The clinical burden of extremely preterm birth in a large medical records database in the United States: complications, medication use, and healthcare resource utilization

Abstract Introduction Approximately 5% of global preterm births are extremely premature (EP), defined as occurring at less than 28 weeks gestational age. Advances in care have led to an increase in the survival of EP infants during the neonatal period. However, EP infants have a higher risk of developing complications such as bronchopulmonary dysplasia (BPD), intraventricular hemorrhage (IVH), and retinopathy of prematurity (ROP). BPD and other respiratory morbidities are particularly prevalent among this population. To understand the healthcare resource utilization (HRU) of EP infants in the United States, the clinical and economic burden of extreme prematurity was examined in this retrospective study of data extracted from electronic medical records in the Kaiser Permanente Northern California (KPNC) health system. Methods The analysis included data from EP infants live-born between January 1997 and December 2016, and focused on complications and HRU up to 3 years corrected age (CA), covering the period up to December 2018. Stillbirths, infants born at <22 weeks gestational age, and infants with major congenital malformations were excluded. Complications of interest (BPD, IVH, and ROP) and medication use were compared by age group (≤1 year, >1 year and ≤2 years, and >2 years and ≤3 years CA). Analysis of HRU included hospital readmissions, ambulatory visits, and emergency room (ER) visits. Results A total of 2154 EP births (0.32% of total live births and 4.0% of preterm births that met the inclusion/exclusion criteria) were analyzed. The prevalence of EP birth showed a declining trend over time. ROP was the most commonly recorded complication during the birth hospitalization (37.1% any stage; 2.9% Stages 3 and 4). BPD was recorded in 34.3% of EP infants. IVH (any grade) was recorded in 22.7% of EP infants (6.4% Grades III and IV). A majority (78.7%) of EP infants were diagnosed with at least one respiratory condition during the first year CA, the most common being pneumonia (68.9%); the prevalence of respiratory conditions decreased over the second and third years CA. During the first 3 years CA, the most common medications prescribed to children born EP were inhaled bronchodilators (approximately 30% of children); at least 15% of children received systemic corticosteroids and inhaled steroids during this period. During the first 3 years CA, at least one hospital readmission was recorded for 16.4% of children born EP; 57.1% of these readmissions were related to respiratory conditions. At least one ER visit was recorded for 33.8% of children born EP, for which 53.1% were due to a respiratory condition. Ambulatory visits were recorded for 54.2% of EP children, for which 82.9% were due to a respiratory condition. Conclusions The short- and long-term clinical burden of EP birth was high. The onset of BPD, IVH, and ROP was common during the birth hospitalization for EP infants. Medication use, hospital readmission, and clinic visits (ER and ambulatory) occurred frequently in these children during the first 3 years CA, and were commonly due to respiratory conditions. Strategies prioritizing the reduction of risk and severity of respiratory conditions may alleviate the clinical burden of EP birth over the long term.


