Report to AHCA 06/2010

Transcription

1 UTILITY OF CURRENT SURVEILLANCE SYSTEMS TO DETECT RESPIRATORY SYNCYTIAL VIRUS SEASONS AND IMPLICATIONS FOR IMMUNOPROPHYLAXIS Report to AHCA 6/21 Christian Hampp, PhD 1 Almut Winterstein, PhD 1, 2 1 Pharmaceutical Outcomes & Policy, College of Pharmacy, University of Florida 2 Epidemiology and Biostatistics, College of Public Health and Health Professions, University of Florida This study was funded by a grant from the Florida Agency of Healthcare Administration, AHCA. It was conducted in collaboration with the University of Florida Center for Medicaid and the Uninsured. We acknowledge the Centers for Medicare and Medicaid Services for the provision of claims data and Vital Statistics, California Department of Public Health; Public Health Statistics, Office of Vital Statistics, Florida Department of Health and Texas Department of State Health Services for the provision of birth and death certificates. We further thank the Florida Department of Health and the Centers for Disease Control and Preventions NREVSS team for the provision of RSV surveillance data. 1

2 EXECUTIVE SUMMARY Background Respiratory syncytial virus (RSV) is the most frequent cause of lower respiratory tract infections among infants and children. Prophylaxis with palivizumab is costly and therefore commonly restricted to periods of high viral activity. To inform timing of immunoprophylaxis, the Centers for Disease Control and Prevention (CDC) monitors RSV activity through its National Respiratory and Enteric Virus Surveillance System (NREVSS). A nationwide sample of laboratories report the number of specimens tested for RSV and the number of positive tests. When the proportion of positive tests exceeds 1% in two consecutive weeks, season onset is assumed in the first week. The relationship between this 1% threshold and the burden of disease as measured, for instance, in RSV related hospital admissions, has never been formally established. This study aimed to validate the CDC s surveillance system. This study further adds detail on RSV epidemiology comparing season onset, offset and extent of seasonality in different states and 5 regions in Florida. Another objective of this study was to investigate past patterns of palivizumab utilization and contrast these patterns with disease occurrence. Finally, since application of the season definition provides only dichotomous information and the burden of disease may differ between months categorized as on season, this study also provides monthly incidences of RSV hospitalizations for children at high risk together with a detailed picture of the numbers needed to treat (NNT) with palivizumab to avoid one RSV hospitalization. Overall, a validated season definition can provide the necessary evidence to guide providers decision making about the optimal timing and duration of RSV immunization. From a payer s perspective, this study can aid in optimizing reimbursement policies as region, monthand age specific NNTs can be used optimize RSV prophylaxis. Part I: Validation of CDC s Current RSV Season Definition For each week, each subject in the Medicaid four state fee for service dataset was categorized as high risk or low risk for RSV infection based on ICD 9 codes, pharmacy claims and birth certificates (Florida only). Subjects had to be continuously eligible for Medicaid coverage from birth and in ambulatory care for 4 weeks before the current week. The statewide weekly incidence rates of RSV hospitalizations were estimated for each risk category and adjusted for RSV prophylaxis. Weeks were categorized as on season if the RSV incidence rate in high risk children exceeded the season peak of the incidence rate in low risk children. Receiver operating characteristics (ROC) curves were used to measure the ability of the median 2

3 proportion of positive (MPP) laboratory tests to discriminate between on season and offseason weeks. Our sample consisted of a total of 72,492,625 subject weeks from 1,591,19 Medicaid fee for service recipients under 2 years of age from California, Florida, Illinois and Texas ( ). In California, Florida, Illinois and Texas, the areas under the ROC curves were.96 (95% confidence interval,.94.98),.91 (.87.95),.88 (.83.92) and.88 (.84.93), respectively, indicating good to excellent accuracy of the NREVSS. Requiring at least 5 positive tests in addition to the 1% MPP threshold further optimized accuracy by reducing the impact of outliers. Season onset according to NREVSS was detected on average 4.5 ( ) weeks apart from hospitalization onset and offset was 3.2 ( ) weeks apart from hospitalization offset. We observed a pattern of very distinct seasons in California with almost no activity outside of the season. Florida experienced longer and less regular seasons with residual activity outside the core seasons. RSV seasons in Illinois and Texas seemed more consistent over time compared to Florida. With the exception of the northwest, none of the regional AUCs reached the statewide AUC in Florida, however they still fall into the good category. Part II: RSV Epidemiology between Four US States and Five Regions in Florida A seasonality index based on the ratio of mean RSV hospitalization rates of on season and off weeks was established for each state/region. We further investigated annual variation in season onset/offset and duration to identify the need for a concurrent surveillance system. We found evidence for a pronounced difference in the seasonality indices between states. California and Illinois had a large seasonality index, indicating that the risk of an RSV infection during a season was more than 1 times the risk of an infection outside of a season. The risk for an on season infection was 9.5 times (95% CI, ) higher in Texas and only 3.7 times (95% CI, ) higher in Florida. The latter was a result of a low on season activity in Florida combined with a high off season activity. The on season activity in Florida only reached half of the activity in California and Texas. The regions in Florida differed widely with regard to their seasonality indices: the northwest and north regions had an off season activity comparable with California and Texas resulting in a seasonality index in the northeast that was only slightly below the Texas estimate. Moving south in Florida increased the off season RSV activity leading to a seasonality index of only 2.7 (95% CI, ) in the southeast, which suggests that the risk for an RSV infection during a season was only increased by 2.7 fold compared to off season periods. With the exception of Florida, annual variation in season onset was limited to about 2 months or nine weeks, but the upper limit of the 95% confidence interval for the range of 3

