Human Factors in Prescription Medication Management

1
Human Factors in Prescription Medication
Management

• Jonathan R. Nebeker MS MD
• VA Salt Lake City GRECC

2
Acknowledgements

• Charlene R. Weir, PhD
• Frank Drews, PhD
• Molly Leecaster, PhD
• Rand Rupper, MPH MD
• Kenneth Boockvar, MD
• Brittany Mallin, MS MPH

• AHRQ R18 HS017186
• VA Salt Lake City GRECC
• VA Salt Lake City IDEAS Center

3
Overview

• The Electronic Health Record context
• Current
• Future
• How theory gets us to future
• Theoretical Framework
• Study design
• Preliminary Findings

4
Current CPRS VistA

• Emphasis on access
• Information siloed in tabs

5
Future CPRS VistA

• Emphasis on control
• Information integrated

6
Goal EHR of future
7
Decision Support v. Sense Making

• Computerized decision support is typically
  normative and targets the right decision.
• The CPRS of the future will emphasize an
  information-rich environment that targets sense
  making to support higher quality decisions in the
  highly variable context of patient care.

8
Progress

• The Electronic Health Record context
• Theoretical Framework
• (The pathway to the future)
• Joint Cognitive Systems or Cognitive Systems
  Engineering
• Contextual Control Model
• Study Design
• Preliminary Findings

9
Towards the Future

• Apply Cognitive Systems Engineering
• Human Factors in this talk
• Not about usability
• About the human-computer system

10
Joint Cognitive Systems

• Erik Hollnagel and David Woods
• System of artifact(s) human(s) that
  accomplishes work.
• Not what do human and computer do best
• Control is a measure of the works quality.
• Examples of JCS
• Scissors
• Fighter jets
• Combat robots

11
Contextual Control Model (CoCoM)

• Performance in context
• Different types of behaviors predict better
  outcomes
• Functional not structural approach
• Not about information processing models Memory,
  programs, etc.
• Used in engineered systems
• ABS at Saab
• Nuclear Power Plants

12
CoCoM Main Concepts

• Competencies possible actions in context
• Constructs assumptions about situation
• Control modes characteristics of performance
  that govern quality of performance
• Feed forward and feedback anticipatory versus
  reactive control

13
Control Cycle in Healthcare
14
Control Modes

• Scrambled
• Lack of purposeful activity
• Opportunistic
• Addressing salient characteristics
• Tactical
• Following procedure, limited scope
• Strategic
• Broader scope and higher-level goals

15
Control Characteristics

• Goal Complexity (Number and Interaction)
• Perceived Time Pressure
• Evaluation of Outcome
• Selection of Action
• Expertise
• Motivation
• Familiarity

16
Progress

• The Electronic Health Record context
• Joint Cognitive SystemsContextual Control Model
• Study Design
• Preliminary Findings
• Control characteristics

17
Study Goals

• Immediate Aim
• Translate CoCoM to medication management for
  chronic diseases
• Explore associations between control
  characteristics and surrogate outcomes
• Next Aim
• Establish validity of adapted CoCoM control
  characteristics as predictor of higher quality
  outcomes through simulation

18
Study Design

• Subjects 40-50 physicians, mid-levels,
  residents, nurses, pharmacists in 5 outpatient
  clinics/4 states. Focus on HTN
• Think-aloud protocol Interview
• Saturation coding for control characteristics
• Content analysis
• Multi-dimensional scaling

19
Preliminary Findings

• Semi-Qualitative
• Stories of control modes
• Scrambled
• Opportunistic
• Tactical
• Strategic

20
Scrambled Mode

• Type Trial and error performance
• Case of the new intern and forgetful patient.
• Low information quality and availability
• Low experience

21
Opportunistic Mode

• Type Reaction to salient characteristics
• Have not seen yet for HTN
• Reaction to SBP only
• Pain syndromes even among experienced
• Poor construct of problem
• Low information quality
• Vague goals difficult to resolve competition
• Vague evaluation of outcome not mentioned, then
  OK.

22
Tactical

• Type Following procedure
• Dominant mode for HTN
• Use of protocol
• Focus on procedure (forget clinical goal)
• Minimal consideration of interacting goals
• Low use of feed-forward control
• Problem with information quality-clinical inertia
• Less common in highly experienced MDs

23
Strategic

• Type Broad consideration of context
• Almost exclusively with experienced MDs
• Awareness of protocols but deviation to
  accomplish conflicting patient goals
• Familiarity with past therapy a key factor
• Feed forward strategies account for physiologic
  and organizational factors
• Still, incomplete use of explicit control limits

24
Conclusions

• CoCoM reveals interesting characteristics of
  system performance.
• High-mode characteristics have face validity for
  predicting better outcomes.
• Implications for software design
• Need to support efficient, rich reconstruction of
  mental model of patient
• Need to highlight interaction of goals and
  therapies
• Need to increase time horizon including feed
  forward

25
(No Transcript)