a Preventive Medicine and Biometrics
d Pediatrics, School of Medicine
b Department of Geography, University of Colorado and Health Sciences Center, Denver, Colorado
c Epidemiology
e Pediatrics, Children's Hospital, Denver, Colorado
| ABSTRACT |
|---|
|
|
|---|
METHODS. A geographic information system was used to assemble data for altitude and demographic variables by zip codetabulation areas. Data then were linked with hospital discharge data for RSV infections. Poisson regression models were developed to explore correlations between hospitalization rates and residential altitude, after adjustment for socioeconomic differences in the underlying population.
RESULTS. RSV-associated hospitalizations averaged 15.9 per 1000 infants who were younger than 1 year and 1.8 per 1000 children who were 1 to 4 years of age per season. A multivariate analysis suggested that the rate of hospitalization for RSV-specific International Classification of Diseases, Ninth Revision, Clinical Modification codes increased 25% among infants who were younger than 1 year and 53% among children who were 1 to 4 years of age for every 1000-m increase in altitude. The risk for RSV-associated hospitalization was highest at elevations above 2500 m.
CONCLUSIONS. High altitude above 2500 m is a modest predictor for RSV-associated hospitalization. Practitioners in these regions should consider additional efforts to educate parents about RSV infection and its prevention and the importance of early treatment.
Key Words: environmental factors epidemiology hospitalization rates respiratory syncytial virus risk factors
Abbreviations: RSVrespiratory syncytial virus CHAColorado Health and Hospital Association ICD-9-CMInternational Classification of Diseases, Ninth Revision, Clinical Modification ZCTAzip codetabulation area GISgeographic information systems RRrate ratio
RESPIRATORY SYNCYTIAL VIRUS (RSV) infection is the foremost cause of bronchiolitis and pneumonia in infants and young children, causing >100000 hospitalizations annually in the United States.1 A number of risk factors for RSV hospitalization have been identified,2 such as prematurity,3 chronic pulmonary4 or congenital heart disease,5 immunodeficiency,68 concurrent birth siblings,9 and Native American heritage.10 Environmental risk factors include exposure to passive smoke,11,12 household crowding,1315 having older siblings in the home,16 and child care attendance.13,17
An unexplored environmental factor for severe RSV disease is residence at high altitude. We recently observed excess respiratory hospitalizations for RSV in a Colorado hospital at
3100 m in elevation.18 A cohort of 57 full-term infants and children were followed prospectively through the 20002001 RSV season. Nineteen (33%) of the children were hospitalized for RSV, and 3 required air-lifting to Denver for treatment. More than two thirds of the cohort needed supplemental oxygen after discharge. These data suggested that infants and children who live at high altitudes may be more vulnerable to severe RSV illness that requires hospitalization. This finding prompted us to explore the effect of altitude on RSV hospitalization in Colorado, where elevations vary from
1093 to 3455 m.
| METHODS |
|---|
|
|
|---|
The variables that were obtained for individual patients from the CHA database included age, zip code of residence, admission dates, and International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) diagnostic codes. Case patients were defined as children who were 4 years and younger and had hospital discharge records with Colorado zip codes and RSV-associated ICD-9-CM code 079.6 (RSV), 466.11 (acute bronchiolitis due to RSV), or 480.1 (RSV pneumonia) recorded in the first 7 of 15 potential diagnostic fields (to produce rates comparable with previous work1). Only 18 (<0.3%) additional cases would have been added if diagnoses from all 15 fields were used. Diagnoses for newborn hospitalization were excluded from the analysis.
In statistical analyses, the cases were stratified by patient ages <1 year and 1 to 4 years. For ensuring that observations most likely represented RSV, hospitalizations were limited to admissions for the 6 months of January through May and December (the Colorado RSV season19) for 5 consecutive years. As the CHA database did not include individual identifiers, it was possible that there were multiple admission records for a single RSV infection. We excluded duplicate admissions for patients who were discharged to an acute care facility by matching records on zip code, birth year, birth month, and gender. Of the original 7089 admission records identified, 113 duplicates from transfers were removed.
