[Skip to Content]
Sign In
Individual Sign In
Create an Account
Institutional Sign In
OpenAthens Shibboleth
Purchase Options:
[Skip to Content Landing]
Figure 1.  Flowchart of Retrospective Cohort to Assess Clinical Outcomes
Flowchart of Retrospective Cohort to Assess Clinical Outcomes

aThe numbers for the individual conditions sum to more than the total because children could be diagnosed with more than 1 type of acute respiratory infection.

Figure 2.  Flowchart of Prospective Cohort to Assess Patient-Centered Outcomes
Flowchart of Prospective Cohort to Assess Patient-Centered Outcomes

aOutcomes assessed: Pediatric Quality of Life Inventory, missed school or day care, caretaker required additional childcare or missed commitments, symptom resolution, and sleep disturbance.

bOutcome assessed: adverse events.

Table 1.  Patient and Clinical Characteristics of Children in the Retrospective Cohort to Assess Clinical Outcomes
Patient and Clinical Characteristics of Children in the Retrospective Cohort to Assess Clinical Outcomes
Table 2.  Clinical Outcomes Among 30 086 Children in the Retrospective Cohort
Clinical Outcomes Among 30 086 Children in the Retrospective Cohort
Table 3.  Patient and Clinical Characteristics of Children in the Prospective Cohort to Assess Patient-Centered Outcomes
Patient and Clinical Characteristics of Children in the Prospective Cohort to Assess Patient-Centered Outcomes
Table 4.  Patient-Centered Outcomes in the Prospective Cohort
Patient-Centered Outcomes in the Prospective Cohort
1.
Chai  G, Governale  L, McMahon  AW, Trinidad  JP, Staffa  J, Murphy  D.  Trends of outpatient prescription drug utilization in US children, 2002-2010.  Pediatrics. 2012;130(1):23-31.PubMedGoogle ScholarCrossref
2.
Hersh  AL, Shapiro  DJ, Pavia  AT, Shah  SS.  Antibiotic prescribing in ambulatory pediatrics in the United States.  Pediatrics. 2011;128(6):1053-1061.PubMedGoogle ScholarCrossref
3.
Grijalva  CG, Nuorti  JP, Griffin  MR.  Antibiotic prescription rates for acute respiratory tract infections in US ambulatory settings.  JAMA. 2009;302(7):758-766.PubMedGoogle ScholarCrossref
4.
McCaig  LF, Besser  RE, Hughes  JM.  Antimicrobial drug prescription in ambulatory care settings, United States, 1992-2000.  Emerg Infect Dis. 2003;9(4):432-437.PubMedGoogle ScholarCrossref
5.
Steinman  MA, Landefeld  CS, Gonzales  R.  Predictors of broad-spectrum antibiotic prescribing for acute respiratory tract infections in adult primary care.  JAMA. 2003;289(6):719-725.PubMedGoogle ScholarCrossref
6.
Hersh  AL, Fleming-Dutra  KE, Shapiro  DJ, Hyun  DY, Hicks  LA; Outpatient Antibiotic Use Target-Setting Workgroup.  Frequency of first-line antibiotic selection among US ambulatory care visits for otitis media, sinusitis, and pharyngitis.  JAMA Intern Med. 2016;176(12):1870-1872.PubMedGoogle ScholarCrossref
7.
Shehab  N, Lovegrove  MC, Geller  AI, Rose  KO, Weidle  NJ, Budnitz  DS.  US emergency department visits for outpatient adverse drug events, 2013-2014.  JAMA. 2016;316(20):2115-2125.PubMedGoogle ScholarCrossref
8.
Kimberlin  DW, Brady  MT, Jackson  MA, Long  SS.  Red Book: 2015 Report of the Committee on Infectious Diseases. 30th ed. Elk Grove Village, IL: American Academy of Pediatrics; 2015.
9.
Tähtinen  PA, Laine  MK, Huovinen  P, Jalava  J, Ruuskanen  O, Ruohola  A.  A placebo-controlled trial of antimicrobial treatment for acute otitis media.  N Engl J Med. 2011;364(2):116-126.PubMedGoogle ScholarCrossref
10.
Hoberman  A, Paradise  JL, Rockette  HE,  et al.  Treatment of acute otitis media in children under 2 years of age.  N Engl J Med. 2011;364(2):105-115.PubMedGoogle ScholarCrossref
11.
Chow  AW, Benninger  MS, Brook  I,  et al; Infectious Diseases Society of America.  IDSA clinical practice guideline for acute bacterial rhinosinusitis in children and adults.  Clin Infect Dis. 2012;54(8):e72-e112.PubMedGoogle ScholarCrossref
12.
Casey  JR, Pichichero  ME.  Meta-analysis of cephalosporin versus penicillin treatment of group A streptococcal tonsillopharyngitis in children.  Pediatrics. 2004;113(4):866-882.PubMedGoogle ScholarCrossref
13.
Coker  TR, Chan  LS, Newberry  SJ,  et al.  Diagnosis, microbial epidemiology, and antibiotic treatment of acute otitis media in children: a systematic review.  JAMA. 2010;304(19):2161-2169.PubMedGoogle ScholarCrossref
14.
Institute of Medicine.  Crossing the Quality Chasm: A New Health System for the 21st Century. Washington, DC: The National Academies Press; 2001.
15.
Gabriel  SE, Normand  SL.  Getting the methods right—the foundation of patient-centered outcomes research.  N Engl J Med. 2012;367(9):787-790.PubMedGoogle ScholarCrossref
16.
Fiks  AG, Grundmeier  RW, Margolis  B,  et al.  Comparative effectiveness research using the electronic medical record: an emerging area of investigation in pediatric primary care.  J Pediatr. 2012;160(5):719-724.PubMedGoogle ScholarCrossref
17.
Shulman  ST, Bisno  AL, Clegg  HW,  et al.  Clinical practice guideline for the diagnosis and management of group A streptococcal pharyngitis: 2012 update by the Infectious Diseases Society of America [published correction appears in Clin Infect Dis. 2014;58(10):1496].  Clin Infect Dis. 2012;55(10):1279-1282.PubMedGoogle ScholarCrossref
18.
Feudtner  C, Christakis  DA, Connell  FA.  Pediatric deaths attributable to complex chronic conditions: a population-based study of Washington State, 1980-1997.  Pediatrics. 2000;106(1 pt 2):205-209.PubMedGoogle Scholar
19.
Gerber  JS, Prasad  PA, Russell Localio  A,  et al.  Variation in antibiotic prescribing across a pediatric primary care network.  J Pediatric Infect Dis Soc. 2015;4(4):297-304.PubMedGoogle Scholar
20.
Fleming-Dutra  KE, Demirjian  A, Bartoces  M, Roberts  RM, Taylor  TH  Jr, Hicks  LA.  Variations in antibiotic and azithromycin prescribing for children by geography and specialty—United States, 2013 [published online July 19, 2017].  Pediatr Infect Dis J. doi:10.1097/INF.0000000000001708PubMedGoogle Scholar
21.
Wald  ER, Applegate  KE, Bordley  C,  et al; American Academy of Pediatrics.  Clinical practice guideline for the diagnosis and management of acute bacterial sinusitis in children aged 1 to 18 years.  Pediatrics. 2013;132(1):e262-e280.PubMedGoogle ScholarCrossref
22.
Szymczak  JE, Barkman  D, Fiks  AG, Conaboy  K, Gerber  JS. Identifying patient and family centered outcomes around antibiotics for acute respiratory tract infections in children. Presented at: Pediatric Academic Societies Annual Meeting; May 9, 2017; San Diego, CA.
23.
Szymczak  JE, Klieger  SB, Miller  M, Fiks  AG, Gerber  JS.  What parents think about antibiotics for their child’s acute respiratory tract infection.  J Pediatr Infect Dis Soc. doi:10.1093/jpids/pix073Google Scholar
24.
Varni  JW. The PedsQL measurement model for the Pediatric Quality of Life Inventory. http://www.pedsql.org/. Accessed November 19, 2017.
25.
Varni  JW, Burwinkle  TM, Seid  M, Skarr  D.  The PedsQL 4.0 as a pediatric population health measure: feasibility, reliability, and validity.  Ambul Pediatr. 2003;3(6):329-341.PubMedGoogle ScholarCrossref
26.
Varni  JW, Limbers  CA, Burwinkle  TM.  Parent proxy-report of their children’s health-related quality of life: an analysis of 13,878 parents’ reliability and validity across age subgroups using the PedsQL 4.0 Generic Core Scales.  Health Qual Life Outcomes. 2007;5:2.PubMedGoogle ScholarCrossref
27.
Varni  JW, Seid  M, Kurtin  PS.  PedsQL 4.0: reliability and validity of the Pediatric Quality of Life Inventory version 4.0 generic core scales in healthy and patient populations.  Med Care. 2001;39(8):800-812.PubMedGoogle ScholarCrossref
28.
Varni  JW, Limbers  CA, Neighbors  K,  et al.  The PedsQL™ infant scales: feasibility, internal consistency reliability, and validity in healthy and ill infants.  Qual Life Res. 2011;20(1):45-55.PubMedGoogle ScholarCrossref
29.
Gerber  JS, Prasad  PA, Localio  AR,  et al.  Racial differences in antibiotic prescribing by primary care pediatricians.  Pediatrics. 2013;131(4):677-684.PubMedGoogle ScholarCrossref
30.
Hansen  BB.  Full matching in an observational study of coaching for the SAT.  J Am Stat Assoc. 2004;99(467):609-618.Google ScholarCrossref
31.
Stuart  EA, Green  KM.  Using full matching to estimate causal effects in nonexperimental studies: examining the relationship between adolescent marijuana use and adult outcomes.  Dev Psychol. 2008;44(2):395-406.PubMedGoogle ScholarCrossref
32.
Austin  PC, Stuart  EA.  Estimating the effect of treatment on binary outcomes using full matching on the propensity score.  Stat Methods Med Res. 2015;0962280215601134.PubMedGoogle Scholar
33.
Allison  PD.  Fixed Effects Regression Models. Newbury Park, CA: Sage; 2009.Crossref
34.
Breslow  NE, Day  NE, Halvorsen  KT, Prentice  RL, Sabai  C.  Estimation of multiple relative risk functions in matched case-control studies.  Am J Epidemiol. 1978;108(4):299-307.PubMedGoogle ScholarCrossref
35.
Hansen  BB, Fredrickson  M, Buckner  J, Errickson  J, Solenberger  P. Optmatch: functions for optimal matching, version 0.9-6. https://github.com/markmfredrickson/optmatch. Accessed November 29, 2016.
36.
Austin  PC, Small  DS.  The use of bootstrapping when using propensity-score matching without replacement: a simulation study.  Stat Med. 2014;33(24):4306-4319.PubMedGoogle ScholarCrossref
37.
Bishop  YM, Fienberg  SE, Holland  PW.  Discrete Multivariate Analysis: Theory and Practice. New York, NY: Springer-Verlag; 2007.
38.
Ding  P, VanderWeele  TJ.  Sensitivity analysis without assumptions.  Epidemiology. 2016;27(3):368-377.PubMedGoogle ScholarCrossref
39.
VanderWeele  TJ, Ding  P.  Sensitivity analysis in observational research: introducing the E-value.  Ann Intern Med. 2017;167(4):268-274.PubMedGoogle ScholarCrossref
40.
Stevens  MW, Hainsworth  KR, Weisman  SJ, Layde  PM.  Health-related quality of life in pediatric minor injury: reliability, validity, and responsiveness of the Pediatric Quality of Life Inventory in the emergency department.  Arch Pediatr Adolesc Med. 2012;166(1):74-81.PubMedGoogle ScholarCrossref
Original Investigation
December 19, 2017

