Objectives. The optimal practice management of highly febrile 3- to 36-month-old children without a focal source has been controversial. The recent release of a conjugate pneumococcal vaccine may reduce the rate of occult bacteremia and alter the utility of empiric testing and treatment. The objective of this study was to determine the cost-effectiveness of 6 different management strategies of febrile 3- to 36-month-old children at current and declining rates of occult pneumococcal bacteremia.
Methods. A cost-effectiveness (CE) analysis was performed to compare the strategies of “no work-up,” “clinical judgment,” “blood culture,” “blood culture + treatment,” “complete blood count (CBC) + selective blood culture and treatment,” and “CBC and blood culture + selective treatment.” A hypothetical cohort of 100 000 children who were 3 to 36 months of age and had a fever of ≥39°C and no source of infection was modeled for each strategy. Our main outcome measures were cases of meningitis prevented, life-years saved compared with “no work-up,” total cost (1999 dollars), and incremental CE ratios.
Results. When compared with “no work-up,” the strategy of “CBC + selective blood culture and treatment” using a white blood cell (WBC) cutoff of 15 × 109/L prevents 48 cases of meningitis, saves 86 life-years per 100 000 patients, and is less costly at the current rate of bacteremia (1.5%). Using the strategy of “CBC + selective blood culture and treatment” with a lower WBC cutoff of 10 × 109/L costs an additional $72 300 per life-year saved. If the rate of bacteremia declines to 0.5%, then the incremental CE ratio of “clinical judgment” compared with “no work-up” is $38 000 per life-year saved; however, strategies that include empiric testing or treatment result in CE ratios greater than $300 000 per life-year saved.
Conclusions. “CBC + selective blood culture and treatment” using a WBC cutoff of 15 × 109/L is cost-effective at the current rate of pneumococcal bacteremia. If the rate of occult bacteremia falls below 0.5% with widespread use of the conjugate pneumococcal vaccine, then strategies that use empiric testing and treatment should be eliminated.
Physicians frequently evaluate and treat young, highly febrile children in the setting of the office and the emergency department. Although the majority of these children have either viral or minor bacterial infections and recover uneventfully, a small proportion are found to harbor occult bacteremias and subsequently develop serious focal sequelae despite a benign clinical appearance.1–3 Thus, in 1993, a consensus panel published guidelines that recommended that practitioners obtain complete blood counts (CBC) and blood cultures and treat children empirically with antibiotics if the white blood cell count (WBC) is ≥15 × 109/L.4 Previous decision analyses regarding optimal management practices for occult bacteremia also supported the strategy of empiric testing and treatment of highly febrile children; however, these guidelines have not been widely accepted.5,,6
Several developments in the background of occult bacteremia have prompted the need to reevaluate these guidelines: 1) the near elimination of Haemophilus influenzae type B as a pediatric pathogen secondary to immunization,7–9 2) the rising incidence of penicillin-resistant Streptococcus pneumoniae,10–13 3) additional studies on the incidence of bacteremia and the sensitivity and specificity of diagnostic studies,8,,14,15 4) increased concerns over the costs and complications of diagnostic testing and empiric antibiotic treatment,16–18 5) continued practice variability in the utilization of these guidelines,19 6) ongoing controversy among experts regarding the optimal approach to the febrile child,20–25 and 7) the recent licensure of a conjugate pneumococcal vaccine.26
We undertook a cost-effectiveness (CE) analysis of the management of febrile 3- to 36-month-old children without a focal infection and compared the strategy of “no work-up” with 5 other strategies: “clinical judgment,” “blood culture,” “blood culture + treatment,” “CBC + selective blood culture and treatment,” and “CBC and blood culture + selective treatment.” We also examined the costs and outcomes using alternative WBC cutoffs to determine whether choosing other criteria to define leukocytosis would be more cost-effective than the currently recommended level of 15 × 109/L. Finally, we examined the effect of varying rates of bacteremia on the CE of different management strategies, because we anticipate a significant decline in the rate of occult bacteremia with the introduction of the conjugate pneumococcal vaccine.
