I’ve been thinking a lot about where to take the discussion that’s exploded on my post about moving my data from PatientSite into Google Health. I’m hardly an IT guru, but as I said, I do work with data at my day job. And before we proceed, there’s something I think is a foundation level truth:
The I in IT stands for Information.
The purpose of putting information into a system is to read it back out.
If the information you get back out does not match what it was intended to represent, #fail (as they say in Twitter).
Let’s let that sit for a few days.
Next post in this series: Encoding information is a key part of I.T.
p.s. Separate issue: remember, we haven’t yet gotten into poking around in Google Health itself, looking for soft spots, nor have I pushed the magic button to try HealthVault. More on that later.
Dave, when I teach data modeling/database programming to others, there something I stress (and was stressed to me when I learned): “Data is only information if you can use it.”
So, often “data” is put into a database, but if it is not retrievable in a viable form, it is not information within that definition.
So, perhaps we aren’t always putting information in. We are putting data which we think will come back out as information. Obviously, that ain’t so. ;)
Keep up the good work, Dave. I appreciate it.
Josh Seidman linked here from the Ix blog. Glad you’re likin’ it, Josh.
“If the information you get back out does not match what it was intended to represent, #fail ”
Are you so sure, Dave?
What if the data in is EXACTLY in the way it was designed and if it gets back out EXACTLY as it is supposed to but still remains completely meaningless to you?
Since you micro-post is a direct result of your earlier masterpiece on the EHR2PHR fiasco I suppose that you mean to talk about health information.
Health IT (HIT) was designed completely by professionals who never had to deal with patients. The fact is that the PHR was born as an afterthought while it should have been build first. HIT is doing a grand job at moving the right bits to the right places. It’s just too bad that neither you or me are of any significance to the designers of EHR.
As I wrote a couple of days ago, we should stop using the term EHR when we talk about the current versions of the HIT data collection systems.
The moment you do that you’ll agree that HIT systems are very powerful and are most definitely great information delivery systems. It’s just that, you know, none of this if for you.
It won’t change until we are able to convince millions of Americans that Healthcare Reform is very serious business, requiring their direct & sustained input and involvement. It should then be trivial to move from the ERR to a more appropriate patient-centered EHR/EMR.
Think about it for a minute. In a world where more and more decisions, at every level, are made via data-rich intermediaries, how can you ever think we are going to achieve patient-centered medicine if we do not have patient-centered information systems responsible for the collection and analysis of the data.
Gilles,
Of course you hit a very sensitive nail on the head. Well done.
I edited that phrase three times right after I posted it, and still wasn’t happy, because it’s such an important and touchy issue: it goes right to the heart of intent, which is the hairiest, non-“computable” thing I personally have ever seen. When I ran an industry standards group a few years ago, the subject of intent almost killed us. It is not solvable with software. The only thing you can do is record it as plain old text, and hope the reader is an intelligent person capable of grasping your meaning. And THAT sure isn’t I.T., that’s skill and common sense.
Adam Bosworth mentioned something similar yesterday. Talking about when he ran Google Health, he said “I focused entirely on getting labs and meds to be generally available because these seemed to be the only two things that were computable and not dependent on the vagueries of the hospitals and EHR’s.”
I mean, data is data, right? We can quantify lab results and my birthdate, but how do you quantify something like this: 18 months ago my friend the magnificent Monique Doyle Spencer was told by her doctor to get a scan. There were no symptoms, but indeed the scan caught some very early new metastases on her spine.
When she asked him “How did you know?” he responded “You just didn’t seem like yourself.”
What’s the ICD9 code for that?
Or SNOMED, or CPT, or anything else? And how could we possibly transfer, from one physician to another, the knowledge of Monique as a person, which allowed her personal doctor to sense that?
I think the answer is that some things can be counted or encoded, and some things can’t. As you say, and as Josh said in his post (linked above), the goal is knowledge. Information can support knowledge, and data can serve information. But not everything is computable, and certainly not if the data is a ridiculous mismatch for reality.
Perhaps we’ll discover that ExR’s of all sorts are a better record of data, a better info tool for a doctor’s arsenal, but that’s all. And they are VERY vulnerable to getting screwed up.
Meanwhile, yeah, if a system is a billing system, it should be called that. Reimbursement Record, Billing Record, sure. Call it that. Not clinical data.
ODLs! Except that it would have to be ODLs from physicians. That’s what you would use to transfer that very important part of medicine (the Art part!). Of course current ERRs are not designed to carry any of that, because the Art of Medicine has no reimbursable value.
Let me quote verbatim one of my heroes, Patty Brennan R.N., Ph.D., professor of Nursing and Industrial and Systems Engineering at the University of Wisconsin-Madison:
““The first round of Project HealthDesign taught us that the needs of patients must be the primary drivers of technological innovation. When it comes to patients being active managers of their own health and health care, we learned that it’s not just important to record observations from the clinical experience; personal observations of health as we go about our daily lives are equally critical. Information about how you feel and what you experience on an ongoing basis needs to be easily integrated into clinical processes because these observations reveal trends that enable care to be vastly improved and tailored to patients’ realities. By allowing this type of information to be shared easily between patients and providers, we believe people will experience a different level of engagement with their health and the health care system. The result might well be better patient-provider relationships and better clinical outcomes.” [emphasis added]
> the Art of Medicine has no reimbursable value
So where does that fit in other data schemes like SNOMED?