Association of Broad- vs Narrow-Spectrum Antibiotics With Treatment Failure, Adverse Events, and Quality of Life in Children With Acute Respiratory Tract Infections

Author Affiliations
  • 1Center for Pediatric Clinical Effectiveness, Children’s Hospital of Philadelphia, Philadelphia, Pennsylvania
  • 2Division of Infectious Diseases, Children’s Hospital of Philadelphia, Philadelphia, Pennsylvania
  • 3Department of Biostatistics, Epidemiology and Informatics, Perelman School of Medicine, University of Pennsylvania, Philadelphia
  • 4Larner College of Medicine, University of Vermont, Burlington
  • 5Division of Patient and Family Experience, Family and Patient Services Department, Children’s Hospital of Philadelphia, Philadelphia, Pennsylvania
  • 6Division of General Pediatrics, Children’s Hospital of Philadelphia, Philadelphia, Pennsylvania
JAMA. 2017;318(23):2325-2336. doi:10.1001/jama.2017.18715
Key Points

Question  Does treatment with broad-spectrum antibiotics result in better clinical or patient-centered outcomes than narrow-spectrum antibiotics for children with acute respiratory tract infections?

Findings  In a retrospective cohort of 30 159 children with acute otitis media, group A streptococcal pharyngitis, and acute sinusitis, treatment with broad-spectrum vs narrow-spectrum antibiotics was not associated with treatment failure but was associated with higher rates of adverse events (3.7% vs 2.7%, respectively). In a prospective cohort of 2472 children, receipt of broad-spectrum vs narrow-spectrum antibiotics was associated with a slightly worse child quality of life and higher rates of adverse events (35.6% vs 25.1%, respectively).

Meaning  Use of broad-spectrum vs narrow-spectrum antibiotics was not associated with better clinical or patient-centered outcomes in children with acute respiratory tract infections.

Abstract

Importance  Acute respiratory tract infections account for the majority of antibiotic exposure in children, and broad-spectrum antibiotic prescribing for acute respiratory tract infections is increasing. It is not clear whether broad-spectrum treatment is associated with improved outcomes compared with narrow-spectrum treatment.

Objective  To compare the effectiveness of broad-spectrum and narrow-spectrum antibiotic treatment for acute respiratory tract infections in children.

