Antibiotic Prescribing at CHOP: Primary Care
Jeffrey S. Gerber MD, PhD, MSCE
Division of Infectious Diseases
The Children’s Hospital of Philadelphia
• Primary Care Pediatrics
Bob Grundmeier, Alex Fiks, Mort Wasserman• General Pediatrics
Lou Bell, Ron Keren• Pediatric Infectious Diseases
Theo Zaoutis, Priya Prasad, Jeff Gerber• Biostatistics/data management
Russell Localio, Lihai Song• PeRC Administrator
Jim Massey
Study Team
Agenda
1. Rationale for assessing antibiotic use
2. Antibiotic prescribing data• across-practice analyses• within-clinician analyses
3. Intervention
Agenda
1. Rationale for assessing antibiotic use
2. Antibiotic prescribing data• across-practice analyses• within-clinician analyses
3. Intervention
AHRQ Goal
To implement and evaluate evidence-based methods or strategies for reducing the inappropriate use of antibiotics in primary care office practices
• must address:1. conditions for which abx are not effective2. broad-spectrum antibiotic use when
narrow-spectrum antibiotics are indicated
Background
• about half of antibiotic use is unnecessary• overuse well-documented in primary care• antibiotic overuse leads to:
bacterial resistance drug-related adverse events increases in health care costs
$20 billion estimated by IOM
Antibiotic Resistance
Resistance Aside. . .
• 5%–25% diarrhea• 1 in 1000 visit emergency department for
adverse effect of antibiotic– comparable to insulin, warfarin, and digoxin
• 1 in 4000 chance that an antibiotic will prevent serious complication from URI
Shehab N. CID 2008:47; Linder JA. CID 2008:47
Antimicrobial Stewardship
• Antimicrobial Stewardship Programs recommended for hospitals
• most antibiotic use (and misuse) occurs in the outpatient setting
• is outpatient “stewardship” achievable?
Agenda
1. Rationale for assessing antibiotic use
2. Antibiotic prescribing data• across-practice analyses• within-clinician analyses
3. Intervention
Study Setting: CHOP Care Network
• 5 urban, academic
• 24 “private” practices
urban, suburban, rural
• common EHR
Case Definitions
• ICD9 codes for common infections (+/- GAS testing, antibiotic use)
verified by chart review and provider feedback
• Excluding:– antibiotic allergy– visit within prior 3 months with antibiotic– concurrent bacterial infection
• AOM, SSTI, UTI, lyme, acne, chronic sinusitis, mycoplasma, scarlet fever, animal bite, proph, oral infections, pertussis, STD, bone/joint
– complex chronic conditions (Feudtner, Pediatrics 2000)
Broad-Spectrum Antibiotics
• amoxicillin-clavulanate• cephalosporins• azithromycin*
*not considered broad-spectrum therapy for pneumonia
Table 1. Demographic characteristics of the study cohort, by site
1,296,517 Encounters
51,421 narrow ABX
29,635 broad ABX
102,102 antibiotic Rx
8,204prior ABX
14,298 ABX allergy
399,793 sick visits
630,502 office visits
363,049 sick visits
230,709 preventive
666,015phone, refills
36,744 visits w/ CCC
260,947no antibiotics
Antibiotic Prescribing for Sick Visits
Excluding: preventive visits, CCCStandardized by: age, sex, age-sex, race, Medicaid
Antibiotic Prescribing: Std for ARTI Dx
Excluding: preventive visits, CCCStandardized by: age, sex, age-sex, race, Medicaid, ARTI Dx
Broad Antibiotic Prescribing
Excluding: preventive visits, CCC, antibiotic allergy, prior antibioticsStandardized by: age, sex, age-sex, race, Medicaid
Broad Antibiotics: Std ARTI Dx
Excluding: preventive visits, CCC, antibiotic allergy, prior antibioticsStandardized by: age, sex, age-sex, race, Medicaid, ARTI Dx
Diagnosis rate of AOM
Excluding: preventive visits, CCC, prior antibioticsStandardized by: age, sex, age-sex, race, Medicaid
Broad Antibiotics for AOM
