Impact and Cost-effectiveness of Respiratory Syncytial Virus Prophylaxis for Kansas Medicaid's High-Risk Children FREE

THE INTRODUCTION of prophylactic agents for the prevention of respiratory syncytial virus (RSV) infections has been met with substantial debate. On the basis of clinical trial data, the 2 agents available, intravenous RSV immune globulin and palivizumab, lower hospitalization rates for RSV-related infections, thereby reducing health care resource use.^{1 - 5} However, the cost of these agents generates questions about their cost-effectiveness. The ranges of cost-effectiveness ratios from modeling studies of palivizumab to prevent a single RSV-related hospitalization are extremely variable and are influenced by assumptions regarding hospitalization and prophylaxis costs.^{6 - 8}

The limited applicability of clinical trial protocols to actual practice leaves open the possibility of suboptimal use of these agents. For instance, concerns about compliance and the impact on effectiveness are paramount with these agents because they are administered intravenously or intramuscularly on a monthly basis throughout an RSV season. Seid and colleagues⁹ reported that compliance with intravenous RSV immune globulin therapy was less than 20% until they resorted to direct efforts to improve completion. Since clinical trial efficacy is based on high rates of compliance,^{1 - 5} effectiveness in practice may be compromised by suboptimal compliance. In addition, effectiveness may be compromised by expanded use of these agents to populations who may not be at high risk for RSV hospitalizations.

There have been 2 observational studies of intravenous RSV immune globulin.^{10 - 11} One reported a decline in the rate of hospitalizations among those who received the drug compared with a historical comparison and a concurrent untreated group, but compliance was high at 81%, and cases were followed up prospectively through a high-risk outpatient clinic.¹⁰ A case-control study in which the rate of compliance was lower at 70.6% demonstrated a nonsignificant decline in the rate of RSV hospitalizations across 2 seasons.¹¹ We were unable to locate any observational studies of palivizumab.

The purpose of this study was to determine the impact of prophylactic therapy for RSV infections on subsequent hospitalizations and related lengths of stay and costs and estimate whether prophylactic treatment was cost-effective. The outcomes examined in this study were lengths of stay and associated costs in RSV-related inpatient hospitalizations. We hypothesized that, after adjustment for differences in risk factors, children who received one of these therapeutic agents would have better outcomes. These outcomes, in turn, should justify the cost of these agents.

The study design was a retrospective cohort analysis conducted for a single RSV season in Kansas: October 1, 1999, through April 30, 2000. Medicaid administrative claims (institutional, medical, and pharmacy) were obtained through the Kansas Medicaid Drug Utilization Review Program. Institutional claims included all services provided in any institution, namely, inpatient services. Medical claims included professional and other health care services provided on an outpatient basis, and prescription data included all outpatient prescription medications. The sampling frame was children who were 10 months or younger at the start of the season or born during the season and who were continuously enrolled in the Medicaid fee-for-service program during the season. We selected children according to criteria typically used in clinical trials of these treatments, eg, those with a diagnosis of prematurity or chronic lung disease. The treated group was defined as children who received either palivizumab or intravenous RSV immune globulin, and their outcomes were compared with those of untreated children. Only claims reported during the RSV season were used in the analysis.

Recipient demographics (age, sex, race or ethnic class) were obtained from Kansas Medicaid eligibility files. Age was determined from the beginning of the study period and divided into 3 groups: born during the season, 0 up to 6 months of age, and more than 6 months to 10 months of age. Race or ethnicity was reported as described in the Medicaid eligibility file; however, because of small numbers of minority children in the study, race or ethnicity was dichotomized as white vs other in the statistical analyses. Children who died during the season were excluded. We identified potential risk factors for RSV infection (chronic lung disease or bronchopulmonary dysplasia, congenital heart disease, and prematurity) by screening all medical and institutional files for the following diagnosis codes (International Classification of Diseases, Ninth Revision, Clinical Modification) in any position in the claim: bronchopulmonary dysplasia (770.7), congenital heart disease (745.x, 746.x, and 747.0-747.29), and prematurity (765.x). We also screened procedural codes in outpatient claims to determine which children received oxygen on an outpatient basis (Current Procedural Terminology codes E0431, E0424, E0443, E0444, E0450, E1390, E1400, and E1403).