We applied a decision analytic model to evaluate 6 different management options (Fig 1) using DATA 3.5 (TreeAge Software, Inc, Williamstown, MA). In Fig 1, there are 2 types of nodes from which branches emanate. The first type of node is a decision node (squares), whose branches represent alternative courses of action. The second type of node is a chance node (circles), where the branches represent different possible outcomes. Each branch of the chance node has a different probability of the outcome occurring. The costs are obtained by summing the individual costs of the branches for each unique pathway.
Structure of the Model
Six different strategies were modeled to compare the management options for children who are 3 to 36 months of age and have a fever of ≥39°C with no identified source of infection. BecauseStreptococcus pneumoniae is the causative pathogen in >90% of cases of occult bacteremia, our model focuses on the management of occult pneumococcal bacteremia and does not include outcomes due to other organisms.8 All of the following management strategies exclude patients who are ill appearing or toxic appearing.
1. No work-up: Patients in this arm receive no testing or treatment. There are no false-positive blood cultures and no antibiotic complications.
2. Clinical judgment: Patients who are judged to be at low risk for occult bacteremia receive no testing or treatment. Low risk is defined as the best possible Yale Observation Score (YOS).27 The YOS is a composite ordinal scoring system that uses 6 variables for the assessment of febrile children who are nontoxic in appearance. These variables include quality of cry, reaction to parent stimulation, state variation, color, hydration, and response to social overtures. Each variable is scored from 1 (normal) to 5 (severe impairment). Previous studies concluded that the YOS is not an accurate means of detecting occult bacteremia; however, objective data are available regarding the sensitivity and specificity of YOS scoring. For the purposes of our analysis, we considered patients to be at risk for occult bacteremia when the YOS was greater than the best possible score of 6. Patients with any elevation in YOS would receive a blood culture and treatment at the initial visit. Those with positive blood cultures receive appropriate follow-up management.
3. Blood culture: Highly febrile patients who are otherwise well appearing receive a blood culture at the initial visit and are sent home without antibiotic treatment. Patients with positive blood cultures are asked to return and receive appropriate follow-up management.
4. Blood culture + treatment: Patients who have high fevers and are otherwise well appearing receive a blood culture at the initial visit and are treated empirically with antibiotics. Children with positive blood cultures are asked to return for follow-up management. Patients with false-positive cultures are evaluated and either hospitalized or sent home. Patients with negative blood cultures do not require further treatment; however, they may experience complications of antibiotic administration.
5. CBC + selective blood culture and treatment: A CBC is performed on all children who are highly febrile and otherwise well appearing with no focal source of infection. If the WBC count at presentation is greater than a specified cutoff, then a blood culture is performed and antibiotics are given. WBC cutoffs of 10, 15, 16, 17, 18, 19, and 20 × 109/L are evaluated. Each WBC cutoff has a unique sensitivity and specificity as determined by previous studies.8 For each WBC cutoff, specific rates of bacteremia exist for each subgroup. If the WBC count is below the specified cutoff, then no further testing or treatment is done. Patients who have undiagnosed bacteremia (no blood culture obtained) may develop complications that necessitate additional evaluation, antibiotic therapy, and hospitalization. If the WBC count is above the cutoff, then a blood culture is performed and antibiotic treatment is given. Patients receive appropriate follow-up management when the blood culture is positive.
6. CBC and blood culture + selective treatment: Children who are highly febrile and appear nontoxic receive a CBC and blood culture at the initial visit. WBC cutoffs of 10, 15, 16, 17, 18, 19, and 20 × 109/L are evaluated. If the WBC count is less than the specified cutoff, then the patient is sent home without treatment. If the blood culture is positive, then follow-up is initiated. If the WBC count is greater than a specified cutoff, then an antibiotic is empirically administered at the initial visit. If the blood culture is positive, then follow-up management is initiated appropriately.