Introduction
An estimated 14.9 million infants worldwide are born preterm each year (defined as birth at <37 weeks gestational age [GA]) [1]. Extremely premature (EP) births (<28 weeks GA) account for $5% of all preterm births globally [1]. Recent cohort studies indicate a wide variation in survival estimates of EP infants across numerous developed countries, which is influenced by cohort selection, place of birth, and differences in the provision of active obstetric and neonatal treatment [2][3][4][5]. Despite the increased survival rate at birth, complications of prematurity are a leading cause of death in children <5 years of age [6]. Infants who survive EP birth are at increased risk for multiple complications, including bronchopulmonary dysplasia (BPD), intraventricular hemorrhage (IVH), and retinopathy of prematurity (ROP) [7,8].
BPD, which is also referred to as chronic lung disease (CLD), is often the first manifestation of lung injury in premature infants, and GA is the most significant single predictor of BPD [9]. The diagnosis and severity of BPD depend on the requirement for oxygen and respiratory support at 36 weeks postmenstrual age [10,11]. Lungs of infants with surfactant deficiency following respiratory distress syndrome were considered representative of "old" BPD. On the other hand, "new" BPD in the past 25-30 years shows histologic characteristics of immature lungs, with disruption to normal pulmonary vascular development and damage in the canicular and saccular lung development stages [12]. Clinical manifestations of BPD usually include airway obstruction and hyperinflation, which impact cardio-pulmonary function and can lead to significant physical and economic burdens beyond infancy [13][14][15]. Rates of BPD have not significantly improved, and have increased in some cases, despite advances in neonatal care and increased survival of premature infants [16].
IVH, another major complication of EP births, occurs after bleeding into the germinal matrix of the brain extends into the ventricular system (resulting in ventricular dilatation or compression on the venous system and periventricular post-hemorrhagic infarction in the surrounding brain). Although management of IVH has improved, the incidence of IVH remains high before 29 weeks GA [17,18]. Preterm infants with IVH are at increased risk for poor neurodevelopmental outcomes and often develop cerebral palsy, developmental delay, deafness, or blindness [19]. The severity of IVH is defined using grading systems [20,21] based on the extent of the bleeding: Grades I and II are considered clinically mild; Grade II has been associated with higher rates of neurodevelopmental impairment including cerebral palsy [22,23]; Grade III and periventricular hemorrhagic infarction (also referred to as Grade IV) are considered severe. Higher rates of neurodevelopment impairment are associated with increasing IVH grade [24].
ROP is the most common cause of blindness and poor visual outcomes in infants. It results from disruption of vascular development in the retina, stemming in part from supplemental oxygen use for treatment of respiratory conditions in premature infants [25,26]. The incidence of ROP increases with decreasing GA and has thus increased in recent years with higher survival rates of EP infants. It is recommended that all infants born <30 weeks GA or with a birth weight 1500 g be screened for ROP [27]. The severity of ROP is classified using stages of progression (1 [presence of demarcation line] to 5 [total retinal detachment]) by zones to indicate location (1-3, with Zone 1 surrounding the optic disk and more serious than Zone 2 or 3) [28].
EP infants have a high healthcare burden [29]; however, there is little evidence of their exact healthcare resource utilization (HRU) requirements, and an increased understanding of the epidemiology and clinical and economic burden of extreme prematurity in the United States is needed. A retrospective, population-based cohort study utilizing data from the Kaiser Permanente Northern California (KPNC) electronic medical record (EMR) database was therefore conducted to analyze the clinical burden of EP birth in terms of complications of prematurity (BPD, IVH, and ROP), medication use, and HRU.

Study design and data source
This was a retrospective, population-based cohort study analyzing data from the KPNC EMR database. KPNC is one of the largest integrated healthcare delivery organizations in the United States. The KPNC membership population consists of 30% of the underlying service area population and repeatedly has been shown to be demographically representative of the underlying community population [30,31]. The KPNC EMR database includes enrollees residing in the Northern California region, including the Greater Bay Area and Central Valley. At the time of this study, 21 hospitals and over 200 outpatient clinics were operated by KPNC, including 15 birth centers and 15 neonatal intensive care units (NICUs; 7 level III, 7 level II, and 1 level I), providing care to approximately 4 million members. There were 30,000-40,000 live births/ year at these centers at the time of the study. Data from the database cover enrollment, demographics, census demographics, provider specialty, utilization (e.g. encounters, procedures, and diagnoses), pharmacy records (internal electronic systems and/or claims data), laboratory results, and mortality (date and cause of death). KPNC implemented all safeguards that were required for the protection of the individuals whose data were included in the study. The study was approved by the KPNC Institutional Review Board.

Patient population
Data from live-born preterm infants born during the study period (between January 1997 and December 2016) were included in this study, with focus on a subgroup of infants born EP (22-<28 weeks GA). Stillbirths, infants born at <22 weeks GA, and infants with major congenital malformations were excluded. Data covering the 22-year analysis period including follow-up to December 2018 were retrieved.

