Quality Improvement Strategies for Antibiotic Prescribing

1
Quality Improvement Strategies for Antibiotic
Prescribing

• Sumant Ranji, M.D.
• February 16, 2005

2
Closing the Quality Gap

• Based on subject areas identified in 2003 IOM
  report, Transforming Health Care Quality
• Identifying activities that increase the rate of
  use of practices that are known to be effective
• Synthesis of QI strategies across diseases and
  topic areas

3
Definitions

• Quality Gap difference between observed
  processes/outcomes and those achievable based on
  current knowledge
• Due to deficiency that could be addressed by
  health care system
• Quality improvement strategy any intervention
  aimed at reducing the quality gap for
  representative patients
• Should attempt to improve care for broad group of
  patients
• May involve patient-, provider-, or system-level
  changes
• Quality improvement target outcome/process/struc
  ture the strategy is intended to influence

4
Antibiotic prescribing background

• Majority of prescriptions in US are for acute
  respiratory infections (ARIs)
• 41 million prescriptions in 1998
• 55 of prescriptions for ARIs are unnecessary
• Successes and failures during 1990s
• Significant decline in prescribing overall
• Especially among children
• Inappropriate script rate still 40 for common
  conditions
• Marked increase in use of broad-spectrum agents
• Quinolones, macrolides, 2nd/3rd gen
  cephalosporins, others

5
Quality of prescribing

• Quality Gap
• Unnecessary prescribing of antibiotics for
  non-bacterial illnesses
• Unnecessary use of broad-spectrum antibiotics
  where narrow-spectrum agents are effective
• Quality improvement target
• Rate of antibiotic prescribing for non-bacterial
  diseases
• Frequency of use of broad-spectrum agents

6
Quality Improvement Strategies

• Provider Education
• Audit and Feedback
• Provider Reminders
• Facilitated relay of clinical data to providers
• Patient Education
• Patient self-management
• Patient reminders
• Organizational change
• Financial and regulatory incentives

7
Theoretical Framework for ABX prescribing
Clinician Factors -sociodemographics -training/spe
cialty -knowledge -judgment and
heuristics -perceived patient expectations
Patient Factors -sociodemographics -past
experiences -expressed expectations -reported
symptoms -illness severity
System Factors -practice setting -health plan
features -visit and pharmacy copay -patient
enabling systems -formularies/restrictions -pharma
ceutical detailing
Clinician's Decision to Prescribe Antibiotics
8
Quality Improvement StrategiesPrescribing
specific

• Patient-directed
• Education
• Self-management (delayed prescriptions)
• Financial and regulatory incentives
• Copayments
• Providers
• Education by different modalities
• Audit and feedback of prescribing practices
• Financial and regulatory incentives
• Capitation, restricted formularies
• Combinations of above strategies

9
Research questions

• Which QI strategies reduce antibiotic prescribing
  for acute non-bacterial illnesses?
• Are particular QI strategies more effective for
  certain target conditions?
• Are these strategies safe for patients?
• Effects on health services utilization, clinical
  outcomes, satisfaction
• What are the consequences of these strategies for
  public health and the health care system?
• Effects on antimicrobial resistance, costs of
  prescribing
• Which QI strategies are most effective in
  improving the selection of recommended
  antibiotics?
• Subtopics as above

10
Inclusion/Exclusion Criteria

• Topic
• Acute outpatient illnesses
• Major contributor to problem
• Different theoretical model for inpatient
  prescribing
• Study design
• Evaluate a QI strategy
• RCT, quasi-RCT, CBA, or ITS
• Outcomes
• Measurement of antibiotic prescribing (overall or
  selection)
• Antimicrobial resistance, disease outcomes, costs
  of prescribing, health services utilization,
  satisfaction with care only abstracted if study
  also measured prescribing

11
A priori hypotheses

• Publication bias
• Smaller, non-randomized trials will have greater
  effects
• Effect of baseline prescribing rate
• Studies done in populations where
  over-prescribing/poor selection is common will
  have greater effects
• Targeting of specific diseases
• Studies targeting prescribing for specific
  diseases will be more effective than those
  targeting a variety of conditions or general ABX
  prescribing
• Multifaceted strategies
• As with prior research, studies using multiple QI
  strategies will be more effective than those
  using a single strategy
• Intensity of intervention
• Studies using interventions repeated over time
  will be more effective

12
Search Strategy
EPOC 537 citations
Hand Search 12 citations
549 citations
382 title exclusions
167 articles
93 full text exclusions
54 articles (74 comparisons)
ABX selection 25 articles (33 comparisons)
ABX prescription 34 articles (41 comparisons)
13
Analysis

