Objectives. New interventions to prevent respiratory syncytial virus (RSV) have recently become available. Clinical decisions about the use of these interventions require a better understanding of the incidence of and risk factors for RSV. We sought to characterize the epidemiology of severe RSV disease among premature infants and to identify high-risk subgroups.
Design. Retrospective cohort.
Setting. Kaiser Permanente Northern California, July 1992 to April 1996.
Participants. One thousand seven hundred twenty-one premature infants born at 23 to 36 weeks who were discharged from a neonatal intensive care nursery (NICU) within 12 months before the December to March RSV season. A secondary analysis included 769 infants discharged during the RSV season.
Outcome Measures. Hospitalization for RSV.
Results. Of 1721 infants already home from the NICU at the start of the season, 3.2% were rehospitalized for RSV. In a multivariate model, risk factors for RSV hospitalization included gestation ≤32 weeks (odds ratio [OR], 2.6), ≥28 days of perinatal oxygen (OR, 3.7), and NICU discharge during September to November (OR, 2.7). Predicted risk of hospitalization varied by subgroup, ranging from 1.2% to 24.6%. Among 769 infants discharged from the NICU during the RSV season, 3.5% were rehospitalized for RSV during the same season; gestation and perinatal oxygen were not associated with admission.
Conclusions. Most premature infants in this population were at less risk of severe RSV disease than previous studies in other populations have suggested. Preterm infants with a lower gestational age, a prolonged perinatal oxygen requirement, and NICU discharge within 3 months of the RSV season were most likely to require hospitalization for RSV disease. Cost-effectiveness analyses are needed to help define the role of available prophylactic interventions.
- respiratory syncytial virus infections
- neonatal infections
- intensive care
- bronchopulmonary dysplasia
- chronic lung disease
- seasonal variation
- passive immunization
- RSV =
- respiratory syncytial virus •
- CLD =
- chronic lung disease •
- NICU =
- neonatal intensive care unit •
- KPNMDS =
- Kaiser Permanente Neonatal Minimum Data Set •
- KPMCP =
- Kaiser Permanente Medical Care Plan (Northern California hospitals) •
- ICU =
- intensive care unit •
- DFA =
- direct fluorescent antibody •
- OR =
- odds ratio •
- CI =
- 95% confidence interval
Respiratory syncytial virus (RSV) is the most important cause of respiratory illness among young infants.1 Infants with chronic lung disease (CLD), prematurity without CLD, and/or young chronologic age are at increased risk of severe RSV disease.2–11 Hospitalization rates for RSV in studies of premature infants with or without CLD have ranged from 2.7% to 37%, with variations explained primarily by differences in the populations studied.2 ,3 ,8 ,10 ,11 However, most studies have been small, weighted toward infants with extreme prematurity and/or CLD, and based primarily at tertiary-care centers. Furthermore, much of our information about the risk of RSV among high-risk infants comes from the control arms of clinical trials, which may not accurately reflect the clinical setting.3 ,8 ,12 No large, population-based cohort study of RSV disease among premature infants has been published, and only one study permits estimation of the risk of RSV among infants with a moderate degree of prematurity, who constitute the majority of all preterm neonates.11 In addition, no study has attempted to separate the effects of degree of prematurity, CLD, and young age on subsequent RSV-related morbidity. A better understanding of the population-based incidence of and risk factors for severe RSV disease is needed to make clinical policy about new interventions.
The recent approvals of RSV immune globulin (Respigam; MedImmune, Inc, Gaithersburg, MD) and humanized monoclonal antibody (palivizumab, Synagis; MedImmune, Inc, Gaithersburg, MD) for prevention of RSV disease among premature infants with or without CLD have provided the first effective prophylactic interventions for high-risk groups. The American Academy of Pediatrics has recommended that prophylaxis be considered for infants and children with CLD who are <2 years of age and are currently receiving or have received supplemental oxygen within the previous 6 months; for infants born at ≤28 weeks who are up to 12 months of age; and for infants born at 29 to 32 weeks who are up to 6 months of age.13 However, the substantial cost of these interventions constrains their wide application. The American Academy of Pediatrics and others have recommended that identification of the subgroups of premature infants most likely to benefit from prophylaxis should be a research priority.14 ,15
The purpose of this population-based retrospective cohort study of a large group of premature infants is to describe how the risk of hospitalization for community-acquired RSV varies with gestational age, duration of perinatal oxygen requirement, and timing of discharge from the neonatal intensive care unit (NICU). These data are necessary to refine recommendations for the use of available prophylactic interventions against RSV disease.
