ORIGINAL www.jpeds.

com • THE JOURNAL OF PEDIATRICS

ARTICLES
A Cost-Utility Analysis of 5 Strategies for the Management of Acute Otitis
Media in Children
Nader Shaikh, MD, MPH1,2, Emily E. Dando, BA2, Mark L. Dunleavy, BS3, Dorothy L. Curran, BS4, Judith M. Martin, MD1,2,
Alejandro Hoberman, MD1,2, and Kenneth J. Smith, MD, MS5

Objective To assess whether antimicrobial therapy in young children with acute otitis media reduces time to reso-
lution of symptoms, overall symptom burden, and persistence of otoscopic evidence of infection. We used a cost-
utility model to evaluate whether immediate antimicrobial treatment seems to be worthwhile, and if so, which antimicrobial
agent is most cost effective.
Study design We compared the cost per quality-adjusted life-day of 5 treatment regimens in children younger
than 2 years of age with acute otitis media: immediate amoxicillin/clavulanate, immediate amoxicillin, immediate
cefdinir, watchful waiting, and delayed prescription (DP) for antibiotic.
Results The 5 treatment regimens, listed in order from least effective to most effective were DP, watchful waiting,
immediate cefdinir, immediate amoxicillin, and immediate amoxicillin/clavulanate. Listed in order from least costly
to most costly, the regimens were DP, immediate amoxicillin, watchful waiting, immediate amoxicillin/clavulanate,
and immediate cefdinir. The incremental cost-utility ratio of immediate amoxicillin compared with DP was $101.07
per quality-adjusted life-day gained. The incremental cost-utility ratio of immediate amoxicillin/clavulanate com-
pared with amoxicillin was $2331.28 per quality-adjusted life-day gained.
Conclusions In children younger than 2 years of age with acute otitis media and no recent antibiotic exposure,
immediate amoxicillin seems to be the most cost-effective initial treatment. (J Pediatr 2017;189:54-60).

See editorial, p 5

A
cute otitis media (AOM) is the most frequent reason that children in the United States receive antimicrobial therapy.
Evidence from 2 recent randomized, placebo-controlled clinical trials in young children diagnosed with AOM using
stringent criteria indicates that antimicrobial therapy compared with watchful waiting (WW) results in faster symp-
tomatic relief, lower rates of treatment failure, and lesser persistence of otoscopic evidence of infection.1,2 Nonetheless, the ques-
tion remains whether risks and costs of antimicrobial therapy, particularly in children younger than 2 years of age, outweigh
its benefits. The lack of clarity results in part from difficulties in integrating respective benefits and harms of the various avail-
able therapies. For example, 1 study compared children 6 months to 10 years of age receiving a delayed prescription (DP) for
antibiotic, mainly amoxicillin, which parents could fill after 72 hours if children failed to improve, with children receiving im-
mediate antimicrobials, albeit at lower dose than currently recommended in the United States. The latter group had, on average,
1.1 fewer days of illness and 0.72 fewer nights with disturbed sleep, but twice as much diarrhea.3
Accordingly, it seemed worthwhile to analyze available data systematically in an effort to determine whether immediate an-
timicrobial treatment seems justified and, if so, which antimicrobial agent is most cost effective. A cost-utility analysis is par-
ticularly well-suited for this task because all outcomes, whether positive, such as reduction of symptoms, or negative, such as
diarrhea, can be combined into a single metric, namely, quality-adjusted life-days (QALDs) accrued, which can be compared
across treatment options.
Accordingly, we constructed a decision-analytic model to compare the cost effectiveness of 5 frequently used options for man-
aging children with AOM. We also explored the impact of inaccurate diagnosis on the relative cost effectiveness of these treat-
ment options.

From the 1Children’s Hospital of Pittsburgh of UPMC,

Division of General Academic Pediatrics; 2Department of
Pediatrics, University of Pittsburgh School of Medicine,
Pittsburgh; 3The Commonwealth Medical College,
AOM Acute otitis media Scranton, PA; 4University of Minnesota School of
Medicine, Minneapolis, MN; and 5Division of General
AOM-SOS scale Acute Otitis Media Severity of Symptom Scale
Internal Medicine, University of Pittsburgh, Pittsburgh, PA
DP Delayed prescription
The authors declare no conflicts of interest.
ICER Incremental cost-effectiveness ratio
QALD Quality-adjusted life-day 0022-3476/$ - see front matter. © 2017 Elsevier Inc. All rights
WW Watchful waiting reserved.
http://dx.doi.org10.1016/j.jpeds.2017.05.047

