Trends in Antibiotic Use by Birth Season and Birth Year
OBJECTIVES: We examined 2 birth cohort effects on antibiotic prescribing during the first year of life (henceforth, infancy) in Denmark: (1) the birth season effect on timing and overall occurrence of antibiotic prescribing, and (2) the birth year effect amid emerging nationwide pneumococcal vaccination programs and changing prescribing guidelines.
METHODS: We linked data for all live births in Denmark from 2004 to 2012 (N = 561 729) across the National Health Service Prescription Database, Medical Birth Registry, and Civil Registration System. Across birth season and birth year cohorts, we estimated 1-year risk, rate, and burden of redeemed antibiotic prescriptions during infancy. We used interrupted time series methods to assess prescribing trends across birth year cohorts. Graphical displays of all birth cohort effect data are included.
RESULTS: The 1-year risk of having at least 1 redeemed antibiotic prescription during infancy was 39.5% (99% confidence interval [CI]: 39.3% to 39.6%). The hazard of a first prescription increased with age throughout infancy and varied by season; subsequently, Kaplan-Meier–derived risk functions varied by birth season cohort. After rollout of a first vaccination program and new antibiotic prescribing guidelines, 1-year risk decreased by 4.4% over 14 months (99% CI: 3.4% to 5.5%); it decreased again after rollout of a second vaccination program by 6.9% over 3 years (99% CI: 4.4% to 9.3%).
CONCLUSIONS: In Denmark, birth season and birth year cohort effects influenced timing and risk of antibiotic prescribing during infancy. Future studies of antibiotic stewardship, effectiveness, and safety in children should consider these cohort effects, which may render some children inherently more susceptible than others to downstream antibiotic effects.
- CI —
- confidence interval
- IRF —
- Institute for Rational Pharmacotherapy
- MBR —
- Danish Medical Birth Registry
- PCV —
- pneumococcal conjugate vaccine
- PCV7 —
- 7-valent pneumococcal conjugate vaccine
- PCV13 —
- 13-valent pneumococcal conjugate vaccine
What’s Known on This Subject:
Widespread antibiotic use leads to bacterial resistance, and unnecessary antibiotic use in early life may cause excess immune dysfunction in childhood. Antibiotic prescribing is influenced by seasonal and temporal factors, but data are limited regarding cohort effects on antibiotic use in children.
What This Study Adds:
In this study, we characterize birth season and birth year cohort effects on antibiotic prescribing during infancy, which merit consideration in pediatric studies of antibiotic stewardship, effectiveness, and safety. We also provide new information on correlation between risk, rate, and burden of antibiotic prescribing.
During early life, children are often prescribed antibiotics for bacterial infections. Approximately 20% to 50% of children’s antibiotic prescriptions are used to treat nonbacterial upper respiratory tract infections,1–3 for which they are largely ineffective.2,4–6 In Western industrialized countries in particular, increasing advocacy for rational antibiotic prescribing has been associated with decreased population-level antibiotic use.7–10 In addition, pneumococcal conjugate vaccination (PCV) programs have been associated with decreased risk of acute otitis media and lower respiratory tract infections,11 2 of the most common indications for antibiotics in children.4,12–16 Taken together, there is a need to better characterize the changing patterns of antibiotic prescribing in children.
Previous studies of trends in early life antibiotic prescribing have focused on rates in cross-sectional population samples,3,8,9,17–23 counting multiple prescriptions per child and tending to present results for coarse age groups (eg, 0–4 years) or groups defined by the year or season when they were observed (eg, during winter 2010). In addition, studies of rates implicitly estimate the frequency of providers prescribing or dispensing antibiotics to groups of children; however, studies of risk aggregate individual-level data to clarify which well-defined cohorts of children receive or redeem a given number of prescriptions over a specific follow-up interval. Currently, data are limited regarding cohort effects on the risk of overall and medication-specific antibiotic use during early childhood.
