Management of Influenza Symptoms in Healthy Children: Title and subTitle BreakCost-effectiveness of Rapid Testing and Antiviral Therapy FREE

Influenza virus infection typically occurs in winter epidemics, affecting 20% to 42% of preschool and school-age children each year.^{1 - 2} Although the disease is usually self-limited in older children, those younger than 3 years face complication rates similar to those in the elderly population.^{3 - 4} For that reason vaccination has been recommended for children aged 6 to 23 months and their close contacts.^{5 - 6} In addition, the economic impact of influenza on all age groups is enormous, resulting in $11 billion to $18 billion annually in direct and indirect costs in the United States.⁷ Each year, between 6% and 28% of children will have office visits for influenza,¹ and one third to half of all parents miss work when their child has influenza.^{8 - 10}

Primary care providers face several decisions when confronted with a child displaying symptoms of influenzalike illness. If the child presents within 48 hours of symptom onset, antiviral treatment can shorten the duration of illness and prevent complications. However, antiviral drugs can be expensive and may cause adverse effects. Moreover, the clinical diagnosis in children can be challenging, as other viral illnesses can mimic influenza. Rapid testing could be used to improve diagnostic accuracy, but testing is expensive and can produce false-positive and false-negative results.

Treatment choices include amantadine hydrochloride and oseltamivir phosphate, the only drugs currently licensed for the treatment of influenza in children younger than 7 years. Zanamivir, an inhaled antiviral drug, is approved for children 7 years and older, but it has been linked to reactive airway disease and is no longer being promoted by its manufacturer. Rimantadine hydrochloride is not licensed for treatment in children. Although similar in efficacy against influenza A, the drugs differ in cost, adverse effects, activity against influenza type B, and induction of viral resistance. The decision is further complicated in children because the cost of medication, probability of complications, and amount of parental work loss vary with the age of the child.

Do the benefits of anti-influenza therapy in healthy children justify the costs and adverse effects? If so, should treatment be based on clinical diagnosis or directed by rapid testing? Should standard therapy include oseltamivir or amantadine? How does cost-effectiveness vary with the age of the child? To answer these questions, we constructed a decision analytic model to determine the cost-effectiveness of empirical vs test-guided antiviral therapy compared with no antiviral therapy for children presenting with symptoms of influenza.

We constructed a simple decision tree (Figure 1) by means of a standard computer program (Decision Maker 7.07; Pratt Medical Group, Boston, Mass) to compare the following strategies: (1) no antiviral therapy, (2) empirical treatment with either amantadine or oseltamivir, and (3) rapid office testing followed by treatment with amantadine or oseltamivir. The outcome measures were cost and quality-adjusted life-years (QALYs) saved.

We considered healthy children at ages 2, 7, and 15 years as representative of toddlers, children, and adolescents. Baseline differences for these age groups include the probability of hospitalization, the cost of weight-based medication, the cost of syrup vs pills, and the need for an adult caretaker to miss work on the final day of illness. In addition, vaccination is now recommended for all children aged 6 to 23 months. Because this recommendation is new, we performed all analyses twice, once for vaccinated children and again for unvaccinated ones. We considered patients presenting to a primary care physician with fever and cough or coryza within 48 hours of symptom onset, during a local influenza outbreak. The model considers the prevalence of influenza, sensitivity and specificity of the tests, and the following adverse events: antiviral adverse effects, otitis media, emergency department visits, hospitalizations, and deaths. Influenza infection could be caused by type A or type B, which we assumed to be of equal severity. We assumed that only oseltamivir is effective in treating influenza B.

Baseline estimates and ranges for sensitivity analyses are provided in Table 1. We performed baseline analyses, sensitivity analyses, and probabilistic sensitivity analysis for each age group for both an influenza A–predominant season and a mixed types A and B season.

Influenza can be difficult to diagnose in children. However, in prospective studies of antiviral drugs conducted when influenza was known to be circulating locally, children with temperature greater than 37.8°C and at least 1 respiratory symptom (cough or coryza) had a 65% to 73% chance of having influenza.^{11 - 12} A clinical prediction rule including cough, headache, and pharyngitis had a positive predictive value of 77% in children.³⁰

Rapid testing can improve diagnostic accuracy. Several tests are available, but, to our knowledge, only one head-to-head comparison has been conducted in a pediatric practice.¹⁷ From this trial we selected 2 tests that are simple enough to be performed in a physician’s office (QuickVue [Quidel Corp, San Diego, Calif] and ZstatFlu [ZymeTx Inc, Oklahoma City, Okla]). Both detect influenza infection but do not distinguish between types A and B.