Follow-up management of children with positive blood cultures depends on their clinical presentation at the follow-up visit. If they are afebrile and have no identifiable source of infection, then blood cultures are repeated, antibiotics are administered, and they are discharged from the hospital.14 If children with a positive blood culture are persistently febrile, then they enter 1 of 6 states in our model: resolved, persistent bacteremia, pneumonia, cellulitis, bone/joint infection, or meningitis. Blood cultures and other indicated evaluations are performed, antibiotics are administered, and patients are admitted for hospitalization. If children have persistent bacteremia, then they may subsequently enter 1 of 5 states in the model: resolved, pneumonia, cellulitis, bone/joint infection, or meningitis.
False-positive blood cultures result in follow-up evaluation. The physician may discharge the patient to home (74%) or admit to the hospital (26%) for intravenous antibiotics until the pathogen is identified as a contaminant.17 The antibiotic complications modeled include anaphylaxis, rash, and diarrhea. If a rash or diarrhea develop, then a follow-up visit is initiated. If anaphylaxis occurs, then all patients require hospitalization and appropriate therapy. Patients who are not bacteremic on initial visit or have spontaneous resolution of bacteremia may develop antibiotic complications.
Baseline estimates for probabilities and costs used in the model are listed in Appendices A and B, respectively. The model was based on assumptions using data available in the literature and costs obtained from our hospital. Because these estimates may be uncertain, sensitivity analyses were performed on these estimates as indicated.
We selected a baseline prevalence of pneumococcal bacteremia of 1.5%.8,,28 The sensitivity of blood culture was assumed to be 100% for this study. For each WBC cutoff, the rates of pneumococcal bacteremia differ for each subgroup; however, the probabilities of focal complications in our model are not affected by the WBC cutoff. We assume that blood cultures allow for earlier follow-up and antibiotic treatment for bacteremic children and that half as many complications have developed when compared with children who do not receive blood cultures or empiric antibiotic treatment at the initial visit.28 Children who are not bacteremic are assumed to improve spontaneously without further complication. We assume that antibiotic administration does not have any effect on the severity of new focal illnesses (eg, neurologic outcome for meningitis) when they occur. We assume that there are no long-term medical costs or sequelae as a result of pneumonia, bone/joint infections, cellulitis, persistent bacteremia, or resolved bacteremia because most of these illnesses are self-limited and do not incur lifelong complications. We do include the long-term medical costs associated with neurologic disability due to meningitis.
Antibiotic therapy consists of 1 dose of intramuscular ceftriaxone. The estimates of antibiotic efficacy are based on several studies that document the rates of infectious complications after bacteremia in treated and untreated children.14,,1529–31 On the basis of these estimates, antibiotic efficacy was calculated. For example, the probability of meningitis in bacteremic patients is 1% in treated patients and 4% in untreated patients. Thus, antibiotic efficacy in preventing meningitis is 75%. Similarly, we estimated the antibiotic efficacies for persistent bacteremia (95%), pneumonia (70%), bone/joint infections (0%), and cellulitis (25%).
The study was performed from the societal perspective. Charges for laboratory studies, hospital rooms, emergency department visits, office visits, and physician fees were obtained from the Handbook of Fees and Fee Committee for the Department of Medicine from our hospital. Average charges for hospitalizations resulting from pneumococcal meningitis (50), pneumococcal pneumonia (27), cellulitis (8), septic arthritis (5), and osteomyelitis (5) were derived from the fiscal database at our hospital from 1991 to 1999 using International Classification of Diseases, 9th Revision, codes. Charges were estimated for hospitalizations as a result of false-positive blood cultures, anaphylaxis, or persistent bacteremia based on the expected utilization of services and goods. Because these charges do not reflect the true opportunity costs to society of the resources used, we applied a cost-to-charge ratio of 0.66 to all charges to obtain a better estimate of costs. The cost of intramuscular ceftriaxone was $32 based on the wholesale price for a 15-kg child. All costs are expressed in 1999 dollars.
