Trends in Mortality and Morbidities for Infants Born 24 to 28 Weeks in the US: 1997–2021 Free

Vermont Oxford Network (VON) is a nonprofit, voluntary worldwide community of practice dedicated to improving the quality, safety, and value of care for newborns.¹²  Local staff at member hospitals collect standardized data on infants’ initial birth hospitalization until discharge home, death, or first birthday, including for infants transferred from the reporting hospital to other hospitals. All data are deidentified and checked for quality and completeness at the time of submission. The University of Vermont Committee on Human Research determined that the use of the VON Research Repository for this analysis was not human subject research.

This study includes infants who were 401 to 1500 g and 24 weeks 0 days to 28 weeks 6 days at birth from January 1, 1997 to December 31, 2021, and were inborn or transferred to the reporting hospital within 28 days of birth in the United States, including those that did not receive active resuscitation.

Mortality applied to all infants including those who died in the delivery room or were transferred. Late-onset sepsis applied to infants who survived to day 4 from birth and included recovery of a bacterial pathogen on a specified list from blood or cerebrospinal fluid, coagulase-negative Staphylococcus from blood or cerebrospinal fluid, or a fungus obtained from a blood specimen; diagnosis of coagulase-negative Staphylococcus also required systemic signs of infection and treatment of 5 days or more with intravenous antibiotics. Necrotizing enterocolitis applied to infants who survived at least 12 hours and was defined as at least 1 clinical sign (bilious gastric aspirate or emesis; abdominal distension or discoloration; occult or gross blood in stool) and at least 1 diagnostic finding (pneumatosis intestinalis; hepato-biliary gas; pneumoperitoneum). CLD was defined as supplemental oxygen at 36 weeks’ postmenstrual age or, for infants discharged at 34 to 35 weeks, supplemental oxygen at discharge, excluding infants at hospitals above 4000 feet in elevation. Intraventricular hemorrhage applied to infants who survived at least 12 hours and were diagnosed with cranial imaging within 28 days of birth with severe intraventricular hemorrhage defined as grades 3 and 4.¹³  Retinopathy of prematurity (ROP) applied to infants who survived at least 12 hours and received an indirect ophthalmologic examination with severe ROP defined as stages 3 through 5.¹⁴  Death or morbidity was defined as death or at least 1 of the morbidities defined above.

The US Census Bureau definitions of race and ethnicity were used for all years. From 1997 to 2011, maternal race was defined as Black, white, Asian or Pacific Islander, Native American, or other. From 2012, maternal race was defined as Black, white, Asian, American Indian or Alaska Native, Native Hawaiian or other Pacific Islander, or other. Abstractors were instructed to obtain the information by personal interview with the mother, review of the birth certificate, or medical record, in that order. Race was combined with ethnicity to form the following groups: black non-Hispanic; white non-Hispanic; Hispanic; Asian, Native Hawaiian, or Pacific Islander non-Hispanic; American Indian or Alaska Native non-Hispanic; other non-Hispanic. We included race and ethnicity in the analysis because the infant distributions changed over the study period. Small for gestational age was defined as a less than the 10th percentile birth weight-for-gestational-age z-score on the Fenton growth curves.¹⁵  Other definitions are found in the Manual of Operations and did not change over the study period.¹⁶ 

We examined descriptive statistics for infant characteristics overall and across 5-year periods (1997–2001, 2002–2006, 2007–2011, 2012–2016, 2017–2021).

For each outcome, we used the National Cancer Institute Joinpoint Regression Program¹⁷  to model log-transformed rates by year of birth using modified Bayesian information criterion (BIC)¹⁸  to select the number and placement of knots. In the final model, each of the joinpoints and its corresponding changes in trend can be interpreted as statistically significant. We fit additional models at each gestational age week using the same number of knots as the overall model but allowing the placement of knots to vary.

We used R (version 4.2.1)¹⁹  to fit segmented relative risk regression models with the knots selected by the Joinpoint Program using robust variance estimators with hospital clusters to estimate the risk-adjusted annual percent change (aAPC) and 95% confidence interval (CI) for each segment. Infant-level covariates were added to the models to estimate the aAPCs for each segment, overall and for each gestational age. aAPCs were adjusted for maternal race and ethnicity, small for gestational age, infant sex, inborn or outborn status, any congenital anomaly, and gestational age at birth in weeks.