Geographic and Demographic Data
The geographic unit for this study was the 2000 US Census 5-digit zip codetabulation area (ZCTA), which approximates US Postal Service zip code boundaries. ZCTAs comprise smaller geographic areas than counties and therefore yield more precise altitude estimates.
Data for household income, household size, education level, and race/ethnicity were obtained for each ZCTA from the 2000 US Census. Because prison populations exerted a strong influence on data for socioeconomic status in some geographic areas,20 we adjusted for the presence of a prison within a ZCTA in our multivariate analyses. Data for 14 zip codes, including 4 cases, were excluded because they could not be located within the study ZCTAs. An additional 29 cases were excluded because they were located in ZCTAs with population estimates of 0 for their corresponding age group. Our geographic analysis was conducted with the remaining 6943 cases. Patients from zip codes without corresponding census demographics were reassigned to the study ZCTA that represented the same geographic area. This accounted for the reassignment of 38 zip codes with 141 cases.
We created a variable with the geographic information systems (GIS) to identify urban and rural ZCTAs. Census 2000 TIGER/Line files with Colorado ZCTA boundaries were overlaid with the Census 2000 Urbanized Areas file using ArcView 8.3 (ESRI Inc, Redlands, CA).
ZCTA boundaries that intersected urban area polygons were selected and coded as urban. The remaining was coded as rural.
The mean elevation for each ZCTA was estimated with the GIS using data from a US Geological Survey digital elevation model (2001) with elevations for the centroids of a grid composed of squares with 30-m sides. To minimize the effect on mean altitude of mountains within ZCTAs, we assumed that there were no residences above 3200 m and excluded altitudes above this level from the calculations of the mean altitude for ZCTAs. For stratified analyses, we used the following predetermined altitude categories: low (<1500 m), moderate (15002500 m), and high (>2500 m) elevations.
Statistical Analysis
Univariate and multivariate Poisson regression models were used to model the association between altitude and RSV-associated hospitalization rates in infants who were younger than 1 year and children who were 1 to 4 years of age. The ZCTAs were the unit of analysis. The dependent variable was the total number of RSV-associated hospitalizations during the 19982002 RSV seasons, counted separately for infants and young children, within each ZCTA. The corresponding 2000 mid-year population estimate of the number of infants and children who lived in each ZCTA were included as an offset variable to model the rate of RSV-associated hospitalizations. The analyses were performed using the SAS 8.2 Genmod procedure (SAS Institute Inc, Cary, NC). Altitude was examined as both a continuous and a categorical variable.
Potential confounders that were considered for inclusion in the multivariate models were differences in the underlying populations measured at the ZCTA level. These included the distribution (the proportion of individuals in each category of each factor) of gender, age, race, ethnicity, education level, and household size; the proportion of households that lived below 150% of the poverty level; the median income for the ZCTA; and indicator variables for whether the ZCTA was urban or contained a prison.
A backward elimination procedure from a model saturated with all potential confounders was used. During each step, the least significant variable was determined for removal using the likelihood ratio
2 test. For the 4 covariates defined by multiple variables (race, age, household size, and education level), the overall significance of the entire set was determined. Either the entire set remained or was removed from the model. The process continued until all remaining variables were significant at an
level of .05. Rate ratios and 95% confidence intervals were obtained for the outcomes of interest. The models were adjusted for overdispersion by multiplying the standard error of each coefficient by the square root of the Pearson
2 goodness-of-fit statistic divided by its degrees of freedom.21 This was performed using the PSCALE option in SAS.
| RESULTS |
|---|
|
|
|---|
|
Unadjusted and adjusted rates by the predetermined altitude exposure categories are reported in Table 2. Infants who were younger than 1 year had higher hospitalization rates compared with children who were aged 1 to 4 years. The highest unadjusted rate was for infants at the lowest altitude at (20.9 per 1000 infants per season). The lowest unadjusted rate for both age categories occurred in the moderate-altitude category.