MATERIALS AND METHODS
We used the Kaiser Permanente Neonatal Minimum Data Set (KPNMDS) to develop a retrospective cohort of preterm infants. The KPNMDS is a dedicated research database that includes clinical and demographic data, collected prospectively by trained research assistants through medical record review, on infants admitted to one of the six primary Kaiser Permanente Medical Care Plan (KPMCP) NICUs in Northern California after July 1, 1992.16 Currently, the KPNMDS captures 100% of Level III (comprehensive perinatal care services for mothers and neonates of all risk categories) and >80% of Level II (neonates with a moderate degree of illness) NICU admissions in the KPMCP Northern California hospitals, a region that serves ∼2.8 million members with an annual birth cohort of 30 000 infants.17
Eligibility criteria for entry into the study consisted of: 1) inclusion in the KPNMDS; 2) gestational age ≤36 weeks; 3) discharge alive from the NICU between July 1, 1992 and March 31, 1996; 4) no diagnosis of congenital heart disease other than patent ductus arteriosus; 5) no diagnosis of cystic fibrosis; and 6) no diagnosis of congenital or acquired immunodeficiency. These comorbidities were excluded to isolate the effect of prematurity and related conditions on subsequent RSV disease, and because the use of prophylaxis is currently not recommended in patients with congenital heart disease.14 No infant in the cohort received RSV prophylaxis.
Gestational age, birth weight, sex, maternal race, dates of nursery admission and discharge, lengths of oxygen and ventilator therapies in the NICU, and presence of multiple gestation were obtained from the KPNMDS database. We defined the RSV season, based on a review of the distribution of hospitalizations, as beginning on December 1 of each year and ending on March 31 of the next year (Fig 1). Our primary analysis (the full-season cohort) assigned infants discharged from the nursery between December 1 of a given year and November 30 of the subsequent year to the following RSV season (ie, an infant discharged from the NICU on December 20, 1993 would be assigned to the December 1, 1994 to March 31, 1995 season). This was done to simulate a model prospective cohort study, in which all infants discharged from the NICU during the year before the start of the RSV season would be followed for RSV hospitalizations for one complete season. Our secondary analysis (the partial-season cohort) was restricted to the subset of infants discharged from the NICU during an RSV season (ie, December 1 to March 31), and included hospitalizations from the date of NICU discharge to the end of the concurrent RSV season.
To determine loss to follow-up (ie, outmigration from the Kaiser Health Plan), we linked the cohorts to the KPMCP membership database. This database permitted determination, for an individual member, of dates of enrollment in the Kaiser Health Plan. Infants were included in the cohort if they were considered to have complete follow-up (ie, if they were continuously enrolled during the RSV season of interest). This was done to avoid including in the denominator infants who, because of disenrollment, would not have been hospitalized at a Kaiser facility had they developed RSV. Depending on enrollment, an infant could be included in the full-season analysis, the partial-season analysis, or both. We were not able to attempt to contact the families of infants who, because of disenrollment, were excluded from the analysis.
We linked the cohort of eligible infants to the KPMCP hospitalization database to identify admissions potentially related to RSV. All admissions between November 1 and April 30 for which an acute respiratory or nonspecific viral diagnosis was listed among the 12 discharge diagnoses in the electronic record were selected for chart review (International Classification of Diseases, 9th Revision, Clinical Modification [ICD9-CM] codes: acute bronchitis and bronchiolitis [466.0–466.1]; pneumonia [480.0–487.8; includes pneumonia secondary to RSV]; asthma/reactive airways disease exacerbation [493.0–493.9]; acute nasopharyngitis ; acute laryngitis and tracheitis [464.0–464.4]; acute upper respiratory infection of multiple or unspecified sites [465.0–465.9]; chronic airway obstruction ; pneumothorax [512.0–512.8]; pulmonary collapse [518.0]; interstitial emphysema [518.1]; other diseases of lung [518.8–518.81]; other respiratory symptoms [786.09]; specified and unspecified viral and chlamydial infections [079.88–079.99]; asphyxia and respiratory arrest [799.0–799.1]).18 To ensure complete case ascertainment, we also linked the cohort to the KPMCP databases that track out-of-plan hospitalization. We did not attempt to identify ambulatory episodes of RSV disease. There were no practice guidelines at KPMCP that might have systematically altered the treatment of RSV-infected infants.