Inpatient visits were extracted from institutional claims on the basis of length of stay. If the length of stay was zero, the claim was considered an outpatient hospital visit and subsequently combined with other outpatient medical claims. Lengths of stay equal to 1 day or greater were considered inpatient admissions. Multiple inpatient claims for the same admission date were aggregated into a single claim by combining diagnosis codes and summing Medicaid reimbursement amounts. Inpatient hospital claims were aggregated by recipient and RSV-related vs non–RSV-related claims summing the lengths of stay, number of admissions, and Medicaid reimbursement amounts. Claims containing any of the following diagnosis codes were denoted as RSV-related inpatient visits: 466.11 (RSV bronchitis), 480.1 (RSV pneumonia), and 079.6 (RSV infection). Most RSV infection codes were located in the primary diagnosis code position for the relevant claim. Since the perspective of all analyses was that of the payer, costs were based on actual Kansas Medicaid payments.

Intravenous RSV immune globulin and palivizumab doses were identified from prescription and outpatient medical claims. Each claim was assumed to be a single dose. The number of doses administered and the amount Medicaid reimbursed were summed. Medication administration costs were included in the medical claims. We determined whether those who received intravenous RSV immune globulin or palivizumab had a sufficient number of doses to provide coverage throughout the entire season. Children born before November 1 of each season should have received 6 doses to complete a series. Children born during the season were expected to have a number of doses equal to 1 less than the remaining number of months in the season. For instance, children born in January were expected to receive 3 doses to complete the series.

One of the limitations of observational study designs is the potential for selection bias. Since the decision to use prophylactic therapy depends on clinical judgment, a number of factors would influence drug use that are not captured with claims data. One method to adjust for this bias in observational research is the use of propensity scores to select untreated subjects matched on factors influencing the likelihood of treatment.^{12 - 13} For this study, the propensity scores allowed us to identify a matched control group with similar risk factors for drug use.

The first step was to perform a logistic regression estimating the probability of palivizumab or intravenous RSV immune globulin use based on identifiable risk factors: age, outpatient oxygen use, bronchopulmonary dysplasia, congenital heart disease, and prematurity. We computed a predicted value for each member of the cohort on the basis of this logistic regression model. Next, we divided the predicted values for the treated cases into approximately 5 equal-sized groups. Using the probability values bounding each group, we randomly selected an equivalent number of untreated subjects for each predicted value group. The number of potential untreated subjects for each of the 5 groups limited us to a one-to-one matching, so there were 137 in the treated group and 137 in the untreated group.

Descriptive statistics are presented to characterize the original population as well as the propensity score cohort. We also present the results of the propensity score logistic regression analysis. Next, we used logistic regression to estimate the effects of palivizumab or intravenous RSV immune globulin use on the probability of any RSV inpatient admission, controlling for the predicted probability from the propensity model. Since no one in the propensity samples (treated or untreated) had more than 1 RSV-related admission, there was no need to model the number of admissions. Multivariate models allowed us to estimate the difference between the treated and untreated groups' RSV inpatient lengths of stay and costs, controlling for the predicted probability from the propensity model. As a final measure of the cost-effectiveness of treatment, we computed Medicaid's average expenditure for palivizumab and intravenous RSV immune globulin and divided the result by the reduction in hospital costs associated with treatment determined from the multivariate model.

The characteristics of the treated and untreated groups for the overall population and the propensity score cohort are given in Table 1. There were 1506 children who were 10 months or younger and had a diagnosis of either prematurity or bronchopulmonary dysplasia. Only 136 children received palivizumab and 1 child received intravenous RSV immune globulin, for a total of 137 in the treated group. There were significant differences in risk factors between the treated and untreated groups in the overall population. The treated children had higher rates of prematurity, congenital heart disease, and outpatient oxygen use. The propensity score matching eliminated most of these differences.