Excluding: preventive visits, CCC, prior antibioticsStandardized by: age, sex, age-sex, race, Medicaid
Broad Antibiotics for Sinusitis
Excluding: preventive visits, CCC, antibiotic allergy, prior antibioticsStandardized by: age, sex, age-sex, race, Medicaid
Broad Antibiotics for GAS pharyngitis
Excluding: preventive visits, CCC, antibiotic allergy, prior antibioticsStandardized by: age, sex, age-sex, race, Medicaid
Broad Antibiotics for Pneumonia
Excluding: preventive visits, CCC, antibiotic allergy, prior antibioticsStandardized by: age, sex, age-sex, race, Medicaid
Summary of variability data
• antibiotic prescribing at sick visits varies significantly across practice sites
• broad-spectrum antibiotic prescribing at sick visits varies significantly across practice sites
• the rate of diagnosis of ARTIs varies significantly across practice sites
• adherence to prescribing guidelines for AOM, sinusitis, GAS pharyngitis, and pneumonia varies significantly across practice sites
Agenda
1. Rationale for assessing antibiotic use
2. Antibiotic prescribing data• across-practice analyses• within-clinician analyses
3. Intervention
Antibiotic Prescribing by Patient Race
• within clinician analyses of antibiotic prescribing and diagnoses in same cohort
• Excluding:– complex chronic conditions– preventive visits, asthma, (allergy, prior antibiotics)
• Adjusted for:– sex, age category (0-1; 1-5; 6-10; 11-18)– Medicaid, site
Antibiotic Prescribing by Patient Race
OR (black) 95% CI Margins P-value
0.764 0.738, 0.790 0.29, 0.24 <0.0001
Receipt of antibiotic prescription per SICK VISIT:
• Excluding: CCC, asthma
• Adjusted for: age category, sex, Medicaid
Visit Rate by Patient Race
Sick visits per year by race:
Primary care Black Non-black
sick visits 1.2 2.0
preventive visits 1.1 1.1
CHOP ED (5 practices) Black Non-black
all ED visits 0.57 0.63
ED visits for ARTI 0.02 0.02
Antibiotic Prescribing by Patient Race
IRR (black) 95% CI P-value
0.64 0.63, 0.65 <0.0001
Receipt of antibiotic prescription per CHILD:
• Excluding: CCC
• Adjusted for: age category, sex, Medicaid
Diagnosis by Patient Race
Diagnosis of various ARTIs:
condition OR 95% CI Margins P-valueAOM 0.767 0.735, 0.801 0.15, 0.12 <0.0001
acute sinusitis 0.817 0.761, 0.877 0.06, 0.05 <0.0001
GAS pharyngitis 0.623 0.576, 0.674 0.05, 0.03 <0.0001
pneumonia 1.058 0.963, 1.163 0.02, 0.02 0.235
UTI 0.985 0.903, 1.074 0.02, 0.02 0.733
• Excluding: CCC, asthma
• Adjusted for: age category, sex, Medicaid
Antibiotic Prescribing by Patient Race
OR 95% CI Margins P-value
0.834 0.781, 0.891 0.36, 0.32 <0.0001
Receipt of broad-spectrum antibiotic (if any antibiotic prescribed)
• Excluding: CCC, asthma, allergy
• Adjusted for: age category, sex, Medicaid
Antibiotic Prescribing by Patient Race
Receipt of broad antibiotics for ARTI:
condition OR 95% CI Margins P-value
AOM 0.737 0.662, 0.821
0.38, 0.31 <0.0001
GAS pharyngitis 0.849 0.569, 1.266
0.08, 0.07 0.421
sinusitis 0.947 0.814, 1.102
0.44, 0.43 0.483
pneumonia 1.003 0.712, 1.412
0.17, 0.17 0.988
• Excluding: CCC, asthma, allergy
• Adjusted for: age category, sex, Medicaid
Summary of race data
• black children receive fewer antibiotic prescriptions per sick visit and per child than non-black children
• black children are diagnosed with less ARTI than non-black children
• when diagnosed with AOM, black children receive more appropriate (i.e. less broad-spectrum) antibiotics
• black children have less sick visits than non-black children (but equal number of well visits)
Why?
• confounding?• difference in epidemiology of disease,
including BOTH prevalence and severity of illness linked with race?
• parental expectations/pressure linked with race?
• perception of parental expectations/pressure linked with race?