4 season onset well exceeded 8 weeks and we can conclude that there was significant annual variation in season onset in each state and in each region in Florida. Between onset, offset and peaks, we observed the largest statewide variation in Florida. A large contributor to this effect were the central and southeast regions of Florida which showed a pronounced variation with regard to season offset, reaching a range of 17 weeks in the central region. Thus, surveillance systems like NREVSS are needed to accurately identify season onset, but the effectiveness of communicating season onset to the medical community to assure timely initiation of prophylaxis needs to be considered (as opposed to having a fixed season onset). Part III: Latitude as a Factor in RSV Epidemiology in Florida We tested whether the regions in Florida differed in week of onset, offset and peak, season duration and peak RSV incidence. Coordinates of centroids of the five surveillance regions contribute latitude to a linear regression model with week of onset, offset and other parameters as a dependent variable. The slope of the regression equation was interpreted as change in timing of onset (offset) relative to change in latitude (1degree ~ 111km/69 miles). Differences in the RSV seasons between the regions of Florida were pronounced. The earliest season onset was observed in the southeast region for the 29 th week of the year (95% CI, ) the latest onset occurred in the northwest on average during the 46 th week (95% CI, ). The earliest offset was observed in the north (7.; 95% CI, ) and the latest in the southwest (18.2; 95% CI, ). Seasons were shortest in the north with an average duration of 18.8 weeks (95% CI, ) and longest in the southeast with an average duration of 38.2 weeks (95% CI, ). The southeast also experienced an early peak week (39.; 95% CI, ) while the other regions peaked closer to the end of the year. Results from the linear regression analysis show that latitude was a strong linear predictor for these differences. With each degree increase in latitude, season onset occurred 3.12 weeks later (95% CI, ). Moving north one degree in Florida was associated with a 1.45 weeks earlier offset (95% CI, 2.78 (.13)). These effects added up to a 4.6 weeks shorter season (95% CI, 5.75, ( 3.46)) for each degree north. Part IV: Timing of Prophylaxis with Palivizumab vs. RSV Seasonality From the Medicaid pharmacy claims dataset separate by state/region and year, the weekly number of palivizumab prescriptions was identified and divided by the number of subjects eligible to the study in that week. Utilization onset was defined as the first of two consecutive weeks where the utilization rate exceeded 5% of the peak utilization rate in the respective geographical area for that season. Utilization offset was defined as the last week 4

5 before two consecutive weeks with a utilization rate below 5% of the peak utilization rate. We compared onset and offset of utilization to (1) onset of RSV hospitalizations, (2) season onset according to RSV surveillance or (3) fixed dates, the latter indicating no adjustment of utilization to concurrent viral activity. A consistent utilization pattern could be observed on a state level with little variation between the years. The northwest and north regions of Florida and to some extent the central region exhibited a similar utilization pattern compared to the states. The southwest showed a less distinct pattern with some utilization year round, however seasons were still recognizable. In the southeast, utilization occurred almost year round in the later seasons. The onset of utilization in the 4 states was closest to a fixed date, namely the first week of October in Florida and Illinois and the first week of November in California and Texas. In all regions of Florida, onset of utilization was closest to early October. Actual season onset according to RSV hospitalizations and onset estimates according to surveillance data were further remote from onset of utilization compared to the fixed dates. With the exception of the central, southwest and southeast regions of Florida, utilization started before the onset of RSV hospitalizations. Utilization offset was observed in California and Illinois in the last week of April. Offset in the regions of Florida coincided with the fixed dates of late March or April. Overall, RSV season offset preceded the offset of palivizumab utilization in all states with the exception southwest region in Florida. These results indicate that timing of immunoprophylaxis did not coincide with viral activity, but rather follow fixed dates every year. Season definitions should be reviewed in order to optimize palivizumab utilization. Part V: Optimizing Timing of Prophylaxis Analogous to weekly analyses, we calculated monthly RSV hospitalization rates for each state/region. Numbers needed to treat indicate how many children need to be immunized to prevent one RSV hospitalization and were calculated as 1/ARR, where the absolute risk reduction (ARR) was calculated based on the monthly RSV incidence rate multiplied with the relative risk reduction associated with palivizumab (5%). The calculation was further broken down by age and calendar month to provide detail on the burden of disease beyond a dichotomous on season/off season rule. Consistent with our findings in part I, we observed a very distinct season in California and we found the high seasonality index from part II confirmed in the high RSV hospitalization rates on season while almost no activity was measured outside of the season. We could further observe a clear age pattern with the oldest age category being at much lower risk for infections compared to the youngest age category. In Florida, the seasonal pattern was less distinct, 5