54
 Volume 189 • October 2017

Children would be categorized as having persistent symptoms

Methods
We compared the clinical and economic outcomes of 5 treat-
ment strategies in a hypothetical cohort of symptomatic chil-
dren younger than 2 years of age who were diagnosed with AOM,
based on symptoms and otoscopic findings. As recommended
by the Panel on Cost-Effectiveness in Health and Medicine,4
we adopted a societal perspective and included both direct and
indirect medical costs in our model. We do, however, also report
results from the payer’s perspective (ie, including only direct
medical costs).5 We constructed and analyzed our decision
tree using TreeAge Pro software (TreeAge, Williamstown,
Massachusetts).
We set a time horizon of 30 days because most of the ben-
efits and harms attributable to an episode of AOM occur within
this period, and because previous studies have failed to find
an association between treatment modality and incidence of
late (beyond 30 days) recurrences of AOM.6-8
We compared 5 treatment strategies: immediate amoxicillin/
clavulanate (90/6.4 mg/kg/day for 10 days), immediate
amoxicillin (90 mg/kg/day for 10 days), immediate cefdinir
(14 mg/kg/day for 10 days), WW, and DP. Parents of chil-
dren in the DP group would be given a prescription for
amoxicillin as rescue treatment that could be filled if their
children’s symptoms did not improve as quickly as the parents
had expected. Parents of children in all other groups would
be asked to return for an office visit if their children were
not improving.
The Figure (available at www.jpeds.com) shows the struc-
ture of the decision tree. For each of the other 4 treatment strat-
egies, we considered 3 possible outcomes within the first 7 days:
improved symptoms, persistent symptoms, or mastoiditis. Chil-
dren in whom symptoms had improved satisfactorily would
continue on the initially prescribed treatment regimen, and
would be reevaluated only if symptoms recurred within 30 days.
if they experienced any of the following during the initial 7
days of therapy: worsening or lack of improvement in scores
on the Acute Otitis Media Severity of Symptom Scale (AOM-
SOS),7-9 otorrhea associated with perforation of the tym-
panic membrane, or hospitalization for an otitis-related reason
other than mastoiditis. In all branches except for the DP branch,
such children would be reevaluated; those with evidence of con-
tinuing infection on examination would be categorized as
having early clinical failure and would be prescribed rescue an-
timicrobial therapy, consisting of amoxicillin, or in children
for whom amoxicillin or amoxicillin/clavulanate had been ini-
tially prescribed, amoxicillin/clavulanate. Children with per-
sistent symptoms but without evidence of continuing infection
would be categorized as having early clinical success and would
continue to be managed according to the original treatment
strategy. In the DP branch, children whose symptoms failed
to improve within 48 hours of diagnosis were assumed to have
been started on antimicrobial therapy without the need for a
consultation; those whose symptoms continued to persist would
be reevaluated by a clinician. Children with acute mastoiditis
would be evaluated initially in the office, then hospitalized and
receive intravenous antimicrobial therapy for an average of 3.2
days.9 After discharge, they would receive antimicrobial therapy
orally for 30 days.10
Children experiencing a second episode of AOM during the
30-day analytic period would be treated with amoxicillin/
clavulanate if the children had received antimicrobial therapy
initially, and with amoxicillin if they had not, for 10 days.
Because few children would be expected to experience more
than 3 episodes of AOM within a 30-day period, we modeled
a maximum of 3 episodes. We did not consider complica-
tions of AOM other than acute mastoiditis because such com-
plications are rare. In Table I, the probabilities of outcomes
in relation to the various treatment strategies considered in
the decision-analytic model are summarized.

Table I. Conditional probabilities of outcomes in relation to treatment strategies used in the cost-utility model
Designated initial treatment strategy
Immediate
DP Immediate amoxicillin WW amoxicillin/clavulanate Immediate cefdinir
Treatment outcomes* Probability (%) of outcome occurrence (range used in the model)
Persistent symptoms days 1-7 45 (37-53) 32 (24-40)† 45 (37-53) 29 (21-37) 34 (26-42)†
Early clinical failure days 1-7 100 16 (5-27)† 37 (25-49) 11 (1-21) 20 (8-32)†
AOM recurrence days 8-30 27 (16-38) 26 (0-59)† 27 (9-45) 25 (0-66) 26 (0-54)†
Early clinical success days 1-7 0 84 (73-95)† 63 (51-75) 89 (79-99) 80 (68-92)†
AOM recurrence days 8-30 0 16 (4-28)† 21 (8-34) 15 (3-27) 17 (5-29)†
Improved symptoms days 1-7 55 (47-63) 68 (60-76)† 55 (47-63) 71 (63-79) 66 (58-74)†
AOM recurrence days 8-30 12 (5-19) 5 (1-9)† 12 (5-19) 4 (0-8) 6 (1-11)†
Mastoiditis§ 0.038 (0.032-0.044) 0.018 (0.015-0.021) 0.038 (0.032-0.044) .018 (.015-.021) .018 (.015-.021)
Diarrhea 12 (7-15)‡ 18 (12-24)¶ 5 (2-8) 24 (17-31) 11 (9-13)¶
Diaper rash¶ 4 (0-10)‡ 6 (1-15)** 2 (0-6) 11 (6-17) 5 (2-8)
Body rash¶ 3 (1-4)‡ 3 (2-4)** 3 (1-4) 5 (3-6) 1 (0-2)

*All data based on Hoberman et al 20112 unless otherwise indicated.

†Calculated using bacteriologic efficacy (see Appendix; available at www.jpeds.com).
‡Calculated from WW branch considering the proportion of children who would have received antimicrobials.
§From Thompson et al.11
¶We systematically reviewed available studies and, if multiple studies were available, used meta-analysis to arrive at these pooled estimates.
**Because of the paucity of data, we used data from studies of both low and high-dose amoxicillin.

55
THE JOURNAL OF PEDIATRICS • www.jpeds.com Volume 189

Table II. Costs per AOM episode and disutility values associated with specific clinical states, as used in the cost-utility
model
Cost ($) Reference
Cost elements and clinical states Best estimate Plausible range* Author Year
Cost elements
Amoxicillin/clavulanate (90/6.4 mg/kg/d) x 10 d 63.58 31.79-95.37 Red Book13 2015
Amoxicillin (90 mg/kg/d) x 10 d 9.81 4.91-14.72 Red Book13 2015
Cefdinir (14 mg/kg/d) x 10 d 50.95 25.48-76.43 Red Book13 2015
Barrier cream for diaper dermatitis (zinc oxide cream) 2.50 1.25-3.75 Red Book13 2015
Topical antifungal (clotrimazole 1% cream) 6.42 3.21-9.63 Red Book13 2015
Diapers 0.32 0.16-0.48 Walmart
Office visit
Physician reimbursement 72.94 36.47-109.41 CMS14 2016
Nonmedical costs per office visit† 19.61 9.81-29.42 Hoberman2 2011
Cost of work lost per office visit‡ 77.59 38.80-116.39 Smith,15 BLS15 2015
Hospitalization for acute mastoiditis§ 7487 3744-11 231 HCUP, CMS, BLS9,14,15 2015, 2016