Design, Setting, and Participants  A retrospective cohort study assessing clinical outcomes and a prospective cohort study assessing patient-centered outcomes of children between the ages of 6 months and 12 years diagnosed with an acute respiratory tract infection and prescribed an oral antibiotic between January 2015 and April 2016 in a network of 31 pediatric primary care practices in Pennsylvania and New Jersey. Stratified and propensity score–matched analyses to account for confounding by clinician and by patient-level characteristics, respectively, were implemented for both cohorts.

Exposures  Broad-spectrum antibiotics vs narrow-spectrum antibiotics.

Main Outcomes and Measures  In the retrospective cohort, the primary outcomes were treatment failure and adverse events 14 days after diagnosis. In the prospective cohort, the primary outcomes were quality of life, other patient-centered outcomes, and patient-reported adverse events.

Results  Of 30 159 children in the retrospective cohort (19 179 with acute otitis media; 6746, group A streptococcal pharyngitis; and 4234, acute sinusitis), 4307 (14%) were prescribed broad-spectrum antibiotics including amoxicillin-clavulanate, cephalosporins, and macrolides. Broad-spectrum treatment was not associated with a lower rate of treatment failure (3.4% for broad-spectrum antibiotics vs 3.1% for narrow-spectrum antibiotics; risk difference for full matched analysis, 0.3% [95% CI, −0.4% to 0.9%]). Of 2472 children enrolled in the prospective cohort (1100 with acute otitis media; 705, group A streptococcal pharyngitis; and 667, acute sinusitis), 868 (35%) were prescribed broad-spectrum antibiotics. Broad-spectrum antibiotics were associated with a slightly worse child quality of life (score of 90.2 for broad-spectrum antibiotics vs 91.5 for narrow-spectrum antibiotics; score difference for full matched analysis, −1.4% [95% CI, −2.4% to −0.4%]) but not with other patient-centered outcomes. Broad-spectrum treatment was associated with a higher risk of adverse events documented by the clinician (3.7% for broad-spectrum antibiotics vs 2.7% for narrow-spectrum antibiotics; risk difference for full matched analysis, 1.1% [95% CI, 0.4% to 1.8%]) and reported by the patient (35.6% for broad-spectrum antibiotics vs 25.1% for narrow-spectrum antibiotics; risk difference for full matched analysis, 12.2% [95% CI, 7.3% to 17.2%]).

Conclusions and Relevance  Among children with acute respiratory tract infections, broad-spectrum antibiotics were not associated with better clinical or patient-centered outcomes compared with narrow-spectrum antibiotics, and were associated with higher rates of adverse events. These data support the use of narrow-spectrum antibiotics for most children with acute respiratory tract infections.

Introduction

Antibiotics are the most common medications prescribed for children.1 The majority of antibiotic use occurs for the treatment of acute respiratory tract infections.2 Acute otitis media, acute sinusitis, and group A streptococcal pharyngitis account for almost all bacterial acute respiratory tract infections. Broad-spectrum antibiotic prescribing for these conditions now accounts for approximately half of all antibiotic prescribing for children.2-6 Overuse of broad-spectrum antibiotics can lead to antibiotic resistance, drug-related adverse events,7 and increased costs.

Current guidance on antibiotic choice for acute respiratory tract infections is inconsistent. The American Academy of Pediatrics recommends penicillin or amoxicillin, which are both narrow-spectrum antibiotics, as first-line therapy for most children with acute otitis media8; however, clinical trials have used amoxicillin-clavulanate, which is a broad-spectrum antibiotic, to compare with placebo for acute otitis media, suggesting that broad-spectrum antibiotics are appropriate.9,10 Likewise, recent guidelines from the American Academy of Pediatrics include amoxicillin as a first-line treatment of acute sinusitis,8 but the practice guidelines from the Infectious Diseases Society of America recommend amoxicillin-clavulanate.11 For group A streptococcal pharyngitis, narrow-spectrum antibiotics have historically been recommended; however, suggestions have arisen that cephalosporins provide added benefit.12 These recommendations toward broad-spectrum antibiotics are influenced by the predicted effects of the changing microbiology associated with pneumococcal vaccination, which may have shifted the balance away from pneumococcus toward more β-lactamase–producing organisms.13 The clinical consequences of this shift in microbiology on patient outcomes, however, remain unclear.

Studies typically assess traditional clinical outcomes such as treatment failure but ignore patient and caregiver perspectives. Patient-centeredness14,15 is particularly relevant for acute respiratory tract infections, which affect the lives of children and their families in ways that might never come to medical attention. Therefore, this study aimed to compare the effectiveness of broad-spectrum and narrow-spectrum antibiotic treatment for acute respiratory tract infections among children by assessing both clinical outcomes and patient-centered outcomes.

Methods
Study Design and Population

We conducted 2 cohort studies. The retrospective cohort study assessed clinical outcomes. The prospective cohort study assessed patient-centered outcomes. Both studies used a network of 31 pediatric primary care practices with a common electronic health record, serving children from diverse communities across southeastern Pennsylvania (4 in Philadelphia, 3 of which were academic teaching practices) and southern New Jersey.16 Infants and children aged 6 months to 12 years were required to have a diagnosis of acute respiratory tract infection (International Classification of Diseases, Ninth Revision, and International Statistical Classification of Diseases and Related Health Problems, Tenth Revision, diagnosis codes appear eTable 1 in the Supplement) and a prescription for an oral antibiotic. For group A streptococcal pharyngitis, a positive rapid streptococcal test also was required. The institutional review board at the Children’s Hospital of Philadelphia approved this study.

Prospective Cohort Study of Patient-Centered Outcomes

Between January 2015 and April 2016, children meeting the inclusion criteria were identified from the electronic health record twice weekly and a sample stratified by acute respiratory tract infection diagnosis, antibiotic spectrum, and practice type (academic teaching practice vs other) was randomly selected. Legal guardians were contacted by telephone 5 to 10 days after diagnosis to confirm eligibility (diagnosis, antibiotic prescribed, and no antibiotic use during the 30 days prior to diagnosis) and that the child started antibiotic therapy. Legal guardians provided verbal informed consent for participation in 2 telephone interviews and to allow use of electronic health record data. Structured interviews occurred 5 to 10 days and 14 to 20 days after diagnosis. At the time of the data analysis, children younger than 3 years diagnosed with group A streptococcal pharyngitis were excluded to reflect the guidelines from the American Academy of Pediatrics.17

Retrospective Cohort Study of Clinical Outcomes

The retrospective cohort was developed by using electronic health record data only, was conducted after the prospective cohort study was completed, and only included children who had acute respiratory tract infections that occurred during 2015. Children were excluded if (1) their legal guardian did not allow use of electronic health record data during the consent process for the prospective cohort, (2) they were diagnosed with multiple bacterial infections, (3) they were prescribed inappropriate therapy (choice or duration), (4) they were treated with antibiotics or diagnosed with an acute respiratory tract infection during the previous 30 days, (5) they were previously diagnosed with a complex chronic condition,18 or (6) they were younger than 3 years and diagnosed with group A streptococcal pharyngitis. The following data were extracted from the electronic health record: health care encounters, telephone calls, and prescriptions from the index visit and the following 30 days. One of 2 registered nurses with more than 10 years of experience in hospital epidemiology reviewed progress notes of select (described below) outpatient and emergency department encounters. A pediatric infectious diseases physician (J.S.G.) reviewed medical records from all hospitalizations.