6 however months with high activity were still distinguishable from months of lower activity, but unlike in California, we did not identify months with virtually no activity. The age difference in risk for RSV infections was also pronounced in Florida with the oldest age category being at comparably low risk. Illinois and Texas exhibited a similar pattern as California, with distinct periods of high and low RSV activity. Comparing NNTs across states, we found that immunoprophylaxis with palivizumab was less beneficial in Florida where NNTs were never below 1 regardless of age or calendar month. Among children months of age, NNTs below 2 are only found in Jan/Feb in Illinois and Texas and never in California and Florida. The statewide picture in Florida was merely a blend of a very different burden of disease in the regions. Although the northwest and north regions showed a pattern more similar to the other states, even here we didn t find months with virtually zero activity as we found in the other states, most pronounced in California. The southwest and southeast regions exhibited a pattern of prolonged viral activity, most obvious in the southeast. However, even in the southeast, we were able to distinguish between months of relatively high and relatively low activity. April through July in the southwest region and May and June in the southeast show NNTs exceeding 85 while the winter months had a peak activity that was comparable to other regions at their peak months. We refer to the discussion section for interpretation of the NNTs in the context of cost for prophylaxis. Conclusions This study validated the CDC s approach to detect seasons of RSV activity based on laboratory surveillance. We recommend the continued use of the 1% MPP threshold, however with the added requirement of 5 positive tests in a given week. Next, our study identified differences in seasonality over time and a different extent of seasonality between the 4 study states and even within Florida, therefore confirming the importance of a surveillance system that offers concurrent and local information on RSV activity. Based on RSV hospitalization data, we confirmed the need to subdivide the state of Florida into 5 regions to appropriately account for differences in RSV epidemiology. We found that historically, palivizumab utilization seemed not primarily triggered by seasons detected with laboratory surveillance; therefore further research is necessary to help understand the acceptance of the NREVSS or how seasonality can be effectively communicated to the (medical) community. Finally, we provided monthly RSV incidence rates for each state and for each region in Florida. The corresponding NNT estimates can provide further detail on the burden of disease to guide reimbursement practices and thus, 6

7 overcome limitations of a dichotomous RSV season definition. Higher NNTs for older children as a result of a lower RSV incidence combined with the need for higher and more costly doses of palivizumab after infancy highlight the reduced benefit of immunoprophylaxis in the second year of life. 7

8 TABLE OF CONTENTS page Executive Summary... 2 Part I: Validation of CDC s Current RSV Season Definition... 2 Part II: RSV Epidemiology between Four US States and Five Regions in Florida... 3 Part III: Latitude as a Factor in RSV Epidemiology in Florida... 4 Part IV: Timing of Prophylaxis with Palivizumab vs. RSV Seasonality... 4 Part V: Optimizing Timing of Prophylaxis... 5 LIST OF TABLES... 1 LIST OF FIGURES Introduction Background Purpose of Study Literature Review Respiratory Syncytial Virus RSV Disease Epidemiology RSV Infections The National Respiratory and Enteric Virus Surveillance System RSV Seasonality RSV Prevention... 2 RSV Risk Factors and Indications for Immunoprophylaxis... 2 Prior Authorization Requirements Methods Datasets NREVSS Florida Department of Health RSV Surveillance Data Medicaid Analytic extract Claims Dataset State Birth Certificates Study Population Part I: Validation of CDC s Current RSV Season Definition Part II: RSV Epidemiology between Four US States and Five Regions in Florida Part III: Latitude as a Factor in RSV Epidemiology in Florida Part IV: Timing of Prophylaxis with Palivizumab vs. RSV Seasonality Part V: Optimizing Timing of Prophylaxis

9 Results Sample Characteristics Part I: Validation of CDC s Current RSV Season Definition Part II: RSV Epidemiology between Four US States and Five Regions in Florida Part III: Latitude as a Factor in RSV Epidemiology in Florida... 4 Part IV: Timing of Prophylaxis with Palivizumab vs. RSV Seasonality... 4 Part V: Optimizing Timing of Prophylaxis Discussion Part I: Validation of CDC s Current RSV Season Definition Part II: RSV Epidemiology between Four US States and Five Regions in Florida Part III: Latitude as a Factor in RSV Epidemiology in Florida Part IV: Timing of Prophylaxis with Palivizumab vs. RSV Seasonality Part V: Optimizing Timing of Prophylaxis External Validity Study Limitations Future Research Summary and Conclusions APPENDIX A. Operational Definitions Palivizumab Exposure Risk Factors for RSV Chronic lung disease Prematurity Congenital heart disease Cystic fibrosis Severe combined or acquired immunodeficiency Down syndrome Asthma Transplant Malignancy Immunosuppression or antineoplastic agents Hospitalizations RSV hospitalization Specific non RSV bronchiolitis or pneumonia Unspecific bronchiolitis or pneumonia B. Supplemental Tables LIST OF REFERENCES