The aAPC is a summary measure of a trend over a fixed interval. It assumes that rates change at a constant percentage of the rate of the previous year. Rates that change at a constant percentage each year change linearly on a log scale. aAPCs standardize trends so they can be compared across outcomes with rates of different magnitudes.

We classified segments based on the aAPC using a modification of the National Cancer Institute method for characterizing trends. Increasing was defined as an aAPC of >0.5 with a 95% CI that did not include 0. Decreasing was defined as an aAPC of <−0.5 with a 95% CI that did not include 0. Stable was defined as a risk-adjusted APC between −0.5 and 0.5 with a 95% CI that did not include 0 or an aAPC with a 95% CI that included 0.

Analyses were based on infants with nonmissing values of outcome variables. Over the 25 year period, less than 0.5% of observations were missing values except for 4.1% missing values for CLD and 1.4% missing values for late infection. Additionally, 4.7% of eligible infants did not receive an eye exam to diagnose ROP and 3.8% did not have imaging to diagnose intraventricular hemorrhage. Missing values for race and ethnicity (0.5%) were imputed as other race for risk-adjustment models.

The primary analyses used data from all centers. Analyses were repeated using data from the centers in VON for at least 20 years and can be found in the supplemental materials.

Overall, 888 US hospitals contributed data for 447 396 infants; 153 contributed for 1 to 5 years, 144 contributed for 6 to 10 years, 306 contributed for 11 to 20 years, and 285 contributed for more than 20 years. Infant characteristics, overall and in 5-year increments, appear in Table 1.

From 1997 to 2021, the unadjusted rate of mortality decreased from 18.1% to 12.4%; late-onset sepsis from 34.2% to 13.4%; necrotizing enterocolitis from 10.0% to 6.8%; severe intraventricular hemorrhage from 11.8% to 10.7%; severe ROP from 14.8% to 9.3%; and death or morbidity from 65.4% to 57.6%. CLD increased from 33.4% to 43.3%. aAPCs with 95% CIs for the entire population appear in Table 2 and by gestational age in Supplemental Table 4. Observed and model-estimated trends for the entire population appear in Fig 1 and by gestational age in Fig 2.

The mortality analysis identified 3 segments: stable from 1997 to 2005 (aAPC −0.3; 95% CI −1.0 to 0.3); decreasing from 2005 to 2012 (aAPC −3.2; 95% CI −3.7 to −2.7); and decreasing more slowly from 2012 to 2021 (aAPC −1.8; 95% CI −2.2 to −1.3). At all gestational ages, aAPCs were stable in segment 1 followed by decreases in segment 2. In segment 3, the aAPCs for infants born at 24, 25, or 26 weeks decreased significantly, whereas those for infants born at 27 or 28 weeks did not.

The late-onset sepsis analysis identified 3 segments: decreasing from 1997 to 2007 (aAPC −1.3; 95% CI, −1.9 to −0.6); decreasing more rapidly from 2007 to 2012 (aAPC −11.8; 95% CI −12.6 to −11.0); and decreasing more slowly from 2012 to 2021 (aAPC −2.9; 95% CI −3.5 to −2.3). Patterns for all gestational ages matched the overall cohort.

The necrotizing enterocolitis analysis identified 4 segments: decreasing from 1997 to 2001 (aAPC −7.4; 95% CI −9.8 to −5.0); increasing from 2001 to 2007 (aAPC 6.3; 95% CI 5.1 to 7.6); decreasing from 2007 to 2015 (aAPC −6.0; 95% CI −6.8 to −5.2); and stable from 2015 to 2021 (aAPC 0.3; 95% CI −1.0 to 1.6). The patterns were similar by gestational age with rates decreasing in segment 1, increasing in segment 2, decreasing in segment 3, and remaining stable in segment 4, except for infants born at 28 weeks’ gestation where rates were stable in the initial segment.

