Objectives: The goal of this study was to describe hospitalizations of infants during the first year of life according to week of gestational age (GA). We hypothesized that odds of any hospitalization would generally decrease with increasing GA, with late preterm infants experiencing additional increased risk of specific hospitalizations, such as hyperbilirubinemia.
Methods: Birth certificates for >6.6 million infants born in California hospitals between 1993 and 2005 and surviving to discharge were linked to hospital discharge records during the first year of life. Odds of any hospitalization and any hospitalization for specific diagnoses during the first year of life were determined for infants 23 to 44 weeks’ GA. Further analysis determined odds of any hospitalization within 14, 30, and 90 days of birth discharge, and observed odds were compared with expected odds obtained through quadratic modeling.
Results: Odds of any hospitalization within the first year of life decreased with advancing GA, but observed odds of any hospitalization exceeded expected odds for 35-, 36-, and 37-week GA infants for all time periods after discharge. Odds of any hospitalization for hyperbilirubinemia were greatest for infants 33 to 38 weeks’ GA (peak odds ratio at 36 weeks’ GA: 2.86 [95% confidence interval: 2.73–3.00]), and a relative peak in odds of any hospitalization for specific infections was observed among infants 33 to 36 weeks’ GA.
Conclusions: Odds of any hospitalization during the first year of life exceeded expected odds of hospitalization for 35-, 36-, and 37-week GA infants. GAs at risk overlapped with, but were not identical to, GAs identified as late preterm infants.
Late preterm infants (LPIs), born between 34 0/7 and 36 6/7 weeks’ gestational age (GA), account for >70% of preterm deliveries in the United States.1 A growing body of literature raises concerns about increased morbidity and mortality in this population during birth hospitalization.2–11 Studies focusing on more long-term outcomes have identified an increased risk of hospitalization and increased associated costs for LPIs compared with term infants (TIs) within the first 14 days after discharge,12–14 28 days after discharge,15,16 and 1 year after discharge.6,14 Most previous studies, however, have compared all LPIs with all TIs, which does not allow a full understanding of the impact of each week of gestation. Recent studies have explored birth hospitalization outcomes according to week of GA and found that each additional week of gestation results in decreased risk of death,9,17 decreased need for respiratory intervention,7,9 and decreased neonatal morbidity.2,7–9 Just as these studies demonstrate differences in birth hospitalization outcomes with each additional week of gestation, examining hospitalizations after birth discharge according to week of GA may clarify the relationship between GA and adverse outcomes after birth hospitalization.
In this study, we examined hospitalizations during the first year of life of early preterm infants (EPIs), LPIs, and TIs according to week of GA to gain a more complete understanding of risks after birth discharge experienced by infants categorized as LPIs in the context of all other GAs. We hypothesized that the odds of any hospitalization during the first year of life after birth discharge would generally decrease with advancing GA. In addition, we examined odds of any hospitalization for specific diagnoses and within specific time periods from discharge, hypothesizing that infants 34 to 36 weeks’ GA will experience greater odds of any hospitalization for hyperbilirubinemia compared with other GAs (due to LPIs’ increased physiologic risk of hyperbilirubinemia compared with TIs, combined with their earlier discharge after birth compared with EPIs) and exploring the impact of GA on other specific diagnoses. Our goal was to describe and contextualize hospitalization risk experienced by 34-, 35-, and 36-week GA infants during the first year of life to further inform policy and interventions targeting these infants.
Study Population and Data Collection
Birth certificates for infants born in California between 1993 and 2005 and surviving to discharge (N = 6 765 494) were linked to discharge records for hospitalizations during the first year of life by the Office of Statewide Health Planning and Development by using previously described methods; the result was successful linkage in >98% of records.18 Infants were excluded if GA was <23 or >44 weeks (116 864 excluded), maternal age was <12 years or >50 years (510 excluded), or birth weight was >5 SDs from expected birth weight for reported GA (1843 excluded) due to concern for data error. GA was determined according to the best estimation of the obstetrician as recorded in the birth certificate data.