10 LIST OF TABLES Table page Table 2 1. Historical landmarks in the NREVSS Table 4 1. Cohort characteristics Table 4 2. Risk factors for RSV hospitalization in Florida Table 4 3. Palivizumab exposure and RSV hospitalizations by state and risk category Table 4 4. Areas under the curve by state and region Table 4 5. Test characteristics at the threshold of 1% median proportion positive laboratory tests* Table 4 6. Test characteristics at optimal thresholds of median proportion positive laboratory tests* Table 4 7. Mean of absolute differences and direction of difference between season onset according to clinical dataset and surveillance dataset under different definitions for season onset Table 4 8. Mean of absolute differences and direction of difference between season offset according to clinical dataset and surveillance dataset under different definitions for season offset Table 4 9. Extent of seasonality and seasonality index in each state and regions in Florida... 5 Table 4 1. Variation in seasons within each state and regions in Florida Table Comparison of season characteristics between regions in Florida Table Linear regression analysis of the effects of latitude on season characteristics in Florida Table Mean of absolute differences and direction of difference between onset of palivizumab utilization and onset of RSV season according to different determinants of RSV season

11 Table Mean of absolute differences and direction of difference between offset of palivizumab utilization and offset of RSV season according to different determinants of RSV season Table 5 1. Cost of prophylaxis per avoided RSV hospitalization... 8 Table B 1. List of counties in Florida Table B 2. Coordinates of Florida regions Table B 3. Week numbers and corresponding calendar months, shown for the year

12 LIST OF FIGURES Figure page Figure 2 1. Map of RSV regions in Florida, 68 reprinted with permission of the Florida Department of Health Figure 3 1. Season detection based on clinical dataset Figure 3 3. Cut off values for areas under the ROC curve Figure 4 1. Flowchart of sample selection and resulting sample size Figure 4 2. RSV hospitalization rates and resulting seasons in A) California, B) Florida, C) Illinois and D) Texas Figure 4 3. RSV hospitalization rates and resulting seasons in the regions of Florida. A) Northwest, B) North, C) Central, D) Southwest and E) Southeast Figure 4 4. NREVSS laboratory tests and resulting RSV season in A) California, B) Florida, C) Illinois and D) Texas. The red line marks the 1% threshold and the green line marks the optimal threshold. MPP: Median proportion of positive RSV laboratory tests Figure 4 5. NREVSS laboratory tests and resulting RSV season in the regions of Florida. A) Northwest, B) North, C) Central, D) Southwest and E) Southeast. The red line marks the 1% threshold and the green line marks the optimal threshold. MPP: Median proportion of positive RSV laboratory tests Figure 4 6. Receiver operating characteristics curves for each state Figure 4 7. Receiver operating characteristics curves for each region in Florida Figure 4 8. Linear effect of latitude on season characteristics in Florida. A) Week of season onset, B) Week of season offset, C) Season duration and D) Peak week Figure 4 9. Palivizumab utilization and RSV seasonality in A) California, B) Florida, C) Illinois and D) Texas Figure 4 1. Palivizumab utilization and RSV seasonality in the regions of Florida. A) Northwest, B) North, C) Central, D) Southwest and E) Southeast Figure A) Monthly RSV hospitalization incidence rates [per 1 subject months], B) NNT (numbers needed to treat) with palivizumab by age in the high risk cohort in California

13 Figure A) Monthly RSV hospitalization incidence rates [per 1 subject months], B) NNT (numbers needed to treat) with palivizumab by age in the high risk cohort in Florida Figure A) Monthly RSV hospitalization incidence rates [per 1 subject months], B) NNT (numbers needed to treat) with palivizumab by age in the high risk cohort in Illinois Figure A) Monthly RSV hospitalization incidence rates [per 1 subject months], B) NNT (numbers needed to treat) with palivizumab by age in the high risk cohort in Texas Figure A) Monthly RSV hospitalization incidence rates [per 1 subject months], B) NNT (numbers needed to treat) with palivizumab in the high risk cohort for each region in Florida... 7 Figure 5 1. Distribution of diagnostic codes for bronchiolitis and pneumonia related hospitalizations

14 CHAPTER 1 INTRODUCTION Background Respiratory syncytial virus (RSV) is the most frequent cause of lower respiratory tract infections among infants and children. In the United States, RSV causes annually up to 125, hospitalizations for bronchiolitis among infants under 1 year. 1 While no vaccination is available, palivizumab (Synagis, MedImmune, Inc., Gaithersburg, MD), a humanized monoclonal antibody, is able to reduce RSV related hospitalizations. 2, 3 According to the label, monthly injections are necessary to provide protection throughout an RSV season. 4 The major limiting factor to the widespread use of RSV prophylaxis is the high drug cost, which often results in expenses of more than $1, to immunize one infant through a six month season. 5 These costs limit prophylaxis to patients at increased risk for infection such as children with chronic lung disease (CLD), congenital heart disease (CHD) and certain preterm infants. 6 Another option for cost containment is the restriction of prophylaxis to a clearly defined RSV season of high viral activity. The Centers for Disease Control and Prevention (CDC) monitors RSV activity through its National Respiratory and Enteric Virus Surveillance System (NREVSS). 7 A nationwide sample of laboratories report the number of specimens tested for RSV and the number of positive tests. These data are updated every week and published on the NREVSS website. While the RSV season peaks in November/December in most countries of the northern hemisphere, regions closer to the equator show less seasonal variability and observe cases year round. Furthermore, even within countries, RSV outbreaks differ based on latitude and proximity to a coast. 8 In the southern United States, the RSV season starts earlier and lasts longer compared to the rest of the nation. 9, 1 Differences in seasonality have been identified even within a single state: Florida s southeast experiences earlier and longer seasons compared to the rest of the state. 11 Critical for the initiation of RSV prophylaxis with palivizumab is not only the knowledge about a patient s risk status but also about the appropriate timing and duration of immunoprophylaxis. The CDC developed an RSV season definition that is based on the proportion of positive RSV tests among all tests in its nationwide sample of laboratories. When this proportion exceeds 1% in two consecutive weeks, season onset is assumed in the first week. While this definition may seem reasonable, the relationship between this 1% threshold and the burden of disease as measured, for instance, in RSV related hospital admissions, has never been formally established. The NREVSS measures RSV activity through the use of population based tests, thus providing season estimates for the population as whole, albeit not necessarily representative. 14