Societal costs include the future impact of disability on a patient's time and resources and the costs incurred by parents, such as time lost from work. We used estimates generated by Lieu et al29 for the work-loss costs for bacteremia ($312) and meningitis ($1492). To obtain work-loss costs for soft tissue ($481) infections, we assumed that cellulitis work-loss costs would equal bacteremia work-loss costs multiplied by the ratio of cellulitis medical costs to bacteremia medical costs. We similarly estimated work-loss costs for bone/joint infections ($1036) and pneumonia ($872). We assumed that the work-loss costs for hospitalization as a result of false-positive cultures or anaphylaxis were the same as the work-loss costs for bacteremia. An annual discount rate of 3% is used for costs that are incurred in the future, such as income lost from neurologic disability, educational needs of neurologically disabled children, and chronic health care needs for neurologically disabled children. Neurologic sequelae include a wide range of conditions from deafness to seizures to vegetative state. Cases of meningitis without neurologic sequelae did not incur any long-term costs in our model.
Our model measures the following outcomes for each strategy: cases of infectious complications (bone/joint infection, cellulitis, meningitis, persistent bacteremia, and pneumonia), life-years saved compared with “no work-up,” cases of antibiotic-associated rash/diarrhea, cases of anaphylaxis, and number of false-positive blood cultures. We also calculated the number of children needed to treat to prevent a case of meningitis. Cases of meningitis and life-years saved were our main outcome measures. Life-years saved were based on the number of cases of fatal meningitis and an assumed life expectancy of 75 years. A discount rate of 3% was applied to long-term costs and the outcome measure of life-years saved.
Incremental CE Ratios
Incremental CE ratios were calculated to assess the additional cost per life-year saved when comparing two strategies. These were calculated according to the formula (cost of strategy A − cost of strategy B)/(effectiveness of strategy A − effectiveness of strategy B). For incremental CE analysis, interventions were ordered by increasing effectiveness and each was compared with the next most effective option. The least effective strategy in each analysis was “no work-up.” Strategies may be eliminated from consideration if they are dominated, either by simple or extended dominance. A strategy is dominated when it is both more costly and less effective than an alternative strategy. Strategies that are ruled out by extended dominance have higher incremental CE ratios than more effective options.30
We can graphically display the CE ratios of different management strategies by plotting effectiveness versus cost. The inverse of the slope of the line connecting 2 different strategies represents the incremental CE ratio. All points to the right and below the line are dominated, either by simple or extended dominance.30
One-way sensitivity analysis was performed on CE ratios to assess the impact of changes in baseline assumptions over the ranges defined in Appendices A and B. Two-way sensitivity analysis was performed to assess the interaction of 2 variables on the results of the model.
We calculated the health outcomes of a hypothetical cohort of 100 000 patients using the different strategies. Table 1 shows that “no work-up” results in 71 cases of meningitis and 0 life-years saved, but there are no false-positive blood cultures or antibiotic complications. At the other end of the spectrum, “blood culture + treatment” leads to 15 cases of meningitis, 101 life-years saved, and 1000 false-positive blood cultures; however, this strategy incurs 50 cases of anaphylaxis and 13 850 cases of antibiotic-associated rash or diarrhea. By using approaches that select high-risk groups, such as “clinical judgment,” “CBC + selective blood culture and treatment,” or “CBC and blood culture + selective treatment,” a smaller proportion of children receive antibiotics and fewer cases of anaphylaxis occur. In addition, “CBC + selective blood culture and treatment” and “clinical judgment” allow for fewer evaluations for false-positive blood cultures.
Table 1 shows that “CBC + selective blood culture and treatment” using a WBC cutoff of 15 × 109/L is the least costly approach at the currently estimated rate of bacteremia. Each of the following approaches becomes successively more expensive: “CBC + selective blood culture and treatment” using a WBC cutoff of 10 × 109/L, “clinical judgment,” “CBC and blood culture + selective treatment” using a WBC cutoff of 10 × 109/L, “blood culture + treatment,” “no work-up,” and “blood culture alone.”