This project was approved by the institutional review boards of The Children’s Hospital of Philadelphia and the Department of Health in California.
The primary outcome was experiencing any hospitalization between discharge from birth hospitalization and day 365 of life. We focused on any hospitalization to avoid having a few infants with extreme numbers of admissions dominate our results. Secondary outcomes included any hospitalization within 14, 30, and 90 days of birth discharge to explore the time frame of postdischarge risk.
In addition, the occurrence of any hospitalization within specific diagnostic categories was examined according to GA and by timing of hospitalization. Each hospitalization was coded as surgical if a surgical procedure was listed as the primary procedure for that admission by using International Classification of Diseases, Ninth Revision, Clinical Modification codes.19 Among remaining nonsurgical hospitalizations, primary diagnoses accounting for ≥0.1% of hospitalizations were reviewed. These 121 diagnoses were grouped into categories. Categories accounting for >2.5% of hospitalizations were kept independent, whereas all remaining nonsurgical hospitalizations were counted as “other.” After this process, 81% of hospitalizations were specifically categorized, leaving 19% of hospitalizations identified as other (Appendix).
When analyzed as GA groups, EPIs were born ≤33 weeks, LPIs 34 to 36 weeks, and TIs ≥37 weeks. Additional predictive factors were included based on possible association with increased hospitalizations,20–25 including infant gender, multiple gestation, small for GA status, maternal age (adolescent [<20 years old] versus adult [≥20 years old]), maternal parity, maternal race, trimester of prenatal care initiation, insurance, median neighborhood income (based on zip code), and birth year.
Analysis was performed by using Stata version 10 (Stata Corp, College Station, TX). For bivariate analysis, the Kruskal-Wallis test was used for continuous variables and the χ2 test for categorical variables.
For multivariate analysis, logistic regression models were constructed to determine odds of experiencing any hospitalization according to GA within specific time frames or for specific reasons after controlling for other infant and maternal characteristics. Independent variables were included in multivariable models if inclusion improved the new model’s likelihood ratio by a significance level of 10%26 or based on a priori determination of potentially relevant factors. Final models included GA, maternal age, infant gender, multiple gestation, maternal parity, maternal education, maternal race, insurance status, small for GA, median neighborhood income, and year of birth. All SE calculations accounted for clustering according to birth hospital by using the Huber-White sandwich estimator. We did not control for birth hospitalization length or comorbidities due to concern that such markers of illness severity are mediators of GA’s impact on hospitalizations and that including these factors could inappropriately minimize the effect of GA on hospitalization risk. We determined expected odds ratios (ORs) for each GA by using quadratic modeling with GA and GA-squared as predictor variables, and we calculated observed/expected ORs for each GA.
General characteristics of the study population according to GA groups are reported in Table 1. The study comprised 6 646 277 infants, including 517 879 LPIs. Compared with TIs, EPIs and LPIs were more likely to be of multiple gestation births. Compared with mothers of TIs, more mothers of EPIs and LPIs were adolescents, had not received first-trimester prenatal care, and had not pursued a college education. Sixteen percent of EPIs experienced ≥1 hospitalization during the first year of life compared with 12.5% of LPIs and 9.5% of TIs (P < .001). During the first 14 days after discharge, EPIs and LPIs experienced similarly increased rates of hospitalization (3.7% and 3.6%, respectively) relative to TIs (2.2%; P < .001).
Odds of any hospitalization for each GA within the first year of life and within 14, 30, and 90 days of birth discharge relative to the 40-week GA reference group after controlling for infant and maternal characteristics are depicted in Fig 1. For all lengths of observation, the most premature infants experienced 3 to 4 times increased odds of any hospitalization compared with 40-week GA infants. In general, ORs for hospitalization declined steadily with increasing GA, with the exception of 34 to 36 weeks’ GA, in which ORs plateaued or even rose with increasing GA. This finding was most notable during the immediate postdischarge period; ORs for any hospitalization within 14 days of discharge declined from 22 to 33 weeks’ GA, then increased to a peak at 36 weeks’ GA (OR: 1.98 [95% confidence interval (CI): 1.91–2.06]) before declining again to the 40-week GA reference group.
