IMPROVING PATIENT SAFETY BY REDUCING MEDICATION ERRORS

1
IMPROVING PATIENT SAFETYBY REDUCING MEDICATION
ERRORS

• Brian L. Strom, M.D., M.P.H.
• Professor of Biostatistics and Epidemiology
• Center for Clinical Epidemiology and
  Biostatistics
• University of Pennsylvania School of Medicine

2
IMPROVING PATIENT SAFETY BYREDUCING MEDICATION
ERRORS THEME

• AHRQ Center of Excellence for Patient Safety
  Research and Practice
• Theme Improving Patient Safety Through Reduction
  of Errors in the Medication Use Process
• PRIME Program for Reduction In Medication Errors

3
OVERALL APPROACH

• Focus - the occurrence of errors anywhere along
  the entire pathway of medication use
• Drugs with ubiquitous use, capacity to lead to
  errors, and severity of consequences
• Different settings and populations
• Both human psychosocial factors and technical
  system factors
• Evaluations in sites prepared to rapidly
  implement the studies findings, which could then
  be evaluated in future studies

4
IMPROVING PATIENT SAFETYBY REDUCING MEDICATION
ERRORSOVERALL ORGANIZATION

• Four projects
• Four cores
• Administrative Core
• Data Collection Core
• Biostatistics and Data Management Core
• Dissemination Core

5
Project One Medication Errors Leading to
Hospitalization Among the Elderly (Joshua Metlay,
MD, PhD--PI)

• Identify predisposing factors for hospitalization
  due to medication errors among the elderly
• Develop a prediction rule to identify high risk
  elderly patients
• Estimate costs associated with hospitalizations
  due to errors

6
DESIGN

• Prospective cohort
• 3 Drug Groups
• Chronic vs. New Users
• 5 Study Cohorts (insufficient numbers for new
  phenytoin)
• One year enrollment
• Two year follow-up

7
HYPOTHESES

• Key risk factors include uncoordinated medical
  and pharmaceutical care, inadequate delivery of
  new medication instructions, visual and cognitive
  impairment, depression
• Risk factors differ across types of drugs and new
  and old users

8
DRUG TARGETS

• Focus on warfarin, phenytoin, digoxin
• High risk drugs, frequently implicated in ADEs
  leading to hospitalization
• Narrow therapeutic windows lead to drug level
  monitoring

9
Medication Grouping Enrolled in Study (surveyed)
N 5569
10
Cohort CharacteristicsN5569
11
Source of Medication Instructions(N5569)
12
Content of Medication Discussions with MD
(N5569)
13
Medication Dispensing Patterns in the Home
(N5569)
14
Project Two Predictors for PoorAdherence to
Warfarin Therapy(Stephen Kimmel, MD, MSCE--PI)

• To determine the clinical, demographic,
  organizational, behavioral, and psychosocial
  predictors of poor adherence
• To develop a predictive index that can identify
  patients at high risk for medication errors
  before starting therapy

15
PROJECT TWOSTUDY DESIGN

• Prospective cohort design, enrolling adult
  patients requiring warfarin who are treated at a
  outpatient pharmacist-managed Anticoagulation
  Clinic (AC)
• Patients presenting to the AC clinic will be
  identified at the start of therapy and followed
  throughout their course
• An addition to a funded NIH study designed to
  examine the effects of genetic polymorphisms and
  adherence on clinical outcomes (INR levels,
  bleeding, and thromboembolism)

16
Project Three Inpatient Medication Errors
Leading to Acute Renal Failure(Harold Feldman,
MD, MSCEPI)

• To evaluate antibiotic monitoring practices that
  may predispose to ARF including
• The failure to use pharmacokinetic monitoring
• Delays in initiating pharmacokinetic monitoring
• Failure to implement recommendations from the
  pharmacokinetic monitoring service
• Pharmacokinetic monitoring service
  characteristics/procedures systems