Results of the propensity score logistic regression model to estimate the likelihood of receiving palivizumab or intravenous RSV immune globulin are given in Table 2. The probability of receiving either agent was influenced by age, prematurity, bronchopulmonary dysplasia, congenital heart disease, and outpatient oxygen use. For instance, children who were 0 through 6 months of age were nearly 6 times more likely to receive one of the agents compared with children who were born during the season. Children older than 6 months to 10 months of age were equally likely to receive the drug, however. Prematurity was the strongest predictor of treatment.

Table 3 displays the outcome variables for the propensity score cohort. The treated group was 0.47 times as likely to be hospitalized for an RSV infection as their untreated counterparts. Mean length of stay and hospitalization costs were lower for the treated cohort, but the differences were not statistically significant. Among children who were hospitalized, hospitalizations in the treated group were 2.3 days shorter and $4190 less expensive than among the untreated children. Approximately one third (46 of 137) of the treated children completed their requisite series of injections.

The next step was the prediction of the probability of any RSV-related inpatient hospitalization (Table 4). After adjustment for the propensity score, the odds of a hospitalization for the treated group were 0.47 vs the untreated group, but this was not statistically significant. The RSV-related hospitalization length of stay and costs were lower in the treated group, but the difference was only statistically significant for length of stay. Length of stay was reduced by nearly one-half day and hospitalization costs were lowered by $703 per admission. Given that Kansas Medicaid spent $642 080 in drug costs for their palivizumab and intravenous RSV immune globulin therapy, the average cost-benefit ratio of using these agents to prevent hospitalizations was 6.67 ($642 080/137 = $4687 per treated child; $4687/703 = 6.67).

The purposes of this study were to determine the impact of prophylactic therapy for RSV infections on subsequent hospitalizations and related lengths of stay and costs and to estimate whether prophylactic treatment was cost-effective. We analyzed claims data for the 1999-2000 RSV season and used a propensity score–matched sample method to adjust for differences that would influence the likelihood of receiving palivizumab or intravenous RSV immune globulin. While these agents did reduce RSV-related hospital length of stay by one-half day, they did not reduce RSV-related hospitalization costs sufficiently to justify the drug costs.

Our results generally concur with clinical trials that have demonstrated a 41% to 63% relative reduction in the number of hospitalizations.^{1 - 5} The average RSV-related inpatient admission length of stay and cost in our sample for the untreated children who were hospitalized were 5.9 days and $7456 per admission. These results for length of stay are in line with results from most modeling studies, but our costs are lower. One report, for instance, estimated a length of stay of 5.9 days and hospitalization costs of $11 083 from a network of university hospitals.⁷ A clinical trial of intravenous RSV immune globulin in previously healthy or low-risk children reported an average length of stay of 5.5 days for placebo and 4.6 days for intravenous RSV immune globulin.³ Costs from a model for ward care only (eg, no intensive care unit bed days) were only $2800. Most models used inpatient cost estimates in the range of $8500 to $12 000, a range higher than our results.^{6 - 8 ,14 - 19} Few of the modeling estimates were drawn from actual data, and those that relied on expert opinion were the most variable and highest.¹⁵

The high cost of therapy that we found to reduce RSV-related hospitalization costs, a cost-benefit ratio of 6.67:1, is consistent with that calculated by others. Previous estimates of prophylactic costs range from $51 000 to $90 000 per hospitalization avoided.^{20 - 22} Our results may reflect a broader range of practice and therefore include children at lower risk, who account for a greater percentage of RSV hospitalizations but have shorter lengths of stay. In addition, only one third of the treated children in our study completed their requisite series of injections. This rate is far below the level of compliance achieved in clinical trials and may undermine effectiveness. Even if compliance with palivizumab or intravenous RSV immune globulin was 100% and RSV-related hospitalizations were eliminated, the unadjusted cost per hospitalization avoided would be $40 130, not including the additional drug costs associated with higher compliance. Previous studies used drug costs ranging from $1000 per dose of intravenous RSV immune globulin up to $2700 to $4500 per course of intravenous RSV immune globulin or palivizumab.^{7 ,14 ,16} The average cost of therapy for children who completed their series in our sample was $8137, nearly twice the estimates of the cost per course of therapy used in previous studies.