As Fig 2 illustrates, infants 35, 36, and 37 weeks’ GA experienced odds of any hospitalization during all observation periods that were significantly increased from odds predicted with quadratic modeling. Calculating observed/expected odds of hospitalization for each GA (Fig 3) demonstrated that 35-, 36-, and 37-week GA infants experienced excess odds of any hospitalization relative to predicted odds within 14, 30, 90, and 365 days, with 36-week GA infants experiencing observed odds of any hospitalization 39% higher than predicted within 14 days of birth discharge.
Figure 4 illustrates odds of any hospitalization for specific reasons within 365 days of life according to GA. Two different trends can be observed. For respiratory, infectious respiratory, gastrointestinal (GI) and feeding-related, surgical, and other hospitalizations, odds of hospitalization declined steadily with increasing GA. Although less than the odds for more preterm infants, odds of hospitalization for 34-, 35-, 36-, and even 37-week GA infants remained significantly higher than the 40-week GA reference group for most of these diagnoses.
In contrast, odds of any hospitalization for hyperbilirubinemia and, to a lesser degree, bacterial infections and nonspecific infections demonstrated peaks including LPI GAs. Odds of any hospitalization for hyperbilirubinemia for infants born at 33 to 38 weeks’ GA exceeded odds for more preterm and TIs, peaking at 36 weeks’ GA (OR: 2.86 [95% CI: 2.73–3.00]). Similarly, for nonspecific infections (including “fever,” “perinatal infection,” and “viral infection”), odds of hospitalization for the most preterm infants were not statistically different from odds for the 40-week GA reference group but increased for 33- to 36-week GA infants, with a peak at 34 weeks’ GA (OR: 1.11 [95% CI: 1.04–1.18]). Hospitalization for any bacterial infections (involving either specific bacterial organisms, such as streptococcal bacteremia, or specific bacterial infections, such as pyelonephritis) demonstrated increased odds among the most preterm infants and then again among infants 33 to 36 weeks’ GA, with a relative peak at 33 weeks’ GA reflecting 12% increased odds compared with the 40-week GA reference group (95% CI: 1.03–1.23)
For sensitivity analysis, regressions were performed excluding infants with prolonged birth hospitalizations (>180 days; n = 628), given the decreased observation time between their discharge and day 365 of life. This calculation did not significantly change our findings, other than slightly reducing adjusted ORs for the most premature infants.
Finally, we determined the percentage of infants hospitalized within each diagnostic category according to GA. During the first year of life, respiratory infection was the most common cause of hospitalization for each GA, resulting in ≥1 hospitalization for 9.3% to 9.5% of infants 23 to 25 weeks’ GA, 3.5% to 4.7% of infants 33 to 37 weeks’ GA, and 2.8% to 2.9% of infants 39 to 41 weeks’ GA. Within 14 days after birth discharge, the most premature infants were most commonly hospitalized for respiratory diagnoses (1.2%–1.6% of infants 23–28 weeks’ GA) and surgical procedures (1.2%–1.8% of infants 23–26 weeks’ GA). For infants >33 weeks’ GA, hyperbilirubinemia was the most common cause of hospitalization within 14 days of birth discharge, resulting in hospitalization for 1.9% to 2.2% of infants 35 to 37 weeks’ GA and 0.8% to 0.9% of infants 40 to 44 weeks’ GA. Median number of days between birth discharge and hyperbilirubinemia hospitalization was 3 (interquartile range: 2–4) across all GAs.