Exposure

Children who were prescribed broad-spectrum antibiotics (full list of antibiotics appears in eTable 2 in the Supplement) at diagnosis, including amoxicillin-clavulanate, cephalosporins, and macrolides, were defined as exposed. Children who were prescribed the narrow-spectrum antibiotics penicillin and amoxicillin were defined as unexposed. The term broad spectrum referred to antibiotics with activity against clinically important colonizing flora beyond pneumococcus (the primary target for the specified acute respiratory tract infections) including Moraxella catarrhalis, Haemophilus influenzae, and Staphylococcus aureus. Although not recommended for the treatment of acute otitis media and acute sinusitis, macrolides were included as broad-spectrum antibiotics because of the frequency of reported use for these conditions in this health care network as well as nationally.2,19,20

Clinical Outcomes

Treatment failure and adverse events were assessed through 14 days (primary outcome) and 30 days after diagnosis. Assessment of treatment failure started 2 days after diagnosis because antibiotic therapy for fewer than 2 days was likely an insufficient amount of time to truly assess the effectiveness of the initial antibiotic.10,17,21

Treatment failure was defined as a child having the same acute respiratory tract infection diagnosis and a new prescription for a systemic antibiotic reported during an in-person or telephone encounter. If the encounter occurred during the effective duration of the initially prescribed (index) antibiotic course (ie, the prescribed duration or twice the prescribed duration for azithromycin to account for its pharmacokinetics), progress notes were reviewed to distinguish between treatment failure (persistence of symptoms or concern for failure of the index antibiotic) and requirement of a new antibiotic due to an adverse event. Encounters with the same acute respiratory tract infection diagnosis and an antibiotic that occurred after the effective duration of the index antibiotic were considered treatment failures (ie, recurrence) and did not require confirmatory record reviews.

Adverse events included diarrhea, candidiasis, noncandidal rash, other or unspecified allergic reaction, other or unspecified adverse event, and vomiting identified from (1) progress notes (described above), (2) electronic health record diagnosis codes, or (3) new antibiotic allergy noted in the electronic health record. Noncandidal rash, other or unspecified allergic reaction, other or unspecified adverse event, and vomiting were assessed during the effective duration of the index antibiotic. Candidiasis was assessed through 1 week after the completion of the index course. Diarrhea was assessed through 30 days.

Patient-Centered Outcomes

Qualitative interviews were conducted with 109 legal guardians and 24 children presenting with acute respiratory tract infection symptoms at 4 practices.22,23 Respondents identified child suffering, missed school and work, child sleep quality, adverse events, and speed of symptom resolution as important outcomes related to treatment. Based on these findings, the following patient-centered outcomes were adopted: child quality of life, disrupted sleep, child missed school, parent missed work, adverse events, and speed of symptom resolution. The measurement model for the Pediatric Quality of Life Inventory (PedsQL)24 (parent-proxy report generic core scales25-27 and parent report infant scales28) was selected to measure health-related quality of life. The score can range from 0 to 100. Higher scores indicate better health-related quality of life and a 4-point difference has been estimated to be clinically meaningful.25

The PedsQL included questions to assess sleep disturbance. An interview guide was developed to assess missed school or day care, missed commitments or need for additional childcare by the parent, adverse events, and persistent symptoms 72 hours after the initial diagnosis. The interview guide was reviewed by members of the Children’s Hospital of Philadelphia family advisory board for clarity and face validity. The eText in the Supplement provides a detailed description of the patient-centered outcomes. Each patient-centered outcome was assessed at the first interview between days 5 and 10. Patient-reported adverse events were assessed at the second interview between days 14 and 20.

Covariates

For both cohorts, the covariates obtained from the electronic health record included age, sex, race/ethnicity (which is routinely documented at the first office visit), public insurance, season, prescription of antibiotic ear drops, practice type (academic teaching vs other), type of clinician training (physician vs nurse practitioner), and years of physician experience (<10 vs ≥10). For the prospective cohort of patient-centered outcomes, additional covariates obtained during the first interview included symptoms at diagnosis, school or day care attendance, and antibiotics prescribed after initial treatment. Race/ethnicity also was obtained during the second interview. The following prespecified categories were provided: American Indian or Alaskan Native, Asian, black or African American, Native Hawaiian or other Pacific Islander, white, or other. Prior work has shown an association between antibiotic choice and race29 and race may be associated with outcomes and documentation or reporting of outcomes.

Analysis

Two complementary analyses were implemented for each cohort: a stratified analysis by clinician and a propensity score–based full matched analysis.30-32 Previous research in this health care network showed that antibiotic choice varied by clinician and that the variability could not be fully explained by differences in patient characteristics.19 Furthermore, outcomes likely vary by clinician through a variety of mechanisms such as the clinician’s communication skills, office setting, and electronic health record documentation. In addition, choice of antibiotic spectrum (broad vs narrow) may be a product of patient characteristics. Variability in antibiotic prescribing by race, even within clinician, was previously documented in this population.29 Outcomes also may be associated with these same patient characteristics. Thus, although clinician and patient characteristics were identified a priori as potentially important confounders, it was not feasible to address both in a single analysis.

Stratification by clinician primarily addressed potential confounding by clinician using an approach comparing the exposure across children who were treated by the same clinician. However, due to the large number of clinicians, the sample size within each clinician stratum precluded simultaneous adjustment for patient-level characteristics. Even though this stratified analysis may indirectly account for patient-level characteristics given the relative homogeneity of children seen by a clinician, it does not fully address confounding by patient-level characteristics.

Alternatively, in the full matched analysis, children were matched using propensity scores to minimize differences between antibiotic spectrum group in patient-level characteristics. This analysis primarily addresses confounding by patient characteristics but does not directly address confounding by clinician because matching stratified by clinician requires large samples of children per clinician. To offset the potential for residual confounding by clinician in this analysis, basic clinician-level characteristics (practice type, clinician type, and years of experience) were included in the propensity score models.

Implementing 2 analyses assessed the robustness of the effect estimates to several sources of confounding. Both analyses were stratified by diagnosis to compare the associations of antibiotics across children within each diagnosis. In the adjusted analysis, children missing race/ethnicity data were excluded.

Stratified Analysis by Clinician

For the PedsQL, the only continuous outcome, fixed-effects linear regression (first differencing method)33,34 was used with strata (diagnosis and clinician) as fixed effects to obtain the health-related quality-of-life score difference between antibiotic spectrum exposure groups. The response model included indicators for the PedsQL questionnaire age groups to account for differences in calibration across questionnaires. For all other outcomes (binary), fixed-effects linear regression was used to estimate risk differences33 and conditional logistic regression was used to estimate odds ratios. Robust variance methods were implemented throughout.