Outcome and summary measures
The prevalence of EP births was reported as a proportion (%) of total births and of preterm births recorded during the period of interest. Maternal and infant characteristics were identified during the birth hospitalization based on pre-specified categorical variables. Complications (BPD, IVH, and ROP) were defined by specific International Classification of Diseases (ICD), Ninth/ Tenth Edition, Clinical Modification (ICD-9/ICD-10 CM) codes (Addendum Table 1).
The number and proportion of EP infants with at least 1 respiratory condition of interest were reported by the corrected age (CA) at diagnosis ( 1 year, >1 year and 2 years, and >2 years and 3 years CA). Of note, some infants born in 2016 may not have reached 3 years CA by the analysis cutoff point of 31 December 2018. The frequency of medication use among infants with at least 1 respiratory condition was reported for 1 year, >1 year and 2 years, and >2 years and 3 years CA. Pre-specified medications of interest included diuretics, inhaled steroids, inhaled bronchodilators, leukotriene receptor antagonists, and systemic corticosteroids.
HRU for the EP cohort included the number and frequency of hospital readmissions, ambulatory visits, and emergency room (ER) visits for all causes occurring during the first 3 years CA and after discharge from birth hospitalization. A hospital readmission was defined as any hospital stay including inpatient stays, same-day hospital discharges, and hospital transfers during hospital admissions. An ER visit was defined as any encounter that involved an emergency department visit, excluding urgent care. An ambulatory visit was defined as any outpatient clinic visit, including ambulatory lab testing or ambulatory radiology-only encounters, and excluding outpatient ER visits. HRU included the frequency of hospital readmissions, ER visits, and ambulatory visits.

Statistical analyses
All data were summarized using descriptive statistics (SAS V R for Windows version 9.3 [SAS Institute, Cary, NC]). The prevalence of EP births was reported as a proportion (%) of total births recorded during the periods of interest and the Spearman's test was used to analyze trends in annual prevalence. Statistical significance was defined as p < .05.

Study sample and prevalence of EP births
A total of 679,157 live births with no major birth defects and GAs of at least 22 weeks were recorded between 1997 and 2016. Of these, 53,253 (7.8%) births were preterm, including 2154 EP births (0.32% of all live births and 4.0% of all preterm births) (Addendum Figure 1). The annual prevalence of EP births was variable; however, there was a general trend of decline during the study period (Spearman's trend test p ¼ .001), with the lowest rate during the last 3 years of the study period (0.28% in 2014; 0.24% in both 2015 and 2016) (Addendum Figure 2). Within the EP cohort, 52.6% of infants were male and 76.6% were singleton births (Addendum Table 2).

Complications
More than half of EP infants (57.0%) were diagnosed with at least one complication of interest during the birth hospitalization, and 30.9% had multiple complications (Table 1). Of the complications of interest, ROP (any stage) was the most commonly reported (37.1%), followed by BPD (34.3%) and IVH (any grade) ( Table 3). Upper respiratory infection (33.0%; 409/1239) was also the most common condition reported in EP children between >2 years and 3 years CA, along with asthma (20.6%) and reactive airway disease (20.4%).

Medication use
Overall, medication use (any) was recorded in 35.3% (598/1695) of EP children 1 year CA, 41.2% (507/ 1231) >1 year and 2 years CA, and 36.0% (443/1230) >2 years and 3 years CA (Addendum Table 4). Inhaled bronchodilators were the most common medications prescribed to EP children over the first 3 years CA (28.6% 1 year CA, 33.4% >1 year and 2 years CA, and 29.7% >2 years and 3 years CA). Systemic corticosteroids and inhaled steroids were also prescribed to at least 15% of children during each of the first 3 years CA, whereas leukotriene receptor antagonists were prescribed to approximately 2% of children.

Healthcare resource utilization
During the first 3 years CA, 16.4% (354/2154) of EP children had at least 1 hospital readmission; of these, 57.1% (202/354) had readmissions associated with respiratory conditions (Table 2). At least one ER visit was recorded for 729/2154 (33.8%) EP children during the first 3 years CA. Among them, 53.1% (387/729) visited the ER due to a respiratory condition, including 315 (43.2%) who had at least 2 ER visits for respiratory conditions. At least one ambulatory visit was recorded for 54.2% (1167/2154) of the EP cohort, including 82.9% (967/1167) of children who visited due to a respiratory condition and 99.0% (1155/1167) who visited due to other conditions. Multiple visits for respiratory conditions were recorded in 76.9% (897/1167) infants who had at least one ambulatory visit.