In recent years, numerous authors have demonstrated that LPIs are not a homogenous group, suggesting that risk of morbidity and mortality associated with prematurity may best be studied by examining GA according to week.2,7–9,17,27 Although studies examining outcomes after birth discharge have shown elevated risk among LPIs,6,12–16 these studies have not generally compared outcomes according to week of GA. Analyzing by GA, we found that odds of hospitalization generally decreased with increasing GA, with the exception of 34 to 36 weeks’ GA. One previous study of infants ≥33 weeks’ GA similarly reported greatest hospitalization risk among 36-week GA infants.27 When compared with predicted odds, infants 35, 36, and 37 weeks’ GA had increased odds of any hospitalization during all observation periods. Importantly, increased hospitalization risk was not homogeneous across GAs typically considered to be LPIs.
As expected, LPIs were at significant risk for hyperbilirubinemia hospitalizations, but infants with increased odds of hyperbilirubinemia hospitalization (33–38 weeks’ GA) extended beyond the LPI classification of 34 to 36 weeks’ GA. This finding is consistent with our hypothesis based on previous studies identifying increased risk among LPIs for hyperbilirubinemia hospitalizations12–16,28,29 and kernicterus, especially in the setting of poorly established breastfeeding.30 One previous study analyzing LPIs and TIs reported similarly highest jaundice admission risk for 35 and 36 weeks’ GA infants.29 Biologic risk for significant hyperbilirubinemia decreases with GA and generally peaks within the first week of life. The risk to LPIs is compounded by current practice, in which LPIs often receive birth hospitalization care similar to TIs (as indicated by a median birth hospitalization length of 2 days in both groups). In contrast, EPIs have more prolonged hospitalizations (median: 14 days; P < .001), such that peak bilirubin levels may occur while hospitalized, and closer feeding monitoring and support may be received. Due to shorter hospitalizations of LPIs relative to EPIs, LPIs are more likely to experience peak bilirubin levels after their hospital discharge and may also receive less in-hospital education and support compared with EPIs. Our study cannot explore specific aspects of care during birth hospitalization, but it does importantly broaden our understanding of the GAs with increased risk of postdischarge hyperbilirubinemia beyond the traditional LPI classification. Although prolonging birth hospitalization may be 1 option to reduce hyperbilirubinemia hospitalizations (due to identification and treatment of hyperbilirubinemia during birth hospitalization and/or improved establishment of feeding), strategies incorporating in-hospital assessment and education, close outpatient monitoring, rehospitalization when indicated, and increased breastfeeding support31 may be more acceptable to families and payers.
Although not as pronounced as hyperbilirubinemia findings, we also found that 33- to 36-week GA infants had a relative peak in odds related to nonspecific infections and bacterial infections, highlighting another area in which LPIs may experience risk after discharge greater than TIs and some EPIs. Infections have previously been identified as an important cause of hospitalizations and emergency department visits among LPIs,15,16,28 but this risk has not been quantified relative to TIs or EPIs. Although increased odds of hospitalizations for nonspecific infections could be influenced by heightened physician caution with LPIs, the increased odds of hospitalization for bacterial infections suggest actual increased vulnerability among LPIs, perhaps again made apparent by shorter birth hospitalizations compared with more immature EPIs. Of note, the GAs of infants with increased odds of any hospitalization for hyperbilirubinemia, nonspecific infections, and bacterial infections did not fit precisely within the GAs generally defined as late preterm, highlighting the utility of assessment according to GA.
In addition to identifying diagnoses in which odds of hospitalization rose during late preterm GAs, our analysis also identified diagnoses in which the odds of hospitalization for LPIs remained significantly higher than TIs (although less than EPIs). Most notably, odds of hospitalization for noninfectious respiratory diagnoses declined steadily with increasing GA but remained significantly elevated among LPIs (ORs ranging from 1.44–1.92) relative to the 40-week GA reference group. Previous studies have demonstrated declining respiratory morbidity with advancing GA during birth hospitalization.7,9,27,32 Although less is known about the impact of week of GA on postdischarge respiratory risk, 1 recent study reported similar direction of findings comparing bronchiolitis hospitalizations for moderately preterm infants, LPIs, and TIs.33 The gradual decrease in odds for respiratory, surgical, GI and feeding concerns, and other diagnoses suggested that risk declines continuously with GA, with no clear threshold at any specific GA. Thus, every additional week of prematurity is associated with increased odds of hospitalization extending beyond the neonatal period, suggesting that hospitalists and ambulatory providers should consider each infant’s specific GA when assessing risk during acute illnesses, rather than relying solely on less specific categories, such as “late preterm.”