Disutility value
Best estimate Plausible range* Author Year
Clinical states
Acute mastoiditis 0.56 0.36-0.76 Coco16 2007
AOM 0.21 0.01-0.41 Oh17 1996
Diarrhea¶ 0.12 0.00-0.32 Melliez18 2008
Body rash** 0.04 0.00-0.24 Garside19 2005
Upper respiratory tract infection†† 0.05 0.00-0.25 Resch20 2012
Diaper dermatitis‡‡ 0.01 0.00-0.21 Garside19 2005

BLS, Bureau of Labor Statistics; CMS, Centers for Medicare/Medicaid Services; HCUP, Healthcare Costs and Utilization Project.
Utility is a term reflecting the health-related quality of life of an individual at a particular point in time. Utility is scaled from 0 to 1, where 0 denotes death and 1 denotes perfect health. Disutility
values are equal to 1 minus the utility values.
*Used 50%-150% of best estimate for costs and ±0.2 for utility values.
†Includes $3 for transportation and/or parking and $15 for childcare (11% of parents had an interim visit and needed childcare for 0.3 days at $20.97/h.).
‡Assumed 3.7 hours per office visit and $20.97/h average wage.
§Includes mean cost of an initial office visit, hospitalization, physician reimbursement, and cost of days lost from work (3.2 d/8 h work lost per day at $20.97/h.).
¶Assumed disutility similar to that of mild rotavirus diarrhea.
**Assumed disutility similar to that of moderate eczema.
††Assumed disutility similar to that of infection with respiratory syncytial virus that does not require hospitalization.
‡‡Assumed disutility similar to that of mild to moderate eczema.

We estimated probabilities for the amoxicillin/clavulanate
branch and for the WW branch from data from the Hoberman
et al2 trial that compared treatment with amoxicillin/clavulanate
with placebo for AOM in children younger than 2 years of age.
That study was well-suited for the present purpose for a number
of reasons: all otoscopic examinations were conducted by ex-
perienced otoscopists; participants were followed for 30 days,
with visits on days 4-6, 10-12, and 20-22; a validated symptom
scale (AOM-SOS)7,8 was used to detail children’s symptoms
during the initial 7 days of treatment and at each subsequent
study visit; parents had 24-hour direct access to the investi-
gators, and for the first 3 days received daily telephone calls;
and children who were not improving were promptly started
on rescue antimicrobial therapy.
We estimated probabilities of clinical efficacy in the
amoxicillin and cefdinir branches based on their reported bac-
teriologic efficacy, using methods described by Marchant et al12
and with extrapolations from the previously reported bacte-
riological efficacy of amoxicillin/clavulanate in relation to its
clinical efficacy as reported in the Hoberman et al trial.2 We
started from bacteriologic rather than clinical efficacy for the
amoxicillin and cefdinir branches because of the heterogeneity
56
of methods and definitions in earlier studies of their clinical
efficacy. Calculations are shown in the Appendix (available at
www.jpeds.com).
We estimated probabilities of clinical efficacy in the DP
branch starting from probabilities in the WW branch and taking
into account the additional number of children in the DP
branch who would have received a rescue antimicrobial.
We estimated the probability of mastoiditis from a report
by Thompson et al,11 which is, to our knowledge the only report
of the rate of mastoiditis in children with AOM in relation to
their receipt of antimicrobial treatment. To estimate the prob-
ability of diarrhea for children in the amoxicillin/clavulanate
and WW branches, we used data from the aforementioned
study by Hoberman et al,2 and for children in the DP branch,
data from children in the WW branch as described. For chil-
dren in the amoxicillin and cefdinir branches, we used pooled
estimates from a systematic review of available studies (un-
published data), and we estimated further that the incidence
of body rash in the amoxicillin branch would be the same as
in the amoxicillin/clavulanate branch.
Base-case estimates represent our determinations of the best
available data or, in the absence of data, our best estimates. The
Shaikh et al
October 2017
estimates used in the model for all costs and all quality-of-
life values associated with individual clinical states are sum-
marized in Table II.
Healthcare costs included costs of visits, medications (an-
tibiotics, barrier cream for diaper dermatitis, and creams used
to treat monilial infection), and diapers. We used 2015 Medi-
care reimbursement data to estimate pediatric costs for out-
patient office visits and hospitalizations. We obtained costs of
medications from the 2015 pharmacy Red Book.13
Nonmedical costs that we estimated for each outpatient visit
included cost of work lost by parents, cost of childcare if needed,
and cost of parking and/or transportation. We estimated the
cost of work lost by multiplying estimated hours of work
missed21 by average wage as reported by the US Bureau of Labor
Statistics.15
We measured quality of life using a utility approach, where
utility reflects the health-related quality of life of an indi-
vidual at a particular point in time. Utility values are scaled
from 0 to 1, where 0 denotes death and 1 denotes perfect health.
Disutility values are obtained by subtracting utility values from
1. Thus, 1 QALD equates to 1 full day in perfect health, whereas
0.5 QALD might reflect either 1 half-day in perfect health or
1 full day lived with a condition having a disutility value of
0.5. We determined specific QALD values by subtracting the
number of QALDs lost attributable to each adverse state from
the maximum total number of QALDs. Because the time horizon
considered was 30 days, the maximum possible number of
QALDs for an individual child was 30. We determined the
number of QALDs lost attributable to each adverse state by
multiplying the time spent in each state by its associated disutility
value. We used colonization and resistance data from an earlier
study in which young children with AOM were treated with
amoxicillin/clavulanate2 to take into account the potential impact
of antimicrobials on the emergence of bacterial resistance.
We estimated the duration of each adverse state based on
data from the study by Hoberman et al.2 We estimated disutility
values for the various adverse states from other, earlier
studies15,21-24 and we summarized these values in Table II. We
then combined the numbers of QALDs lost with medical costs
to arrive at the cost per QALD, and we used this measure to
assess the cost effectiveness of each treatment.
Because, as noted, all examinations in the Hoberman et al
study2 that served as the basis for most of the estimates we used
had been conducted by validated otoscopists, we wondered
whether our findings might differ substantively in settings where
the accuracy of examiners’ otoscopic diagnoses might be more
representative of the diagnostic accuracy of primary care pe-
diatricians. Accordingly, we constructed a model that incor-
porated, as an element, misdiagnosis of AOM of some degree.
In this model, for each treatment strategy we included a hy-
pothetical subset of children who were diagnosed errone-
ously as having AOM and thus received treatment, and a subset
of children who were diagnosed erroneously as not having AOM
and thus failed to receive treatment. For this exercise, we used
data from an earlier study22 in which we found that, on average,
a group of primary care pediatricians diagnosed 84% of cases
of AOM and 67% of non-AOM cases correctly.
A Cost-Utility Analysis of 5 Strategies for the Management of Acute Otitis Media in Children
ORIGINAL ARTICLES
Statistical Analyses
Our model took into account costs and disutility values for
each of the 5 treatment strategies. We compared the strate-
gies concerning their incremental cost-effectiveness ratios
(ICERs), defined as the extra cost of a more expensive strat-
egy over the cost of the nearest less expensive strategy, divided
by the extra clinical benefit conferred by the more expensive
strategy. We used a willingness-to-pay ceiling, defined as a cost-
effectiveness ratio that denotes the most that society has been
considered willing to pay for an incremental gain in health,
namely, $274 per QALD gained (derived from $100 000 per
quality-adjusted life-year).23 We conducted 1-way sensitivity
analyses for all variables by varying baseline estimates within
clinically plausible ranges. Whenever raw data were available,
we based ranges on 95% CIs calculated using the Clopper-
Pearson formula. We also conducted probabilistic sensitivity
analyses in which we varied all measures simultaneously. In
these analyses, values from each probability distribution are
randomly selected during each of 10 000 iterations, and the
percentage of iterations for which a given strategy is favored
is tracked.
Results
Table III shows the costs, utilities, and ICERs for each of the
5 treatment options. The cost of each treatment strategy, listed
in order from least costly to most costly, was as follows: DP,
immediate amoxicillin, WW, immediate amoxicillin/clavulanate,
and immediate cefdinir. The effectiveness of each strategy, listed
from least effective to most effective, was as follows: DP, WW,
immediate cefdinir, immediate amoxicillin, and immediate
amoxicillin/clavulanate. DP was the least costly and also the
least effective treatment strategy. Because WW and immedi-
ate cefdinir were more expensive and less effective than im-
mediate amoxicillin, they were not considered in calculating
the ICERs. The ICER of immediate amoxicillin compared with
DP was $101.07 per QALD gained, whereas the ICER of im-
mediate amoxicillin/clavulanate compared with immediate
amoxicillin was $2331.28 per QALD gained. The projected in-
cidence of mastoiditis did not influence our findings substan-
tially. Results of the probabilistic sensitivity analysis were
consistent with the previous analysis. Adopting the payer’s per-
spective when calculating costs yielded similar results (Table III).
Table IV shows the number of antimicrobial prescriptions
and number of visits in each branch for every 100 children
entered into the model. WW resulted in the fewest antimi-
crobial prescriptions (34) but required the most medical visits
(168). DP resulted in the second fewest antimicrobial pre-
scriptions (62) and the fewest medical visits (123). In a 1-way
sensitivity analysis, we attempted to identify variables (from
Table I) that changed the preferred strategy from immediate
amoxicillin to either DP or immediate amoxicillin/clavulanate.
One such variable was the probability of improved symp-
toms with amoxicillin; above 61% amoxicillin was preferred,
and below 61% amoxicillin/clavulanate was preferred. Values
for the only 2 variables that had any effect are shown in the
Appendix.
57
THE JOURNAL OF PEDIATRICS • www.jpeds.com Volume 189