Discussion
Despite significant advances in the care of preterm infants, chronic respiratory diseases remain a common complication of preterm birth into early childhood and beyond, particularly among EP infants and those with BPD [32]. During the first 3 years of life, EP children from this study had substantial rates of respiratory conditions (78.7% at year 1-46.0% at year 3 CA), and over one-third of these cases were prescribed medications to treat their conditions. These findings are in agreement with others which show that respiratory problems are a key contributor to the large healthcare burden associated with EP birth. Chen et al. described a correlation between the severity of BPD and poor respiratory prognosis in preterm infants [33], and Mowitz et al. found that EP infants with BPD had significantly longer stays in hospital following birth, incurred higher total costs, and were more likely to have two or more hospital encounters following birth hospitalization than those without BPD [29]. In contrast, respiratory burden is markedly lower for infants born at term; respiratory morbidity was reported to occur in $6% of term births, with most cases being mild and transient, and severe cases often related to infections [34,35]. In a study of term infants born between 1991 and 2014 at a single center in Israel and who were followed for up to 18 years of age, $5% of infants were hospitalized due to a respiratory morbidity [36]. Among infants born between 2003 and 2013 in a US cohort, the rate of hospitalization within 1 year due to respiratory syncytial virus was 0.9-1.6% for full-term infants versus 1.7-3.8% for infants born <29 weeks GA [37]. More than half of EP infants in this study were diagnosed with at least one complication during the birth hospitalization, and nearly one-third had multiple complications, with the most common being ROP (any stage) (37.1%), BPD (34.3%), and IVH (any grade) (22.7%). Of note, 25% of EP infants were diagnosed with retrolental fibroplasia or unspecified stage of ROP; it is expected that most of these cases represented Stage 1 or 2 ROP, which often regresses spontaneously without long-term consequences [38].
Comparison of morbidity rates across studies is complicated by the use of different definitions, GA cutoffs, and inclusion/exclusion criteria. In a 2011/2012 study of neonatal morbidity among infants born <28 weeks GA in 11 European countries, variation in severe morbidities between regions was wide (16-40%). The most prevalent morbidities were severe BPD, ROP stage !3, and IVH grades III/IV, with 36.9% of infants having severe BPD as well as at least one additional severe morbidity [39]. In a study of infants  Percentage values for HRU by cause (i.e. respiratory condition or other condition) are based on the total number of cases with the respective HRU of interest. A hospital readmission was classified as any acute inpatient hospital stay including inpatient stays, same-day hospital discharges, or hospital transfers when the patient was admitted into the hospital. An ER visit indicates any encounter that involved an emergency department visit, excluding urgent care. An ambulatory visit included all outpatient visits, excluding ER visits. All visits (hospital readmission, ER, or ambulatory visits) occurred after the initial discharge date. There were no exclusions based on mortality; (i.e. an infant could have died during the initial hospital stay and would not have been able to incur additional encounters). For this table, that infant would be included in the no hospital readmissions group and the 0 ambulatory and 0 ER visit groups. c Patients could have respiratory condition, "other" condition, and/or "unspecified" condition during inpatient stays, ER visits, and ambulatory visits, and therefore could be classified under one or more visit subsets. A visit was classified as "for respiratory condition" if the principal diagnosis for that visit was a respiratory condition. If the principal diagnosis was not a respiratory condition, the hospital readmission was classified as "for other condition". If the visit did not have a principal diagnosis, the primary diagnosis was examined to determine if it pertained to a respiratory or other diagnosis. If the visit did not have a principal or primary diagnosis but only had one diagnosis, that single diagnosis was used to determine the cause. Any visit without a principal or primary diagnosis and at least one diagnosis was considered to be "unspecified". born 22-28 weeks GA between 2003 and2007 [8] the prevalence of severe IVH was 16% and ROP (based on ophthalmologic examinations) was 59%. The authors noted that the rate of BPD diagnoses varied due to differences in the definitions of BPD. In addition, this study excluded infants born before 22 weeks GA and did not adjust for the high early mortality rate among EP infants, which is attributed in a large part to respiratory complications [40,41]. The short-term and long-term clinical burden of EP birth in this study was high. By 3 years CA, 16.4% of EP infants had required hospital readmission, and 28.5 and 53.5% had at least 2 ER and ambulatory