One of us (S.J.) reviewed the medical records of all hospitalizations matching any of the specified ICD9-CM codes to confirm the diagnosis of RSV and to describe the course of the illness. Transfers outside KPMCP facilities were reviewed as well. Charts were abstracted using a standardized protocol and abstraction form. Information recorded included demographic and administrative data, preexisting conditions, timing and results of RSV tests, durations of ICU care, mechanical ventilation and supplemental oxygen, therapeutic interventions, and clinical diagnosis of bronchiolitis. Where a hospitalization consisted of two or more discrete segments (ie, interhospital transfer), the durations of hospitalization, ICU care, and so forth, were manually summed. Readmissions for the same problem within 24 hours of discharge were considered to be continuations of the initial hospitalization. If RSV infection was incidental to rather than the cause of the admission, the hospitalization was not considered as a case.
We defined an illness as attributable to RSV if the infant was hospitalized for a respiratory indication and an RSV direct fluorescent antibody (DFA) test performed between 7 days before and 3 days after admission was positive. RSV testing was performed at the discretion of the treating clinician; no institutional policy was in place. Viral cultures were rarely performed. All RSV tests were performed at the KPMCP regional virology laboratory. Before May 1994, the Ortho DFA reagent (Ortho Diagnostic Systems, Inc, Raritan, NJ) was used. Since May 1994 the Bartels DFA test kit has been used (Bartels, Inc, Issaquah, WA).
The outcome of interest was hospitalization for laboratory-proven RSV. Infants with two or more admissions for RSV during the same season were considered to have a single positive outcome. Primary risk factors evaluated included gestational age (≤28 weeks, 29–32 weeks, and 33–36 weeks), oxygen therapy in the NICU for ≥28 days, and month of NICU discharge. Oxygen therapy for ≥28 days was chosen as a proxy for CLD, because there was no systematic way to identify which infants met other clinical criteria for CLD.19 In particular, there was no registry of infants who were currently receiving, or had previously received, home oxygen therapy. We chose time since NICU discharge, stratified into 3-month blocks, as our measure of an infant's maturity because of its ease of recall and because, in contrast to chronologic age, it was uncorrelated with gestation. For example, most infants born at ≤28 weeks remained in the NICU for >3 months. Had we considered chronologic age, there would have been virtually no infants in the cohort who were born at ≤28 weeks and were ≤3 months old at the start of the RSV season. Additional risk factors considered were birth weight, length of mechanical ventilation in the NICU, sex, maternal race, and multiple gestation.
Statistical analysis was performed using STATA 4.0 for Windows (Stata Corp, College Station, TX). The full- and partial-season cohorts were analyzed separately. Bivariate analyses were performed using Pearson χ2 or Student's t test, unless otherwise indicated. All bivariate comparisons were two-tailed. Backwards stepwise multiple logistic regression was then used to evaluate the simultaneous role of multiple risk factors and to assess interactions. Predictor variables that were significant at P < .10 in the bivariate analyses were included in the initial regression model; variables that were not significant at P < .05 were sequentially dropped. Interaction terms between the three primary risk factors were also evaluated.
The study was approved by the KPMCP Institutional Review Board.
Infants Discharged Before the Beginning of the RSV Season
Of the 2341 eligible infants, 26.5% were excluded because of disenrollment during the season of interest. Singletons were more likely than infants from a multiple gestation to be excluded (27.5% vs 22.8%, P = .03). In addition, black infants (39.6%,P < .01) and Latino infants (30.4%, P= .04) were more likely than white infants (24.8%) to be excluded, whereas Asian-American infants (18.2%, P = .02) were less likely than whites to be excluded. Included and excluded infants did not differ by gestation, birth weight, length of oxygen or ventilator therapies in the NICU, sex, or timing of NICU discharge.
Among the 1721 included infants, 55 (3.2%) were hospitalized for laboratory-proven RSV disease between December 1 and March 31. Three infants were hospitalized twice for RSV, and 1 infant was hospitalized three times, for a total of 60 RSV-related hospitalizations. Thirty admissions for a respiratory illness were associated with a negative RSV test, and no RSV test was performed during an additional 9 respiratory admissions. Thus we found a total of 99 hospitalizations for respiratory illness. There were no episodes of nosocomial RSV.