In Denmark, recent policy and guideline changes may have induced changes in antibiotic prescribing. The Danish Health and Medicine Authority’s Institute for Rational Pharmacotherapy (IRF), a government institute, issued a bulletin in 2007 with rational antibiotic prescribing guidelines.24 In 2007, the Danish childhood vaccination program introduced the 7-valent PCV (PCV7), replacing it with the 13-valent PCV (PCV13) in 2010.25–27 No studies have evaluated the impact of these events on antibiotic prescribing in Denmark during the first year of life (henceforth, infancy).
Our objectives were to estimate birth season cohort effects on antibiotic prescribing during infancy in Denmark, and birth year cohort effects on antibiotic prescribing potentially brought about by the IRF bulletin and PCV programs.
This cohort study included all live births in Denmark from 2004 to 2012, identified by using the Danish Medical Birth Registry (MBR), which has registered all live births in Denmark since 1973.28 We used each child’s unique 10-digit personal registration number to link their individual-level data across multiple registries. After excluding children with inconsistent identification information (Supplemental Information), we examined the remaining cohort from birth for 365 days or until death or emigration.
To identify birth cohorts, we grouped children by week, month, season (winter: December to February; spring: March to May; summer: June to August; autumn: September to November), and year of birth. We ascertained demographic information from the MBR and Civil Registration System29 and used the MBR to identify clinical characteristics of the mother, pregnancy, and birth event.
Data on Antibiotics
The Danish National Health Service Prescription Database contains nationwide records for all prescriptions redeemed in community pharmacies and hospital-based outpatient pharmacies onward from 2004.30 We used Anatomical Therapeutic Chemical (ATC) J01 codes to identify prescriptions for systemic antibiotics redeemed from January 1, 2004, to December 31, 2013, and classified antibiotics by chemical substance (henceforth, medication).
Risk, Rate, and Burden of Antibiotic Prescribing
To examine differences by birth year cohort in overall and medication-specific prescribing, we estimated 3 measures of antibiotic prescribing: (1) 1-year risk of at least 1 redeemed antibiotic prescription during infancy by using Kaplan-Meier methods to account for censoring at death or emigration; (2) incidence rate of redeemed antibiotic prescriptions, allowing for multiple redeemed prescriptions per infant (henceforth, rate); and (3) 1-year burden of antibiotic prescriptions, on the basis of the total days supplied for redeemed antibiotic prescriptions throughout infancy. For rate and burden, we also computed each medication’s share of overall prescribing, ie, the proportion of all antibiotic prescriptions (for the rate) or all days supplied (for the burden).
Birth Season Cohort Effect on Time to First Redeemed Antibiotic Prescription
We assessed the impact of birth season on age at first redeemed antibiotic prescription during infancy by using Kaplan-Meier methods to account for censoring at death or emigration. Infants were stratified by birth season cohort and assessed using age as the time scale. We estimated the hazard (ie, instantaneous risk) function for first redeemed antibiotic prescription and the Kaplan-Meier–derived risk function over 1 year for having at least 1 redeemed antibiotic prescription during infancy.
Interrupted Time Series Analysis
To estimate trend changes in antibiotic prescribing over time, we used segmented linear regression analysis of interrupted time series data,31–34 a common quasi-experimental method to assess trend changes after clearly defined events. Separately for each birth week cohort, we estimated the 1-year risk of redeeming at least 1 prescription for any antibiotic, with subanalyses for amoxicillin and penicillin V.
We identified 5 time-based interruptions when a population-level change in Denmark could have altered antibiotic prescribing in infants born before versus after the interruption (Table 1). We negatively lagged all interruptions to match each policy or guideline change with the infants subject to its potential impact. To control confounding by seasonality in our segmented linear regression model, we used a transformed cosine periodic function.35–37 Between any 2 interruptions, the trend estimate (ie, the change in 1-year risk during that specific time interval) resulted from the accumulation of the baseline trend and the effect of each policy or guideline change on the underlying trend. For example, for segment #3 (July 1, 2007, until January 19, 2010) the estimate was equal to the baseline trend plus the trend changes occurring at interruptions 1 and 2 (Supplemental Information).
