Skip to content Skip to navigation

Solid Clinical Recommendations

October 23, 2015
| Reprints
Click To View Gallery

Solid Clinical Recommendations


Now that we have broad deployment of electronic health records, how do we make recommendation systems trustworthy?


In a recent article, Peter Bregman described a rapid escalation of anger between a father (Bregman) and his daughter over the messiness of a project that happened to involve child play with sand. He was making the point that parents, and managers in general, should adopt a simple and effective approach to communication by verbally


1.    Identify the problem

2.    State what needs to happen

3.    Offer to help


He provides an example in a business context as follows:


Fred, this presentation made six points instead of one or two. I’m left confused. It needs to be shorter, more to the point, and more professional looking. Would it help if we talk about the point you’re trying to make?


Absent this kind of approach, defensive posturing and angry, heated and defensive arguments are the rule. The participants rage fully abandon the dialogue. There is an important lesson here for informatics.


The connection is that decision support is often a statement of what needs to happen. Content, suggestive, and prescriptive analytics may often help to better predict diagnoses and therapies, and lead to better outcomes. This all presupposes a minimal discipline that, when absent, deeply erodes trust. And with the erosion comes the abandonment of the solution. 


For example, earlier this year I was working with a team using natural language processing and Medicare rules for risk adjustment of Medicare Advantage enrollees’ diagnosis lists. The goal was to surface relevant diagnoses that weren’t otherwise explicitly and adequately documented. We identified lots of amazing opportunities. In our zeal to not miss anything, we also recommended diagnoses we really shouldn’t have. Asking a doctors to sign-off on a naïve recommendation will erode confidence, with the risk of solution abandonment.


If, on your first use of a GPS, you were wrongly led into a dead end alley, you’d feel burned. The accuracy of recommendations needs to be right 99 percent of the time, adequately informed, and transparent so that trust can be quickly established and maintained.


What follows is a list of the five minimal elements necessary to earn and maintain that trust:


1.    Expose the known information

2.    Raise associated considerations

3.    Assure important contextual considerations by summarizing #1 and #2

4.    Think it forward and backward

5.    Apply sanity checks


In the first writing of this post, when I explained each of the five sequential elements above, my editor told me it was too tedious. So I’ll simply elaborate the fifth element, sanity checks, with this example shown in the graphic beginning with “a=b” and ending with “2=1.”   


Obviously, the conclusion is wrong. The algebraic operations are each legitimate. But what’s missing is the wisdom that the last step, while symbolically legitimate, requires division by zero, which is practically illegitimate. Any set of decision support rules can produce results that are illegitimate in some real clinical context. That’s why we must always screen for them.




In conclusion, healthcare can be complex, involving the input from the patient and many care givers over months, even years. Computer systems, often employing thousands of rules, need to be appropriately constructed to serve the needs of Population Health Management, quality Improvement, and delivering the highest quality care to people in need. 


The five steps outlined here are essential to accurately synthesize and summarize the considerations to achieve our shared goal. That is, to ensure that everyone can act at the level that the best of us do. And we need to trust the systems that make that possible, because are proven to be trustworthy in the hands of competent users. That translates to making the output of those five steps transparent.


Bregman describes that doing the right thing will feel inauthentic. Here, this means “stating the key facts, describing the necessary action, and offering to help the user take that action.” It’s not likely this is the designer’s instinct or first impulse. However, it needs to be, and the implications are clear. 


Solid clinical recommendations require this simple framing. Otherwise, we’re simply automating ineffective behaviors that lead to anger, defensiveness, and abandonment of immature solutions. As Bregman shares in his closing, having the right dialogue leaves the recipient of the support feeling positive and grateful.


What do you think?    




Joe, I'd suggest adding humility as a key component of any automated recommendation. The computer knows only a tiny fraction of what's going on with the patient or the health care provider at any given time. The recommendation may be much less important than some of those other things that are going on, or it may fail to take into account something that would be obvious to a person. Recommendations mustn't be delivered as though nothing else could be more important.

I'd also quibble with your 99% accuracy threshold, not because I disagree with your basic point, but because there actually is a minimal acceptable accuracy number (for any given context), but we haven't bothered to measure it, or even to estimate it reasonably well. We're so far away from adequately accurate recommendations (most of the time) that our systems act like we've stopped caring if the recommendations are worth blurting out.


Thanks for your comment.  I agree completely with

  - the humility mindset,

  - the reminder that "all of the necessary information is often not available",

  - it often contains incorrect observations,

  - and that the accuracy criterial, e.g. 99% was somewhat randomly chosen.

    (The difference between offering "education" and "prescribing medial advice" is

    extermely important in informatics and in the informatics law.  For more, read

    about "FDASIA" and HIT devices, starting here: )


In reflection, I think the whole post was about establishing expectations.  

This is even more important in the digital clinical world, and especially when it is faceless.


I would encourage interested readers to read this (inspired by Scott):


Thanks again for your terrific insights, Scott.




Hi Joe and Scott - I agree with clarity in expectations and humility in the execution. As I read Peter Bregman's article it seems he is taking the Parent-Child interaction style, appropriate with his daughter and less so with Adult-Adult interactions. Another way would be for the speaker to own his/her perception of the presentation and then ask what the presenter could do to alter that perception? Sometimes direct work and sometimes, well, less so. Thank you for sharing this model, Joe.

Wow.  Thanks Margaret (Maracry).

I was perhaps too quick to adopt Breman’s Parent-Child interaction style as appropriate and necessary.  I implicitly assumed his points were true to understand them.  I did not go back and decide whether to accept they were true.  You helped me remember to do that and it is very relevant.

If our goal is to help the user gain self awareness, and do it respectfully, taking an Adult-Adult interaction style is far more appropriate and necessary.  It is respectful and may require doing each of the five steps mentioned in the post.  Most importantly are summarizing known and associated considerations (step 3) and not proposing something insulting and/or trust-eroding (step 5).

As per the discussion with Dr Scott Finley in the comments here, there are other interaction models to consider.  Often, enthusiastic technologist inadvertently use an “I am great- and you are not” interaction style.  Readers may recognize this as “Stage 3” Tribal Leadership, succinctly summarized by the authors Dave Logan, John King and Halee Fischer-Wright, book and highly recommended.  Here’s the relevant text:

In Stage 3, the dominant culture in half of U.S. workplace tribes, the theme is “I’m great” or, more fully,

“I’m great, and you’re not.”

In this culture, knowledge is power, and so people hoard it, from client contacts to gossip People at this stage have to win, and winning is personal. They’ll out-work, think, and maneuver their competitors. The mood that results is a collection of “lone warriors,” wanting help and support and being disappointed that others don’t have their ambition or skill. What holds people at Stage 3 is the “hit” they get from winning, besting others, being the smartest and most successful [jb: a frequent component historically of the medical profession]. … ”

There are options for interaction styles, and that is what this post was all about. 

The Butler-Master interaction style, described here, incorporates the humilty that most systems today need to be cognizant that there is almost always relevant missing information.  That information makes style essential. 

To re-iterate my conclusion, anything less than all five elements in the analysis and presentation of clinical recommendations is substandard.  Building trust demands attention to style, communication, and context.