Complications requiring antibiotics, most often otitis media, occurred in 15% to 28% of healthy children in the placebo arms of antiviral and vaccine trials.^{11 - 12 ,14} More serious complications include pneumonia, encephalopathy, and death. To calculate the probability of hospitalization after an office visit, we divided the excess number of hospitalizations by the excess number of office visits as estimated by Neuzil et al.³ Mortality per hospitalization came from the Healthcare Cost and Utilization Project database.¹⁵

We reviewed all trials of anti-influenza drugs in children. Two trials^{18 - 19} of amantadine showed that it shortened febrile symptoms by 12 and 28 hours (43% and 44% shorter than in the placebo group). One trial^{11 ,16} of oseltamivir showed that it shortened fever by 25 hours (35% shorter than in the placebo group), returned children to normal activities 45 hours sooner, and decreased antibiotic prescriptions by 44%. In a meta-analysis, oseltamivir was shown to prevent hospitalization in adults,³¹ but the study in children was too small to assess this; the study of asthmatic patients has not been published.³² We assumed that both drugs would be equally effective in shortening illness and preventing otitis media, but that neither drug would prevent hospitalization. We tested these assumptions in sensitivity analysis.

Compared with placebo, amantadine appears better tolerated in children than oseltamivir. In trials involving 599 children, those randomized to amantadine experienced less dizziness, stimulation, or insomnia than children taking placebo (1.9% vs 2.4%), as well as less nausea or vomiting (6% vs 10% for placebo).^{18 - 19} Five additional studies of amantadine prophylaxis that included more than 1200 children also found no adverse effects.^{20 - 24} Oseltamivir, however, was associated with an increase in nausea and vomiting (14% in the oseltamivir group vs 9% in the placebo group, n = 695).¹¹ Adverse effects rarely resulted in discontinuation of the drug. We assumed that adverse effects would begin on the second day of treatment and last 4 days.

Because neither medication has been proved to decrease mortality, the benefits of antiviral therapy must come from improved quality of life. Quality-adjusted life expectancy, usually expressed in QALYs, is calculated by multiplying the years (or, in this case, days) in a given health state by the utility of that state. Utilities, which are used to make the quality adjustments, express patient preference for various health states on a scale of 0 to 1, with 0 representing death and 1 representing full health. For example, 4 days in a hypothetical health state with a utility of 0.4 translates into 1.6 quality-adjusted life-days—a reduction in quality-adjusted life expectancy of 2.4 days or 0.0066 QALY. Converting all outcomes in a single unit allows for direct comparison of disparate health outcomes such as influenza illness, otitis media, and medication adverse effects. Our utility estimates for influenza were based on unpublished data from zanamivir trials in children that used the EQ-5D instrument from the EuroQOL group.³³ Estimates were similar to those seen in healthy adults.³⁴ Our utility for otitis media combined estimates from parents, pediatricians, and community members.^{25 ,35}

We took a societal perspective in keeping with the recommendations of the Panel on Cost-effectiveness in Health and Medicine.³⁶ Thus, we considered direct medical costs, including physician visits, diagnostic tests, medications, and hospitalizations, as well as indirect costs in the form of lost productivity from parents staying home to care for ill children. On the basis of survey data from 3 Virginia elementary schools, we estimated that half of parents with children younger than 15 years would miss work to care for children with influenza.⁸ Lost productivity was valued by means of average total employee compensation for US civilian workers.²⁹ Physician fees were based on Medicare relative value units for an established-patient low-complexity office visit (Current Procedural Terminology³⁷ code 99213). Retail costs were used for rapid tests. Medication costs reflected the average wholesale price for a 5-day course plus a $5 acquisition cost.²⁶ Hospitalization costs reflect Healthcare Cost and Utilization Project data.¹⁵ All costs were updated to 2003 US dollars by means of the medical care component of the Consumer Price Index.³⁸

Because the value of some variables is uncertain and others vary among patients, we conducted 1-, 2-, and 3-way sensitivity analysis for all variables in Table 1 to evaluate the impact on cost and effectiveness. We also conducted N-way probabilistic sensitivity analysis. All variables, except the cost of the medications, the probability of influenza, and the proportion of influenza B, were entered as probability distributions based on the 95% confidence intervals. We then performed 1000 Monte Carlo simulations. For each simulation, new variable values were randomly selected from within each of the probability distributions, and the associated costs and quality-adjusted life expectancy were calculated. We used β-distributions for variables between 0 and 1 (eg, test sensitivity or antibiotic efficacy), a uniform distribution for influenza utilities, and normal distributions for the remaining variables (eg, efficacy of antiviral drugs in shortening disease duration).

