quality improvement

Epic Medication Reconciliation Fail: A True Story

January 15, 2016

Today I want to share a true story that I’ve been mulling over recently, as I ask myself when will we start to see more substantial gains in health care quality.

It’s the story of a 94 year old woman who was sent from her memory-care residential unit to the emergency room, due to nausea and vomiting. She ended up being hospitalized for about 48 hours, for UTI.

(Sad but true aside: her family has asked for hospitalization to be avoided unless absolutely necessary for comfort. But the facility feels they have to send her to the ER if she falls, vomits, or otherwise looks seriously ill. Argh.)

Before hospitalization, she was only taking vitamin D and a daily aspirin and a multivitamin. She’d been in the memory unit for years due to dementia, and on minimal meds since “graduating” from a stint in hospice a few years ago. Because she’s chair-bound and it’s hard for her to leave the facility, she hasn’t been to her PCP’s office in years. Instead, I come and see her at the facility once or twice a year.

Guess how many medications she was discharged from the hospital on? Fourteen.

As in, fourteen new daily medications to be taken indefinitely! (There was also a course of antibiotics for UTI plus a few new PRN medications.)

I thought I was hallucinating when I saw her med sheet at the facility. [Read more…] about Epic Medication Reconciliation Fail: A True Story

Medicine in Denial: A book on how to really leverage technology to improve healthcare

May 10, 2013

[This post was first published on The Health Care Blog on 5/4/13, titled “Medicine in Denial.”]

“Any system of care that depends on the personal knowledge and analytic capabilities of physicians cannot be trusted.”

Finally, I’ve come across a really spot-on analysis of what ails healthcare, and some proposals that have serious potential to improve healthcare for people like my patients. Come to think of it, implementing these proposals would surely improve care for all patients.

The analysis, and the proposed fixes, are detailed by Dr. Lawrence Weed and his son Lincoln Weed, in their book “Medicine in Denial.” (The quote above is from this book.)

The book is a little long, but for those who are interested in leveraging technology to make healthcare more consistent and more patient-centered, I’d say it’s a must-read and must-discuss. (I’m a bit surprised that this book doesn’t seem to have many reviews, and that Dr. Weed’s ideas are not more often cited by those advocating for digital health and patient empowerment.) In particular, the Weeds’ book provides:

An excellent description and analysis of two huge fundamental problems in healthcare. One is the way we persist in relying on fallible physician minds to manage the process of evaluating, diagnosing, and managing medical problems. The other is our lack of standards for consistently documenting and organizing information related to our evaluation and management of patients. Both lead to idiosyncratic, disorganized healthcare that generally serves patients poorly, especially those who are medically complex or have multiple chronic conditions.
A proposed method of using computers and technology to consistently connect patient data to medical knowledge, leading to better diagnosis and medical management.
A proposed method of reorganizing of medical records and clinical data. This “problem-oriented medical record” would provide a fundamental level of organization and transparency to the practice of medicine, and would allow better management of multiple problems over time.
A vision of healthcare focused on empowering patients, and on enabling healthcare to be tailored to each patient’s needs, rather than driven by provider idiosyncracy or the blunt tools of evidence-based (aka population-based) medicine.

The book also covers several other topics, such as related problems in medical education and credentialing, and redefining competence in medicine. But the points above are the ones that resonated most deeply with me and my frustrations with the healthcare system.

“The concept of a physician as we know it is not viable”

The Weeds point out the obvious: there exists far too much medical information for the human brain to keep it all in mind, and apply it in a consistent and thorough fashion during a medical encounter.

This creates serious problems when it comes to the core medical work of diagnosis and providing treatment recommendations. To being with, when a patient comes to a physician with a complaint, the physician invariably does not collect enough data. (Just take a look at any UpToDate topic – or JAMA clinical review article — on evaluating a common complaint, and ask yourself if clinicians usually inquire about everything they should. We don’t.) Instead, clinicians ask questions somewhat idiosyncratically, depending on factors such as their initial hunch, their specialty habits, etc.

Next, physicians do a highly imperfect job of matching the patient’s data – i.e. the positive and negative findings – with medical knowledge. This results in a diagnostic conclusion that is often wrong, or in a differential which is incomplete.

As the Weeds point out, a patient with a medical concern can go see three different doctors and emerge with three different diagnoses. And of course, just as clinicians are idiosyncratic in their diagnostic processes, they are also idiosyncratic in how they recommend further evaluation, or in prescribing a management plan.

Doctors will call this “clinical judgement,” but the Weeds consider this unacceptable human-generated variation in medical practice, and I have to say that I agree with them.

To make matters even worse, not only are clinicians applying idiosyncratic human processes to diagnosis and management, but they then go on to document their findings and thought-processes in spotty idiosyncratic ways. This leaves the patient without a good record of his or her medical findings, and makes it difficult for subsequent clinicians – or the patient, for that matter — to reliably build upon the efforts of the initial clinician.