17
PROJECT THREE STUDY DESIGN

• Hospital-based case-control study nested within a
  cohort of hospitalized patients receiving
  aminoglycosides
• Cases of ARF (defined by elevations in serum
  creatinine) occurring among patients receiving
  aminoglycoside antibiotics to be compared to
  random sample of controls not experiencing ARF
• Data collection structured review of medical
  records and evaluation of their interaction with
  the pharmacokinetic monitoring service prior to
  the occurrence of ARF for the cases, or during an
  analogous exposure time for controls

18
Project Four Medication ErrorsRelated to
Workplace Stressors(Ross Koppel, PhD--PI)

• To determine if, and to what extent, the
  organization of work within a hospital, (e.g.,
  schedules, shifts, workloads) affects
  houseofficers risk of medication errors
• To determine if houseofficers experience of
  workplace stress affects the risk of medication
  errors
• To determine how stressors interact with
  psychological profiles to influence the risk of
  medication errors

19
PROJECT FOURSTUDY DESIGN

• A series of cross sectional studies
• Data collection
• 1) analysis of house-officers workloads, shifts,
  schedules 2) surveys of houseofficersat several
  points in their trainingabout workplace
  stressors and strain 3) one-on-one interviews
  with housestaff, pharmacists, IT staff 4)
  focus groups 5) observations on hospital floors,
  in pharmacy and meetings 6) psychometric
  personality inventory

20
PROJECT FOUR OUTCOMES

• The near misses for medication errors detected
  by experienced pharmacistsin relation to
  houseofficers workloads, fatigue, schedules,
  rotations, shifts, experience, etc
• Self-reported strains and errors in relation to
  workplace stressors
• Analysis of the physician computer ordering
  system in preventing and in, perhaps,
  facilitating error

21
PROJECT FOUR QUALITATIVE SUMMARY OUTCOMES

• An emerging theme focused on the errors created
  by technological solutions designed to reduce
  errors
• Several examples illustrate the unintended harms
  caused by the commercial CPOE system (TDS)

22
AHRQ Center of Excellence for Patient Safety
Research and Practice

• Now that we have developed hypotheses for
  possible interventions, who will evaluate them????

23
Center for Excellence in Patient Medication
SafetyInvestigative Teams

• Brian L. Strom, MD, MPH PI Director, Admin Core
  
• Dir, Data Collection Core
• Joshua Metlay, MD, PhD Co-PI Proj Leader, Proj
  1
• Co-Dir, Admin Core
• David Asch, MD, MBA Dir, Dissemination Core
• Lily Cheung, PharmD Co-investigator, Project 3
• Abigail Cohen, PhD, MA Senior Project Manager
• Dean Cruess, PhD Co-inv, Project 2
• John Farrar, MD, MSCE, PhD Co-inv, Data
  Collection Core
• Harold Feldman, MD, MSCE Project Leader, Project
  3

24
Center for Excellence in Patient Medication
SafetyInvestigative Team

• Sean Hennessy, PharmD, Co-inv, Project 1,
  Project 3,
• MSCE, PhD Data Collection Core
• Stephen Kimmel, MD, MSCE Project Leader, Project
  2
• Robert Gross, MD, MSCE Co-inv, Project 2
• Ross Koppel, PhD Project Leader, Project 4
• Russell Localio, JD, MS Dir, Biostats Data
• Management Core
• Sandra Norman, PhD Co-Dir, Data Collection Core
• Daniel Polsky, PhD Co-inv, Project 1

25
METHODS OF PROVIDING MEDICATION INSTRUCTIONS
26
Adherence to Warfarin Sodium Using Electronic
Pill-Cap Monitoring and the Millon Behavioral
Medicine Diagnostic Inventory

• Adherence assessed with multiple techniques,
  electronic monitoring and psychosocial instrument
  that predict adherence.
• Electronic monitoring device (MEMS caps)
• Makes use of a microprocessor in the cap of the
  pill bottle to record exact dates/times the
  bottle is opened
• A standard cutoff of gt 25 of days on which the
  prescribed dose was not taken was used to
  designate non-compliant participants
• A 140-item T/F questionnaire (MBMD) that produces
  a number of psychosocial indices, including
  Problematic Compliance
• Non-compliance was designated for participants
  with scores gt 75 on the Problematic Compliance
  Index