We estimated rates and burdens with corresponding 99% confidence intervals (CIs), risks with 99% pointwise CIs, and all trend estimates with 99% CIs. Hazard functions were smoothed by using a 20th-order polynomial function and plotted with 99% confidence bands. Data management and statistical analyses were conducted by using SAS version 9.4 (SAS Institute, Inc, Cary, NC), and graphics were created by using R version 3.3.2 (R Foundation for Statistical Computing, Vienna, Austria).
This study received ethics approval from the Danish Data Protection Agency (2013-41-1790), Danish Statens Serum Institut (FSEID-00001450), and University of North Carolina at Chapel Hill (13-3155).
The final study population included 561 729 live births in Denmark occurring from 2004 to 2012. Antibiotic prescriptions accounted for 46.0% of all drug prescriptions redeemed during infancy; among 333 298 infants with a redeemed prescription for any drug, 66.2% had at least 1 antibiotic prescription.
In Table 2, we report descriptive data for the study population along with subgroup-specific estimates of the 1-year risk for having at least 1 redeemed antibiotic prescription during infancy. The overall 1-year risk was 39.5% (99% CI: 39.3% to 39.6%). Boys had a higher risk than girls, first-born children had a lower risk than younger siblings, and risk decreased with increasing maternal age. Maternal smoking, pregravid BMI, and prenatal clinic visit frequency were associated with higher risk. Risk increased with increasing birth weight but was not associated with gestational age. Children whose mother requested cesarean delivery had a higher risk than the general population. Geographically, risk was highest in Zealand and Southern Denmark and lowest in Central Denmark; higher population density was associated with lower risk (Fig 1).
The overall rate was 72 redeemed antibiotic prescriptions per 100 infant years of follow-up, and the overall burden was 67 daily doses per 10 000 infant days (Table 3). Amoxicillin and penicillin V accounted for 89% of redeemed antibiotic prescriptions out of 22 antibiotic medications prescribed to the study population. One-year risks and rates decreased over time; however, because of increasingly concentrated prescribing to infants with at least 1 prescription, antibiotic burden among infants remained stable over time, ranging from 61 to 75 daily doses per 10 000 infant days (Supplemental Tables 4–6). Over time, the share of the overall antibiotic rate and burden over time increased for amoxicillin and decreased for penicillin V.
Birth Season Cohort and Time to First Redeemed Antibiotic Prescription During Infancy
In Fig 2, we illustrate the joint effects of birth season cohort, seasonality of prescribing, and age on the time to first redeemed antibiotic prescription during infancy. Each birth season cohort experienced a unique order of the seasons during infancy (eg, winter-spring-summer-autumn versus summer-autumn-winter-spring), inducing intractable age differences between birth season cohorts in any given season of the year (Fig 2A). The hazard (ie, instantaneous risk) of a first redeemed antibiotic prescription varied with the seasons and peaked in February (white diamonds in Fig 2A, snowflakes in Fig 2B). Hazard also increased with increasing age through infancy (Fig 2B, black diamonds), thus leading hazard functions to differ by birth season cohort.
Each birth season cohort consequently had a unique risk function for having at least 1 redeemed antibiotic prescription during infancy (Fig 2C). Interpreting risk comparisons between birth season cohorts therefore depends partly on age. For example, at 6 months of age, infants born in autumn had 4.9% (bootstrapped 99% CI: 4.7% to 5.2%) higher risk than infants born in spring (11.5% vs 6.5%). At 12 months, the effect reversed such that infants born in autumn had 10.6% (bootstrapped 99% CI: 10.1% to 11.0%) lower risk than those born in spring (34.2% vs 44.8%).