Table III. Comparative cost-effectiveness of 5 treatment strategies for AOM in children less than 2 years of age
Treatment strategy*
Immediate Immediate Immediate
Analytic components DP amoxicillin WW amoxicillin/clavulanate cefdinir
Costs, $
Mastoiditis 2.85 1.35 2.85 1.36 1.36
Diapers and diaper dermatitis creams 4.70 5.64 1.79 6.37 3.52
Antibiotics 12.81 18.79 5.78 70.37 61.83
Visits 31.96 69.86 105.58 62.19 75.54
Total costs 59.21 95.58 115.97 140.19 142.14
QALDs lost to
Mastoiditis 0.00 0.00 0.00 0.00 0.00
Diarrhea 0.09 0.10 0.04 0.13 0.07
Diaper or body rash 0.01 0.01 0.01 0.02 0.01
AOM symptoms 1.34 0.96 1.32 0.91 1.00
Total QALDs lost 1.43 1.07 1.37 1.06 1.08
QALDs experienced in 30-d period 28.57 28.93 28.63 28.95 28.92
ICER (ie, total cost/QALD in 30-day period) comparing treatment strategy N/A 101.07 Dominated‡ 2331.28 Dominated‡
with next most effective strategy from a societal perspective†

N/A, not applicable.

*The strategies are listed in order from least expensive to most expensive.
†The intervention costs more and is no more effective than the comparator.
‡The corresponding results from the payer's perspective were N/A, 44.36, dominated, 2530.80, and dominated. The corresponding results excluding nonmedical costs (ie, cost of work lost, cost
of childcare if needed, and cost of parking and/or transportation, cost of diapers) were N/A, 48.86, dominated, 2536.84, and dominated.