Incremental CE Analysis for Varying Rates of Bacteremia
Incremental CE ratios were calculated for each of the 6 major strategies. We used 3 different estimates of the prevalence of bacteremia to determine the costs and outcomes associated with the use of different strategies. For the strategies that use different WBC cutoffs, we eliminated strategies by dominance (more costly and less effective than an alternative strategy) and extended dominance (higher incremental CE ratio than another strategy). Incremental CE ratios were calculated for the remaining strategies to determine the additional cost incurred for each life-year saved. Figure 2 shows a graphical representation of the results of incremental CE analysis at 3 different rates of bacteremia. The lines represent efficient strategies that provide increasing effectiveness at increased cost. Points that lie to the right and below the lines are dominated and are excluded from incremental CE analysis. At a rate of bacteremia of 1.5%, “no work-up” is dominated when compared with “CBC + selective blood culture and treatment” using a WBC cutoff of 15 × 109/L. Using a lower WBC cutoff of 10 × 109/L results in 12 additional life-years saved at a cost of an additional $875 000 for an incremental CE ratio of $72 300 per life-year saved. “Blood culture + treatment” costs an additional $1.6 million per life-year saved and has a flat slope, shown in Fig 2.
As the rate of bacteremia declines to 1%, “no work-up” is dominated when compared with “clinical judgment.” Using “CBC + selective blood culture and treatment” with a WBC cutoff of 15 × 109/L may be considered incrementally cost-effective at $30 800 per additional life-year saved. However, using the same strategy with a lower WBC cutoff of 10 × 109/L results in a CE ratio of $236 500 per life-year saved, whereas “blood culture + treatment” costs an additional $2.4 million per life-year saved. Finally, at a rate of bacteremia of 0.5%, “clinical judgment” has an incremental CE ratio of $38 000 per life-year saved when compared with “no work-up.” Using other strategies, such as “CBC + selective blood culture and treatment,” results in CE ratios of >$300 000 per life-year saved.
One-Way Sensitivity Analysis
One-way sensitivity analysis was performed on each of the rates and costs shown in Appendices A and B to see how the results of our CE analysis change with varying estimates. The conclusions of our model are sensitive to change for the following variables.
Probability of Meningitis
We varied the baseline estimate of the probability of meningitis after bacteremia from 0% to 20%. If the probability of meningitis is higher than 5%, then “no work-up” is dominated by “CBC + selective blood culture and treatment” using a WBC cutoff of 10 × 109/L. All other strategies either are dominated or their CE ratios exceed $100 000 per life-year saved. As the probability of meningitis goes from 3% to 2%, “no work-up” remains dominated by “clinical judgment.” However, the incremental CE ratio of “CBC + selective blood culture and treatment” using a WBC cutoff of 15 × 109/L increases from $160 to $124 700 per life-year saved. At a probability of meningitis of 1%, all strategies other than “no work-up” result in CE ratios of >$100 000 per life-year saved.
Probability of Neurologic Complication
We varied the probability of neurologic complication from 0% to 90%. As the probability exceeds 40%, “no work-up” is dominated by “CBC + selective blood culture and treatment” using a WBC cutoff of 10 × 109/L. All other strategies cost >$1 million per life-year saved. If the probability of neurologic complication drops below 10%, then “clinical judgment” costs an additional $90 100 per life-year saved when compared with “no work-up,” whereas all other strategies result in CE ratios of >$180 000.
Probability of Antibiotic Efficacy in Meningitis
We used a baseline estimate of 75% and varied the probability from 0% to 100%. If antibiotics are >60% effective in reducing the risk of meningitis, then “no work-up” is dominated by “CBC + selective blood culture and treatment” using a WBC cutoff of 15 × 109/L. The additional cost of using a lower WBC cutoff of 10 × 109/L ranges from $18 600 to $118 300 per life-year saved. If antibiotic efficacy is between 20% and 50%, then “no work-up” is dominated by “clinical judgment.” Using the next most effective strategy of “CBC + selective blood culture and treatment” using a WBC cutoff of 15 × 109/L results in CE ratios of $11 800 to $250 000 per additional life-year saved. If antibiotic efficacy is <10%, then all strategies other than “no work-up” result in CE ratios of >$80 000.