Interrupted Time Series Analysis
In Fig 3, we show 1-year risk and birth season–adjusted trend estimates from our interrupted time series analysis. The overall 1-year risk of having at least 1 redeemed antibiotic prescription decreased from 40.7% (births in 2004) to 34.6% (births in 2012). Amoxicillin risk increased by 4.8% (99% CI: 3.7% to 5.9%) from January 2004 to April 2006, but penicillin V risk remained stable. After rollout of the PCV7 “catch-up” vaccination program and publication of the IRF bulletin, overall risk decreased by 4.4% over 14 months (99% CI: 3.4% to 5.5%), decreasing equally for amoxicillin and penicillin V. After standard PCV7 enrollment began, increasing amoxicillin prescribing drove increases in overall prescribing. After the childhood vaccination program’s replacement of PCV7 with PCV13, overall risk decreased by 6.9% over 3 years (99% CI: 4.4% to 9.3%). Penicillin V prescribing changed little for children born after mid-2007 but decreased after standard PCV13 enrollment began. The effects of the bulletin and vaccination programs were 1 order of magnitude smaller than birth season cohort effects (Supplemental Table 7).
A sensitivity analysis, which added a 7-month lag to relax assumptions about the timing and duration of the effects of the bulletin and the PCV7 catch-up program, affected the trend for amoxicillin but not for penicillin V (Supplemental Information). After remaining level during the lag through January 1, 2007, the amoxicillin trend decreased precipitously through July 1, 2007, by 0.58% per month (99% CI: 0.79% to 0.37%). Trend interpretation did not change in all other sensitivity analyses (Supplemental Information).
In this nationwide cohort study of children born in Denmark, we observed decreases over time from 2004 to 2012 in the proportion of infants who received antibiotics and the overall rate of antibiotic prescriptions per infant year. For those infants who received antibiotics, there was a less pronounced decrease in the rate over time, accompanied by an increase over time in days supplied. Taken together, the increasing concentration of antibiotic prescribing in a shrinking proportion of the infant population resulted in little change over time in overall antibiotic burden. Amoxicillin, which has a more extended spectrum of antibacterial activity than penicillin V, became increasingly prominent over time. We observed that infants’ first antibiotic prescriptions occurred more frequently with increasing age and during the winter months. As a result, the association between birth season and risk of having a redeemed antibiotic prescription during infancy depended partly on age. In our interrupted time series analysis, we found that 1-year risk of at least 1 redeemed amoxicillin prescription was dynamic, decreasing after the IRF bulletin and the PCV7 catch-up program, and again after rollout of the PCV13 program.
Our risk and rate estimates corroborate previous findings from substantially smaller studies on antibiotic prescribing among infants in Europe.20,39 Previous studies8,9,17–21 of antibiotic prescribing have focused on estimating rates. We used granularly defined time scales to provide new information on overall and medication-specific antibiotic prescribing patterns, seasonal effects, and secular trends. We described rates (prescriptions ÷ person time) and burden (days supplied ÷ person time) but focused primarily on estimating risks. Risks provide aggregated information about individual infants with and without redeemed antibiotic prescriptions by posing the question, “in a well-defined cohort, which children had a redeemed antibiotic prescription in their first year, and when was their first prescription redeemed?” In contrast, rate and burden estimates require additional assumptions40,41 and tend to shift the research focus to a less strictly defined group, by asking, “how many prescriptions (or days supplied) were redeemed for children in a given subgroup, defined by calendar time or another characteristic?” We recognize that each of these questions has merit depending on the setting; however, if an investigator wishes to assess individual-level determinants of antibiotic prescribing, then the risk estimator is of primary importance.
Our results for birth season cohort effects corresponded with patterns of acute otitis media, a prominent indication for antibiotic treatment, which peaks in winter and for children >6 months old.12 Researchers for two previous studies18,19 have considered seasonal differences in antibiotic prescribing in children, also showing peak prescribing in winter. However, because those studies focused on rates, interpreting seasonality was tied to time periods of peak usage rather than to intrinsically different cohorts of children (ie, birth season cohorts). Other prominent studies of antibiotic prescribing in children1,9,17,20,39 have not included explicit assessments of seasonal differences.