Also since we keep drawing new audiences, could you say a bit about ODLs? I know they’re observations of daily living, but I have no clue about their significance in the context of health records. (I keep hearing people tweet about them but the ugly truth is, je n’ais pas d’clue.)
Oops! ODLs are Observations of Daily Living. I’ll reproduce what I wrote about them a while back:
“We have been stumped for too many years to explain what are some of the activities taking place on the active ACOR eCommunities. Reports of ODL is really what is taking place and conversations that evolve from these minute ODL reports have generated big changes for many of the communities susbcribers, including for example patient-generated research about drug dosage and chronotherapy. So my hat to the RWJF’s funded Project HealthDesign and their wonderful ODL & PHR Primer
Many times, here and on other blogs, we comment about the need to include patient-generated reports when discussing medical cases with physicians. We all know that this idea runs contrary to the medical education received by most MDs because this data has not gone through “expert peer” reviewers. The serious studies funded by Project HealthDesign may very well become the most important resources needed to change the mindset of the majority of doctors.
So, I’m still not sure what’s in an ODL. From my own ACOR kidney cancer patient list, I don’t recall anyone logging what they do during their day. What are you referring to?
This is the first time I’ve heard of ODL’s, too. (Gilles, I did read your post from last June just now.) It seems they may be skewed by the subjective nature of any observation. Am I understanding that right?
Dave
I’m the Publicity Director for the IAIDQ (http://iaidq.org). I stumbled across your posts via a google alert for Information Quality (the irony is not lost on me).
You are profoundly and exceptionally right to remind people that the I in IT is Information… and regardless of the technology it is the Information which is used to make decisions and inform courses of action and if the information is of poor quality, all decisions stemming from it are affected.
I disagree, however, that the purpose of putting information into a system is to read it back out. The purpose of putting information into a system is to enable it to be used to inform decisions and courses of action – you read it back out to do that but the read is for a purpose.
What seems to have happened with GoogleHealth is information that was created/input for Purpose A was viewed as a valid surrogate source of information for Purpose B without any validation of the quality and ‘fitness for purpose B’ of that information. #Fail.
In one of the first books on Information Quality, written by Larry P. English (http://infoimpact.com), the example is given of a medical insurance provider who believed that the most common ailment claimed for by their clients was a broken leg, because that is what the data told them. Analysis uncovered that “Broken Leg” was item 1 on a drop down list and was set as a system default.
Giving patients access to their records to make corrections is only feasible if the information is in a format that is accessible and understandable to a lay person – again, it is in effect a change of purpose when you present the information to a patient as opposed to a clinician as the audience is different with different understandings.
I also like Gille’s quote from Pat Brennan… it is a truism of all quality management systems that the focus of the process and the innovation must be the customer (in this case the patient).
Although it seems information is put into a database, it really may only be data. For example: In a well-designed database, birthdate is put into a database, not age. Why? Because birthdate is not volatile, while age is. Thus, when retrieving information, the system used needs to be able to derive age from birthdate.
Databases are typically designed for codes. That is partly due to the need to minimize redundancy. (With redundancy, information can be very conflicting and if changed in one place may not be in another.) Another reason is the need to make it “easy” to use. And the third is for ease of queries that give sensible output (i.e. “information”). All data are not necessarily information.
Again, I’m not sure “correcting” as much as giving input/perspective is what I want to see as a patient. Transparent, collaborative, and participatory are key words that come to mind.
I appreciate your thought-provoking comments. I love how you found Dave’s post, and hope you will read the others on this site and comment. Great input.
Hey Daragh, good to meet you! Are you the author of the terrific piece tonight on IQTrainwrecks.com?
I work with data but I don’t have the disciplined background of either Robin or you, so I’m grateful for what you’re bringing.
I want to shine a light on one thing you said:
In my ears, this brings a level of rigor to the discussion that clarifies something I sorta knew but didn’t know how to articulate: validation of quality and fitness for purpose. It’s superb, and I’ll use it in some ongoing discussions I’m having offline.
Is there a well defined process for doing that? Can you steer me to some classic reference about it?
Your note is bolstered by a comment from another blog that I quoted last night on the Encoding chapter of this series:
=========
1. The US took a system which was largely created for mortality reporting and made a clinical modification so it would be useful for morbidity too.
2. It is required to be reported for billing.
3. Payment policy is designed around the codes.
4. True use of the codes changes dependent upon the payment policies of the insurance company (not supposed to happen under HIPAA, but check with your billing office).
5. ICD-9 is now over 30 years old because ICD-10 implementation, which would help, was delayed for a decade because the transition would be too hard.
6. Now its supposed to somehow be used for true clinical condition feedback to the patients.
=========
So, like, um, you data quality guys are MORE than welcome to come help us do this right. Join the teaching. Especially since you seem to have the knack for speaking Ordinary People language. :–)