Compared with infants who were not hospitalized for RSV, hospitalized infants were more likely to have been ≤32 weeks' gestation, to have required ≥28 days of supplemental oxygen in the NICU, to have had a lower birth weight, to have required longer periods of mechanical ventilation in the NICU, and to have been discharged from the NICU during the months of September, October, or November. Among infants with gestation ≤32 weeks, there was a nonsignificant trend toward increased hospitalization for RSV among those infants born at ≤28 weeks. This trend disappeared, however, when the comparison between infants born at 23 to 28 weeks and those born at 29 to 32 weeks was stratified by the presence of an oxygen requirement for ≥28 days (Mantel-Haenszel relative risk = 0.9, P = .8). Race, sex, and multiple gestation did not differ between hospitalized and nonhospitalized infants (Table 1).
In the logistic regression model, infants with gestational age 23 to 32 weeks were at greater risk of hospitalization for RSV than those with gestation of 33 to 36 weeks (odds ratio [OR] = 2.6; 95% confidence interval [CI], 1.4–5.1; P = .003). An oxygen requirement in the NICU for ≥28 days was independently associated with an increased likelihood of hospitalization (OR = 3.7; 95% CI, 1.8–7.6; P < .001), compared with infants who required <28 days of oxygen. Discharge from the NICU during the months of September to November was also a predictor of admission for RSV (OR = 2.7; 95% CI, 1.6–4.7; P< .001), compared with discharge between December and August. Birth weight and length of mechanical ventilation in the NICU were not risk factors in the multivariate model. No significant interactions among length of oxygen therapy, gestation, and month of NICU discharge were found. Predicted probabilities of hospitalization for RSV are shown inTable 2. The results of the analysis were virtually unchanged when only infants with gestation ≤35 weeks were considered (data not shown).
Infants Discharged During the RSV Season
Between December and March, 950 eligible infants were discharged from the NICU. Of these, 19.1% had a period of disenrollment between the date of discharge from the NICU and the end of the concurrent RSV season and were considered lost to follow-up. More black than white infants were excluded (36.5% vs 15.8%, P < .01), whereas there was no difference among Asian-American, Latino, and white infants. In addition, more singletons than infants from a multiple gestation were considered lost to follow-up (21.3% vs 10.8%,P < .01). Included and excluded infants did not differ by gestation, birth weight, length of oxygen or ventilator therapies in the NICU, sex, or timing of NICU discharge.
Among 769 infants with complete follow-up who were discharged during the RSV season, 27 (3.5%) were admitted before March 31 for laboratory-proven RSV disease. No infant had more than one hospitalization for RSV. Twenty-five additional infants admitted for respiratory illness had a negative RSV DFA test, and 4 infants hospitalized for respiratory disease were not tested for RSV, for a total of 56 hospitalizations for respiratory illness. No infant had nosocomial RSV.
In contrast to our findings among infants discharged before the start of the season, the risk of hospitalization for RSV among infants discharged during the December 1 to March 31 season was not influenced by gestational age, birth weight, length of mechanical ventilation, or length of oxygen therapy in the NICU. The only factor associated with increased risk was discharge during the first half of the RSV season (6.1% vs 1.8%, P < .001, Fisher's exact test). Race, sex, and multiple gestation did not differ between infants hospitalized for RSV and those not hospitalized for RSV.
Severity of Illness
The median length of stay for RSV disease was 4 days. Infants who were discharged from the NICU and rehospitalized during the same RSV season tended to have longer admissions than those discharged before the beginning of the season. Infants discharged during the season also required more days of oxygen therapy for RSV, had a greater likelihood of admission to the ICU, and were more likely to require mechanical ventilation for RSV disease than those discharged before the start of the season (Table 3). Meaningful comparisons of duration of ICU care and mechanical ventilation were limited by the small number of infants who required these interventions. Infants requiring ICU admission for RSV had been discharged from the NICU more recently (mean, 66 days vs 143 days;P = .01) than those not requiring ICU care. There were no deaths from RSV.