These new findings on birth season cohort effects inform future researchers focused on antibiotics as the exposure in 2 ways: (1) antibiotic exposure status might differ meaningfully between children born in different seasons, and (2) birth season differences in age at first antibiotic use might modify intended or unintended effects of antibiotic treatment given some children’s increased vulnerability to microbial insults in early life.42,43
This is the first study in which the impact of the IRF bulletin (in 2007) and PCV programs (from 2007 onward) on pediatric antibiotic prescribing patterns in Denmark is explicitly evaluated. Ecologic data on PCV receipt among children in Denmark have revealed evidence of sharp increases in coverage from ∼50% (children born in 2006) to sustained levels >85% for children born from 2007 to 2012.44–46 PCV programs have been associated with marked decreases in invasive pneumococcal disease in Danish children.47 Our results therefore provide potential evidence of a downstream effect of vaccination on antibiotic dispensing brought about by decreases in the proportion of children presenting with bacterial infections requiring antibiotic treatment.
It is unclear why amoxicillin trended upward after implementation of the standard PCV7 program. Potential explanations include (1) changes in circulating illnesses from 2007 to 2010, particularly among children born in late 2009 whose elevated risk appears unique compared with the overall trend in the 2 previous years; (2) a temporary minimum threshold effect32 from 2007 to 2008; and (3) limited impact of PCV7 on infection prevention.
The delayed decrease in the trend for amoxicillin after allowing a 7-month lag could result from a stronger lagged (versus immediate) effect of the bulletin or the PCV7 catch-up program, or both. If the bulletin caused the lagged decrease, then it would have affected children born after January 1, 2007, more than those born earlier. This would signify a larger effect on infants <3 months old when the bulletin was published compared with infants who were 3 to 12 months old at that time. Given that hazard increased with age during infancy, a discernible impact of the bulletin would likely have occurred some months after it was published. At the same time, a strong lagged effect of the PCV7 catch-up program would be plausible if the standard PCV7 program had been associated with decreasing risk. Researchers for future studies should assess infection trends and antibiotic prescribing patterns after 2012, to examine the potential ongoing effect of the PCV program.
This study has limitations to consider when interpreting results. We lacked data on indication because general practitioners (GPs) (who are not mandated to record a diagnosis to issue prescriptions48) administer the large majority of antibiotic prescribing for infants. This feature of the health care system limited our ability to explore infection trends coincident with antibiotic prescribing changes over time or across population subgroups. This limitation is exemplified by the differences between our observations of antibiotic dispensing trends compared with a previous description17 of infection trends in Denmark for Mycoplasma pneumoniae and Bordetella pertussis. We also lacked individual-level data on vaccination receipt, which limited inference in the interrupted time series analysis. We note limited ability to compare results to a control population in the interrupted time series analysis33,49 because the IRF bulletin and PCV programs pertained to all children born in Denmark. In addition, our ability to observe stronger effects of the bulletin and PCV programs on antibiotic dispensing may have been limited by ongoing stewardship efforts and lower antibiotic utilization in Denmark compared with most other European countries during the study period.8,23 Antibiotic prescribing trends over time are driven by the prevalence and infectiousness of circulating illnesses; although we controlled for seasonality and carefully selected relevant interruptions and their onset times, our results are limited by our inability to account for underlying differences in circulating illness from year to year and other unmeasured or unknown temporal influences.