15 Subjects tested for RSV are, naturally, patients with a suspected infection. The majority of infections, however, occur among children at low risk, 12 simply because they constitute the largest part of the infant population. Yet, the definition is used to provide immunization recommendations for high risk infants. Susceptibility at periods of lower viral activity may account for a different temporal pattern in RSV hospitalizations among high risk children. Such a scenario could occur if infants experience longer RSV seasons than older children. Finally, state specific surveillance suggests that the RSV season in Florida differs from the rest of the nation with regard to onset and duration, which can be attributed to differences in latitude and climate. Of note, when the CDC introduced its season definition, none of the 74 originally contributing laboratories were based in Florida. 13 Seasonality in Florida has been described as almost year round in the southeast using the 1% definition, though with a large variation in the absolute number in positive tests. 11 This variation is likely correlated with pronounced temporal differences in the burden of disease, but the actual RSV incidence rate for each of the months above this threshold is unclear. 14 Purpose of Study By establishing a validated RSV season definition, this study can maximize the acceptance of the CDC s surveillance data. A definition with face and construct validity can help health plans and providers select the optimal timing for RSV prophylaxis. The application of a validated definition to different states provides knowledge about regional differences in the onset and duration of the RSV season. Physicians can use this state specific information to prevent RSV infections most efficiently with immunoprophylaxis at the appropriate time. 15 Since April 28, the Florida Medicaid program has required prior authorization (PA) for palivizumab. Florida Medicaid accounts for regional differences in seasonality by allowing different periods of prophylaxis in different regions. 16, 17 However, the regional validity of the NREVSS and the 1% threshold may differ from the statewide validity due to a smaller number of labs and therefore, RSV tests. This study provides information about the regional validity of the surveillance system and the appropriateness of the different regional immunization recommendations in Florida. Another objective of this study was to investigate past patterns of palivizumab utilization and contrast these patterns with disease occurrence as measured with the current season definition based on surveillance data, clinical information from claims data and a fixed immunization schedule ignoring surveillance data. Therefore, we provide insight about whether providers based timing of immunoprophylaxis on NREVSS data or on their clinical observations of disease burden, or neither of these. Since application of the season definition provides only dichotomous information and the burden of disease may differ between months categorized as on season, this study also 15

16 provides monthly incidences of RSV hospitalizations for children at high risk together with a detailed picture of the numbers needed to treat (NNT) with palivizumab to avoid one RSV hospitalization. In addition, we provide NNTs for different age groups to further inform about differences in the need for prophylaxis even among children with indications. Overall, a validated season definition can provide the necessary evidence to guide providers decision making about the optimal timing and duration of RSV immunization. From a payer s perspective, this study can aid in optimizing reimbursement policies as region, monthand age specific NNTs can be used optimize RSV prophylaxis. 16

17 CHAPTER 2 LITERATURE REVIEW Respiratory Syncytial Virus RSV Disease Epidemiology RSV is the most frequent cause of lower respiratory tract infections among infants and children. By one estimate, RSV causes annually up to 125, hospitalizations for bronchiolitis or pneumonia among children younger than 1 year in the United States. 1 However, a more recent study reports an annual number of 57,275 RSV hospitalizations for children under the age of 5 years. 18 The same study further estimates that 2.1 million children under the age of 5 experience RSV infections each year; 3% of these infections lead to hospitalizations, 25% are treated in emergency departments and 73% in pediatric practices. Annual RSV related mortality for underlying pneumonia and influenza deaths has been estimated as 3.1 per 1, infants younger than 1 year in the US, amounting to approximately 124 RSV related deaths yearly. 19 RSV is associated with all cause death in 5.4 per 1, infant years or 214 deaths in this age group. Another study found that the majority of RSV related deaths is not associated with typical RSV risk factors. Specifically, among infants who died from RSV related causes before reaching the age of 5 years, only 9.9% had underlying congenital heart disease, 5.5% chronic lung disease and 4.2% were born prematurely. Consequently, the authors concluded that immunoprophylaxis of high risk infants would not prevent the majority of RSV related deaths. 2 The incidence of RSV related deaths is lower for older children, adolescents and adults, but increases with older age. 19 Pneumonia and influenza deaths associated with RSV occur at a rate of 7.2 per 1, person years above the age of 65 which amounts to 2,388 annual deaths. RSV associated all cause mortality was estimated at 29.6 per 1, person years or 9,812 annual deaths between the seasons 199/91 and 1998/99. Eighty eight percent of RSVrelated pneumonia or influenza deaths occurred in the elderly; however, this may be due to their overall higher mortality rate. RSV Infections The clinical picture of RSV infections appears differently in neonates and infants compared to older children and adults. 21 Infections in the first 4 6 weeks of life are rare, which may be related to the presence of maternal antibodies. 21, 22 Infected children between 6 weeks and 2 years of age usually develop symptomatic lower respiratory tract infections, including bronchiolitis and pneumonia, but also acute otitis media. Symptoms typically appear after an incubation period of 5 days and resolve after a few days to one week. 21 Older children and 17