Sensitivity of “Clinical Judgment”
We varied our estimate of the sensitivity of “clinical judgment” from 0% to 100%. If the sensitivity of “clinical judgment” is 50%, then “no work-up” is dominated by “clinical judgment.” The CE ratios of “CBC + selective culture and treatment” using WBC cutoff of 15 × 109/L and 10 × 109/L are $51 100 and $72 300 per life-year saved, respectively. As the sensitivity rises above 60%, “clinical judgment” remains cost-effective whereas all other strategies either are dominated or have CE ratios of >$135 000.
Cost of CBC and Blood Culture
If the cost of a CBC is >$55, then “no work-up” is dominated by “clinical judgment.” “Blood culture + treat” costs an additional $14 000 per incremental life-year saved. If the cost of a blood culture is between $5 and $40, then “no work-up” is dominated by “CBC and blood culture + selective treatment.” All other strategies cost >$100 000 per life-year saved. If the cost of a blood culture is between $50 and $200, then “no work-up” is dominated by “CBC + selective blood culture and treatment” using a WBC cutoff of 15 × 109/L. As the cost of a blood culture continues to rise above $50, “CBC + selective blood culture and treatment” using a WBC cutoff of 10 × 109/L becomes less cost-effective. The incremental CE ratio of using a WBC cutoff of 10 × 109/L changes from $24 500 to $104 300 per life-year saved as the cost of a blood culture goes from $50 to $80. If the cost of a blood culture is <$40, then using a lower WBC cutoff of 10 × 109/L is cost saving compared with using a WBC cutoff of 15 × 109/L.
Two-Way Sensitivity Analysis
Two-way sensitivity analyses on the incremental CE ratios were performed on the variables that seemed to be interdependent and the most sensitive to variation in 1-way analysis (Figs 3A and 3B). The lines indicate the incremental CE threshold of $100 000 per life-year saved. For this threshold, the areas between the lines represent preferred strategies given this particular societal threshold. Overall, as the rates of bacteremia and meningitis decline, the CE of “CBC + selective blood culture and treatment” using a WBC cutoff of 15 × 109/L changes from cost saving to $1 971 100 per life-year saved. The CE of “clinical judgment” improves as the rates of bacteremia and meningitis initially decline. However, as the rates of bacteremia and meningitis approach 0, the CE of this approach also is reduced.
When sensitivity analysis is performed on the rate of bacteremia versus antibiotic efficacy, we see a similar trend as both rates decline. If the rate of bacteremia declines to 0.5% and antibiotic efficacy is 100%, then strategies that use testing and treatment result in CE ratios of >$185 000. As antibiotic efficacy declines, the CE of “clinical judgment” is reduced significantly.
In previous analyses of the management of febrile 3- to 36-month-old children, the most cost-effective approach was blood culture plus empiric antibiotic treatment.5,,6 However, the emergence of new data and changes in the background of occult bacteremia, particularly the decline in invasive H influenzae type B disease and the recent licensure of a conjugate pneumococcal vaccine, have led us to reexamine the question of the optimal management of febrile children who are at risk for occult bacteremia. In addition, recent commentaries highlight the concern over negative aspects of diagnostic testing and empiric treatment of all febrile children.20–25 To address these concerns, we structured our decision tree to include the complications of empiric antibiotic therapy and unnecessary evaluations for false-positive blood cultures. We chose to focus our analysis on the outcomes of cases of meningitis and life-years saved, as we realize that the goal of most practitioners is to prevent meningitis and death, not bacteremia itself.
Our analysis did not include the strategy of empiric treatment of all highly febrile children without using diagnostic testing as we believe that this strategy is not appropriate clinically in the face of national efforts to limit the use of antibiotics.31–33 In addition, without appropriate diagnostic testing before the initiation of treatment, the clinician may not be able to identify the cause of serious bacterial infections on reevaluation.