Overall results were similar when we applied the model to
the hypothetical primary care setting described. Although the
relative cost effectiveness of amoxicillin compared with DP de-
creased (ICER increased from $118.08 to $148.59), amoxicillin
nonetheless remained the most cost-effective option.
Discussion
We sought to determine whether immediate antimicrobial
therapy, given its benefits, adverse effects, and costs, is cost ef-
fective in children younger than 2 years of age with AOM and,
if so, which antimicrobial agent is most cost effective.
Of the 5 strategies we considered, we found that immedi-
ate amoxicillin treatment was the most cost effective. The cost
for the immediate-amoxicillin treatment branch was $36.37
more than the cost for the DP branch (Table III), mainly
because of extra visits incurred by children with persistent
symptoms who had received amoxicillin. The incremental gain
in QALD was 0.36 days. Given that the disutility value of a day
with AOM is 0.21 (Table II), this represents 1.7 fewer days (ie,
0.36/0.21) with symptoms of AOM for the immediate-
amoxicillin strategy compared with the DP strategy, at an ad-
ditional cost of $101.07 per QALD gained. This is less than the
$274 per QALD gained that is generally considered cost
effective.23 In contrast, compared with immediate amoxicillin,
immediate amoxicillin/clavulanate costs an additional $2331.28
per QALD gained. This high ICER is a function of the limited
gain in QALDs (0.02 QALD over the 30-day period) when using
amoxicillin/clavulanate rather than amoxicillin. Thus,
amoxicillin/clavulanate as initial therapy does not represent the
most cost-effective option. Amoxicillin/clavulanate became the
most cost-effective first-line treatment option only when the
probability of symptom improvement with amoxicillin was pro-
jected to be quite low, at less than 61%, compared with our
base-case estimate of 68% (Appendix). This circumstance may
occur, for example, when infection with a nonsusceptible patho-
gen is suspected, such as in children with persistent symp-
toms despite treatment with amoxicillin, or in children with
recurrence who had recently been treated with antimicrobial
therapy. Our results thus do support the use of amoxicillin/
clavulanate as a second-line treatment.
Immediate treatment with cefdinir, attractive because of the
drug’s once daily administration and favorable taste, was the
most expensive strategy and was less effective than amoxicillin.
WW resulted in the lowest number of antimicrobial prescrip-
tions, and accordingly, may represent a favorable option in cir-
cumstances where limiting antimicrobial use seems of
paramount importance and limiting costs is less important.

Table IV. Number of antimicrobial prescriptions and number of medical visits per 100 children in a 30-day period for
each treatment strategy
Treatment strategy
Immediate Immediate Immediate
Cost elements DP amoxicillin WW amoxicillin/clavulanate cefdinir
No. of antimicrobial prescriptions for 100 children in a 30-d period 64 113 34 110 115
No. of medical visits in addition to the initial visit for 100 children in a 30-d period 23 41 62 37 44