Overall, it seems reasonable to expect that the physiology of infants born early places them at increased risk for hospitalization relative to TIs, which our data found across most reasons for hospitalization and all time periods. This finding supports ongoing recommendations to prolong gestation when medically appropriate for the fetus and mother. However, we also identified time periods and diagnostic categories in which 33- to 38-week GA infants experience increased risk relative not only to TIs but also to a portion of more premature infants, which cannot be explained purely through physiology but rather through the interplay between current practice and physiology. For some hospitalization diagnoses, careful outpatient monitoring and selective hospitalization may be preferable to prolonging nursery stays; our results do not allow comparison of these strategies. For hyperbilirubinemia and feeding-related hospitalizations in particular, increased breastfeeding assessment and support during and after birth hospitalization may be beneficial, as previous work has shown that breastfeeding affects the risk of hospitalization for LPIs.15,16,34 In many states, additional postdischarge programs may be available based on GA to EPIs but not LPIs, which could further modify risk across GAs. Although our data have no clear threshold effect at California’s High Risk Infant Follow-up program’s GA-based threshold of 32 weeks,35 many infants ≥32 weeks’ GA may also receive services through diagnosis-based eligibility. Overall, decisions about optimal care for these infants before and after discharge from birth hospitalization will require careful consideration of risks, benefits, and costs. Given the significant number of infants born annually within at-risk GAs, practices and policies related to care of these infants have the potential for a large public health impact.
Although our database included all infants born in California during our study period, the population of California may not be generalizable to other states. As an observational study, we were unable to control for all desired variables, such as breastfeeding status. To focus our analysis, we did not consider secondary diagnoses, which may have underestimated the impact of secondary issues (eg, feeding concerns in hyperbilirubinemia hospitalizations). We also did not have information on specifics of care received during birth hospitalization or after discharge. Using data from 1993 through 2005 means that our study included infants born during years when care for hyperbilirubinemia and LPIs was evolving; however, we controlled for year of delivery in our models. In addition, our study was limited by our use of an administrative data set, which may contain inaccuracies, although we would not expect any specific directional bias to be present.
Examining all births in California from 1993 to 2005 allowed robust analysis of hospitalizations experienced by infants of all GAs and insurance types, as well as quantification of odds of hospitalization experienced by all infants, including LPIs. During the most immediate postdischarge time period, 34- to 36-week GA infants experienced increased odds of hospitalization relative not only to TIs but also to a subset of more preterm infants. Comparing observed odds of hospitalization with expected odds of hospitalization, 35- to 37-week GA infants exceeded expected odds of hospitalization during all observation periods. Infants 33 to 38 weeks’ GA had increased odds of any hospitalization for hyperbilirubinemia relative to other GAs, and infants 33 to 36 weeks’ GA had a relative peak in odds of any hospitalization for bacterial and nonspecific infections. Examining outcomes according to week of GA allowed more detailed understanding of hospitalizations experienced after birth discharge by infants of all GAs, including those categorized as LPIs.
The authors acknowledge Corinne Ahlberg, MS, for her assistance with database management.