Our data demonstrate that at the currently estimated rate of bacteremia, “CBC + selective blood culture and treatment” using a WBC cutoff of 15 × 109/L, is a cost-effective strategy compared with “no work-up.” The incremental CE ratio of “CBC + selective blood culture and treatment” using a WBC cutoff of 10 × 109/L is $72 300 per life-year saved, which is slightly more expensive than other accepted medical interventions. For example, interventions that are commonly regarded as cost-effective include using tissue plasminogen activator compared with streptokinase in patients with acute myocardial infarction ($32 678 per life-year saved), medical therapy compared with no therapy for severe hypertension ($20 000 per life-year saved), and coronary bypass surgery compared with medical therapy for left main coronary artery disease ($7000 per life-year saved).34,,35Although some may consider the use of a lower WBC cutoff of 10 × 109/L to be cost-effective at the current rate of bacteremia, its CE will be significantly reduced as the rate of bacteremia declines.
Sensitivity analysis allows us to examine the CE of different management strategies at varying rates of bacteremia. If the rate of bacteremia in a population is >1.5%, then performing a diagnostic work-up at the initial visit may be considered a cost-effective strategy. At a lower rate of bacteremia of 0.5%, “clinical judgment” is a cost-effective option when compared with “no work-up.” All other strategies have CE ratios of >$300 000 per life-year saved, which is relatively expensive when compared with other life-saving interventions. Depending on the epidemiology of the population, management strategies may be altered accordingly.
Our sensitivity analyses also reveal several other important findings. First, “clinical judgment,” which is defined in our model as any elevation in the YOS, is useful at low rates of bacteremia, because a high-risk population may be selected at minimal additional cost and complication when compared with “no work-up.” Before this analysis, “clinical judgment” had not been considered a beneficial strategy because of its poor sensitivity.5,,36 Second, varying the rates and costs of anaphylaxis and false-positive blood cultures over a broad range does not change the results of our analysis. Thus, arguments over the economic impact of false-positive blood cultures or antibiotic reactions are not significant enough to change the practice management of febrile children at current rates of bacteremia. Third, our model is sensitive to different assumptions about the costs of diagnostic testing. If the cost of a CBC is >$55, then “blood culture + treatment” is a cost-effective strategy because the cost of performing a diagnostic test to identify high-risk groups outweighs the cost of treating all patients. If the cost of a blood culture is <$40, then “CBC and blood culture + selective treatment” using a WBC cutoff of 10 × 109/L is cost-effective. Again, if the cost of performing a diagnostic test is low enough, then strategies that use testing are preferred because fewer cases of meningitis are missed.
The results of our 2-way sensitivity analyses depict a dynamic model in which the CE of strategies changes depending on the changing epidemiology of occult bacteremia. As rates of bacteremia and meningitis decline, strategies that involve empiric testing and treatment become significantly less cost-effective. A similar shift in strategies is seen when the rate of bacteremia and the efficacy of antibiotics in meningitis are varied.
The results of our CE analysis may differ from previous analyses because our outcomes were measured as cases of meningitis prevented or life-years saved. Lieu et al5 focused on preventing all major infectious complications; thus, their model favored strategies that included empiric treatment. Downs et al6 measured quality-adjusted life-years, which included the outcomes of death and permanent disability; however, their model did not consider the costs associated with alternative strategies and did not include negative consequences of testing and treatment. Thus, their model favored the strategy of blood culture plus empiric treatment, because fewer cases of meningitis occurred with this strategy.
There has been significant variation in the management practices of physicians who care for young febrile children. The difference is particularly striking when comparing the practices of emergency physicians and office-based pediatricians.19 CE analysis may be used as a tool to aid clinicians in their decision making regarding the approach to young febrile children. On the basis of our analysis at the currently estimated rate of bacteremia, we conclude that “CBC + selective blood culture and treatment” using a WBC cutoff of 15 × 109/L is a reasonable approach. At a rate of bacteremia of 1%, “clinical judgment” becomes a useful strategy, although “CBC + selective culture and treatment” using a WBC cutoff of 15 × 109/L results in substantially more life-years saved at little additional cost. Finally, if widespread use of the conjugate pneumococcal vaccine reduces the overall rate of occult bacteremia to 0.5%, then “clinical judgment” remains a cost-effective strategy in the management of young febrile children, whereas strategies that include empiric testing or treatment have CE ratios of >$300 000 per life-year saved.