58 Shaikh et al
October 2017
The results of this analysis were robust to the extent that
immediate amoxicillin remained the most cost-effective treat-
ment option even after we altered the variables in the model
over clinically plausible ranges. The results of those sensitiv-
ity analyses, however, suggested that DP might be a reason-
able strategy in certain situations, for example, in a child
relatively free of symptoms at the time of diagnosis, such that
the disutility value of AOM would be lower than in our origi-
nal calculation. Further supporting our analysis, immediate
amoxicillin remained the most cost-effective option in the
model in which lower accuracy in diagnosing AOM was pro-
jected, although it was relatively less cost effective than in our
original model (ICER $148.59 compared with $101.07). Given
the large number of prescriptions for antimicrobial drugs that
are written for children with AOM, it seems likely that im-
provements in the diagnostic accuracy of primary care pedia-
tricians could result in a substantial reduction in the cost of
care for children with AOM.24,25
A strength of our study was our use of systematic reviews
to obtain values for most measures. Our study had several limi-
tations. The main limitation of this study is that we did not
take into account the potential impact of antimicrobial-
induced bacterial resistance at the population level, that is, the
possible transmission of nonsusceptible respiratory flora
between children or from children to adults. We used the data
from a single study to arrive at many of the estimates used here.
This may limit the generalizability of our findings. Neverthe-
less, the study we used is the only placebo-controlled trial to
date in which a validated scale was used to follow the symp-
toms of children on a daily basis. Further, we modeled the bac-
teriologic efficacy for cefdinir and amoxicillin from single
studies conducted before pneumococcal vaccination was
routine. Accordingly, shifts in serotypes might have taken place
that could have altered the outcomes we considered. The wide
ranges used in our sensitivity analyses may have mitigated this
limitation. In contrast, if later studies were to find the effi-
cacy of amoxicillin lower than the base-case value we used, im-
mediate amoxicillin/clavulanate could then become the most
cost-effective strategy.
In summary, for children less than 2 years of age with AOM
and no recent antibiotic exposure, immediate treatment with
amoxicillin seems to be the most cost-effective initial treat-
ment strategy currently. ■
We would like to thank Jack L. Paradise, MD, Professor of Pediatrics
from the University of Pittsburgh, for his review and helpful editing of
an earlier draft of this manuscript.
Submitted for publication Dec 9, 2016; last revision received Mar 30, 2017;
accepted May 17, 2017
Reprint requests: Nader Shaikh, MD, MPH, Children’s Hospital of Pittsburgh,
General Academic Pediatrics, 4401 Penn Ave, Pittsburgh, PA 15224. E-mail:
nader.shaikh@chp.edu
References
1. 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;364:116-26.
A Cost-Utility Analysis of 5 Strategies for the Management of Acute Otitis Media in Children
ORIGINAL ARTICLES
2. Hoberman A, Paradise JL, Rockette HE, Shaikh N, Wald ER, Kearney DH,
et al. Treatment of acute otitis media in children under 2 years of age. N
Engl J Med 2011;364:105-15.
3. Little P, Gould C, Williamson I, Moore M, Warner G, Dunleavey J. Prag-
matic randomised controlled trial of two prescribing strategies for child-
hood acute otitis media. BMJ 2001;322:336-42.
4. Weinstein MC, Siegel JE, Gold MR, Kamlet MS, Russell LB. Recommen-
dations of the panel on cost-effectiveness in health and medicine. JAMA
1996;276:1253-8.
5. Sanders GD, Neumann PJ, Basu A, Brock DW, Feeny D, Krahn M, et al.
Recommendations for conduct, methodological practices, and report-
ing of cost-effectiveness analyses: second panel on cost-effectiveness in
health and medicine. JAMA 2016;316:1093-103.
6. Le Saux N, Gaboury I, Baird M, Klassen TP, MacCormick J, Blanchard
C, et al. A randomized, double-blind, placebo-controlled noninferiority
trial of amoxicillin for clinically diagnosed acute otitis media in chil-
dren 6 months to 5 years of age. Can Med Assoc J 2005;172:335-
41.
7. Mygind N, Meistrup-Larsen KI, Thomsen J, Thomsen VF, Josefsson K,
Sorensen H. Penicillin in acute otitis media: a double-blind placebo-
controlled trial. Clin Otolaryngol Allied Sci 1981;6:5-13.
8. van Buchem FL, Dunk JH, van’t Hof MA. Therapy of acute otitis media:
myringotomy, antibiotics, or neither? A double-blind study in children.
Lancet 1981;2:883-7.
9. Healthcare Cost & Utilization Project (HCUP). Agency for Healthcare Re-
search and Quality. http://www.ahrq.gov/data/hcup/. Accessed January 25,
2015.
10. Wald E. Acute mastoiditis in children: Treatment and prevention: UptoDate;
2016 [updated Jan 5, 2015 June 30, 2016]. http://www.uptodate.com/
contents/acute-mastoiditis-in-children-treatment-and-prevention. Ac-
cessed November 12, 2016.
11. Thompson PL, Gilbert RE, Long PF, Saxena S, Sharland M, Wong IC. Effect
of antibiotics for otitis media on mastoiditis in children: a retrospective
cohort study using the United kingdom general practice research data-
base. Pediatrics 2009;123:424-30.
12. Marchant CD, Carlin SA, Johnson CE, Shurin PA. Measuring the com-
parative efficacy of antibacterial agents for acute otitis media: the “Pol-
lyanna phenomenon”. J Pediatr 1992;120:72-7.
13. Red Book: Pharmacy’s Fundamental Reference: PDR Network; 2015.
14. Centers for Medicare and Medicaid Services. License for use of Current
Procedural Terminology. 4th ed. 2016 [updated February 12]. https://
www.cms.gov/apps/physician-fee-schedule/search/search-criteria.aspx. Ac-
cessed November 12, 2016.
15. Bureau of Labor and Statistics. Total private average hourly earnings of
prod. and nonsup. employees - seasonally adjusted 2015. http://data.bls.gov/
cgi-bin/surveymost?bls. Accessed November 14, 2015.
16. Coco AS. Cost-effectiveness analysis of treatment options for acute otitis
media. Ann Fam Med 2007;5:29-38.
17. Oh PI, Maerov P, Pritchard D, Knowles SR, Einarson TR, Shear NH. A
cost-utility analysis of second-line antibiotics in the treatment of acute
otitis media in children. Clin Ther 1996;18:160-82.
18. Melliez H, Levybruhl D, Boelle PY, Dervaux B, Baron S, Yazdanpanah Y.
Cost and cost-effectiveness of childhood vaccination against rotavirus in
France. Vaccine 2008;26:706-15.
19. Garside R, Stein K, Castelnuovo E, Pitt M, Ashcroft D, Dimmock P, et al.
The effectiveness and cost-effectiveness of pimecrolimus and tacrolimus
for atopic eczema: a systematic review and economic evaluation. Health
Technol Assess 2005;9:iii. xi-xiii,1-230.
20. Resch B, Sommer C, Nuijten MJ, Seidinger S, Walter E, Schoellbauer V,
et al. Cost-effectiveness of palivizumab for respiratory syncytial virus in-
fection in high-risk children, based on long-term epidemiologic data from
Austria. Pediatr Infect Dis J 2012;31:e1-8.
21. Smith KJ, Cook RL, Ness RB. Cost comparisons between home- and clinic-
based testing for sexually transmitted diseases in high-risk young women.
Infect Dis Obstet Gynecol 2007;2007:62467.
22. Shaikh N, Stone MK, Kurs-Lasky M, Hoberman A. Interpretation of tym-
panic membrane findings varies according to level of experience. Paediatr
Child Health 2016;21:196-8.
59
THE JOURNAL OF PEDIATRICS • www.jpeds.com
23. Laupacis A, Feeny D, Detsky AS, Tugwell PX. How attractive does a new
technology have to be to warrant adoption and utilization? Tentative guide-
lines for using clinical and economic evaluations. CMAJ 1992;146:473-
81.
24. Grijalva CG, Nuorti JP, Griffin MR. Antibiotic prescription rates for acute
respiratory tract infections in US ambulatory settings. JAMA 2009;302:758-
66.
25. McGrath LJ, Becker-Dreps S, Pate V, Brookhart MA. Trends in antibi-
otic treatment of acute otitis media and treatment failure in children, 2000-
2011. PLoS ONE 2013;8:e81210.
26. Arguedas A, Dagan R, Leibovitz E, Hoberman A, Pichichero M, Paris M.
A multicenter, open label, double tympanocentesis study of high dose
