Knowledge Crystallization and Clinical Priorities

Research • January - May 2014

Knowledge Crystallization Model

PROJECT OVERVIEW

This research was conducted for my Master's Capstone project in 2014. I teamed up with a Pediatric Nephrologist at Seattle Children's Hospital, an Internal Medicine Physician, and a Professor at the University of Washington to investigate information seeking and synthesis within clinical care environments. I graduated from the University of Washington with Master of Science in Information Management in 2014.

 
 

Accolades

Our culminating work was published in JAMIA, the Journal of the American Medical Informatics Association. I presented our findings at the Computing Research Association's Committee on the Status of Women in Computing Research (CRA-W) at their Grad Cohort Workshop in Santa Clara, California in April, 2014; and again in May, for the Founding Board at UW's Information School. My colleagues presented it at AMIA's Annual Symposium in Washington, D.C. in November, 2014. 

TEAM

Ari Pollack, MD, Carolyn Tweedy, MS, Katherine Blondon, MD, PhD, Wanda Pratt, PhD

Tools Used

Atlas.ti, Keynote, Omnigraffle, Excel

ABSTRACT

Information seeking and synthesis are time consuming processes for physicians. Although systems have the potential to simplify these tasks, future improvements must be based on an understanding of how physicians perform these tasks during clinical prioritization. We engaged physicians in semi-structured focus groups, and found that data is collected to categorize and prioritize patients according to expected clinical course. When data does not support these expectations, or when categorization indicates potential for morbidity, physicians increase efforts to act or re-categorize patients. We identified a standard prioritization workflow that differs slightly by medical specialty. These findings lay a foundation to advance information displays that facilitate information processing by physicians in clinical care environments.

JUSTIFICATION

In 1999 the Institute of Medicine released a report “To Err is Human” which estimated that hospital errors led to an estimated 98,000 deaths each year. More recent estimates have pushed this number even higher, with some making the claim that medical error is the third leading cause of death in the United States behind heart disease and cancer.

This is an information problem. These errors occur because the right information is not available to the right people at the right time, in the right place, or within the right context. 

This is also a tools problem. There is no dearth of opportunity for UX improvement within Electronic Medical Records (EMR) systems. Today, clinicians must wade through volumes of information to find meaning, assign value, and take action, which is especially complicated when they have many sick patients, and not a lot of time to figure out who is the sickest, and who they should go see first.

So, how do they do it? That's what we wanted to find out. 

Our oBJECTIVEs

  1. To understand how physicians collect, process, and utilize data during the clinical prioritization process 
  2. To develop a model that describes this process
  3. To identify opportunities to improve the delivery and utilization of information in the clinical care environment, and within future medical information systems

METHODS

We went through the Institutional Review Board (IRB) approval process at Seattle Children's Hospital, and recruited 23 participants from a variety of specialties. 

Participant Demographics

Participant Demographics

 

Participant Medical Specialties

 

We held semi-structured focus groups and used fictional cases design to represent a clinician's typical case load. These cases helped frame and guide the discussion. We asked participants to describe their typical prioritization process, and recorded the sessions for later transcription and coding. We used Atlas.ti for qualitative coding and analysis, and created specific categories for actions that physicians indicated that they typically performed while prioritizing their patients.

RESULTS

It was clear from the results that physicians are constantly juggling two overarching categories of information: clinical and non-clinical (e.g. the physical location of the patient, social dynamics including the presence of family members, etc.). The clinical elements included acuity (how sick the patient is), the problem or diagnosis, and most interestingly, the concept of change. In this context, change can mean variance in a lab value or vital sign, or change in diagnosis or clinical course. 

Standouts from the qualitative coding and analysis of focus group transcripts

Clinical considerations: acuity vs. change

NOTABLE QUOTES FROM FOCUS GROUP Participants

 
"... I'm looking for a change in the pattern or in the trend which maybe I wasn't expecting.... And that stops me in my tracks, why is this not making sense? ..."
-- Participant 2.2
 
"So if you're sure of the diagnosis and you're sure [the patient is] on the right therapy even if they're worse, not such a big deal. If all of a sudden you say, ‘Wait, this doesn't make sense anymore.’  You have to go back to square one... are we treating the wrong thing?" -- Participant 2.4
 
 
"We have plenty of patients who get worse, but sometimes that's part of pattern. And that actually doesn't freak you out anywhere near as much as when somebody gets even just a little bit worse, but that's not what they're supposed to do. It's when they do what they're not supposed to do that you think something weird is going on, you have to regroup."
-- Quote from Participant 2.2

SYNTHESIS

Our adaptation of Card, Mackinlay, and Schneiderman's Knowledge Crystallization Model

We recognized that clinicians were repeatedly describing classic knowledge seeking problems, and we sought out existing frameworks to help us make sense of their prevailing mental model. 

We happened upon a model first described by Stuart Card, Jock Mackinlay, and Ben Shneiderman in the 1990s called the Knowledge Crystallization framework, which was designed to help people make sense of information problems.  We were able to leverage this framework as a foundation and adapted it to fit the domain of clinical prioritization. 

iMPLICATIONS

Through application of information visualization and machine learning techniques, our revised model of Knowledge Crystallization has great potential to simplify existing knowledge acquisition processes in healthcare settings, and as a result, improve clinical outcomes. These possibilities also have the potential to facilitate better understanding the cost structure of information seeking, which has broad implications for healthcare information systems.

Rough sketch: potential for visualization of clinical data

Rudimentary visualization: status of clinical data for a physician's case load

Rough sketch: potential for clustering algorithms

Rough sketch: potential for visualizing prognoses

Rough sketch: potential for visualizing patient trajectories