In short, the Weeds argue that medicine is plagued by a culture of severe, pervasive disorder. We are not orderly in how we evaluate patients, we are not orderly in how we match their data to our existing knowledge base, and we are not orderly in how we document our clinical processes and data.

The Weeds attribute much of this to medicine’s habit of valorizing the individual physician’s intellect and autonomy. Because of this, we persist in organizing healthcare around the efforts of fallible physician minds. The authors declare that the profession of medicine is in terrible denial.

I found myself agreeing, yet again, with them.

The computer-assisted alternative

To counter the existing sorry state of affairs, the Weeds propose a “standardization of inputs,” and argue that clinical judgement should be applied after we use computers and technology to complete two key tasks. The first task is to reliably identify and collect the necessary information from patients, via standardized questionnaires that are tailored to the complaint in question. The second is to use a “knowledge coupler” to analyze the patient’s responses and propose a list of diagnostic possibilities.

Only then should clinical judgement really enter the picture, and according to the Weeds, this should be applied in order to tailor the next clinical steps to the patient’s preferences and individual circumstances. (Hear hear! I like it.)

Presumably the reflexive response of many physicians will be to decry this as cookbook medicine.

Is it? Having been dismayed by the spotty clinical work that many physicians produce under today’s usual rushed outpatient conditions, I’m not sure a little cookbook structure is such a bad thing. As the Weeds point out, the purpose is to start with a solid, consistent foundation, and *then* proceed to individualizing:

“Decision-making must begin with a simple, mechanical process of association between data and knowledge, conducted without reliance on the practitioner’s mind. Thereafter, the processes of care must remain highly organized and explicit. Care would become highly standardized at the front end, and medical decisions at the back end would become highly individualized – precisely the opposite of the status quo, where physicians have broad discretion during the intial patient encounter but are expected to conform to standardized, “evidence-based” guidelines in their ultimate decisions.”

Being a junkie for order and completeness, I found myself quite attracted to the concept of standardizing inputs and applying a knowledge coupler before bringing in a physician’s clinical judgement. (The Weeds call this the “combinatorial” approach, as compared to the now predominant “judgemental” approach, which relies almost entirely on clinical judgment.)

How fantastic would it be if my elderly patients complaining of falls could have worked through a nice thorough questionnaire and computer-assisted differential, all before I even sat down to hold their hand. And it would be even better if such digital assistance would enable the non-geriatricians to identify orthostasis and medication side-effects as source of falls in the elderly.

But is it actually feasible to apply questionnaires and knowledge coupling to most older patients? I couldn’t help thinking that it would take my patients a long time to go through the questionnaires, and that they would probably need someone’s assistance.

The Weeds do address likely objections to the combinatorial approach. They point out that “comprehensive does not mean exhaustive” (but actually it does, when it comes to geriatrics). They also note that even if a standardized initial data collection is time-consuming, this should be considered time well-spent if it leads to better quality diagnosis and management. (On this I agree.)

Still, I couldn’t help but wonder if detailed data collection might not be more overwhelming for patients and providers than they admit. It certainly would’ve helped if the Weeds had provided an actual example of a sample questionnaire for one or more common complaints in an older adult.

For example, for shortness of breath, I presume an older person with history of CHF, CAD and COPD will require a more detailed questionnaire than a young adult with no significant past medical history. What would such a questionnaire actually look like? And how long would it take to complete?

In short, I found myself easily persuaded by the theoretical case for a technology-assisted combinatorial approach, rather than today’s terribly error-prone judgmental approach. But I was left uncertain as to how feasible it actually would be to implement in the case of complex elderly patients.

[See Part Two and Part Three of this commentary, which address some of the other key concepts discussed in “Medicine in Denial.”]

See here for comments to this post at The Health Care Blog.

Tweet and You Might Receive: Social Media, Serendipity, and Process Improvements

February 7, 2013

It all started with my sending a tweet.

Actually, that’s not quite true. The way it really started was with my frail elderly patient calling me in mid-January, to tell me he thought he had a UTI. But that part of the story is not new and novel; I’ve often had patients contact me with similar concerns.

I did what I usually do: ordered a UA and UCx. (I know, in theory better to check a UA, and if it looks suspicious, send for UCx. In practice, that’s logistically difficult if you don’t have the patient in clinic and aren’t able to dip the urine right then and there.)

The trouble was, the patient had called me on a Friday morning. “Bummer,” I thought, “I’ll probably get the UA on Saturday but the culture might not be back until Monday.”

Sure enough, on Saturday I checked my fax queue and there was a preliminary report: lots of white cells and nitrite in the urine.

High-risk elderly patient with symptoms. He needed empiric treatment started before the urine culture results would be available.

