Influencing Patient Safety in Rural Primary Care Clinics through Computerized Order Entry

1
Influencing Patient Safety in Rural Primary Care
Clinics through Computerized Order Entry

• Matthew Samore, MD
• VA Salt Lake City Health Care System
• Professor of Internal Medicine
• Adjunct Professor of Biomedical Informatics
• University of Utah

2
Acknowledgments

• Kim Bateman, MD
• Frank Drews, PhD
• Wu Xu, PhD
• Brian Sauer, PhD
• Shobha Phansalkar, PhD
• Charlene Weir, PhD
• Jonathan Nebeker, MD

• Amyanne Wuthrich
• Jose Benuzillo
• Warren Pettey
• Marjorie Carter
• Marci Fjelstad
• Rui Saito
• Shuying Shen

• Partners University of Utah, Healthinsight,
  CaduRx
• Funding
• INFORM study AHRQ R01 HS15413

3
Outline

• Background
• Conceptual framework
• Related work
• Our active studies in patient safety
• Patient safety and health information technology
  in rural settings the INFORM study
• Methods
• Participating clinics
• Health information technology
• Measurements
• Results to date
• Summary and future directions

4
Conceptual framework

• Natural vs. engineered systems
• Natural systems
• Living organisms
• Weather systems
• Engineered systems
• Aircraft
• Computers

5
Conceptual framework

• Commonalities
• Defining properties and functional capabilities
  are causal
• Internal component relationships are
  deterministic
• Potentially predictable relationship between
  internal components of system

6
Conceptual framework

• Differences
• Adaptability
• How does system deal with change
• Natural systems evolved
• No spontaneous change of defining parameters
• Adaptation as a result of unstable previous
  states, to meet the changing demands placed upon
  by environment
• Technical systems
• Intentional design and manufacturing to provide
  solution to a specific, practical problem
• Suitable to perform a number of specific tasks

7
Conceptual framework

• Differences
• Transparency
• Observers ability to observe and understand the
  algorithms that govern a systems operation
• Natural systems
• Non-transparent
• Algorithms describing component interactions and
  state variables have to be deduced and are not
  fully understood
• Models describe and mimic behavior
• Engineered systems
• Designed for transparency
• Relationships are planned and calculated during
  design process
• Despite transparency have complex systems
  sometimes unpredicted and unanticipated emergent
  behavior
• Transparency affects the ability to cope with
  changes of system

8
Conceptual framework

• Differences
• Linearity
• Mathematical description of the relationship
  between output of system to its input
• Natural systems
• Often display non-linear behavior
• Engineered systems designed to be linear
• Makes it possible to solve linear problems
  analytically
• Linear systems output can be predicted from input
  
• Allows application of reductionistic approach

9
Conceptual framework

• Differences
• Predictability
• Allows for anticipation of future state of system
• Natural systems
• More challenging to understand and modify
• Partly due to adaptive and non-linear nature
• Even simple natural systems difficult to predict
• Engineered systems
• Easier to understand and maintain
• Higher complexity increases operator demand
• Maintenance and prediction are possible

10
Conceptual framework

• Clinical care delivery
• Complex natural system that varies across myriad
  settings and care
• Health information technology implementation
• Engineered system meets natural system

11
Our active studies in patient safety

• Medication management
• Real-time detection of prescribing problems
• Medicaid population (Nebeker, Xu, and Sauer)
• VA health care system (Nebeker and Weir)
• Clinical decision support systems in rural
  settings
• Primary care (Samore and Bateman)
• Nursing homes (Rubin)
• Optimal laboratory monitoring intervals (Sauer)
• Error-producing conditions in intensive care
  units, including interruptions and task ambiguity
• Medical device problems
• Use of observation (Drews and Samore)

12
Methods

• INFORM study
• Clinic-randomized trial to evaluate impact of
  computerized clinic order entry tool on clinical
  practice and office efficiency
• Features of computerized clinic order entry tool
• Web-based writing of outpatient orders, including
  prescriptions, immunizations, laboratory tests,
  X-ray studies, and work notes.
• Accessible via any type of computer.
• Electronic transmission of orders to pharmacies
  and other vendors or to local printers
• Common access to records across providers who
  share patients while maintaining strict levels of
  confidentiality.

