Kaufman, David Ross

Rapid ethnography and data mining approaches have been used individually to study clinical workflows, but have seldom been used together to overcome the limitations inherent in either type of method. For rapid ethnography, how reliable are the findings drawn from small samples? For data mining, how accurate are the discoveries drawn from automatic analysis of big data, when compared with observable data? This paper explores the combined use of rapid ethnography and process mining, aka ethno-mining, to study and compare metrics of a typical clinical documentation task, vital signs charting. The task was performed with different electronic health records (EHRs) used in three different hospital sites. The individual methods revealed substantial discrepancies in task duration between sites. Specifically, means of 159.6(78.55), 38.2(34.9), and 431.3(283.04) seconds were captured with rapid ethnography. When process mining was used, means of 518.6(3,808), 345.5(660.6), and 119.74(210.3) seconds were found. When ethno-mining was applied instead, outliers could be identified, explained and removed. Without outliers, mean task duration was similar between sites (78.1(66.7), 72.5(78.5), and 71.7(75) seconds). Results from this work suggest that integrating rapid ethnography and data mining into a single process may provide more meaningful results than a siloed approach when studying of workflow.

Medication reconciliation (MedRec) is a mission-critical process which can serve to reduce adverse drug events (ADEs) in surgical settings. However, providing quality care is limited by current health information technology (IT), which is often inefficient and unintuitive due to poor usability, resulting in high cognitive burden. We have been characterizing EHR mediated workflow in the Mayo Clinic enterprise prior to a system-wide electronic health records (EHR) conversion in order to harmonize workflows. We compared and evaluated MedRec processes in pre-operative nursing assessments across two different EHRs in place in different locales at baseline. The interfaces differed both in their modes of interaction and cognitive support. Analyses surfaced interface elements that were unintuitive and inefficient, creating unnecessary complexities in clinicians’ interactive behavior. Keystroke level models (KLM), a modeling tool for predicting task completion time, showed that to access medication lists required a different series of operations across the two systems. Different designs can differentially mediate task performance, which can aid in the mitigation of errors for complex cognitive tasks. Identification of barriers in EHR-mediated workflow and barriers to interface usability could lead to system redesigns that minimize cognitive load while improving patient safety and efficiency.

Background Type 1 diabetes (T1D) care requires multiple daily self-management behaviors (SMBs). Preliminary studies on SMBs rely mainly on self-reported survey and interview data. There is little information on adult T1D SMBs, along with corresponding compensation techniques (CTs), gathered in real-time. Objective The article aims to use a patient-centered approach to design iDECIDE, a smartphone application that gathers daily diabetes SMBs and CTs related to meal and alcohol intake and exercise in real-time, and contrast patients' actual behaviors against those self-reported with the app. Methods Two usability studies were used to improve iDECIDE's functionality. These were followed by a 30-day pilot test of the redesigned app. A survey designed to capture diabetes SMBs and CTs was administered prior to the 30-day pilot test. Survey results were compared against iDECIDE logs. Results Usability studies revealed that participants desired advanced features for self-tracking meals and alcohol intake. Thirteen participants recorded over 1,200 CTs for carbohydrates during the 30-day study. Participants also recorded 76 alcohol and 166 exercise CTs. Comparisons of survey responses and iDECIDE logs showed mean% (standard deviation) concordance of 77% (25) for SMBs related to meals, where concordance of 100% indicates a perfect match. There was low concordance of 35% (35) and 46% (41) for alcohol and exercise events, respectively. Conclusion The high variability found in SMBs and CTs highlights the need for real-time diabetes self-tracking mechanisms to better understand SMBs and CTs. Future work will use the developed app to collect SMBs and CTs and identify patient-specific diabetes adherence barriers that could be addressed with individualized education interventions.

Patient-centered cognitive support has been shown to be critically important to facilitate the effective use of health information technologies (HIT). There is a well-documented need to better understand HIT-mediated clinical workflow. Current technologies can burden clinicians’ cognitive resources, which is associated with patient safety risks and medical errors. We sought to employ a distributed cognition approach to examine how information flows across the activity system to support clinicians’ problem-solving. Specifically, we studied the propagation of representational states across media, conversations, actors and time in the coordination of patient-care processes. We examined multiple instances of work and information flow in a real-world setting, revealing problems in information flow: a) use of paper artifacts has limitations to facilitating coordination of care, b) clinicians challenged in developing shared awareness, c) responsibility of representing patient states is distributed across documents, d) clinical reasoning that informed care plans was absent from documents. Findings surface a challenge to automated monitoring of care goals; much of the information is present only in clinicians’ minds and in informal documents.