APPENDIX Diagnosis Categories by ICD-9 Code
|Medical Diagnosis Categorya||ICD-9-CM Codes|
|Infectious respiratory hospitalizations||033.0, 033.9, 079.6, 464.4, 465.9, 466.0, 466.1, 466.11, 466.19, 480.1, 480.9481, 482.9, 485, 486, 487.1, 490|
|Noninfectious respiratory hospitalizations||493.00, 493.01, 493.02, 493.90, 493.91, 493.92, 507.0, 518.81, 518.82, 519.8, 748.3, 770.7, 770.81, 770.89, 772.3, 769, 786.03, 786.09|
|GI/feeding hospitalizations||008.61, 008.8, 276.1, 276.5, 276.51, 530.81, 558.9, 579.8, 779.31, 775.5, 777.9, 783.3, 783.4, 787.03, 783.41, 787.91|
|Bacterial infection hospitalizations||038.0, 038,1, 038.42, 038.9, 079.9, 320.2, 376.01, 382.9, 590.10, 590.80, 599.0, 682.0, 771.4, 771.81, 771.82, 790.7|
|Other (nonspecific, viral) infection hospitalizations||047.8, 047.9, 079.89, 079.99, 322.9, 771.8, 771.89, 778.4, 780.6, V29.0|
|Hyperbilirubinemia hospitalizations||773.1, 774.2, 774.39, 774.6|
|Other medical hospitalizations||All remaining medical hospitalizations|
ICD-9-CM, International Classification of Diseases, Ninth Revision, Clinical Modification.
↵a Surgery hospitalizations determined by identification of any hospitalization listing a surgical procedure as the primary procedure for that hospitalization. Nonsurgical diagnostic procedures (eg, otoscopy [18.11], colonoscopy [45.23], diagnostic radiology studies [87.xx]) were not counted as surgical procedures.
Dr Ray’s current affiliations are Division of General Academic Pediatrics, Children’s Hospital of Pittsburgh, Pittsburgh, PA, and University of Pittsburgh School of Medicine, Pittsburgh, PA.
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
FUNDING: Funding in part by the Maternal and Child Health Bureau (R40 MC05474-01-00) and the Agency for Healthcare Research and Quality (R01 HS 015696).
- confidence interval
- early preterm infant
- gestational age
- late preterm infant
- odds ratio
- term infant
- Pulver LS,
- Denney JM,
- Silver RM,
- Young PC
- Bird TM,
- Bronstein JM,
- Hall RW,
- Lowery CL,
- Nugent R,
- Mays GP
- Gouyon JB,
- Vintejoux A,
- Sagot P,
- Burguet A,
- Quantin C,
- Ferdynus C
- Shapiro-Mendoza CK,
- Tomashek KM,
- Kotelchuck M,
- et al
- De Luca R,
- Boulvain M,
- Irion O,
- Berner M,
- Pfister RE
- Escobar GJ,
- Joffe S,
- Gardner MN,
- Armstrong MA,
- Folck BF,
- Carpenter DM
- Escobar GJ,
- Greene JD,
- Hulac P,
- et al
- McLaurin KK,
- Hall CB,
- Jackson EA,
- Owens OV,
- Mahadevia PJ
- Young PC,
- Glasgow TS,
- Li X,
- Guest-Warnick G,
- Stoddard G
- 19.↵US Department of Health and Human Services, Centers for Disease Control and Prevention, and the Centers for Medicare and Medicaid Services. International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM). Updated September 10, 2009. Available at: www.cdc.gov/nchs/icd/icd9cm.htm. Accessed July 21, 2010.
- Chamberlain JM,
- Joseph JG,
- Patel KM,
- Pollack MM
- Koehoorn M,
- Karr CJ,
- Demers PA,
- Lencar C,
- Tamburic L,
- Brauer M
- Burgos AE,
- Schmitt SK,
- Stevenson DK,
- Phibbs CS
- Colin AA,
- McEvoy C,
- Castile RG
- 35.↵California Department of Health Care Services. High risk follow up. Available at: www.dhcs.ca.gov/services/ccs/Pages/HRIF.aspx. Accessed October 27, 2012.
- Copyright © 2013 by the American Academy of Pediatrics