Being the clinical decision-support junkie that I am, I decided to take a quick peek at empiric treatment recommendations on UpToDate.com, where I confirmed that the recommended treatment is TMP-SMX. Unless, that is, there is local resistance >20%.

Or I could prescribe a fluoroquinolone. But, notes UpToDate, “increased resistance is mitigating the usefulness of the fluoroquinolone class.”

Hm. I found myself noticing that choices do, in fact, induce decision-fatigue. (No wonder so many docs just prescribe whatever they’re used to prescribing.)

Clearly, the task at hand — selecting a suitable antibiotic for empiric treatment of UTI — would be much easier if I knew what the local resistance antibiotic resistance patterns have been recently.

So, I decided to call the lab itself, Quest Diagnostics, thinking that maybe they’d be able to tell me about local resistance patterns.

The staff answering Quest’s results line seemed quite perplexed by my inquiry. They transferred me to the microbiology lab in San Jose, where they were equally perplexed. Sorry doctor, your culture results won’t be available for 1-2 days. And *then* we can tell you what the resistance pattern is for YOUR submitted sample.

I kept telling them I’m not asking about my sample, I’m asking about recent resistance on all urine samples from community patients.

I kept being told that resistance results will soon be available for MY sample.

Finally they transferred me to a supervisor, who told me that she sees what I’m getting at, and no, they don’t provide this information.

“But you must be culturing 1000 urines per week,” I pointed out. “You must have a sense of how much resistance there is to certain drugs.”

She laughed. “We run more like 10,000 urines.” But, she went on, this didn’t mean they had general antibiotic resistance data to share with doctors. Instead, she recommended I try the public health office. I didn’t bother to point out that they wouldn’t be open on a Saturday.

I hung up, picked an antibiotic to prescribe, and sent my request to the patient’s pharmacy

And then I sent out a tweet, commenting that although Quest does all these cultures, somehow the resistance data isn’t available to community docs like me.

A tweet heard across the country

To my surprise, my tweet was noticed by another doc, @HenryWeiMD, who addressed a follow-up tweet directly to @QuestDx — something I hadn’t thought to do — urging Quest to help @GeriTechBlog.

A few hours later, I received an email from Henry, addressed to me and someone at Quest, in which Henry introduced me to a contact at Quest and pointed out that if this kind of antibiotic resistance data isn’t yet being made available to community doctors, then it really should be as this would be a “HUGE low-lying apple/big win for public health.”

It turns out that Henry was a Presidential Innovation Fellow, though he’d been very much acting in a personal capacity when he followed up on my tweet. His tweet had prompted someone from Quest to send him a message offering assistance.

A week later, I was notified that my issue had been referred to Quest executives, who would be following up with me.

And then a few weeks later, I finally received a phone call from a very nice Quest VP.

“You’re the first to have asked”

Yep, that’s what he told me. Which I find a little hard to believe, but it’s certainly possible that this is the first time that such a request has made it up the command chain. (Seems unlikely that the local microbiology supervisor would be forwarding inquiries such as mine.)

But here is what is really really exciting: now that I’ve asked — and miraculously been heard — Quest is willing to work on producing local antibiograms for community clinicians!

Now, it’s not available yet. We’ll have to be a little patient and let them figure out how to do it. It’s one thing to have the raw data, and another to collect it, organize it, and present it to a clinical audience. But if all goes well, eventually community docs will be able to access local antibiotic resistance data from Quest.

Woo-hoo! Smarter antibiotics prescribing, here we come!

Will docs actually use this info? Who knows, it’ll probably depend on how easy it is to access. For instance, if local antibiograms end up printed at the end of every abnormal UA report, I’d expect many doctors to incorporate this into their prescribing decision. However, if one has to call Quest or look it up online, then the information will likely be used less often. Still, better to have such info online rather than not at all.

In the meantime, as someone with a background in quality improvement, I’m intrigued by the twist that social media brings into all of this. In the past, we practicing doctors have not had easy ways to make ourselves heard and noticed. Now we can tweet and blog, although if you’re a small fry like me it’s also helpful to get a boost from someone with a little more clout and connections. (Thanks Henry!)

So what conclusions have I drawn from this so far?

Clinicians should be vocal about specific things we need in order to practice according to guidelines.
Social media can connect you to allies and like-minded others.
It helps to know people who know people.

Summing it up

By commenting on a sensible clinical issue — my needing local antibiograms for better empiric UTI treatment — via Twitter, and getting echoed by another doctor with more visibility and connections, I found my request being considered by a senior executive at Quest Diagnostics.

I’m left concluding that we clinicians should be vocal and specific in pointing out things we need in order to practice care according to guidelines or best practices. Social media offers some good opportunities to do this.

I’m also very grateful to the leadership at Quest Diagnostics for engaging with this issue. If they can start providing clinicians with local antibiotic resistance data, they’ll be doing patients and providers a really good, useful service.