Because our analysis involved a Medicaid population, the ability to generalize conclusions across populations is limited. However, low-income children are as much as 8 times more likely to have an RSV-related hospitalization, and Medicaid recipients account for up to 59% of RSV infections, so this is an appropriate target population.²³ Although access to claims data allowed us to examine all sources of direct medical care, our study relied on the validity of the diagnosis codes submitted by billing providers, and we could not adjust for variation within diagnosis, eg, severity of bronchopulmonary dysplasia or degree of prematurity. The adjustment made by using the propensity score eliminated substantial differences in risk factors between the treated and untreated groups, but this method is limited by additional unmeasured factors that may vary between the 2 groups and influence their outcomes, eg, other environmental risk factors such as household smoking exposure, day care, or multiple siblings.^{12 - 13} We also acknowledge that we considered only the costs incurred by the payer, Kansas Medicaid, since the analysis was conducted from that perspective. In practice, there are other costs borne by society, such as work loss costs of parents and the potential loss of lifetime earnings for children who die of their infections. A cost-benefit ratio of 6.67:1 suggests that, even with a substantial level of error, it would be unlikely that the cost of these therapies would break even.

Despite the demonstrated efficacy of intravenous RSV immune globulin and palivizumab in clinical trials, use of these agents in actual practice has not generated expected outcomes in the Kansas Medicaid program. The actual use of intravenous RSV immune globulin and palivizumab was not cost-effective in children we identified as at an increased risk for RSV hospitalizations. A consensus opinion on immunoprophylaxis with palivizumab has emphasized the use of selection criteria to identify infants at risk for increased disease severity.²⁴ Our findings and prophylaxis cost estimates of others^{20 - 22} mentioned previously underscore the importance of identifying only those at highest risk for RSV hospitalization as candidates for prophylactic therapy. Even then, the drug may not be cost-effective.

Accepted for publication August 8, 2002.

This project was supported by a grant from the Kansas Department of Social and Rehabilitation Services, Topeka, and the University of Kansas School of Pharmacy, Lawrence, through the Kansas Medicaid Drug Utilization Review Program.

We thank G. R. Gordon-Ross, PharmD candidate, University of Kansas School of Pharmacy, Lawrence, for his assistance in cleaning and aggregating the data. We also thank our anonymous reviewers for their suggestions for improving the manuscript.

The conclusions drawn in this study do not necessarily reflect the opinions of the Kansas Department of Social and Rehabilitation Services, the University of Kansas School of Pharmacy, or the Kansas Medicaid Drug Utilization Review Program.

Corresponding author and reprints: Theresa I. Shireman, PhD, Department of Pharmacy Practice, University of Kansas School of Pharmacy, 1251 Wescoe Hall Dr, Lawrence, KS 66045-7582 (e-mail: shireman@ukans.edu).

Randomized clinical trials of intravenous RSV immune globulin and palivizumab have demonstrated significant reductions in the number of hospitalizations and length of hospital stays. However, clinical trials often poorly reflect actual practice, and lower levels of efficacy may be seen in uncontrolled settings. Cost-benefit and cost-effectiveness studies modeled after the clinical trials have been inconclusive. There are only 2 observational studies of intravenous RSV immune globulin and none of palivizumab documenting their use in practice.

This study quantifies Kansas Medicaid's experience with RSV-related treatment and prophylaxis. We used propensity scores to derive an untreated group matched on risk factors with treated children for the likelihood of receiving one of these agents to adjust for a key limitation to observational study designs. Despite their demonstrated efficacy in randomized clinical trials, the use of these agents in actual practice did not generate expected outcomes in the Kansas Medicaid program. Although use was associated with a reduction in the odds of hospitalization, length of stay, and inpatient costs, these benefits were not sufficient to overcome the costs associated with drug therapy. In short, palivizumab and intravenous RSV immune globulin were not cost-effective in practice. Given the rarity of adequately designed trials that examine clinical practice, this observational analysis yields important information for both Kansas Medicaid and clinicians regarding the role of these agents.