13
Methods participating clinics

• Rural primary care clinics with minimum of two
  providers were recruited
• None had pre-existing electronic health records
• Eighteen clinics randomly assigned to two groups
• Early implementation (launch in year 1)
• Delayed implementation (launch in year 2)
• Post-launch, 2 clinics closed

14
Methods participating clinics
15
Methods health information technology user
interface
Fumbling for his recline button, Ted
unwittingly instigates a disaster.
16
Methods provider home page
List of all refill requests your staff has
submitted for your approval
Pt Search for a patient record
17
Methods refill requests
Select drug to review request and to edit,
approve or deny
Select patient name to review medication history
18
Methods patient home page

• A brief tour
• Status bar
• Navigation commands Special functions

• Active drug list
• Filters
• Status icons

19
Methods drug favorites
20
Methods drug interactions
21
Methods mild drug interactions
22
Methods measurements

• Observation of office processes
• Efficiency of refill process
• Survey
• Theoretical framework of Information Technology
  Adoption Model (ITAM)
• Focus groups
• Time trials using scenarios
• Comparison handwritten and electronic
  prescription writing
• Time-to-complete task
• Assessment of completeness, legibility, errors
• Electronic prescriptions
• Chart review (Fall, 2007)

23
Methods experimental scenarios

• Instructions for time trials
• For these scenarios, imagine that youre with a
  real patient and work at your normal pace.
• Example
• Next you see a regular patient, Abigail B. Cook.
  She is 74-year old with a history of rheumatoid
  arthritis. She also complains of pain when she
  urinates. The urine culture reveals that the she
  has a UTI due to E. coli that is resistant to
  fluoroquinolones. Prescribe an antibiotic to
  treat her UTI

24
Results impact on workflow

• Medication refill system in one participating
  clinic before implementation of computerized
  clinic order entry

25
Results

• Post-implementation of computerized clinic order
  entry

26
Results survey

• Response to one of the questions pertinent to
  patient safety
• Integrating drug reference and drug interaction
  look-up with prescription writing is useful
• 57 (26 of 45) strongly agree
• (answered 6 or 7 on 7 point Likert scale)
• 9 (4 of 45) strongly disagree
• (answered 1 or 2 on 7 point Likert scale)

27
Results focus groups

• Favorable
• We see the interactions. That helps quite a bit.
  So, that has made a difference a couple of times.
  Essentially changed my mind
• One thing I liked, that there was a medicine
  that I dont prescribe very often and didnt know
  about the dose

28
Results focus groups

• Unfavorable
• its like getting those stupid things from the
  pharmacy. It commented on too many menial,
  insignificant things that just waste your time.
  Way too sensitive.
• the drug to drug reaction says major and you
  Look at other look at ePocrates later and no
  reaction at all.

29
Results time trials
30
Results drug interaction scenario

• Frequency with which trimethoprim/sulfa was
  prescribed to elderly woman on methotrexate for
  rheumatoid arthritis
• Handwritten
• 11 of 13 instances
• Computerized clinic order entry tool
• 3 of 12 instances
• Recommendation of the Multum database
• Generally avoid

31
Results electronic prescriptions

• Analysis of instances of significant
  drug-interactions is in progress
• Altogether
• 430,000 electronic prescriptions
• 52,000 unique patients
• Coumadin interactions
• 7 patients received coumadin and amiodarone
  concomitantly
• 29 patients received coumadin and NSAID
  (excluding celecoxib)

32
Summary and future directions

• Implementation is clinical care system redesign
• Measures of impact
• Perceived usefulness versus average effect
• Experiment versus experience in the wild
• Errors versus adverse events
• Transition to health information technology that
  enhances patient-centered care