We considered 3 different-aged children presenting to a primary care provider during a local influenza outbreak with either a high or a low proportion of influenza B (Table 2). Empirical therapy with antiviral medication resulted in the greatest quality-adjusted life expectancy in all age groups. Compared with not treating, antiviral therapy improved quality-adjusted life expectancy by 0.003 QALY in all age groups by shortening the duration of illness and preventing otitis media. Vaccinated children had lower costs and better health outcomes than nonvaccinated children, regardless of strategy chosen. Whether antiviral therapy also saved money and which antiviral drug was preferred depended on the prevalence of influenza B and the age of the child.

When influenza A predominated, empirical amantadine therapy provided the best health outcomes and was also the least expensive strategy in children younger than 15 years, because the savings from parents returning to work sooner far exceeded the cost of the medication. Compared with no treatment, amantadine therapy saved $121 per child. For children old enough to stay home alone on the last day of illness, not treating was the least expensive strategy. Empirical amantadine therapy also improved life expectancy in this group by 0.003 QALY at an additional cost of $5, giving an incremental cost-effectiveness ratio of $1800 per QALY saved. Compared with amantadine, empirical oseltamivir cost more and provided worse outcomes, because oseltamivir has more adverse effects than amantadine.

During seasons in which influenza B was common, empirical oseltamivir replaced amantadine as the least expensive option in 2-year-olds; for older children, the incremental cost-effectiveness of oseltamivir relative to amantadine ranged from $7400 per QALY saved in 7-year-olds to $47 000 per QALY saved in 15-year-olds. Treating 7-year-olds was less cost-effective than treating 2-year-olds, because children weighing more than 15 kg required larger doses of medicine.

Rapid influenza testing was not helpful during either influenza season, because the pretest probability of influenza was high and the tests were not sufficiently sensitive. Testing was both more expensive and less effective than empirical therapy.

Throughout the ranges tested, empirical amantadine therapy was always either cost-saving or inexpensive (less than $20 000 per QALY saved) relative to not treating. The remaining question, whether the extra benefit of oseltamivir justified its higher cost, depended on the prevalence of influenza B alone (Figure 2). When influenza B was rare, amantadine was always preferred. However, when influenza B was common, the following variables affected the choice of antiviral in children older than 3 years: the probability of influenza, the cost and specificity of the QuickVue test, the efficacy of antivirals in shortening the course of illness, the short-term morbidity of influenza, the cost of a workday, and the cost of oseltamivir. Of these, patient age, proportion of influenza caused by type B, probability of influenza, and short-term morbidity as represented by severity of illness can be assessed by the clinician at presentation. Table 3 shows the optimal strategy, defined as that which provides the greatest number of QALYs at a marginal cost-effectiveness ratio of $50 000 or less per QALY saved, for all combinations of these variables.

Rapid testing was not cost-effective during seasons that were predominantly influenza A. In mixed influenza seasons, however, rapid testing saved money by avoiding costly courses of oseltamivir when influenza prevalence was low to moderate (30%-50%), or if the test cost less than $10.

Results of the Monte Carlo analysis, using a societal willingness-to-pay threshold of $50 000 per QALY saved, supported the base-case results. During influenza A–predominant seasons, empirical amantadine was the preferred strategy in all age groups, with a maximum incremental cost-effectiveness of $9600 per QALY saved.

In mixed influenza seasons, empirical oseltamivir was the preferred strategy in 2-year-olds, 7-year-olds, and 15-year-olds in 99.9%, 80%, and 42% of the simulations, respectively. Empirical amantadine was usually the least expensive strategy but was rarely preferred, except in 15-year-olds, when it was favored in 51% of simulations. Increasing the willingness-to-pay threshold made oseltamivir attractive for 15-year-olds as well (Figure 3).

This cost-effectiveness analysis demonstrates that during local influenza outbreaks, children with symptoms of influenzalike illness benefit from antiviral therapy if it is initiated within 48 hours of symptom onset. At the same time, antiviral therapy saves money if parents return to work sooner. In that case, there is no trade-off between cost and effectiveness. When parents do not miss work, antiviral therapy still improves quality-adjusted life expectancy, but at a monetary cost. The cost-effectiveness of antiviral therapy, especially with amantadine, compares favorably with that of other recommended interventions such as vaccinations against varicella,³⁹ pneumococcal meningitis,⁴⁰ and hepatitis B.⁴¹

The preferred antiviral drug depends almost entirely on the proportion of influenza cases caused by type B influenza. Oseltamivir is much more expensive than amantadine, but little is known about their relative efficacy against influenza A infection, because there are no comparative trials. When influenza A predominates, both drugs appear equally effective, but amantadine may be better tolerated and is less expensive. Because amantadine is not active against influenza B, however, oseltamivir will be more effective when influenza B is prevalent. To practice in a cost-effective manner, clinicians must be aware of the proportion of influenza B, which is available weekly from state health departments and online from the Centers for Disease Control and Prevention¹³ and others.