18 adults often experience mild to moderate upper respiratory tract infections which are consequences of recurrent RSV infections, suggesting an incomplete acquired immunity after a primary infection. 23 Following RSV bronchiolitis, infant patients often exhibit recurrent wheezing up to ten years after infection, and an increased incidence of asthma in children subsequent to RSV hospitalization has been reported. 21, 24, 25 Conversely, a recent study found that RSV infections increased the incidence of asthma up to 8 fold only during the first 2 months after RSV hospitalization, and there was no increased risk one year after the infection. 26 The National Respiratory and Enteric Virus Surveillance System The NREVSS monitors temporal and geographic activity of respiratory and enteric viruses. These viruses include RSV, human parainfluenza viruses, respiratory and enteric adenoviruses, rotavirus, and since 27, rhinovirus, enterovirus and human metapneumovirus. Influenza specimen information, also reported to NREVSS, is integrated with CDC influenza surveillance data. Collaborating university, community hospital, commercial, and state and county public health laboratories report virus detections, isolations, and electron microscopy report results on a weekly basis. Annual summaries from NREVSS are published in Morbidity and Mortality Weekly Reports (MMWR). 27 The first reference to RSV surveillance by the CDC appeared in an MMWR from 1984 with qualitative rather than quantitative information and unclear temporal and geographic coverage. 28 A more systematic approach coincided with the first mentioning of the NREVSS in 199 including data from 95 laboratories in 49 states. 29 The same publication announced for the season a switch from a monthly postcard reporting system to a weekly telephonebased reporting system through a computer polling service with automatic tabulation allowing the publication of available results in the following week. 29 The information supplied by each participating laboratory consists of the weekly number of specimens tested for RSV and the number of positive tests. In 1993, the CDC provided a first definition for virus activity: weeks with 1% of specimens tested positive. 3 This definition was refined in the following season as: onset is the first of two consecutive weeks when at least half of participating labs reported any RSV detections or isolations. 31 An update to this definition from 1998 is still in use: when at least half of labs report any RSV detections for at least 2 consecutive weeks and when greater than 1% of all specimens ( ) are positive. 32 Further changes to the system include treating Florida separately from 25 due to its unique RSV pattern 33 and using inclusion criteria for participating labs for the 26 7 season, requiring that each lab reported 3 weeks of data, tested 15 specimens per week during winter months and reported 2% of specimens positive annually. Lastly, from 27 8, the system has experienced a substantial increase in the number of participating laboratories by using commercially available data from Surveillance 18

19 Data, Inc., which is supported by MedImmune, Inc., the manufacturer of palivizumab (see table 2 1). RSV Seasonality The RSV season spans from late fall to spring with peaks in November/December in most countries of the northern hemisphere, but regions closer to the equator show less of a seasonal variability with year round case occurrences. Even within countries, RSV outbreaks differ based on latitude and proximity to the coast. 8 In the southern United States, the RSV season starts earlier and lasts longer than in the rest of the nation. 9, 1, 34 More specifically, the impact of latitude and proximity to the coast may even be detectable within single states such as Florida, with a considerable north south extension and large coast line. In fact, longer seasons have been reported for southeast Florida compared to other parts of the state based on surveillance data. 11 Immunization recommendations are further complicated by annual variation in season onset and duration in the United States, necessitating current information on viral activity. 9, 34 A biennial pattern of RSV seasonality has been indentified in Sweden, with earlier seasons alternating with later seasons and higher hospital admission rates in earlier seasons. 35, 36 A similar pattern was found in Germany 37 and Finland 38, however this biennial pattern has not been shown for the US. The CDC s season definition has been widely applied. 9 11, 13, 34, It has been acknowledged that the 1% threshold is arbitrary 39 and that its suitability for defining months for RSV prophylaxis is unclear. 43 Two studies supported by MedImmune applied a slightly modified season definition ( 1% of tests positive in a given month) to the Florida DoH RSV surveillance system 11 and test data from three hospitals. 39 These studies identified RSV activity almost year round in Florida, reaching epidemic levels in most months. An editorial published with the former study found that even across time periods above this threshold, the absolute number of cases is highly variable. 44 More specifically, we found and expressed that, in selected months (May through August 21) the absolute statewide number of positive tests reported in that study was less than 1 as compared to 143 in January indicating a large difference in the burden of disease. 14 A study from 1998 using laboratory data from two hospitals in Jacksonville, Florida claimed to have used the CDC s definition, however looking at the first of 2 months 1% instead of the first of 2 weeks as the CDC does. 1 This study reaches similar conclusions with almost year round RSV activity in Florida; yet again, many summer months show less than one tenth of positive tests of some winter months and still exceed the 1% epidemic threshold. All these observations highlight the problem that a dichotomous categorization into on season and off season periods can mask large differences in the burden of disease even during season. Whether the potentially much lower incidence of RSV in the summer months identified in the three quoted studies warrants prophylaxis, despite exceeding the 1% threshold, is not clear. 19