This study also has several strengths. The registry databases facilitated our implementation of a large, well-defined, nationwide, population-based cohort study of all children born in Denmark during this 9-year period when antibiotic prescribing practices were in transition. The tax-supported health care system for the entire Danish population includes free access to medical care and partial reimbursement of prescribed medications,29 leading to minimal disparity in access to health care services in this study. In our study, we linked individual-level data across multiple registries to compare infants’ redeemed antibiotic prescriptions with demographic and health-related characteristics. Furthermore, the registries that we used for this study contain accurate data on the date and medication type of redeemed antibiotic prescriptions, date of birth, residence, and other variables we assessed.28–30
Our estimates for the total birth season cohort effect on antibiotic prescribing are unlikely to be confounded because season of birth is not affected by other risk factors for antibiotic use during infancy (eg, birth order, sex, gestational age at birth).50 Throughout the study period, population-level characteristics of children born in Denmark did not change, no new antibiotic formulations were introduced, and the government’s administrative system for prescribing and dispensing antibiotics to infants did not change. Stability of the study population and data systems during the study period limits potential for the interrupted time series study to be biased by confounding, measurement error, or cointervention effects.32,49
Children’s season of birth impacted both their overall risk of having a redeemed antibiotic prescription during infancy and their age at first redeemed antibiotic prescription. Antibiotic prescribing was dynamic over the study period but decreased after a bulletin with guidelines on rational antibiotic use in general practice and rollout of 2 nationwide PCV programs. The approaches used in this study provide a working example of how “big data” can enable researchers to extract new information on risk, rate, and burden measures of antibiotic prescribing in infants, including the correlation between these measures. Finally, these birth cohort effects have implications for future studies of antibiotic stewardship, effectiveness, and safety in children, because some children may be inherently more susceptible than others to downstream effects of antibiotics.
- Accepted June 9, 2017.
- Address correspondence to Alan C. Kinlaw, PhD, Cecil G. Sheps Center for Health Services Research, University of North Carolina at Chapel Hill, CB#7590, Chapel Hill, NC 27599. E-mail:
FINANCIAL DISCLOSURE: Dr Kinlaw reports receiving grant support for this study from the Eunice Kennedy Shriver National Institute of Child Health and Human Development (T32 HD052468-05); a National Service Research Award Postdoctoral Traineeship from the Agency for Healthcare Research and Quality sponsored by the Cecil G. Sheps Center for Health Services Research at the University of North Carolina at Chapel Hill (5T32 HS000032-28); the Graduate School, Department of Epidemiology, and Gillings School of Public Health at the University of North Carolina at Chapel Hill; and the Center for Pharmacoepidemiology, which receives financial support from GlaxoSmithKline, UCB BioSciences, and Merck. Dr Stürmer reports receiving salary support as director of the Center for Pharmacoepidemiology (current members: GlaxoSmithKline, UCB BioSciences, and Merck) and research support from pharmaceutical companies (Amgen, AstraZeneca) to the Department of Epidemiology at the University of North Carolina at Chapel Hill; Dr Stürmer does not accept personal compensation of any kind from any pharmaceutical company. Dr Stürmer owns stock in Novartis, Roche, BASF, AstraZeneca, Johnson & Johnson, and Novo Nordisk. Dr Lund reports receiving grant support from a Research Starter Award from the Pharmaceutical Research and Manufacturers of America Foundation to the University of North Carolina Department of Epidemiology. Dr Pedersen, Ms Frøslev, and Dr Sørensen did not report receiving fees, honoraria, grants, or consultancies; the Department of Clinical Epidemiology is, however, involved in studies with funding from various companies as research grants to (and administered by) Aarhus University. Dr Kappelman reports receiving grant support and consulting fees from GlaxoSmithKline and Johnson & Johnson and stock ownership in GlaxoSmithKline and Johnson & Johnson. Dr Mack reports being a full-time employee at Quintiles, Inc, and as such consults for large pharmaceutical companies. Dr Daniels has indicated she has no financial relationships relevant to this article to disclose.
FUNDING: Supported by the Eunice Kennedy Shriver National Institute of Child Health and Human Development (T32 HD052468-05); a National Service Research Award Postdoctoral Traineeship from the Agency for Healthcare Research and Quality sponsored by the Cecil G. Sheps Center for Health Services Research at the University of North Carolina at Chapel Hill (5T32 HS000032-28); the Graduate School, Department of Epidemiology, and Gillings School of Public Health at the University of North Carolina at Chapel Hill; and the Center for Pharmacoepidemiology, which receives financial support from GlaxoSmithKline, UCB BioSciences, and Merck. The funding sources had no role in the design or conduct of the study, data analysis or interpretation, manuscript preparation, or the decision to submit the manuscript for publication. Funded by the Agency for Healthcare Research and Quality. Funded by the National Institutes of Health (NIH).