27
PROJECT TWOStudy Design Participants

• 44 participants, (28 men, 16 women, mean age
  51.514.7 years) beginning warfarin therapy at an
  anticoagulation clinic
• Indications for warfarin included thrombosis,
  atrial fibrillation/flutter embolism, and
  myocardial infarction.
• Pill cap adherence was monitored and averaged
  over the time of pill-cap use (minimum of 7
  days)
• Participants completed the MBMD inventory, along
  with other psychosocial measures towards the
  start of pill-cap monitoring

28
PROJECT TWO DATA COLLECTION

• Data collection 1) demographics, 2) clinical
  characteristics, 3) health-care structure
  characteristics, 4) pill taking practices, 5)
  psychosocial variables, 6) study outcomes
• The primary outcome is adherence, to be measured
  using an electronic data monitoring system

29
PROJECT TWOResults

• Table 1. Adherence results

Adherent gt 75 of days with correct dose
taken for pill cap measurement and MBMD scores
of lt 75 on the Problematic Compliance Index
Table 2. Agreement of adherence results
(k.062)
30
PROJECT TWO Discussion

• While the pill-cap monitoring and MBMD
  problematic compliance index produced similar
  percentages of adherent and non-adherent
  participants, a closer examination of the data
  found that the two measures did not have adequate
  statistical agreement.
• It appears that the MBMD inventory assesses
  general medical compliance, while pill-cap
  monitoring limits compliance to pharmacological
  regimens.

31
PROJECT THREE Schematic Protocol
32
PROJECT THREEAbstraction Form
33
First Wave SampleDemographics (n261)

• Female 44.8
• White 66.3
• Mean Age 29.6 yrs
• PGY1 32.6
• PGY2 32.2
• PGY3 35.2

34
First Wave Specialty (n261)

• Medicine 47.1
• Neurology 10.7
• General Surgery 10.0
• Emergency Medicine 09.6
• Ob/Gyn 06.9
• Otorhyno 06.1
• Family Medicine 05.7
• Urology 02.3
• Neurosurgery 01.5

35
Computerized Order Entry Survey ResultsWave One
(n261)
36
Computerized Order Entry Survey ResultsWave One
cont. (n261)
37
CPOE vs. Paper-Based SystemsThe Benefits of
CPOE (part 1)

• 1) free of handwriting identification problems
• 2) faster to reach the pharmacy
• 3) less subject to error associated with similar
  drug names
• 4) more easily integrated into medical records
  and decision support systems
• 5) less subject to errors caused by use of
  apothecary measures
• 6) easily linked to drug-drug interaction
  warnings
• 7) more likely to identify the prescribing
  physician
• 8) able to link to ADE reporting systems

38
CPOE vs. Paper-Based SystemsThe Benefits of
CPOE, Part II

• 9) able to avoid specification errors, such as
  trailing zeros
• 10) available and appropriate for training and
  education
• 11) available for immediate data analysis,
  including post marketing reporting.
• Also, with on-line prompts CPOE systems can
• 12) link to algorithms to emphasize
  cost-effective medications
• 13) reduce under-prescribing and
  over-prescribing and
• 14) reduce the incorrect choice of drugs.

39
Increasing Interest in CPOE

• CPOE adoption has, perhaps, gathered such strong
  support because its promise is so great, the
  effects of medication error so distressing, the
  circumstances of medication error so preventable,
  and the studies of CPOE so reassuring, albeit
  preliminary.

40
Our Studys Genesis

• Project objective The role of hospital workplace
  stressors (e.g., shifts, sleeplessness, new
  rotations) on housestaff medication prescribing
  errors.
• The CPOE system emerged as a study focus when
  housestaff repeatedly told us it caused stress
  and error

41
Our CPOE Study Methods

• Face-to-face intensive interviews with physicians
  
• 5 focus groups with physicians
• Shadowing docs as they entered orders and at
  handoffs
• Executive interviews with leaders of Nurses,
  Medicine/Surgery, Pharmacy, IT
• Face-to-face interviews with nurses and
  nurse-managers
• Shadowed nurses and pharmacists processing
  orders, interacting with CPOE system
• 72 item questionnaire to 90 sample of housestaff
  (9 questions about CPOE)

