadenced, without top-down manager intervention. As for results, in short it works — all physician engagement (i.e., >60% weekly, >80% monthly active usership), physician buy-in (e.g., relative lack of negative attribution feedback), and practice change (>70% of client-insight pairs spanning platform have significant improvement).
All cards on the table, we even have a “reverse case study” after non-administrative users stopped receiving insights yet the data feed and app (i.e., a nice if behind the scenes dashboard) were kept on for both administrative and frontline clinical users for a year. Practice patterns reverted, there was effectively no engagement from non-admins, yet medical directors logged in hundreds of times (effectively representing a “best-possible” dashboard comparator). Unsurprisingly, given our hospitalist focus, that “dashboard” was better than anything produced before or after internally, even with more informatics resources available. So, candidly, there just is not a realistic internal substitute, at least as pertains to hospitalist opportunity and need for user engagement and practice improvement.
Why would we not build this ourselves?
In addition to all previous comments, it is worth noting that Agathos’ library of hospitalist insights is the results of eight years of R&D with hospitalist advisors and users across the country in collaboration with data scientists and Epic data engineers. While many organizations have one or more hospitalist dashboards, to build something of similar depth and breadth (again, both data and user experience) to Agathos would take many years and millions of dollars. So far, out of all the organizations that said they would build their own, none have done so. And all of that with all due respect, we are merely in different businesses. The question is ultimately, would you like to see physician engagement and practice improve soon or years from now, and what are you willing to invest in the difference? We grant that dollars are the easy part relative to time, focus, etc. For most, the sooner they deliver higher value care, the better. Agathos projects pay for themselves in a clear business case that we commit up front.
Are there benchmarks already established in the literature or by other companies with whom you compare yourself?
A key axiom to our experience and differentiation (relative to some other established benchmarking products) is that local benchmarks (same facility, or perhaps facilities in a given region) at the level of process metrics (actions, inactions, decisions) amongst known and relevant, respected peers are more compelling and motivating references than national benchmarks about outcomes. Often our metrics in libraries lack direct guidelines or numerical targets that would be appropriate to visualize, much less manage toward. That said, we are beginning to approach a scale where comparison of at least families of insights (note: local customization of attribution, exclusions, parameters is one of our differentiating features, and thus confounds platform benchmarking efforts) across our platform can be both interesting and helpful. For instance, even if just internally, we leverage platform benchmarks to appraise baseline opportunity and recommend insight send sequences.
Do you have other questions? Responses or rebuttals to these? We would love to hear from you (click below), and thank you for your investment in our mission: organize health data and generate care insights that improve patient outcomes.
About Author
