Without consulting the Great Google, answer this:
What is iHealth?
I would have guessed the ‘I’ stood for Information or Informatics. But last week I had the good fortune to attend AMIA’s iHealth 2017 Clinical Informatics Conference in Philadelphia, PA and learned that it is not just one ‘I.’
More than just a play on words or an alternative to the overused term ‘eHealth,’ the use of so many I’s in the conference description provides insight into the psychological state of the field, not unlike how using I- and me-focused words do the same for humans. In fact, iHealth can be characterized by the confluence of ‘informatics,’ ‘innovation,’ and ‘impact’ within healthcare. Comprised mostly of clinical informaticists, this meeting covered a spectrum of issues including how these I’s play into patient engagement, clinical research, and population health.
I is for Instruct
Education is a big part of the clinical informatics conference (and any AMIA event). We presented a pre-meeting workshop to 30+ attendees, ‘Understanding and Communicating Risk: Critical Competencies for the Healthcare Workforce.’ The reality is, data literacy, numeracy, and statistical literacy are competencies that are essential for the modern healthcare enterprise. Without a firm understanding of these methods and their interpretation, data-driven strategies to improve healthcare will fall short of their potential impact. To that end, the goals of this workshop were to empower clinical informaticists to:
- Interpret risk
- Make better decisions about the viability of using risk scores
- Explain risk concepts to others
We covered topics such as how and why we commonly misunderstand risk; the methodologies behind calculating risk; techniques for interpreting and evaluating risk as produced by others; and best practices for visualizing risk.
I is for Informatics
The use of big data and advanced technology was at the center of most presentations. Clinical informatics is indisputably a key enabler of modern, innovative care practices. But for as much as they enable, informatics systems can also detract from the efficient provision of high-quality care. The Wednesday afternoon panel on ‘Immediate Adaptability’ shared EHR usability problems that add speedbumps and distractors to workflow. Much of this has to do with how information is displayed:
- Display visibility: Too much scrolling needed to see all components within an activity (e.g. placing a drug order)
- Navigation: Inability to easily backtrack to a previous screen that may have critical patient context (e.g. allergies)
- Consistency: Unclear signals for where to enter data which cause components to be overlooked
- Labeling: Clinical decision support descriptors not informative enough to cause pause (e.g. best practice alerts)
These issues can create a new source of medical error– e-iatrogenesis, where a patient is harmed by errors generated in the use of information technology. During the panel, it was noted that many commercial systems require far too many clicks across different screens to perform a clinical review– up to 37 clicks for an admitted patient within one EHR.
So what do we do about this? Unfortunately, the feverish pace of EHR deployment often leaves organizations with so much pressure to go-live that wide usability gaps remain unclosed. The post-deployment EHR optimization phase needs to be an ongoing project to maintain and improve the quality of tool usage. Informaticists need to listen to end users of healthIT and use their influence on leadership and vendors to force through change. Placing the problem in the context of measurable patient harm or care inefficiency will be essential to bend those ears.
I is for Innovation
Penn Medicine was well-represented and not just because of their proximity to the conference. Some of the most exciting work comes from the efforts of Johannes Eichstaedt and Lyle Unger and they shared the clever ways they are able to use social media data to learn more about patient outcomes. Lyle Ungar presented one study showing how public Twitter data were able to predict county-level heart disease rates across the country using geocoded information associated with tweets. In fact, Twitter was more accurate than all of the regularly employed predictors (e.g. demographics; chronic disease status) combined.
Another fascinating insight came from an analysis of Yelp reviews where patients reported their experiences. While these data are biased to be negative in nature, the word cloud showing the most common words in negative reviews was very telling. That word is ‘told’, with ‘worst,’ ‘rude,’ ‘not,’ and ‘hours’ rounding out the top five. These results are different than the dissatisfiers that are often cited in hospital HCAHPS surveys. I find these results fascinating because it suggests that communication practices play a large part in dissatisfaction, which many of us might predict as patients but somehow forget when playing a role in care. Patients don’t like being told what to do or left out of the loop of information. Sensitivity training may be a key component of improving satisfaction scores.
I is for Impact
It’s one thing to develop innovation in informatics, but it is a whole other sphere of activity to make those innovations impactful. In my opinion, this is where AMIA can improve in supporting clinical informatics. Much of AMIA’s membership serve in more academic-focused circles and may not have the insight or influence to affect operationalization. I was pleased to see this topic raised in a number of sessions.
The panel session on Drug-Drug Interaction Seriousness raised the issue of clinical decision support methods and how to reduce nuisance alerts. Dr. Bimal Desai from the Children’s Hospital of Philadelphia shared a figure from a recent paper showing how override rates for drug-drug interaction and drug-allergy alerts hover north of 75%. His sober assessment was, ‘this is the most depressing graphic I’ve seen in a long time’. He made a clever analogy to phone calls and asked how many of us answer an unknown number given it could be a sales call (e.g. a false alarm). Dr. Desai’s position is that alert fatigue is a biostatistical problem and not just one of poor usability optimization. He shared that by taking some measures to tweak alert logic to be more patient-specific and reflect a Precision Medicine philosophy, nuisance alerts were reduced and adherence to the good alerts increased.
We also focused on the concept of impact when giving our pre-conference workshop ‘Understanding and Communicating Risk: Critical Competencies for the Healthcare Workforce’. Here we challenged the audience to really dig into what difference expressions of medical risk scores mean and how that affects decision-making.
For example, Therapeutic Illusion is the tendency to overestimate the impact of medical treatment due to misunderstanding risks and benefits. In a prior blog during American Heart Month, I mentioned how the Number Needed to Treat (NNT) for a variety of common therapies is often far higher than people would expect. (The NNT is the number of patients that must be treated to achieve the desired outcome for a single patient). Many times, analytic products used as clinical decision support place some sort of risk score into clinician workflow. This can be fraught with numerous cultural, logistical, and process challenges. Overcoming these requires the adoption of an analytic product lifecycle that ensures involvement of key stakeholders, adoption of agile project management techniques, and training on how to interpret the risk scores.
Would I Attend the Clinical Informatics Conference Again?
I also would have enjoyed more networking time that could have focused on discussing a topic or some other engagement exercise. The poster session was dominated by student submissions and was not well-attended. I am sure that session was impacted by the cost of a glass of wine, which was more than a mini-bottle from your in-room minibar, which didn’t require you to have cash. I think a lot of people went to the hotel bar instead. I do feel that it was a little cruel not to have free-flowing coffee during all sessions, especially since there was not a coffee shop or similar within the venue. They did, however, provide a number of ways to bling up your badge, which was entertaining.
While not even the Great Google has analytics yet good enough to predict future travel budgets, I enjoyed the size of the conference and would say the risk of reattending is good.
For those of you who could not join us in Philadelphia, we are hosting a webinar, “Developing Your Analytic Intuition: Techniques for Understanding Healthcare Risk Scores” summarizing our workshop on Thursday, May 18th at 1 PM ET. Click here to register and join the conversation.