|
The adjusted models suggested a positive linear relationship between rates of hospitalization and ZCTA elevation. However, when altitude was examined categorically, an increased risk was observed only for ZCTAs at the highest elevations, above 2500 m. Infants who were living in ZCTAs at high altitude had a 30% higher rate of hospitalization (RR: 1.30; P = .018) than those who were living at moderate elevation, and 1- to 4-year-olds who were living at high altitude had an 80% increase in their hospitalization rate (RR: 1.80; P < .001). Infants who resided at high altitude had a rate that was 22% higher (RR: 1.22; P = .122) than those at low altitude, whereas children at high altitude had a 62% increase in their rate of hospitalization (RR: 1.62; P = .004) when compared with residents at low altitude. There were no significant differences between hospitalization rates for those who were living at moderate altitude as compared with those who were living at low altitude for infants (RR: 0.94; P = .480) or for children (RR: 0.90; P = .393). Maps (Fig 1) display the geographic distribution of adjusted hospitalization rates by ZCTA, with the high-altitude region outlined.
|
1600 m, the altitude of metropolitan Denver, where more than half of Colorado's population resides. The rates did not begin to increase considerably until
2500 m, and additional analyses suggested that infants and children who were living in ZCTAs with mean elevations of 2750 m or above had the highest rates of RSV-associated hospitalization. Compared with infants who were living in ZCTAs below 2500 m, infants who were living in ZCTAs at 2500 to 2750 m had a 14% increased hospitalization rate (RR: 1.14; P = .331) and a 54% increased rate (RR: 1.54; P = .006) in ZCTAs above 2750 m. The observed association was stronger for children, with a 47% increase (RR: 1.47; P = .024) in hospitalization rate for ZCTAs at 2500 to 2750 m and more than doubling of rates (RR: 2.36; P = <.001) in ZCTAs above 2750 m.
|
| DISCUSSION |
|---|
|
|
|---|
The effect of altitude on hospitalization rate was stronger in children than in infants. The explanation for this observation may lie in the shorter length of exposure to altitude for infants, as compared with older children, and their subsequent ability to compensate for altitude exposure. In a small cohort of neonates who were born at 3100 m, Niermeyer et al23 described normal to moderately elevated indices for pulmonary artery pressures and lower arterial oxygen saturation readings in the first week of life. These values returned to normal ranges within a few months. However, because infants decompensate more readily with illness than older children, there may be a lower threshold for admission to a hospital and, hence, less of a difference in hospitalization rates at different altitudes for infants.
The normal physiologic effects of altitude on the respiratory tract of infants and children are multiple. Baseline oxygen saturation values decrease for infants and young children who live at increasing altitudes,2426 an expected response to lower barometric pressure and decreased fraction of inspired oxygen. Nasal obstruction and impaired ciliary activity, exhibited at high altitude,27 in conjunction with low humidity, can impair the ability to clear respiratory secretions associated with RSV. Hypoxemia stimulates pulmonary vasoconstriction, leading to increased vascular permeability and pulmonary congestion.28 This pulmonary vasoconstriction may be enhanced with RSV infection as illustrated in animal models.
RSV infection in a nonhuman primate model (the bonnet monkey Macaca radiata) demonstrates severe bronchiolar edema and alveolitis along with an exuberant peribronchovascular inflammation.29 In these animals, there seems to be pulmonary arteriolar hyperplasia 7 days after infection. These findings were subsequently described in newborn rats.30 Besides the pathologic changes, rat pups are much more susceptible to pulmonary edema when exposed to hypoxemia if the animals were infected with Sendai virus compared with controls (Sendai virus infection in rats mimics RSV infection in humans).
This observation has also been extended to children, and an association has been found with preexisting inflammatory illnesses including upper respiratory tract infection in children who develop high-altitude pulmonary edema at altitudes above 2500 m.31 Finally, a small but significant lower mean oxygen saturation was seen in infants and toddlers with upper respiratory infection at an altitude of 1500 m.32
For both infants and young children, we found no difference in adjusted rates of hospitalization at moderate-altitude compared with the low-altitude categories, yet the comparison between high and moderate altitude demonstrated a significantly elevated risk. The exploratory analysis to determine an altitude cut point for increased rates showed a significant relationship above 2500 m. Possibly, altitudes lower than 2500 m do not exert substantial physiologic effects to influence the course of RSV infection. The lowest mean elevation by ZCTA in Colorado is
1092 m.