42
Measuring Success

• CPOE systems are currently found in only 5 to 9
  of hospitals
• CPOE systems efficacy (17 to 81 error
  reduction) usually focus on their advantages and
  are generally limited to single outcome studies,
  potential error reduction, or physician
  satisfaction

43
PROJECT FOUR QUALITATIVE SUMMARY OUTCOMES -1

• Sometimes serious delays were caused because of
  the re-approval process for antibiotics done by
  infectious disease fellows
• While a sticker is placed on chart the day before
  a renewal is needed, most rely on the computer
  system (TDS) which may not be updated due to the
  difficulty of working with the system to re-enter
  data

44
PROJECT FOUR QUALITATIVE SUMMARY OUTCOMES-2

• Doctors sometimes use computer displays to judge
  lowest dose range of doses
• TDS is not designed to illustrate dose or range
  information for clinical decisions
• The dosages displayed reflect purchasing and
  warehousing considerations by the pharmacy
• Computer gives a false sense of accuracy

45
PROJECT FOUR QUALITATIVE SUMMARY OUTCOMES-3

• Doctor must go though 15 or more computer screens
  to discontinue a medication causing problems with
  missed information
• If houseofficer is interrupted or in hurry,
  process maybe postponed and forgotten

46
PROJECT FOUR QUALITATIVE SUMMARY OUTCOMES-4

• Nurses often don't record the administration of
  medications in the TDS system due to cumbersome
  and time consuming nature of interface
• Doctors must assume medications given when
  ordered or find nurse and ask
• System is updated only at the end of a shift to
  reconcile records
• Accuracy questionable

47
PROJECT FOUR QUALITATIVE SUMMARY OUTCOMES-4

• When a patient enters OR all medications stopped 
• Surgeons must enter continuing and new
  medications into TDS for use after post-op 
• When patient released from the post anesthesia
  care unit, the nurse pre-activates meds
• But doctor must re-re-activate meds
• This last step is too often forgotten

48
PROJECT FOURPROGRESS AND PRELIMINARY FINDINGS

• Creation and administration of house officer
  stressor inventory
• Findings on differing causes of stress vs. causes
  of medication error
• Findings on the role of CPOE systems as
  facilitator of error (in addition to its
  preventive role)
• Findings of greater stress and error among
  residents compared to interns

49
Findings 22 medication error risks facilitated
by CPOE

• I) Information Errors generated by fragmentation
  of data and failure to integrate the hospitals
  several computer and information systems
• II. Human-Machine Interface Flaws Reflecting
  machine rules that do not correspond to work
  organization or usual behaviors

50
Ex Assumed Minimum Dose Assumed Dose Range
Information

• 73 of housestaff use CPOE displays to determine
  low doses
• 82 used CPOE displays to determine range of
  doses
• 40 used CPOE to determine dosages at least a few
  times weekly 10 to 14 daily

51
Ex. Patient Selection--Screen Design
Inconsistent Color/Font Coding

• 55 of housestaff Difficulty identifying the
  patient they were ordering for because of
  fragmented CPOE displays
• 23 say this happened a few times weekly or more
  frequently

52
Our RECOMMENDATIONS concentrateon organizational
factors

• Focus primarily on the organization of work not
  on technology. CPOE must only determine clinical
  actions if they improve care.
• Aggressively examine the technology in use.
  Problems obscured by workarounds, medical problem
  solving ethos, and low housestaff status.
• Substitution of technology for people is a
  misunderstanding of both. Aggressively fix
  technology when shown to be counter-productive.
  Failure to do so engenders alienation and
  dangerous workarounds.

53
Our RECOMMENDATIONS concentrateon organizational
factors

• Episodic and incomplete error reporting are
  standard. Management belief in these reports
  obfuscates and compounds problems. Pursue errors
  second stories.
• Plan for continuous revisions and quality
  improvement recognizing that all changes
  generate new error risks.

54
CPOE Tremendous Promise

• But do not bend hospitals and clinicians around
  the CPOE system make CPOE work with other
  systems and with clinicians.
• Our future research New CPOE systems may be
  better, but face the same and new challenges of
  integration with workflow, humans, and
  organizations.