We hope that the introduction and widespread use of the conjugate pneumococcal vaccine will lead to a rapid and sustained drop in the rate of occult pneumococcal bacteremia. If the rate of bacteremia declines to 0.5%, then clinicians should reevaluate their approach to the highly febrile child and eliminate strategies that use empiric testing and treatment.
This project was supported in part by Grant T32 HS00063 from the Agency for Health care Research and Quality, US Department of Health and Human Services.
- Received September 25, 2000.
- Accepted February 23, 2001.
Reprint requests to (G.M.L.) Division of Infectious Diseases, Children's Hospital, Boston, 300 Longwood Ave, Boston, MA 02115. E-mail:
- CBC =
- complete blood count •
- WBC =
- white blood count •
- CE =
- cost-effectiveness •
- YOS =
- Yale Observation Score
- McCarthy P,
- Grundy G,
- Spiesel S,
- Dolan T
- Schoendorf K,
- Adams W,
- Kiely J,
- Wenger J
- Kaplan S,
- Mason E,
- Barson W,
- et al.
- Doern G,
- Pfaller M,
- Kugler K,
- Freeman J,
- Jones R
- Tan T,
- Mason E,
- Kaplan S
- Waltzman M,
- Harper M
- Timmermans D,
- Sprij A,
- De Bel C
- Kramer M,
- Shapiro E
- Bauchner H,
- Pelton S
- Schriger D
- Baraff L,
- Bass J,
- Fleisher G,
- Klein J,
- McCracken G,
- Powell K
- ↵Black S. Efficacy of heptavalent conjugate pneumococcal vaccine in 37,000 infants and children: impact of pneumonia, otitis media, and an update on invasive diseases—results of the Northern California Kaiser Permanente efficacy trial. Interscience Conference on Antimicrobial Agents and Chemotherapy. San Francisco, CA; September 1999
- Alpern E,
- Alessandrini E,
- Bell L,
- Shaw K,
- McGowan K
- ↵Gold M, Siegel J, Russell L, Weinstein M. Cost-Effectiveness in Health and Medicine. New York, NY: Oxford University Press; 1996
- Butler J,
- Hofmann J,
- Cetron MS,
- Elliott JA,
- Facklam RR,
- Breiman RF
- ↵Mark D. Economic analysis methods and endpoints. In: Califf R, Mark D, Wagner G, eds. Acute Coronary Care in the Thrombolytic Era. 2nd ed. St Louis, MO: Mosby-Yearbook; 1995:167–182
- Bradley J,
- Kaplan S,
- Tan T,
- et al.
- Baron M,
- Fink H
- Rothrock S,
- Harper M,
- Green S,
- et al.
- Baraff L,
- Oslund S,
- Prather M
- Carroll W,
- Farrell M,
- Singer J,
- Jackson M,
- Lobel J,
- Lewis E
- Jacobs N,
- Lerdkachornsuk S,
- Metzger W
- Korones D,
- Shapiro E
- Lebel M,
- McCracken G
- Richardson M,
- Reid A,
- Tarlow M,
- Rudd P
- Arditi M,
- Mason E,
- Bradley J,
- et al.
- Wenger J,
- Hightower A,
- Facklam R,
- Gaventa S,
- Broome C,
- Group BMS
- Fortnum H,
- Davis A
- Grimwood K,
- Anderson V,
- Bond L,
- et al.
- Centers for Disease Control and Prevention. Active Bacterial Core Surveillance (ABCs) Report: Emerging Infections Program (EIP) Network. Streptococcus pneumoniae. Available:http://www.cdc.gov/ncidod/dbmd/abcs/survreports/spneu98.pdf Accessed February 14, 2001
- ↵Waitzman N, Scheffler R, Romano P. The Cost of Birth Defects. Lanham, MD: University Press of America Inc; 1996
- Copyright © 2001 American Academy of Pediatrics