Agenda
1. Rationale for assessing antibiotic use
2. Antibiotic prescribing data• across-practice analyses• within-clinician analyses
3. Intervention
Specific Aim
• To determine the impact of an outpatient antimicrobial stewardship bundle within a pediatric primary care network on antibiotic prescribing for common ARTI:1. Antibiotic prescribing for viral infections
2. Broad-spectrum antibiotic prescribing for conditions for which narrow-spectrum antibiotics are indicated.
Study Design
• cluster-randomized controlled trial• bundled intervention vs. no intervention• unit of observation will be the practitioner
but randomized at practice level– natural distribution of physicians– avoids intra-practice contamination
Intervention
1. guideline development
2. education
3. audit and feedback
Why Might Unnecessary Prescribing Occur?
Prescribing Awareness
Antibiotic Prescribing
Parental Expectations
Knowledge Gaps
Diagnostic Challenges
Time Constraints
Parental Expectations
Diagnostic Challenges
Time Constraints
Knowledge Gaps
Prescribing Awareness
Why Might Unnecessary Prescribing Occur?
Antibiotic Prescribing
Hypotheses
1. clinicians have incomplete knowledge of the data regarding the effectiveness of antibiotics for respiratory tract infections
GAS and broad spectrum antibiotics antibiotic activity against pneumococcus prevention of bacterial superinfection role of moraxella and Hflu in disease
2. clinicians are unaware of/have not been presented with data regarding their own prescribing of antibiotics
Education
• on site, interactive sessions for each practice randomized to the intervention– present the purpose of the study– discuss guideline development/contents– instruct how to access guidelines– explain audit & feedback– present baseline data– gather feedback
Guidelines
• review AAP and Red Book guidelines• pediatric primary care/ID/clinical pharmacy• modified if necessary• generate benchmarks
GAS: Rationale for penicillin/amox
• GAS resistance to pcn has NEVER been seen • azithromycin and cephalosporins
have NOT been shown to be superior for pharyngitis or for prevention of sequelae
data does not support increased patient compliance over oral penicillin or amoxicillin.
exposure promotes resistance to these and other antibiotics.
AAP/Red Book endorsed
Guideline Access
• email (pdf)• EPIC link:
linked to chief complaint NOT decision support optional no workflow interruption
PARTI
Study Setting: CHOP Care Network
5 urban, academic
24 “private” practices urban suburban rural
VIRALcommon coldURIacute bronchitistonsillitispharyngitis (non-strep)
Outcomes
no antibiotics
BACTERIALacute sinusitisStrep pharyngitispneumonia
penicillin/amoxicillin
Case Definitions
• ICD9 codes for common infections (+/- GAS testing, antibiotic use)
verified by chart review and provider feedback
• Excluding:– antibiotic allergy– visit within prior 3 months with antibiotic– concurrent bacterial infection
• AOM, SSTI, UTI, lyme, acne, chronic sinusitis, mycoplasma, scarlet fever, animal bite, proph, oral infections, pertussis, STD, bone/joint
– children with complex chronic diseases
Data Collection
• EPIC EMR• ICD9 coding
– diagnoses– chronic medical conditions
• antibiotic orders• telephone encounters• age, race/ethnicity, sex, insurance, allergies• provider: degree, yr grad, sex, % effort, practice
volume, support staff
Analysis/Sample Size
• descriptive analysis of changes within and among sites.
• multivariable repeated measures analysis using generalized linear models
• 140 clinicians; 70 each arm• power > 0.9 to detect 10% improvement in
prescribing
Randomization
• 22 of 24 Enrolled (18 “sites”)• 143,254 patients; 512,943 encounters
– 49.5% female– 69% White
• each site enumerated by location and volume• block-randomized 9 sites to each arm
Intervention: Timeline
12 months ofaudit/feedback
12 months afterfeedback ends
12 monthsbaseline data
Site presentation
Feedback reports
**
*
*
Some Limitations
• ICD9 codes– misclassification of outcome– intervention may change coding
• contamination of intervention• lack of “buy-in” by practitioners• generalizability
Future Directions
• complete analysis• assess durability of effect (if there is one)• gather qualitative data from providers
• predictors of prescribing• clinical pathways/decision support?