20 RSV Prevention While no vaccination is available for RSV, 22, 45 two preventive agents, respiratory syncytial virus immune globulin intravenous (RSV IGIV, RespiGam, MedImmune, Inc., Gaithersburg, MD) and palivizumab (Synagis, MedImmune, Inc.) are indicated to reduce the risk of RSV related hospitalization. RSV IGIV was approved by the US Food and Drug Administration in January The second generation product, palivizumab, was approved in Palivizumab is a humanized monoclonal antibody and can thus avoid the risk for infections potentially associated with the older, pooled human blood product RSV IGIV. Also, the smaller injection volume and less burdensome intramuscular application of palivizumab contributed to its replacement of intravenous RSV IGIV from the market As a consequence, the manufacturer discontinued the production of RSV IGIV at the end of A new agent for the prevention of RSV infections, motavizumab (MedImmune, Inc.), has reached phase III of drug development, and its sponsor submitted a Biologics License Application to the FDA in January At this time, no published clinical trial results have been identified as evidence for its efficacy. Currently, palivizumab is the only available pharmaceutical product for the prevention of RSV related hospitalizations. Palivizumab has proven its efficacy in clinical trials, with varying estimates for different indications. IMPACT RSV reports a relative risk reduction for RSV hospitalizations (RRR) of 39% for children with CLD, 78% for premature infants and 55% for the combined group. 2 In another trial, children with CHD experienced a 45% reduction in RSVrelated hospitalizations. 51 To date, a reduction in RSV related mortality has not been 6, 52 demonstrated for palivizumab. The cost of RSV prevention with palivizumab is significant. The average wholesale price of one 5mg vial is $926.48, 53 sufficient for one monthly dose for a 3 kg infant at 15mg/kg. With increasing age and body weight, higher monthly doses at higher total cost are required. One study found that the average dose in a Medicaid population of 2 year old children costs about $1,7. 5 If 6 doses are administered as is common in Florida, total cost of immunoprophylaxis for one child averages more than $1, for a single season and significantly more if palivizumab is administered year round. Costs per avoided hospitalization are often found to exceed expenses of the actual hospitalization by far, suggesting unfavorable cost benefit. 5, 54 In the light of these considerations, recommendations limit immunoprophylaxis to patients at highest risk for infection during seasons of high viral activity. 6 RSV Risk Factors and Indications for Immunoprophylaxis The American Academy of Pediatrics has defined indications that describe children at high risk for RSV infections 6 and recommends immunoprophylaxis for: Children less than two years of age with chronic lung disease; 2

21 Children with a gestational age of less than 28 weeks if they are not more than 12 months old at season onset (for this study: Prematurity I); Children with gestational age of weeks if they are not more than 6 months old at season onset (Prematurity II); Children with a gestational age of weeks if additional risk factors such as day care attendance or smoking parents are present (Prematurity III); Children with hemodynamically significant cyanotic and acyanotic congenital heart disease; Children with cystic fibrosis (CF); and Children with severe immunodeficiencies which include severe combined immunodeficiency (SCID) and acquired immunodeficiency syndrome (AIDS). The last two indications are not based on strong evidence for effectiveness (expert opinion only) and more carefully phrased in the guideline. These recommendations are supported by several studies that report increased RSV incidences in some of the above risk groups, yet to a varying extent. For instance, Boyce et al. (2) report a risk for RSV hospitalization in infants younger than 6 months with CHD of 12.8 per 1, infant years of RSV season compared to 44.1/1 for low risk infants in the Tennessee Medicaid population. In contrast, Duppenthaler et al. (24) estimate a rate of only 25 RSV hospitalizations per 1, infant years in the same age group with CHD compared to 18/1 without CHD in Switzerland. 12, 55 Not included in the AAP guideline, the presence of Down syndrome has been suggested to be associated with increased risk for RSV infections. 56 Furthermore, a few small sample studies suggest a more severe course of RSV infections in immunocompromised children with liver transplant 57, 58 59, 6 or after chemotherapy for cancer. Whether malignancy actually increases the infection risk for RSV is disputed, 61 and conclusive evidence is lacking. Internationally, RSV prevention guidelines differ in scope; for example, the Swedish guideline 62 is more restrictive than the rather inclusive AAP guideline in the US. 6 The former recommends immunoprophylaxis for infants with extreme prematurity up to 26 weeks gestational age only up to 6 months of age or up to 24 months of age with CLD and prematurity up to 36 weeks gestational age. Prior Authorization Requirements To limit expenditure for RSV immunoprophylaxis, third party payers typically restrict reimbursement for prophylaxis to children who meet certain criteria and to a limited time period of high risk. A common instrument is a requirement for PA where providers have to confirm the presence of risk factors to the third party payer to request reimbursement. Reimbursement policies in the 4 study states during the study period are detailed here: 21