cefdinir in children with acute otitis media at high risk of persistent or
recurrent infection. Pediatr Infect Dis J 2006;25:211-8.
27. Howie VM, Ploussard JH. Efficacy of fixed combination antibiotics versus
separate components in otitis media. Effectiveness of erythromycin
estrolate, triple sulfonamide, ampicillin, erythromycin estolate-triple sul-
fonamide, and placebo in 280 patients with acute otitis media under two
and one-half years of age. Clin Pediatr (Phila) 1972;11:205-14.
60
Volume 189
28. Piglansky L, Leibovitz E, Raiz S, Greenberg D, Press J, Leiberman A, et al.
Bacteriologic and clinical efficacy of high dose amoxicillin for therapy of
acute otitis media in children. Pediatr Infect Dis J 2003;22:405-13.
29. Martin JM, Hoberman A, Paradise JL, Barbadora KA, Shaikh N, Bhatnagar
S, et al. Emergence of streptococcus pneumoniae serogroups 15 and 35
in nasopharyngeal cultures from young children with acute otitis media.
Pediatr Infect Dis J 2014;33:E286-90.
30. Paradise JL, Campbell TF, Dollaghan CA, Feldman HM, Bernard BS,
Colborn DK, et al. Developmental outcomes after early or delayed in-
sertion of tympanostomy tubes. N Engl J Med 2005;353:576-86.
31. Shaikh N, Hoberman A, Rockette HE, Kurs-Lasky M, Paradise JL. Toward
an improved scale for assessing symptom severity in children with acute
otitis media. J Pediatric Infect Dis Soc 2015;4:367-9.
32. Spiro DM, Tay K-Y, Arnold DH, Dziura JD, Baker MD, Shapiro ED. Wait-
and-see prescription for the treatment of acute otitis media: a random-
ized controlled trial.[see comment]. JAMA 2006;296:1235-41.
33. Wald ER, Guerra N, Byers C. Upper respiratory tract infections in young
children: duration of and frequency of complications. Pediatrics
1991;87:129-33.
Shaikh et al
October 2017
Appendix
1. Calculations to arrive at the clinical efficacy of amoxicillin
and cefdinir
First, we conducted a systematic review of the literature on
available studies describing the bacteriologic efficacy of the
various antimicrobials under consideration. Pooled bacterio-
logic efficacy values were 94%, 83%, 72%, and 28% for
amoxicillin/clavulanate, amoxicillin, cefdinir, and placebo, re-
spectively. Of note, for amoxicillin, cefdinir, and placebo, we
only found only 1 available study.26-28
Next, we solved a series of equations to arrive at the clini-
cal efficacy for amoxicillin and cefdinir. We followed the meth-
odology described by Marchant et al,12 where the observed
clinical success rate for an antimicrobial would be a function
of bacteriologic success rate times a constant (x) plus bacte-
riologic failure rate times another constant (y). In this equa-
tion, x and y correct for the fact that clinical success can occur
despite bacteriological failure and vice versa; x is the rate of
clinical success in children with bacteriologic success and y is
the rate of clinical success in those with bacteriologic failure.
Because we knew that the probability of early clinical success
for amoxicillin/clavulanate and placebo from the study by
Hoberman et al (89% and 63%, respectively), we were able
to solve for x and y (equation for amoxicillin/clavulanate:
94x + 6y = 89; equation for placebo: 28x + 72y = 63). Using these
x and y values and the bacteriologic efficacy of amoxicillin
and cefdinir from previous studies, we estimated the prob-
ability of early clinical success for these 2 antimicrobials
(eg, for amoxicillin: 83*0.91 + 17*0.52 = 84%; cefdinir:
72*0.91 + 28*0.52 = 80%). We determined other probabili-
ties (early symptom improvement, AOM recurrence) using
the same methodology.
We know that pneumococcal vaccination has altered the NP
flora and perhaps the ratio of Staphylococcus pneumoniae to
Haemophilus influenzae. Thus, the current bacteriologic effi-
cacy rates could be different from what we used. The only
double tympanocentesis study using amoxicillin, from which
the 83% efficacy was derived, was conducted in 2003 in Israel
(no pneumococcal conjugate vaccine vaccination).28 However,
the organism mix in that study (58% H influenzae and 37%
S pneumoniae, 2% Moraxella catarrhalis, with 34% of the H
influenzae producing B lactamases and 75% of the S
pneumoniae exhibiting intermediate or full resistance to peni-
cillin)28 does not seem to be substantially different from those
in recent reports.29 The only study on cefdinir was from 200626;
14% had received pneumococcal conjugate vaccine. The or-
ganism mix was 47% H influenzae, 41% S pneumoniae, and
9% M catarrhalis, with 48% of the S pneumoniae exhibiting
intermediate or full resistance to penicillin (which seems high)
and 17% of the H influenzae producing beta lactamases (which
seems low). It is unclear if pneumococcal conjugate vaccine
would have had an impact on the efficacy rates, but even if we
assume that the efficacy we estimated for cefdinir is low in this
analysis, and use a value of 76% instead of 72%, the clinical
efficacy would increase from 80% to 82%, which was well inside
A Cost-Utility Analysis of 5 Strategies for the Management of Acute Otitis Media in Children
ORIGINAL ARTICLES
the ranges of 68% to 92% that we allowed in the tree. In con-
clusion, we acknowledge that the efficacy rates used in the de-
cision tree are only estimates, and accordingly we used wide
CIs in the tree to account for this.
2. Other assumptions
• Although residual middle ear effusion often persists for longer
than the 30-day time horizon selected, it is largely asymp-
tomatic and generally does not lead to long-term sequelae.30
Accordingly, we did not include middle ear effusion in our
model.
• We assumed that 50% of children with diaper dermatitis
would develop monilial superinfection necessitating pre-
scription of a topical antifungal. We could not find any evi-
dence to support the notion that the rate of monilial infection
varies according to the type of antibiotic used, indepen-
dent of the rate of diarrhea. No studies have reported the
rate of monilial infection in children taking cefdinir for AOM.
Two small studies reported the rate of monilial dermatitis
in children taking amoxicillin, but this rate was higher than
the reported for amoxicillin/clavulanate. Because of the lack
of empiric data, we feel that it is more prudent to assume
the rate of monilial infection is a function of the rate of
diarrhea.
• We estimated the number of extra diapers required as 16,
14, 7, and 4, for children receiving amoxicillin/clavulanate,
amoxicillin, cefdinir, and DP, respectively, based on data from
Hoberman et al2 and on other reports of antimicrobial-
related diarrhea (unpublished data).
• The definitions we used to define symptom persistence were
based on the reported minimally important difference in
scores on the AOM-SOS (version 4.0)31 as follows.
• Worsening is an AOM-SOS score of 9 or greater on at least
2 instances in children with a baseline score of 9 or greater,
or 2 scores higher than baseline plus at least 3 points in
children with baseline scores of greater than 9.
• Lack of improvement is a relative change of less than 55%
in the AOM-SOS score compared with the baseline score
on 2 consecutive occasions during days 4-7.
• We assumed the incidence of mastoiditis to be equivalent
in all strategies in which immediate antimicrobials were pre-
scribed, and also equivalent in children in all strategies where
antimicrobials were delayed.
• We assumed, that among children with AOM, those diag-
nosed also with an upper respiratory infection would ex-
perience adverse events at the same rate as children in the
WW strategy.
• We assumed that all episodes of AOM would be treated in
an outpatient setting and that all episodes of mastoiditis
would require hospitalization.
• We assumed that 9.9% of children (range, 4.1%-15.6%) in
the DP branch requiring an office visit during the first 7
days.32
• We assumed that outpatient visits occurring because a child
had persistent symptoms would be of low complexity (level
3; Current Procedural Terminology code: 99213). To calcu-
late costs of hospitalization for mastoiditis, we used the sum
60.e1
THE JOURNAL OF PEDIATRICS • www.jpeds.com Volume 189

