Convert your FHIR JSON -> XML and back here. The CDA Book is sometimes listed for Kindle here and it is also SHIPPING from Amazon! See here for Errata.

Monday, April 15, 2013

Precision is the enemy of accuracy

I'm sure to get reamed for this post, but if so, it will only prove my point unless those doing the reaming are both accurate and precise.

I've been watching the growing eterna-thread* about the proper way to record allergies, intollerances and adverse reactions over on the HL7 Patient Care list for the past few weeks.  It simply raises up the same old issues, and I think we are still no closer to having a better model for recording allergies than CCD used more than six years ago.

The issues that keep coming up have to do with the "correct" way of doing things.  Which as far as I can tell, no two professionals completely agree.  As far as I can tell, I am actually one step ahead of many, since I don't claim to know much, whereas others who know much more disagree quite robustly.

What seems to have been forgotten in this discussion is the difference between precision and accuracy.  Current models for reporting allergies and intolerances are accurate, they just lack precision.  And because they lack precision, and because healthcare professionals understand that they lack precision, the data about allergies is used appropriately.

Because we can be more precise, there is a great deal of desire to do so.  Being more precise gives physicians more treatment options.  It enables better use of available resources.  It avoids overstatement.  It allows for a more precise assessment.

But, in so doing, it requires others to act with a greater degree of precision as well.  I'm not so happy about providing physicians with a very precise statement of my allergies and intolerances if it means that they become paralyzed trying to understand it, or are unable to make use of it with the degree of precision and accuracy necessary.

Where we are now enables us to make a generalized statement about allergy and intolerances, and invites further exploration by the healthcare provider when greater precision is necessary.  This seems to be the safest path forward.  But what do I know?  I'm not an expert.  I just have to implement what they tell me.  Once they all agree on what that is...


* An eterna-thread is a discussion thread that had been going on (or will go on) seemingly for eternity.  The way I identify eterna-threads is if I can write a rule for filing it based on keywords in content.


  1. For example, "normal" body temperature is 98.6°F and people sometimes get worried if their own reading is off ±0.5°F. What that does not take into account is normal metabolic variance, time of day, whether a person is lying down or sitting up, clothing or other covering, the presence of other physical symptoms, lack of calibration of the thermometer, etc. Context is everything.

  2. This ethera-thread is also confounded by a classic design engineering problem that HL7 is overrun with. This is the need to separate gathering Needs from the users, from the Systems Design step. Yes the users needs need to be understood very clearly, and Validated often (depending on your design methodology). But the design process should be done away from the users. Too often the users think that the design is wrong, without understanding the full system design. Such is the case often in Interoperability, where there is a fallacy that the encoding in the interoperability layer is the User Interface. These two are different, the user interface is a different design step. The user interface needs to be supported by the interoperability layer, and the interoperability layer needs to be supported by the user interface; but they are different steps.