The current study of RSV disease among premature infants is the first to use a population-based cohort of infants that fairly represents the spectrum of prematurity in the United States. In this large group of infants, we have shown that three factors independently increased the risk of hospitalization for RSV: 1) gestational age ≤32 weeks, 2) a perinatal oxygen requirement of ≥28 days, and 3) discharge from the NICU during the 3 months before the RSV season. Once a prolonged oxygen requirement was included in the analysis, there was no further effect of increasing prematurity in the group ≤32 weeks. This is consistent with the findings of the IMpact-RSV trial.12
Overall, the probability of rehospitalization in this study among most subgroups was relatively low. This is compatible with the results of Nachman et al.11 Among our higher-risk subgroups, the likelihood of hospitalization was consistent with previous cohorts that focused on sicker patients.8 ,9 ,12 The more current rates are lower than others previously reported,2 ,3 possibly because of referral bias in the previous studies and secular decreases in hospitalization rates during the last decade.20
Among infants discharged from the NICU during the RSV season, only discharge during the first half of the season increased the risk of hospitalization. This was presumably as a result of exposure to the January to February peak.1 Neither gestational age nor length of oxygen therapy in the NICU was related to risk of hospitalization for RSV. These surprising findings, which we plan to retest in a subsequent KPNMDS cohort, contrast with the strong associations among increasing prematurity, prolonged oxygen dependence, and hospitalization for RSV among infants who were discharged before the start of the season. The unexpected absence of hospitalization for RSV among extremely premature infants makes us hesitate to draw conclusions about readmission in the group of infants discharged from the NICU during the RSV season.
Although infants discharged from the NICU during the RSV season had a similar risk of hospitalization to that among infants discharged before the start of the season, they were more likely to require intensive care and mechanical ventilation. The lack of association between severity of illness and gestational age or length of oxygen requirement in the NICU was notable.
We made every effort to limit sources of bias. To ensure an accurate denominator, we included only those infants who had continuous membership in the KPMCP. Although the number of infants excluded from the study raises the possibility of selection bias, observed differences between included and excluded infants were few and were unrelated to risk of hospitalization within the study cohort. Although unmeasured differences, such as lower socioeconomic status, may have resulted in a higher risk of hospitalization among excluded infants, the effect on the study's findings could not have been large. For example, even if excluded infants were at 50% greater risk than included infants, the likelihood of hospitalization in the highest-risk subgroup would only have increased from 24.6% to 27.8%.
As a result of the population-based nature of the KPNMDS, there is little referral bias in the sample. However, the demographic characteristics of KPMCP members may differ from those of the general population because infants of lower socioeconomic status may be underrepresented.21 Such infants may be at increased risk of RSV.1 Possible differences between KPMCP members and nonmembers and between included and excluded infants are an important potential limitation to the generalizability of the study.
Several factors may have resulted in missed cases, thereby lowering the apparent risk of rehospitalization for RSV in the KPMCP cohort. First, RSV tests were not performed in 7% to 9% of respiratory admissions. Second, the Bartels DFA test kit used in most patients has a 90% to 95% sensitivity in clinical practice, resulting in occasional false-negative tests.22 ,23 Third, infants with duplicate health insurance might, unbeknownst to us, have had hospitalizations for RSV at non-KPMCP facilities that were covered by other payers. A 1992 internal KPMCP survey of parents of 13-month-old infants revealed 5% to 10% double coverage. Fourth, our decision to exclude November and April from our definition of the RSV season resulted in the exclusion of 3% of cases from the analysis (Fig 1). Fifth, although we selected for chart review hospitalizations with a wide range of potentially related diagnostic codes, the use of administrative data to identify potential hospitalizations for RSV may have resulted in a few missed cases because of coding errors. We estimate that accounting for all five factors would result, at most, in a 15% to 20% increase in the proportion of infants hospitalized for RSV, and that this effect would be uniform across risk categories.
Our records did not permit determination of which infants had clinically active or recent CLD at the start of the RSV season. We were therefore unable to determine the probability of and risk factors for readmission for RSV among infants with CLD. Although we used ≥28 days of perinatal oxygen as a proxy for CLD, not all infants who require ≥28 days of oxygen in the NICU ultimately meet criteria for CLD.19 These results should therefore be used with caution to guide recommendations for the use of RSV prophylaxis among infants with CLD.