visits, respectively. Studies have shown that preterm and low birth-weight infants have some of the highest healthcare expenditures of any patient population. Walsh et al. [42] and Mangham et al. [43] illustrated that birth hospitalization costs were responsible for most of the overall costs per surviving preterm infant. A retrospective cohort study of spending using a large national claims database of commercially insured individuals found that costs over the first 6 months after birth increased with shorter GA, ranging from $356,839 for infants born at 27-28 weeks GA to $603,778 for infants born at 24 weeks GA [44]. A high proportion of EP infants who utilized healthcare services in our study had principal diagnoses for respiratory conditions, but many were also treated for other complications, such as ROP and IVH, which contribute a large proportion to healthcare expenditures in preterm infants [44].
In this study of almost 700,000 live births between 1997 and 2016 in an integrated healthcare delivery system in the United States, the prevalence of EP birth was approximately 0.3% overall, or 4.0% of preterm births. In comparison, in a meta-analysis of international birth data, EP births represented a mean of 5.2% of preterm births across 41 countries from approximately 1990 to 2010 [1]. The slightly lower prevalence of EP births observed here could have been attributed to improvements in prenatal care in the later time period that was studied, higher availability of prenatal care in the KPNC population compared with lower income countries, the exclusion of EP births with congenital malformations in this analysis, and differences in the racial and ethnic makeup of the study populations. Indeed, the rate of preterm birth in the United States in 2019 was shown to be higher among African-American women (14.4%) compared with White or Hispanic women (9.3% and 10%, respectively) [45], and the proportion of White women was likely much higher in the KPNC population than in the international study.
The KPNC EMR database is a large, high-quality source of data that spans many years and is maintained for research purposes, and it represents a major strength of this study. The extensive clinical data system collects information directly entered by medical professionals on all aspects related to clinical care delivery. Data from the KPNC EMR are more accurate and comprehensive than claim-based data sources. The KPNC population is comprised of 30% of the underlying population in the service area communities in the Northern California region and has been repeatedly shown to be representative of the underlying population with a slightly lower presence of both the low and high ends of income level [30,31]. The racial/ ethnic composition of the KPNC database closely reflects the distribution of the United States as a whole (with the exception that the Asian population is slightly over-represented and the Hispanic population is slightly under-represented).
Limitations of this study include a possible underestimation of the prevalence of comorbidities among infants who died soon after birth, due in part to deaths that occurred prior to diagnosis. Diagnoses were based on ICD codes, which may be subject to errors in interpretation and recording; for example, claims codes may not accurately reflect NICU diagnoses or definitions. In addition, a full 3 years of followup data were not available for infants born after 2015 because data collection for the study ended in December 2018. However, these partial follow-up data comprised less than 5% of births in 2016 and a re-analysis that excluded these infants without complete 3 years' follow-up data did not change the overall results (data not shown).
In conclusion, this study illustrates that extreme prematurity is associated with a high prevalence of complications during the early years of life. The related short-and longer-term clinical burden on EP children is high, with many requiring continual use of respiratory medication, readmittance to hospital, or ER or ambulatory visits. Strategies to reduce the risk and severity of conditions and complications, particularly respiratory conditions and IVH, may alleviate the longer-term clinical burden of EP births. editing also was provided by Excel Medical Affairs. Shire, a Takeda company, provided funding to Excel Medical Affairs for support in editing this manuscript. The interpretation of the data was made by the authors independently.

Disclosure statement
Csaba Siffel is an employee of Takeda Development Center Americas, and owns stock/stock options in Takeda. Andrew K. Hirst, Hong Chen, Jeannette Ferber, Michael W. Kuzniewicz, and De-Kun Li are employees of Kaiser Permanente Northern California, which was contracted by Takeda to perform this study. Sujata P. Sarda was an employee of Takeda at the time the study took place.

Funding
This work was supported by Shire (a Takeda company), now a member of the Takeda group of companies (grant number is not applicable).

Data availability statement
Participants of this study did not provide consent for their data to be shared publicly. Thus, supporting data are not available.