Propensity Score–Based Full Matched Analysis

The full matched analysis had 2 steps. In the first step, the propensity score models were estimated and stratified by diagnosis using logistic regression with antibiotic spectrum as the dependent variable and patient and clinician characteristics included as independent variables (eText in the Supplement). Children prescribed narrow-spectrum antibiotics were optimally matched to children prescribed broad-spectrum antibiotics based on the propensity score (R package Optmatch35). The advantage of a full matched analysis is that it minimizes overall distance (ie, differences) across children by creating matched sets of different sizes and includes all children. The ratio of exposed to unexposed children was restricted within matched sets to prevent large weights.30 Weights within each set were calculated so that the sum of the weights in each exposure group equaled the total number of children in the set for estimation of the average treatment effect.32 After the groups were matched, weighted means were calculated for each covariate by exposure category to check for any remaining imbalance of covariates. In the retrospective cohort, the acute otitis media stratum was too large to implement matching without subsetting, so these data were stratified by sex and public insurance. Because practice type remained imbalanced after matching for group A streptococcal pharyngitis in the retrospective cohort, matching was repeated stratified by practice type. For some patient-centered outcomes (described in results), remaining imbalances were observed by race, insurance, and practice type. Based on the geography of the practice types, the remaining imbalances by race and insurance were likely driven by practice-type imbalance; therefore, practice type, diagnosis, and an interaction term were included in response models for the patient-centered outcomes for the full matched analysis.

In the second step, weighted response models were fit. For the PedsQL used in the prospective cohort, the score difference was obtained from linear regression, which included the questionnaire age categories. For binary outcomes, logistic regression estimated the odds ratios and marginal standardization transformed the results to risk differences. The parametric-based variances accounted for matching.36 The risk difference variances were computed using the δ method.37

Sensitivity Analyses

In the prospective cohort, the degree of confounding from an omitted covariate38 necessary to shift the PedsQL result so that broad-spectrum antibiotics would be superior by 4.0 points (clinically superior) was assessed. For clinical outcomes assessed in the retrospective cohort, a similar sensitivity analysis based on the e value considered hypothetical confounders that could hide a broad-spectrum risk reduction of 1 percentage point.39 A second analysis for patient-reported adverse events used inverse-probability weighting to account for missing responses in the prospective cohort (eText in the Supplement). Two post hoc analyses also were conducted in the prospective cohort for the PedsQL outcome.

In a 3-level hierarchical analysis, a mixed-effects model was fit that included each covariate as a fixed effect, random intercept for clinician, and random intercept and slope for practice site. An additional full matched analysis also was implemented that used exact matching on the number of symptoms reported at baseline (0, 1, ≥2). Stratified response models were fit to assess the potential for a ceiling effect, especially among children without symptoms at baseline.

The statistical analyses were performed using Stata version 14.2 (StataCorp) and R version 3.1.1 (R Foundation for Statistical Computing). The 95% CIs were calculated to determine whether broad-spectrum antibiotic treatment was superior to narrow-spectrum antibiotic treatment. Statistical tests were 2-sided and the significance threshold was P < .05. Although there was a large number of end points, all comparisons were prespecified except for the results stratified by diagnosis and post hoc analyses. Because there was no adjustment for significance to address potential type I error, conventional levels of statistical significance may not be applicable.

Results
Clinical Outcomes

After applying the inclusion and exclusion criteria, 30 159 children comprised the retrospective cohort (19 179 with acute otitis media, 6746 with group A streptococcal pharyngitis, and 4234 with acute sinusitis; Figure 1) and 4307 (14%) were prescribed broad-spectrum antibiotics. Differences by treatment group were noted for race/ethnicity and public insurance for acute otitis media and group A streptococcal pharyngitis (Table 1). Children missing race/ethnicity data (73; 0.2%) were excluded, leaving 30 086 children for the primary analysis assessed at 14 days after diagnosis. In the stratified analysis, there were 336 clinicians (median, 54 children per clinician; range, 1-442 children per clinician) resulting in 798 diagnosis strata (304 children diagnosed as having acute otitis media, 202 as having acute sinusitis, and 292 as having group A streptococcal pharyngitis). In the full matched analysis, covariates stratified by diagnosis were generally successfully balanced (eTable 3 in the Supplement).

The results for the clinical outcomes appear in Table 2 (the odds ratios appear in eTable 4 in the Supplement). In the 14 days after diagnosis, 956 children (3.2%) experienced treatment failure and 852 (2.8%) experienced adverse events. Through 30 days, 2454 children (8.2%) experienced treatment failure and 1038 (3.5%) experienced adverse events. Broad-spectrum antibiotics were not superior to narrow-spectrum antibiotics for any clinical outcomes in the stratified analysis or in the full matched analysis. Receiving broad-spectrum antibiotics was associated with a higher risk of adverse events requiring clinical care compared with receiving narrow-spectrum antibiotics (risk at 14 days, 3.7% for broad-spectrum antibiotics vs 2.7% for narrow-spectrum antibiotics; risk difference for stratified analysis, 0.9% [95% CI, 0.3%-1.6%]; risk difference for full matched analysis, 1.1% [95% CI, 0.4%-1.8%]).

The sensitivity analyses for the 14-day treatment failure outcome demonstrated that an unobserved confounder would have to be strongly associated with both outcome and antibiotic type (relative risk >2.4) to shift the results in favor of broad-spectrum antibiotics by 1 percentage point. Similarly, for adverse events at 14 days, the association of an unobserved confounder with both outcome and treatment group would require a relative risk of 4.3.

Results for outcomes on day 14 stratified by diagnosis appear in eTable 5 in the Supplement. Broad-spectrum antibiotics were not superior to narrow-spectrum antibiotics for acute otitis media or acute sinusitis, but were associated with reduced risk of treatment failure for group A streptococcal pharyngitis (risk of 1.0% for broad-spectrum antibiotics vs 2.4% for narrow-spectrum antibiotics; risk difference for stratified analysis, −1.7% [95% CI, −2.5% to −0.9%]; risk difference for full matched analysis, −1.3% [95% CI, −2.2% to −0.3%]). In the pooled matched analysis to assess the interaction between exposure and diagnosis, the interaction was not significant at conventional significance levels for either treatment failure (P = .06) or adverse events (P = .06).

Patient-Centered Outcomes

For the prospective cohort, 58 488 children met the inclusion criteria. Of these, 10 296 children were randomly selected. The legal guardians of 2981 children (29%) were successfully contacted (Figure 2) and 2472 children were enrolled (1100 had acute otitis media, 705 had group A streptococcal pharyngitis, and 667 had acute sinusitis); each enrolled child had a primary caretaker who completed the first interview between day 5 and 10. Enrolled children had similar characteristics to sampled children who were not enrolled (eTable 6 in the Supplement). Broad-spectrum antibiotics were prescribed for 868 children (35%). The second interview between day 14 and 20 was completed for 2096 children (85%). White children and those with public insurance were less likely to have both interviews completed (eTable 7 in the Supplement). Differences by treatment group for race/ethnicity and public insurance were noted for acute otitis media and group A streptococcal pharyngitis (Table 3), which reflect the population that sought care at academic teaching practices where clinicians had a strong preference for narrow-spectrum antibiotics.

The first interview revealed that 77 children (3.1%) had received a new antibiotic after the initial prescription. For 33 children, the new antibiotic was a different spectrum than the original. For 22 children, the parent did not report the name of the new antibiotic. Therefore, 55 children (2.2%) at most were switched to a different antibiotic spectrum prior to assessment of the patient-centered outcomes.