of an initial office visit, hospital costs, and physician reim-
bursement costs.
• To calculate doses of medications, we assumed a weight of
10 kg (mean weight of a 1-year-old child).
• Because Hoberman et al6 had observed no difference in an-
algesic use between the amoxicillin/clavulanate and placebo
groups in their clinical trial, we did not model analgesic costs.
• We estimated the numbers of diapers needed for children
in the amoxicillin/clavulanate and WW branches based on
the numbers of stools per day in the amoxicillin/clavulanate
and placebo groups in the Hoberman et al study.2 We
estimated diaper use in children in the other branches based
on the probability of diarrhea in those children. We assumed
that children with diaper dermatitis would be treated with
a barrier cream and that those with monilial infection would
also receive topical antifungal cream. Based on data from
Hoberman et al,2 we estimated that one-third of children who
developed a generalized rash would require a change in an-
timicrobial therapy.
• We estimated costs of childcare based on data from the study
by Hoberman et al,2 and we assumed parking and or trans-
portation would cost an average of $3.

3. Other variables used in the tree

Variables Days* Source

Duration of hospital stay in child with acute mastoiditis† 3.2 HCUP9
Duration of diarrhea if no immediate antimicrobials 1.9 Hoberman2
Duration of diarrhea if immediate antimicrobials 4 Hoberman2
Duration of diaper rash 9 Hoberman2
Duration of body rash 2.5 Hoberman2
Days with AOM symptoms in children with persistent symptoms 7 Hoberman2
Days with AOM symptoms in children with improved symptoms if received immediate antimicrobials 2.5 Hoberman2
Days with AOM symptoms in children with improved symptoms if received delayed or no antimicrobials 3.5 Hoberman2
Days with AOM symptoms if AOM recurs in 30-day period from onset of symptoms to resolution 7 Estimate‡
Duration of uncomplicated URI symptoms after office visit 4 Wald33§
Days with AOM symptoms in a child with AOM who is mistakenly diagnosed as having only a URI 6 Estimate
For children whose treatment regimen was DP, days parents waited to pick up a prescription 2 Spiro32

HCUP, Healthcare Cost and Utilization Project; URI, upper respiratory tract infection.
*All values varied ±1 day in the sensitivity analysis.
†We assumed that children with mastoiditis spent 3.2 days in the mastoiditis state, 10 days in the AOM state, and took antimicrobials for 30 days (per current treatment guidelines).
‡Children with an AOM recurrence who were treated with rescue antimicrobials were assumed to have 7 additional days of AOM symptoms from onset to resolution (3 days before office visit and
4 days after).
§Because symptoms of an uncomplicated URI typically last approximately 7 days33 and because the initial office visit occurs around day 3, children with AOM who were misdiagnosed as having
a URI were assumed to have 4 days of URI symptoms.

4. Sensitivity analysis

Base-case Threshold Preferred strategy for values Preferred strategy for values
Variable* value value below threshold value above threshold value
Probability of improved symptoms days 1-7 with amoxicllin 0.68 0.61 Amoxicillin/clavulanate Amoxicillin
Disutility of AOM 0.21 0.08 DP Amoxicillin

*Only variables which changed the preferred strategy are shown.

60.e2 Shaikh et al
October 2017 ORIGINAL ARTICLES

Figure. Structure of the decision tree for each treatment modality (except for DP). Oval nodes represent chance nodes. For
children diagnosed with AOM, the first chance node considered 3 possible early outcomes: improved symptoms, persistent symp-
toms, or mastoiditis. Later nodes considered the possibility of recurrence of AOM. We determined specific QALD values by
subtracting the QALDs lost owing to each adverse state from the maximum possible QALDs. For example, for the branch labeled
“persistent symptoms,” we subtracted the utility lost owing to AOM symptoms (disutility of AOM times the days with AOM), the
utility lost owing to diarrhea if it occurred (disutility of diarrhea times the days of diarrhea), the utility lost owing to diaper der-
matitis if it occurred (disutility of diaper dermatitis times the days of diaper dermatitis), and utility lost owing to body rash if it
occurred (disutility of body rash times days of body rash). To calculate the costs for this branch we added the cost of any an-
tibiotic administered, the expected cost of diapers, and the average cost of topical medication required to treat diaper derma-
titis if it occurred.

A Cost-Utility Analysis of 5 Strategies for the Management of Acute Otitis Media in Children 60.e3