POTENTIAL CONFLICT OF INTEREST: Dr Kinlaw reports receiving grant support for this study from the Eunice Kennedy Shriver National Institute of Child Health and Human Development (T32 HD052468-05); a National Service Research Award Postdoctoral Traineeship from the Agency for Healthcare Research and Quality sponsored by the Cecil G. Sheps Center for Health Services Research at the University of North Carolina at Chapel Hill (5T32 HS000032-28); the Graduate School, Department of Epidemiology, and Gillings School of Public Health at the University of North Carolina at Chapel Hill; and the Center for Pharmacoepidemiology, which receives financial support from GlaxoSmithKline, UCB BioSciences, and Merck. Outside this study, Dr Stürmer reports receiving investigator-initiated research funding and support as principal investigator (R01 AG023178) from the National Institute on Aging and as coinvestigator (R01 CA174453, R01 HL118255, R21-HD080214), National Institutes of Health (NIH); Dr Stürmer also receives salary support as director of the Comparative Effectiveness Research Strategic Initiative, the North Carolina Translational and Clinical Sciences Institute, the University of North Carolina Clinical and Translational Science Award (UL1TR001111), and as director of the Center for Pharmacoepidemiology (current members: GlaxoSmithKline, UCB BioSciences, Merck) and research support from pharmaceutical companies (Amgen and AstraZeneca) to the Department of Epidemiology, University of North Carolina at Chapel Hill; Dr Stürmer does not accept personal compensation of any kind from any pharmaceutical company; Dr Stürmer owns stock in Novartis, Roche, BASF, AstraZeneca, Johnson & Johnson, and Novo Nordisk. Outside this study, Dr Lund reports receiving grant support from the University of North Carolina Oncology Clinical Translational Research Training Program (K12 CA120780), and from a Research Starter Award from the Pharmaceutical Research and Manufacturers of America Foundation to the University of North Carolina Department of Epidemiology. Outside this study, Dr Pedersen, Ms Frøslev, and Dr Sørensen did not report receiving fees, honoraria, grants or consultancies; the Department of Clinical Epidemiology is, however, involved in studies with funding from various companies as research grants to (and administered by) Aarhus University. Outside this study, Dr Kappelman reports receiving grant support and consulting fees from GlaxoSmithKline and Johnson & Johnson, and stock ownership in GlaxoSmithKline and Johnson & Johnson. Outside this study, Dr Mack reports being a full-time employee at Quintiles, Inc, and as such consults for large pharmaceutical companies; and Dr Daniels has indicated she has no financial relationships relevant to this article to disclose.
COMPANION PAPER: A companion to this article can be found online at www.pediatrics.org/cgi/doi/10.1542/peds.2017-1695.
- Vaz LE,
- Kleinman KP,
- Raebel MA, et al
- American Academy of Pediatrics Subcommittee on Management of Acute Otitis Media
- Thompson PL,
- Spyridis N,
- Sharland M, et al
- Tregoning JS,
- Schwarze J
- Pedersen ASS,
- Gahrn-Hansen B
- Andersen PH
- Valentiner-Branth P,
- Andersen P,
- Simonsen J, et al
- Shadish WR,
- Cook TD,
- Campbell DT
- Choudhry NK,
- Fischer MA,
- Avorn JL, et al
- Stolwijk AM,
- Straatman H,
- Zielhuis GA
- Nam J
- The Danish Health Data Authority
- Statens Serum Institut
- Statens Serum Institut
- Statens Serum Institut
- Rothman KJ,
- Greenland S,
- Lash TL
- Copyright © 2017 by the American Academy of Pediatrics