The observation that discharge from the NICU between September and November placed infants at increased risk of RSV should be interpreted with attention to local context. The RSV season in Northern California lasted from December to March during each of the 4 study years. This is compatible with the findings of Gilchrist et al.24 Because the timing of the RSV season may differ significantly from place to place and, to a lesser degree, from year to year, the particular 3-month period of NICU discharge associated with elevated risk of subsequent hospitalization for RSV will likely vary as well.
Most preterm infants are at lower risk of rehospitalization for RSV than previously believed. Infants with a lower gestational age (≤32 weeks), a prolonged perinatal oxygen requirement (≥28 days), and NICU discharge ≤3 months before the start of the RSV reason are most likely to develop severe RSV disease. Infants with all three risk factors have a predicted probability of hospitalization for RSV of 25%, whereas the likelihood of hospitalization for RSV among other groups of premature infants ranges from 1% to 11%. The identification of these risk factors should enable interventions including RSV immune globulin and palivizumab, as well as future vaccines against RSV, to be targeted to those infants most likely to develop severe disease. To permit clinical policy recommendations, however, the relative benefits, risks, and costs of RSV prophylaxis for these subgroups should ideally be evaluated in a formal cost-effectiveness analysis.
Funding for the Neonatal Minimum Data Set database was provided by the Kaiser Foundation Health Plan, Inc (Grant 115-6137).
We thank Veronica M. Gonzales and Marla N. Gardner for their editorial assistance and Priscilla T. Branch for her programming and help in preparing datasets for analysis.
- Received December 3, 1998.
- Accepted March 25, 1999.
Reprint requests to (G.J.E.) Division of Research, Kaiser Permanente Medical Care Program, 3505 Broadway, Oakland, CA 94611. E-mail:
Steven Joffe is currently at the Division of Pediatric Hematology/Oncology, Children's Hospital, and the Dana-Farber Cancer Institute, Boston, MA. Dr Lieu is currently at the Division of Ambulatory Care and Prevention, Harvard Community Health Plan and Harvard Medical School, Boston, MA.
- Groothuis JR,
- Gutierrez KM,
- Lauer BA
- Groothuis JR,
- Simoes EAF,
- Hemming VG,
- et al.
- Cunningham CK,
- McMillan JA,
- Gross SJ
- Law BJ,
- De Carvalho V,
- and the Pediatric Investigators Collaborative Network on Infections in Canada
- The PREVENT Study Group
- Wang EEL,
- Law BJ,
- Stephens D,
- et al.
- ↵Wang EEL, Law BJ, Robinson JL, et al. PICNIC (Pediatric Investigators Collaborative Network on Infections in Canada) study of the role of age and respiratory syncytial virus neutralizing antibody on respiratory syncytial virus illness in patients with underlying heart or lung disease. Pediatrics. 1997;99(3). URL: http://www.pediatrics.org/cgi/content/full/99/3/e9
- ↵Nachman S, Navaie-Waliser M, Qureshi M. Rehospitalization with respiratory syncytial virus after neonatal intensive care unit discharge: a 3-year follow-up. Pediatrics. 1997;100(6). URL: http://www.pediatrics.org/cgi/content/full/100/6/e8
- The IMpact-RSV Study Group
- American Academy of Pediatrics, Committee on Infectious Diseases and Committee on Fetus and Newborn
- American Academy of Pediatrics, Committee on Infectious Diseases, Committee on Fetus and Newborn
- Storch GA
- Escobar G,
- Fischer A,
- Kremers R,
- et al.
- ↵Freeman RK, Poland RL, eds. Guidelines for Perinatal Care. 3rd ed. Elk Grove Village, IL: American Academy of Pediatrics and American College of Obstetricians and Gynecologists; 1992
- ↵International Classification of Diseases, 9th Revision, Clinical Modification. 5th ed. New York, NY: McGraw-Hill, Inc; 1996
- ↵US Bureau of the Census. Statistical Abstract of the United States: 1995. 115th ed. Washington, DC: US Government Printing Office; 1995
- ↵Tison DL, Blixt D, Jouglard J. Performance of a commercial fluorescent antibody conjugate for the direct identification of respiratory syncytial virus from nasopharyngeal secretions. Abstracts of the Annual Meeting of the American Society of Microbiology. 1987;C52
- Halstead DC,
- Todd S,
- Fritch G
- Copyright © 1999 American Academy of Pediatrics