We found no role for rapid diagnostic testing during local outbreaks, because clinical diagnosis is highly predictive,³⁰ whereas rapid testing leads to frequent false-negative results,¹⁷ thus adding cost without improving outcome. Similar conclusions have been drawn from analyses in adults.^{34 ,42} Rapid testing is useful when the probability of influenza is less than 50%, especially during influenza B outbreaks, because oseltamivir is expensive relative to the cost of testing, and sensitivity is less important when the pretest probability is low (Table 3).

There are several obstacles to implementing our model in practice. First, some practitioners will be uncomfortable prescribing empirical therapy. Such caution avoids unnecessary adverse effects but leads to undertreatment, even if rapid testing is used. Because the rate of adverse effects from oseltamivir and the false-negative rate for QuickVue are identical, a child is more likely to be helped than harmed by empirical oseltamivir whenever the probability of influenza is greater than 50%. Amantadine, which has been associated with confusion in the elderly, appears to be much better tolerated in children, with adverse effects comparable to those of placebo. As a result, the benefits of empirical therapy with amantadine extend to children with much lower probabilities of disease.

Second, our analysis was conducted from a societal viewpoint. Parents paying for medications out of pocket may be unwilling to purchase oseltamivir, especially if they anticipate no lost wages. Generic amantadine, on the other hand, should always be attractive.

Finally, we did not include development of viral resistance in our model, because the implications of resistance are not well understood. Resistance to oseltamivir is uncommon,¹¹ but resistant viruses emerge rapidly during amantadine therapy and can be transmitted in a closed environment such as a nursing home.⁴³ Resistant viruses are not more virulent and have not caused epidemics.⁵ Nevertheless, if treatment were to result in widespread amantadine resistance, amantadine would no longer be useful for prophylaxis or treatment. Unlike antibiotics, antivirals cannot induce resistance in individuals who are not currently infected with the virus. As a result, empirical treatment, which involves treating some patients without influenza, is no more likely to induce resistance than test-directed therapy. If widespread use of short-course therapy with amantadine is adopted, the transmission of resistant viruses should be monitored, and a large clinical trial would be welcome.

Our model neglected some potential benefits of antiviral therapy. If antiviral therapy were to reduce hospitalizations, it could be cost-saving for high-risk children at any age. However, the only trial of oseltamivir for asthmatic children did not show a significant decrease in duration of illness, and it has not been published.³²

Antivirals might also decrease transmission of influenza to other household members, especially unvaccinated infants or the elderly. Even a small decrease in transmission would have large economic and health implications. On the basis of a pooled analysis, Couch et al⁴⁴ reported that antiviral therapy resulted in a 30% reduction in transmission compared with placebo, but these results have never been published. Two randomized trials compared postexposure prophylaxis of households with oseltamivir. In one trial,⁴⁵ the index case received no antiviral therapy, and 21% of individual contacts receiving placebo developed influenza. In the other,⁴⁶ the index case also received antiviral therapy, and only 13% of contacts receiving placebo developed influenza (a 38% reduction). A prospective trial is needed to see whether treatment of the index case decreases spread.

Our analysis may overestimate the benefit of therapy for patients with influenza B, which may be milder than influenza A. On the basis of a small study sample, we assumed that oseltamivir would be equally effective against influenza A and B.^{31 ,47} If oseltamivir is less effective against influenza B than A, then oseltamivir will be less cost-effective as well.

Another obstacle to widespread use of antiviral drugs in children is lack of familiarity. Despite the drugs having been available for years, few physicians routinely prescribe them, and patients rarely present early enough to receive treatment. Practitioners who prescribe oseltamivir suspension or amantadine syrup may be disappointed to find that local pharmacies do not carry it.

The severity of the influenza epidemic of 2003-2004 in some areas of the United States has awakened interest in influenza as a cause of morbidity and mortality in children.⁴⁸ Although the new vaccination effort should somewhat decrease the economic and health burden of influenza, it will by no means eliminate it. As is frequently the case, there are very limited data on the efficacy of antiviral medications in children. There are no comparative trials, no studies of high-risk children, and no studies of amantadine in 20 years. The existing evidence, however, is not contradictory. Antiviral therapy shortens disease, decreases otitis media, and may decrease spread to other persons. Our analysis shows that it could also save money.

Our model offers a framework for decision making. Younger patients, especially the unvaccinated, benefit most, but even older children can be treated cost-effectively with amantadine when influenza A is circulating. When influenza B is prevalent, the decision to test or treat should include careful consideration of the probability of disease, the anticipated work loss, and the severity of illness.

Correspondence: Michael B. Rothberg, MD, MPH, Division of General Medicine and Geriatrics, Baystate Medical Center, 759 Chestnut St, Springfield, MA 01199 (Michael.Rothberg@bhs.org).

Accepted for Publication: July 21, 2005.