The unadjusted rates were highest in the low-altitude category, which is the eastern portion of the state. Once adjusted, these rates were similar to those for the moderate-elevation category. The majority (84.6%) of Colorado's population lives in the urban, moderate altitude, where they likely have better access to health care and affiliated social services. Access to preventive and primary care health care services in low-income rural areas is limited. We attempted to adjust for increased distance from health care providers by creating a proxy for "rural" ZCTAs but may not have accounted for this confounding adequately.
The elevated incidence of unadjusted and adjusted RSV-associated hospitalizations in the low-altitude region of the state is somewhat puzzling. Multivariate analysis accounted for the racial, educational, and income differences. The disparity in the eastern region of the state (Fig 1) may be attributed to poorer baseline health status in the population. Both this region and the San Luis Valley, an area that is at moderate altitude and has elevated rates of RSV hospitalizations, have higher proportions of Hispanic individuals. These findings may be consistent with a culture-associated delay in access to care.3335 In addition, barriers reported by Hispanic parents to gaining access to care for children are transportation difficulties, language problems, and long wait times.35 Undocumented immigrants may be at additional risk for delays in gaining access to health care.
The population in the low-altitude region was poorer and less educated compared with populations in areas at higher elevations. The concentration of poverty and educational level was a significant contributor to the final models. Poverty on American Indian reservations10,15,36 has been implicated in higher rates of bronchiolitis hospitalizations. In other studies, however, low socioeconomic status and low maternal education have not been found consistently to be independent risk factors for RSV hospitalization.13,14,17
Limitations of the Study
The use of a GIS to assign mean altitudes for ZCTAs on the basis of self-reported zip codes for patients was dictated by privacy restrictions that were placed on the hospital discharge data. Access to residential address and more detailed demographic data for patients would have allowed us to assign accurate altitudes to residences and analyze the data set with individual records, rather than using altitude and demographic data for ZCTAs as surrogates for data specific to individual patients. Because of these restrictions, this study would not have been possible without a GIS.
We were unable to account for household smoke and ambient levels of air pollutants as potential confounders. We did not adjust for preexisting medical conditions, although it seems unlikely that young children with chronic heart or lung conditions would live in greater proportions at higher elevations. We partially addressed the possible confounding of premature births by excluding hospital admissions for newborns. We could not account directly for differences in physician admitting practices. Rural practitioners may have a lower threshold for hospital admission for infants and young children who live some distance from emergency facilities. Adjusting for residence in rural and urban areas may not have accounted adequately for this possible confounding.
In this study, we used ICD-9-CM codes for identifying RSV infections, with no information on whether the diagnoses were confirmed by viral identification. Diagnosis of RSV infection without laboratory confirmation may be related to other factors that are correlated with altitude, such as income, race, and health insurance coverage. These factors may not have been accounted for adequately with our demographic variables.
| CONCLUSIONS |
|---|
|
|
|---|
| ACKNOWLEDGMENTS |
|---|
| FOOTNOTES |
|---|
Address correspondence to Eric A.F. Simoes, MB, BS, DCH MD, Children's Hospital, Section of Pediatric Infectious Diseases, 1056 E 19th Ave, Box B055, Denver, CO 80218. E-mail: eric.simoes{at}uchsc.edu
The authors have indicated they have no financial relationships relevant to this article to disclose.
| REFERENCES |
|---|
|
|
|---|
This article has been cited by other articles:
![]() |
M Yanney and H Vyas The treatment of bronchiolitis Arch. Dis. Child., September 1, 2008; 93(9): 793 - 798. [Abstract] [Full Text] [PDF] |
||||
![]() |
J. M. Mansbach, S. Clark, N. C. Christopher, F. LoVecchio, S. Kunz, U. Acholonu, and C. A. Camargo Jr Prospective Multicenter Study of Bronchiolitis: Predicting Safe Discharges From the Emergency Department Pediatrics, April 1, 2008; 121(4): 680 - 688. [Abstract] [Full Text] [PDF] |
||||
![]() |
L. O'Keefe Special article in Pediatrics outlines laws affecting children of same-gender parents AAP News, July 1, 2006; 27(7): 1 - 21. [Full Text] |
||||
Read all P3Rs
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||