The CLD analysis identified 3 segments: increasing from 1997 to 2001 (aAPC 10.0; 95% CI 8.2 to 11.8); decreasing from 2001 to 2012 (aAPC −1.7; 95% CI −2.1 to −1.3); and increasing from 2012 to 2021 (aAPC 0.6; 95% CI 0.1 to 1.0). The pattern was similar by gestational age with rates increasing in segment 1, decreasing in segment 2, and stable in segment 3 for all gestational ages except infants born at 27 weeks where rates increased in the final segment.

The analysis for severe intraventricular hemorrhage identified 3 segments: increasing from 1997 to 2003 (aAPC 3.4; 95% CI 2.3 to 4.6); decreasing from 2003 to 2015 (aAPC −2.3; 95% CI −2.7 to −1.8); and stable from 2015 to 2021 (aAPC 0.3; 95% CI −0.5 to 1.0). Rates increased in segment 1 at all gestational ages, decreased for all gestational ages in segment 2, and remained stable for all gestational ages in segment 3.

The analysis for severe ROP identified 3 segments: increasing from 1997 to 2002 (aAPC 3.9; 95% CI 1.9 to 5.9); decreasing from 2002 to 2011 (aAPC −6.6; 95% CI −7.4, −5.8); and decreasing more slowly from 2011 to 2021 (aAPC −0.9; 95% CI −1.6 to −0.1). Severe ROP increased in segment 1 for infants born at 24-, 25- and 26-weeks’ gestation but remained stable for infants born at 27 and 28 weeks, decreased in segment 2 for all gestational ages, and remained stable in segment 3 for all gestational ages.

The analysis for death or morbidity found 3 segments: increasing from 1997 to 2003 (aAPC 1.2I; 95% CI 0.8 to 1.7); decreasing from 2003 to 2015 (aAPC −1.8; 95% CI −1.9 to −1.6); and stable since 2015 (aAPC 0.2; 95% CI −0.1 to 0.5). Death or morbidity increased for all gestational ages in the first segment and decreased for all gestational ages in the second segment. In the third segment, death or morbidity decreased for infants born at 24 weeks’ gestation, increased for infants born at 27 weeks and was stable for infants born at 25, 26, and 28 weeks.

We describe changes in outcomes for infants born at 24 to 28 weeks' gestation from 1997 to 2021 that extend our understanding of neonatal mortality and morbidity with a current description of trends at a larger number of hospitals than previous studies revealing that improvements in mortality and morbidity have slowed, stalled, or reversed in recent years. These data provide benchmarks for units to use in identifying opportunities for improvement and for informing discussions with parents.

From 1997 to 2021, mortality decreased from 18.1% to 12.4%, late onset infection from 34.2% to 13.4%, necrotizing enterocolitis from 10.0% to 6.8%, severe intraventricular hemorrhage from 11.8% to 10.7%, severe ROP from 14.8% to 9.3%, and death or morbidity from 65.4% to 57.6%. CLD increased from 33.4% to 43.3%. The detailed time course for each of the outcomes is different as shown by the aAPCs. Many of the outcomes had a period of relatively rapid improvement with either slower or no improvement in the final segments. Mortality, late-onset sepsis, and severe ROP decreased, but more slowly than in previous years; necrotizing enterocolitis and severe intraventricular hemorrhage were stable since 2015; and CLD increased since 2012. The distributions of infants by week of gestational age remained stable over the period, and trends by gestational age generally mirrored those of the overall outcomes.

Despite the variation in when the inflections occurred, the outcomes had some improvement followed by slower improvement, stagnation, or worsening in the most recent years. To regain the pace of improvement seen in earlier years, we propose a 3-part strategy: research to develop innovative new therapies; quality improvement to optimize the effectiveness of available interventions; and a commitment to follow through addressing the social determinants of health.

A potential explanation for the lack of progress in recent years is that we may have reached the limit of effectiveness of available therapies. In 2021 over 85% of infants born at 24- to 28-weeks’ gestation were exposed to antenatal steroids,²⁰  an intervention known to reduce mortality and morbidity for preterm infants.³  Increasing antenatal steroid exposure is not likely to further improve outcomes. Since most infants likely to benefit from surfactant in this population now receive it, we may have reached the limit of this therapy as well.²¹  Additionally, despite a shift to noninvasive ventilation expected to reduce the risk of CLD,²²  its occurrence has increased, a trend mirroring that of postnatal steroid exposure, which decreased in the late 1990s.⁸  Research to develop innovative new therapies and understand existing ones is the first pillar in our proposed strategy.