There were missing race/ethnicity data for 41 children (1.7%) who were excluded, leaving 2431 for the analysis. The stratified analysis included 265 clinicians (median, 7 children per clinician; range, 1-47 children per clinician) resulting in 592 diagnosis strata (220 children diagnosed as having acute otitis media, 221 as having group A streptococcal pharyngitis, and 151 as having acute sinusitis). In the full matched analysis, imbalance by practice type, race/ethnicity, and insurance remained after matching for group A streptococcal pharyngitis (eTable 8-eTable 12 in the Supplement).

Broad-spectrum antibiotics were associated with a slightly worse quality-of-life score and more adverse events compared with narrow-spectrum antibiotics, but no association was found with other patient-centered outcomes (Table 4). The odds ratios appear in eTable 13 and an explanation of the sample sizes appears in the eText in the Supplement. The results were consistent across the stratified analysis and the full matched analysis. The overall mean PedsQL score was 91.1 (SD, 9.8) and it was 90.2 (SD, 10.5) for broad-spectrum antibiotics and 91.5 (SD, 9.4) for narrow-spectrum antibiotics (Table 4 and eFigure 1 in the Supplement). In the stratified analysis, broad-spectrum antibiotic treatment was associated with a score difference of −1.6% (95% CI, −2.8% to −0.5%) compared with narrow-spectrum antibiotic treatment. The results were similar for the full matched analysis with a score difference of −1.4% (95% CI, −2.4% to −0.4%) (Table 4).

In the post hoc analyses, results from the hierarchical analysis were consistent and the results from the analysis restricted to children with parent-reported symptoms at baseline also were consistent but not statistically significant (eText in the Supplement). In a sensitivity analysis, a potential unobserved confounder would have required a relative risk of 2.7 for the association with both types of antibiotic spectrum exposure and the outcome to shift the estimate from the matched analysis to favor broad-spectrum antibiotics by 4.0 points for the PedsQL score (eText in the Supplement).

Receiving broad-spectrum antibiotics was associated with a higher risk of adverse events compared with receiving narrow-spectrum antibiotics (risk of 35.6% with broad-spectrum antibiotics vs risk of 25.1% with narrow-spectrum antibiotics; risk difference for stratified analysis of 11.6% [95% CI, 6.0%-17.2%]; risk difference for full matched analysis of 12.2% [95% CI, 7.3%-17.2%]). The results of the analyses that accounted for missing data were similar (eText in the Supplement). Among the 599 children (28.9%) with reported adverse events, 69.6% were due to diarrhea; rash, 40.1%; upset stomach, vomiting, or both, 21.4%; and more than 1 adverse event, 27.8%.

Results stratified by acute respiratory tract infection diagnosis were consistent across diagnoses (eTable 14 in the Supplement); however, there was a limited difference in adverse events by antibiotic spectrum for children diagnosed with acute sinusitis.

Discussion

This study used a diverse, pediatric primary care network to compare the 2 prevailing treatment strategies for the most common childhood infections. Broad-spectrum antibiotics did not perform better than narrow-spectrum antibiotics on clinical or patient-centered outcomes, but were associated with a higher rate of adverse events than narrow-spectrum antibiotics. These data support the use of narrow-spectrum antibiotics for the treatment of acute respiratory tract infections in most children.

Assessing outcomes identified as important by legal guardians of children with acute respiratory tract infections distinguished this work.22 Clinical studies of different therapeutic approaches that rely exclusively on clinician-observed outcomes risk missing differential rates of symptom resolution and adverse events that are meaningful to families. These events might not be recorded in the medical record either because no medical encounter occurred (eg, diarrhea manageable at home) or because the parent did not communicate or the clinician did not elicit or document the effect on the child’s or family’s quality of life (eg, sleep disruption).

The consistency of findings across clinical and patient-derived outcomes was notable. This remained true in the subanalyses by individual acute respiratory tract infection, with the lone exception of group A streptococcal pharyngitis in the retrospective electronic health record-based cohort, in which a lower rate of treatment failure was observed with broad-spectrum antibiotic therapy. Although statistically significant, the clinical significance of this finding is questionable because (1) the risk difference favored broad-spectrum antibiotics by less than 2% for both the stratified and full matched analyses, (2) the overall success rate was greater than 97% for narrow-spectrum antibiotics, and (3) this was not observed in the prospective cohort with the patient-centered approach. Although adverse events rates were higher for broad-spectrum antibiotics in both cohorts, the overall rates of adverse events identified in the prospective cohort were 10.3 times the rate identified in the retrospective cohort, which reveals that (1) the majority of adverse events resulting from antibiotic use are not reported to (or documented by) prescribing clinicians and (2) assessing patient-centered outcomes might inform the risk-benefit ratio of antibiotic use in children.

Limitations

This study had several limitations. Because children were identified based on clinician diagnosis plus an antibiotic prescription to identify bacterial acute respiratory tract infections, some children likely had viral infections. This misclassification could have blunted the ability to detect a difference between antibiotic choices. This approach, however, reflects the true incidence of antibiotic prescribing for acute respiratory tract infections and, therefore, captures its association with symptom resolution, adverse events, and quality of life. Residual confounding likely persists, but sensitivity analyses suggest that this is unlikely to explain the study findings. Although conducted in a large and diverse primary care network, results might not be generalizable outside this setting. Because children who had received antibiotics within the previous 30 days were excluded from the study, the potential benefits of broad-spectrum antibiotics for these children could not be assessed. All pooled comparisons were prespecified, but because of the multiplicity of end points, comparisons of these results using conventional levels of statistical significance should only be made cautiously.

Specific to the prospective cohort study, the PedsQL was used to capture the outcome of child quality of life. This outcome was identified as a key concern by legal guardians and children in the qualitative study. Although the PedsQL is a validated instrument that captures social, physical, emotional, and school functioning and it has been demonstrated to be feasible and have good construct and discriminant validity and responsiveness in measuring short-term outcomes after minor acute episodes of illness or injury,25,40 it has not been specifically validated for acute respiratory tract infections. In addition, prior work estimated that a difference of more than 4 points on the PedsQL was a clinically meaningful difference.25 Thus, the 1.4-point higher score observed in children who received narrow-spectrum antibiotics, though statistically significant, does not necessarily support a clinically superior outcome for narrow-spectrum antibiotics. Instead, this finding rejects the notion that broad-spectrum antibiotics provide a clinically meaningful advantage.

Although the relatively high overall PedsQL scores suggest a potential ceiling effect of this measure, the distribution shift was in favor of narrow-spectrum antibiotics, which was consistent across most outcomes and analyses, and the sensitivity analysis indicated that there would need to be very strong residual confounding to shift the PedsQL results in favor of broad-spectrum antibiotics. In addition, because the PedsQL was assessed at only a single time point, we could not assess the effect of antibiotic spectrum on the change in score from baseline.

The relatively low rate of contact (29%) for the prospective cohort could have biased estimates if the chances of contact were related to the antibiotic prescribed. However, when the full sample of eligible participants was compared with the children included in the cohort, differences in patient characteristics were negligible. Fifteen percent of families who completed the first interview between day 5 and 10 did not complete the second interview between day 14 and 20. The assessment of only 1 outcome (adverse events) and 2 covariates (race and ethnicity) at the second interview, however, minimized the effect of these missing data and a sensitivity analysis accounting for the missing adverse event outcome yielded similar results as the primary analysis.