22 California: California Medicaid has always required PA for palivizumab since its market introduction in Requirements changed with updates in AAP guidelines (personal communication, California Medicaid Pharmacy Benefit Division, 2/23/29). The current PA guideline is in agreement with the AAP guideline, with the exception that it allows for utilization up to the age of 48 months in immunodeficient children. 63 It allows for the administration of up to six doses between October and May. Florida: Prior to 28, Florida Medicaid did not restrict access to palivizumab for children under the age of 2 years, neither with regard to risk factors nor timing of immunization. Since April 28, Florida has used a PA system that allows palivizumab utilization for high risk children according to the AAP guideline for different time periods in 5 regions and up to year round in the southeast of Florida (personal communication with Anne C. Wells, Bureau Chief, Florida Medicaid Pharmacy Services, 2/2/29). 17 The 5 regions are illustrated in figure 2 1; an overview of the county composition of each region is provided in the appendix (table B 1). Illinois: Until 25, no PA requirement was in place for children under 4 years of age, and immunoprophylaxis was not restricted to season months. During this time, PA was required for children older than 4 years of age. Of note, palivizumab has even been used for ventilated children up to the age of 18 years. Since 25, the PA requirement mirrors the AAP guideline (personal communication with Brad Berberet, Assistant Director, Prior Authorization, Department of Pharmacy Practice, University of Illinois at Chicago, 2/23/29). Texas: Texas Medicaid has had a PA requirement in place during the study period and beyond. The PA was closely modeled after AAP guidelines, allowing prophylaxis for children with CLD, prematurity or CHD during season months according to NREVSS (personal communication with Judy Devore, Special Assistant to the Medicaid/CHIP Medical Director, Texas Health & Human Services Commission, 2/24/29). 22

23 Table 2 1. Historical landmarks in the NREVSS Season Update to the surveillance system MMWR Reference RSV surveillance first mentioned in available MMWR, 1/ no explicit season definition Expanded Surveillance System : first reference to NREVSS with 11/ participating labs Change from monthly postcard to weekly telephone based system announced for 199/ Definition of activity: weeks with >1% of specimens positive 1/ Onset: first of two consecutive weeks when at least half of 12/ participating labs reported any RSV detections or isolations Offset: no definition Definition as above, This definition generally indicates a mean 12/ percentage of specimens positive by antigen detection in excess of 1% Onset: when at least half of labs report any RSV detections for 12/ at least 2 consecutive weeks and when greater than 1% of all specimens ( ) are positive Same as above, in addition: 12/ Community outbreaks: greater than 2 consecutive weeks with greater than 1% positive tests, by city Florida is added as a separate region. National activity: defined as above ( ) Regional activity: median date that indicates the first of 2 consecutive weeks a participating lab reports >1% ( ) positive ( ) and the last week >1% positive tests preceding 2 consecutive weeks of <1% positive tests. Labs have to meet inclusion criteria: reported 3 weeks of data, tested 15 specimens per week during winter months and reported 2% of specimens positive annually. National and Regional season Onset: first of 2 consecutive weeks during which the median percentage of specimens positive for RSV antigen is 1% Offset: last of 2 consecutive weeks during which the median percentage of positive specimens is 1%. Expansion of available lab data by inclusion of data from Surveillance Data, Inc. for the season 27 28, with support from MedImmune, Inc. Labs have to meet inclusion criteria: reported >3 weeks and averaged >1 antigen detection tests per week. 12/26 12/27 12/

24 Figure 2 1. Map of RSV regions in Florida, 68 reprinted with permission of the Florida Department of Health 24

25 CHAPTER 3 METHODS Datasets This study was based on two datasets, a surveillance dataset and a clinical, patient level dataset based on medical and pharmacy claims. The surveillance dataset was comprised of NREVSS RSV surveillance data for the states of California, Illinois and Texas, and DoH surveillance data for the state of Florida. The clinical dataset consisted of Medicaid data for the same four states, commonly referred to as Medicaid Analytic extract (MAX) data, provided by the Centers for Medicare and Medicaid Services (CMS). The claims datasets for California, Florida and Texas were merged with Vital Statistics birth certificates data to obtain gestational age estimates. Each dataset covered the years , thus including 5 RSV seasons. The choice of the 4 states provided a large sample size as well as geographic diversity with the intent to increase external validity. Characteristics of the source datasets are described below. NREVSS The RSV component of the NREVSS collects weekly information on RSV tests from a sample of laboratories as described in greater detail above. The resulting dataset consists of a laboratory identification number and the city and county where the lab is located. Laboratories report the test type, the number of total tests and the number of positive tests recorded for a given week. Florida Department of Health RSV Surveillance Data As mentioned above, Florida had not been a separate part of the NREVSS before 25, and only a small number of Florida based laboratories reported to the CDC before that. To investigate seasonality on a regional level within the state, these data had to be supplemented. The Florida DoH has its own RSV surveillance system in place, using the same laboratory survey approach as the CDC, but with a larger sample in Florida. Geographic detail is provided at the city level and for 5 regions within the state (figure 2 1, see appendix for the regional classification of counties). 68 Medicaid Analytic extract Claims Dataset For each of the four states, eligibility, inpatient, outpatient and pharmacy datasets for Medicaid recipients 2 years of age were requested. Eligibility and demographic information was updated for each month. The MAX dataset has already been reconciled by CMS to display for each transaction a final action claim which eliminates the need to remove duplicate claims 25