It is reasonable to speculate that adoption of quality improvement by neonatal care teams contributed to the initial improvements that we observed. Vermont Oxford Network conducted the first quality improvement collaborative²³  in neonatology in 1994. Since then, over 700 multidisciplinary teams have participated in 1 or more Vermont Oxford Network collaboratives. Hundreds of teams have participated in collaboratives organized by state perinatal organizations,²⁴  children’s hospitals,²⁵  and neonatology practice management groups.²⁶  The dramatic reductions in late onset infections coincided with widespread quality improvement efforts to reduce catheter associated blood stream infections.²⁷  The implementation of targeted quality improvement interventions have been reported to reduce the risks of CLD²⁸  and intraventricular hemorrhage.^29,30 

Teams may face challenges implementing quality improvement efforts, demonstrated by persistent marked variation in outcomes across units.⁷  However, this variation suggests that teams can still realize improvements. Applying meticulous focus to optimize the effects of available interventions and using local data to drive improvements will help level the playing field, allowing all units to achieve outcomes currently achieved only by the best performers.⁷  Quality improvement to optimize the effectiveness of available interventions is the second pillar in our proposed strategy.

The ultimate measures of success for neonatal intensive care are the neurodevelopmental outcomes of the infants and the long-term health and wellbeing of the infants and their families. Although we do not report on neurodevelopmental outcomes, the rates of neurodevelopmental impairment did not change significantly between 1990 and 2020.^9,31  Morbidity experienced in the neonatal period is only 1 contributing factor to long-term outcomes. The social determinants of health play a significant and perhaps predominant role.^32,33  To achieve better outcomes for infants and families, we must accept our responsibility to follow through, improving the health and well-being of infants and families by addressing the social determinants of health and promoting equity.^34,35  Follow through, the third pillar in our proposed strategy, may provide the greatest opportunity in the near term for improvement for the infants and families we serve. Preterm infants’ long birth hospitalizations³⁶  offer a unique opportunity for intervention, such as with screening and referral.^37–39  The impact of early-life disparities on long-term outcomes^32,40,41  make follow through imperative.

Although our data do not permit causal explanations for the trends we observed, the impact of the severe acute respiratory syndrome coronavirus 2 pandemic on maternal and newborn infections and disruptions of hospital operations^42–49  should be considered. The start years for the final segments of the time series for mortality and individual morbidities ranged from 2011 to 2015; therefore, it is unlikely that the pandemic can explain the trends we observed, although it is possible that the pandemic contributed to these trends in the last 2 years of the study. Further studies are needed to understand how the pandemic affected outcomes for preterm infants. More importantly, the severe acute respiratory syndrome coronavirus 2 pandemic continues to stress hospital operations and burden health professionals and families, emphasizing the need for continued commitment to quality improvement and follow through to improve outcomes.

Our study has potential limitations. The number of hospitals that participated in Vermont Oxford Network increased from 1997 to 2021. However, the findings are nearly identical in a sensitivity analysis restricted to hospitals that participated for 20 years. Our study only includes members of Vermont Oxford Network. However, since the Vermont Oxford Network database now enrolls approximately 90% of infants born 24 to 28 weeks in the United States each year, it is likely that our results are representative of practices and outcomes overall. Missing data may bias the results in ways that are challenging to quantify; however, the rates of missing data are low and did not change over time.

Improvements in mortality and morbidity for infants with gestational ages from 24 to 28 weeks in the United States have slowed, stalled, or reversed in recent years. To regain the pace of improvement seen in earlier years, we propose a 3-part strategy: research to develop innovative new therapies, quality improvement to optimize the effectiveness of available interventions, and a commitment to follow through addressing the social determinants of health to improve the long-term health and well-being of infants and their families.

We are indebted to our colleagues who submit data to Vermont Oxford Network on behalf of infants and their families. The list of hospitals contributing data to this study is in Supplemental Table 4.

aAPC

adjusted annual percent change

CI

confidence interval

CLD

chronic lung disease

ROP

retinopathy of prematurity

VON

Vermont Oxford Network