Recall bias might have affected data collected by telephone interview, although most outcomes and covariates were obtained only 5 to 10 days after diagnosis and it is unlikely that antibiotic choice was related to the parent’s ability to accurately report these data. The specific questions used to elicit the outcomes from legal guardians by telephone interview were not validated or tested for reliability. They were, however, tested for feasibility and family partners confirmed the face validity and clarity of these questions. Some imbalance of race, insurance, and practice type remained after matching for group A streptococcal pharyngitis. This imbalance was addressed by including practice type, which is the likely driver of imbalance in race and insurance, in the response models. Remaining imbalance in season was both restricted to a single season and inconsistent and, thus, likely to be an artifact. Although this could have resulted in uncontrolled confounding, results were robust to multiple analyses.

Conclusions

Among children with acute respiratory tract infections, broad-spectrum antibiotics were not associated with better clinical or patient-centered outcomes compared with narrow-spectrum antibiotics, and were associated with higher rates of adverse events. These data support the use of narrow-spectrum antibiotics for most children with acute respiratory tract infections.

Back to top
Article Information

Corresponding Author: Jeffrey S. Gerber, MD, PhD, Children’s Hospital of Philadelphia, 2716 S St, Philadelphia, PA 19146 (gerberj@email.chop.edu).

Accepted for Publication: November 10, 2017.

Author Contributions: Dr Gerber had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Gerber, Ross, Bryan, Localio, Wasserman, Barkman, Conaboy, Bell, Zaoutis, Fiks.

Acquisition, analysis, or interpretation of data: Gerber, Ross, Bryan, Localio, Szymczak, Wasserman, Odeniyi, Fiks.

Drafting of the manuscript: Gerber, Ross, Localio, Szymczak, Odeniyi.

Critical revision of the manuscript for important intellectual content: Gerber, Bryan, Localio, Szymczak, Wasserman, Barkman, Conaboy, Odeniyi, Bell, Zaoutis, Fiks.

Statistical analysis: Ross, Bryan, Localio, Fiks.

Obtained funding: Gerber, Localio, Szymczak, Barkman, Conaboy.

Administrative, technical, or material support: Ross, Szymczak, Barkman, Conaboy, Odeniyi, Bell, Fiks.

Supervision: Gerber, Localio, Wasserman, Bell, Zaoutis, Fiks.

Conflict of Interest Disclosures: The authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Dr Zaoutis reported serving as a consultant to T2 Biosystems and Nabriva. Dr Fiks reported receiving a Pfizer independent research grant. No other disclosures were reported.

Funding/Support: The research was funded through award CE-1304-7279 from the Patient-Centered Outcomes Research Institute (PCORI).

Role of the Funder/Sponsor: PCORI had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Disclaimer: The statements presented in this publication are solely the responsibility of the authors and do not necessarily represent the views of PCORI or its board of governors or methodology committee.

Additional Contributions: We thank the network of primary care clinicians, the patients, and their families for their contributions to this project and the clinical research facilitated through the Pediatric Research Consortium at the Children’s Hospital of Philadelphia. We thank Pediatric Research Consortium Director Jim Massey, RN, for his work on this project. Mr Massey received administrative support for this project, which included compensation from PCORI for help with recruiting practices for the intervention. We also thank the family advisory council (not compensated) for their support and guidance on this project from conception through execution; Susan L. Rettig, BSN, RN, and Eva Tezsner, RN, CIC (paid consultants from the Children’s Hospital of Philadelphia Department of Infection Prevention and Control), for assistance with manual review of medical records; Svetlana Ostapenko, MS (paid consultant from the Children’s Hospital of Philadelphia Department of Biomedical and Health Informatics), for her assistance with extraction of electronic health record data; and Robert W. Grundmeier, MD (Children’s Hospital of Philadelphia, Division of General Pediatrics; not compensated), for critical review of the manuscript.