• Measured outcomes
• ABX prescribing
• visits at which patient received ABX
  prescription
• Prescriptions per person-year
• Prescriptions per provider
• ABX selection
• of total prescriptions written for recommended
  agent
• compliance to clinical guideline for
  prescribing
• Prescriptions for recommended/nonrecommended ABX
  per person
• Prescriptions for recommended/nonrecommended ABX
  per provider

14
Analysis

• Quantitative
• N31 for ABX prescribing, N19 for ABX selection
• Meta-regression planned but failed
• Random effects meta-analysis
• However, extreme heterogeneity (I2 90)
• Median effects semi-quantitative analysis
• Limitations no weighting by sample
  size/variance
• Necessitates stratified analyses
• Does allow preservation of natural study units
• Qualitative
• N10 for ABX prescribing, N14 for ABX selection
• Systematic review format, complementary to above

15
Key Findings(a work in progress)

• Overall effectiveness of QI strategies
• Possible benefit of self-management
• Variable methodologic quality of studies
• No benefit from more intense interventions
• Possible benefit of multifaceted strategies

16
Results studies suitable for quantitative
analysis

• ABX prescription (N31)
• 8 US, 23 non-US
• 26 target prescribing for ARIs
• 18 RCT, 13 CBA
• ABX selection (N19)
• 3 US, 16 non-US
• 12 target choice for ARI, 7 for UTI
• 7 RCT, 12 CBA

17
Study quality

• Failure to properly document intervention
• Rationale for study methodology not explained
• Key study components described inadequately
• Duration and intensity of intervention
• Short follow-up
• Minimal reporting of outcomes beyond prescribing
• Failure to report key data
• e.g. number of patients in study
• Inappropriate statistical analyses
• Unit of analysis errors
• Lack of accounting for temporal trends in
  prescribing

18
Overall results

• QI strategies overall beneficial
• Prescribing Median reduction of 9.0 (IQR
  -16.6 to -3.4) in prescribing of ABX when not
  indicated
• Selection Median increase in prescribing of
  recommended ABX of 13.8 (IQR 4.6 - 19.7)

19
Comparative effects on ABX prescribing
20
Comparative effects on ABX selection
21
Median effect on Prescribing Stratified by study
size and design
22
Publication Bias

• Larger effects among smaller trials
• Less effect of study design type

23
Baseline prescribing rate

• Prescribing studies
• Would expect that studies with high baseline
  prescribing rate may show larger effects
• Not found in our sample, but skew issues
• Selection studies
• Expect baseline low compliance to correlate with
  higher effects
• Also not demonstrated

24
Targeting of specific diseases

• Hypothesize that studies targeting prescribing
  for specific conditions may be more likely to
  show effects
• Not found in our analysis for either prescribing
  or selection studies
• Confounding by sample size?

25
Multifaceted strategies

• Previous work (DM) did not reveal benefit for
  multifaceted strategies
• Possible benefit in prescribing studies
• Median effect -12.0 (IQR -16.0 to -1.7) for
  multifaceted studies (N15)
• Median effect -8.8 (IQR -16.9 to -5.9) for
  single-faceted studies (N16)
• Selection studies single-faceted studies all
  provider education only no difference seen

26
Repeated Interventions

• Complicated by poor description of study details
• No difference found for either prescribing or
  selection studies

27
Other outcomes

• Antimicrobial resistance
• Only assessed in 2 studies both showed no
  effect, but short duration of followup
• Health services utilization
• Assessed in 6 studies in prescribing group
• No increase in return visits, hospitalizations
  seen
• ? Effect on duration of illness
• Patient satisfaction
• Assessed only in delayed prescribing studies
  (N6)
• No significant effect seen
• Costs of prescribing
• Assessed in 7 studies 15-30 reductions seen,
  but in short-term (
• Mostly non-US

28
Conclusions and Hypotheses

• No clear benefit for any single QI strategy
• Possible exception of patient self-management
  (delayed prescribing)
• Poor quality of studies limits interpretation of
  results
• However, overall trials are effective at reducing
  prescribing and improving selection
• Future analyses
• Stratified analyses effects of QI strategies in
  relation to sample size, baseline prescribing,
  study design, disease targeting, country
• Preliminarily no major effects
• Nonparametric (rank-sum) tests for differences
  between groups
• Further attempts at meta-regression
• Common outcome measure for dichotomous and
  continuous studies

29
Thanks

• Stanford
• Vandana Sundaram
• Robyn Lewis
• Kathy McDonald
• Doug Owens
• UCSF
• Ralph Gonzales
• Mike Steinman
• Ottawa
• Kaveh Shojania