References
1.
Chai  G, Governale  L, McMahon  AW, Trinidad  JP, Staffa  J, Murphy  D.  Trends of outpatient prescription drug utilization in US children, 2002-2010.  Pediatrics. 2012;130(1):23-31.PubMedGoogle ScholarCrossref
2.
Hersh  AL, Shapiro  DJ, Pavia  AT, Shah  SS.  Antibiotic prescribing in ambulatory pediatrics in the United States.  Pediatrics. 2011;128(6):1053-1061.PubMedGoogle ScholarCrossref
3.
Grijalva  CG, Nuorti  JP, Griffin  MR.  Antibiotic prescription rates for acute respiratory tract infections in US ambulatory settings.  JAMA. 2009;302(7):758-766.PubMedGoogle ScholarCrossref
4.
McCaig  LF, Besser  RE, Hughes  JM.  Antimicrobial drug prescription in ambulatory care settings, United States, 1992-2000.  Emerg Infect Dis. 2003;9(4):432-437.PubMedGoogle ScholarCrossref
5.
Steinman  MA, Landefeld  CS, Gonzales  R.  Predictors of broad-spectrum antibiotic prescribing for acute respiratory tract infections in adult primary care.  JAMA. 2003;289(6):719-725.PubMedGoogle ScholarCrossref
6.
Hersh  AL, Fleming-Dutra  KE, Shapiro  DJ, Hyun  DY, Hicks  LA; Outpatient Antibiotic Use Target-Setting Workgroup.  Frequency of first-line antibiotic selection among US ambulatory care visits for otitis media, sinusitis, and pharyngitis.  JAMA Intern Med. 2016;176(12):1870-1872.PubMedGoogle ScholarCrossref
7.
Shehab  N, Lovegrove  MC, Geller  AI, Rose  KO, Weidle  NJ, Budnitz  DS.  US emergency department visits for outpatient adverse drug events, 2013-2014.  JAMA. 2016;316(20):2115-2125.PubMedGoogle ScholarCrossref
8.
Kimberlin  DW, Brady  MT, Jackson  MA, Long  SS.  Red Book: 2015 Report of the Committee on Infectious Diseases. 30th ed. Elk Grove Village, IL: American Academy of Pediatrics; 2015.
9.
Tähtinen  PA, Laine  MK, Huovinen  P, Jalava  J, Ruuskanen  O, Ruohola  A.  A placebo-controlled trial of antimicrobial treatment for acute otitis media.  N Engl J Med. 2011;364(2):116-126.PubMedGoogle ScholarCrossref
10.
Hoberman  A, Paradise  JL, Rockette  HE,  et al.  Treatment of acute otitis media in children under 2 years of age.  N Engl J Med. 2011;364(2):105-115.PubMedGoogle ScholarCrossref
11.
Chow  AW, Benninger  MS, Brook  I,  et al; Infectious Diseases Society of America.  IDSA clinical practice guideline for acute bacterial rhinosinusitis in children and adults.  Clin Infect Dis. 2012;54(8):e72-e112.PubMedGoogle ScholarCrossref
12.
Casey  JR, Pichichero  ME.  Meta-analysis of cephalosporin versus penicillin treatment of group A streptococcal tonsillopharyngitis in children.  Pediatrics. 2004;113(4):866-882.PubMedGoogle ScholarCrossref
13.
Coker  TR, Chan  LS, Newberry  SJ,  et al.  Diagnosis, microbial epidemiology, and antibiotic treatment of acute otitis media in children: a systematic review.  JAMA. 2010;304(19):2161-2169.PubMedGoogle ScholarCrossref
14.
Institute of Medicine.  Crossing the Quality Chasm: A New Health System for the 21st Century. Washington, DC: The National Academies Press; 2001.
15.
Gabriel  SE, Normand  SL.  Getting the methods right—the foundation of patient-centered outcomes research.  N Engl J Med. 2012;367(9):787-790.PubMedGoogle ScholarCrossref
16.
Fiks  AG, Grundmeier  RW, Margolis  B,  et al.  Comparative effectiveness research using the electronic medical record: an emerging area of investigation in pediatric primary care.  J Pediatr. 2012;160(5):719-724.PubMedGoogle ScholarCrossref
17.
Shulman  ST, Bisno  AL, Clegg  HW,  et al.  Clinical practice guideline for the diagnosis and management of group A streptococcal pharyngitis: 2012 update by the Infectious Diseases Society of America [published correction appears in Clin Infect Dis. 2014;58(10):1496].  Clin Infect Dis. 2012;55(10):1279-1282.PubMedGoogle ScholarCrossref
18.
Feudtner  C, Christakis  DA, Connell  FA.  Pediatric deaths attributable to complex chronic conditions: a population-based study of Washington State, 1980-1997.  Pediatrics. 2000;106(1 pt 2):205-209.PubMedGoogle Scholar
19.
Gerber  JS, Prasad  PA, Russell Localio  A,  et al.  Variation in antibiotic prescribing across a pediatric primary care network.  J Pediatric Infect Dis Soc. 2015;4(4):297-304.PubMedGoogle Scholar
20.
Fleming-Dutra  KE, Demirjian  A, Bartoces  M, Roberts  RM, Taylor  TH  Jr, Hicks  LA.  Variations in antibiotic and azithromycin prescribing for children by geography and specialty—United States, 2013 [published online July 19, 2017].  Pediatr Infect Dis J. doi:10.1097/INF.0000000000001708PubMedGoogle Scholar
21.
Wald  ER, Applegate  KE, Bordley  C,  et al; American Academy of Pediatrics.  Clinical practice guideline for the diagnosis and management of acute bacterial sinusitis in children aged 1 to 18 years.  Pediatrics. 2013;132(1):e262-e280.PubMedGoogle ScholarCrossref
22.
Szymczak  JE, Barkman  D, Fiks  AG, Conaboy  K, Gerber  JS. Identifying patient and family centered outcomes around antibiotics for acute respiratory tract infections in children. Presented at: Pediatric Academic Societies Annual Meeting; May 9, 2017; San Diego, CA.
23.
Szymczak  JE, Klieger  SB, Miller  M, Fiks  AG, Gerber  JS.  What parents think about antibiotics for their child’s acute respiratory tract infection.  J Pediatr Infect Dis Soc. doi:10.1093/jpids/pix073Google Scholar
24.
Varni  JW. The PedsQL measurement model for the Pediatric Quality of Life Inventory. http://www.pedsql.org/. Accessed November 19, 2017.
25.
Varni  JW, Burwinkle  TM, Seid  M, Skarr  D.  The PedsQL 4.0 as a pediatric population health measure: feasibility, reliability, and validity.  Ambul Pediatr. 2003;3(6):329-341.PubMedGoogle ScholarCrossref
26.
Varni  JW, Limbers  CA, Burwinkle  TM.  Parent proxy-report of their children’s health-related quality of life: an analysis of 13,878 parents’ reliability and validity across age subgroups using the PedsQL 4.0 Generic Core Scales.  Health Qual Life Outcomes. 2007;5:2.PubMedGoogle ScholarCrossref
27.
Varni  JW, Seid  M, Kurtin  PS.  PedsQL 4.0: reliability and validity of the Pediatric Quality of Life Inventory version 4.0 generic core scales in healthy and patient populations.  Med Care. 2001;39(8):800-812.PubMedGoogle ScholarCrossref
28.
Varni  JW, Limbers  CA, Neighbors  K,  et al.  The PedsQL™ infant scales: feasibility, internal consistency reliability, and validity in healthy and ill infants.  Qual Life Res. 2011;20(1):45-55.PubMedGoogle ScholarCrossref
29.
Gerber  JS, Prasad  PA, Localio  AR,  et al.  Racial differences in antibiotic prescribing by primary care pediatricians.  Pediatrics. 2013;131(4):677-684.PubMedGoogle ScholarCrossref
30.
Hansen  BB.  Full matching in an observational study of coaching for the SAT.  J Am Stat Assoc. 2004;99(467):609-618.Google ScholarCrossref
31.
Stuart  EA, Green  KM.  Using full matching to estimate causal effects in nonexperimental studies: examining the relationship between adolescent marijuana use and adult outcomes.  Dev Psychol. 2008;44(2):395-406.PubMedGoogle ScholarCrossref
32.
Austin  PC, Stuart  EA.  Estimating the effect of treatment on binary outcomes using full matching on the propensity score.  Stat Methods Med Res. 2015;0962280215601134.PubMedGoogle Scholar
33.
Allison  PD.  Fixed Effects Regression Models. Newbury Park, CA: Sage; 2009.Crossref
34.
Breslow  NE, Day  NE, Halvorsen  KT, Prentice  RL, Sabai  C.  Estimation of multiple relative risk functions in matched case-control studies.  Am J Epidemiol. 1978;108(4):299-307.PubMedGoogle ScholarCrossref
35.
Hansen  BB, Fredrickson  M, Buckner  J, Errickson  J, Solenberger  P. Optmatch: functions for optimal matching, version 0.9-6. https://github.com/markmfredrickson/optmatch. Accessed November 29, 2016.
36.
Austin  PC, Small  DS.  The use of bootstrapping when using propensity-score matching without replacement: a simulation study.  Stat Med. 2014;33(24):4306-4319.PubMedGoogle ScholarCrossref
37.
Bishop  YM, Fienberg  SE, Holland  PW.  Discrete Multivariate Analysis: Theory and Practice. New York, NY: Springer-Verlag; 2007.
38.
Ding  P, VanderWeele  TJ.  Sensitivity analysis without assumptions.  Epidemiology. 2016;27(3):368-377.PubMedGoogle ScholarCrossref
39.
VanderWeele  TJ, Ding  P.  Sensitivity analysis in observational research: introducing the E-value.  Ann Intern Med. 2017;167(4):268-274.PubMedGoogle ScholarCrossref
40.
Stevens  MW, Hainsworth  KR, Weisman  SJ, Layde  PM.  Health-related quality of life in pediatric minor injury: reliability, validity, and responsiveness of the Pediatric Quality of Life Inventory in the emergency department.  Arch Pediatr Adolesc Med. 2012;166(1):74-81.PubMedGoogle ScholarCrossref
×