Burden of Comorbidities and Healthcare Resource Utilization Among Medicaid-Enrolled Extremely Premature Infants

Background: The effect of gestational age (GA) on comorbidity prevalence, healthcare resource utilization (HCRU), and all-cause costs is significant for extremely premature (EP) infants in the United States. Objectives: To characterize real-world patient characteristics, prevalence of comorbidities, rates of HCRU, and direct healthcare charges and societal costs among premature infants in US Medicaid programs, with respect to GA and the presence of respiratory comorbidities. Methods: Using International Classification of Diseases, Ninth/Tenth Revision, Clinical Modification codes, diagnosis and medical claims data from 6 state Medicaid databases (1997-2018) of infants born at less than 37 weeks of GA (wGA) were collected retrospectively. Data from the index date (birth) up to 2 years corrected age or death, stratified by GA (EP, ≤28 wGA; very premature [VP], >28 to <32 wGA; and moderate to late premature [M-LP], ≥32 to <37 wGA), were compared using unadjusted and adjusted generalized linear models. Results: Among 25 573 premature infants (46.1% female; 4462 [17.4%] EP; 2904 [11.4%] VP; 18 207 [71.2%] M-LP), comorbidity prevalence, HCRU, and all-cause costs increased with decreasing GA and were highest for EP. Total healthcare charges, excluding index hospitalization and all-cause societal costs (US dollars), were 2 to 3 times higher for EP than for M-LP (EP 74 436vsM−LP27 541 and EP 28 504vsM−LP15 892, respectively). Conclusions: Complications of preterm birth, including prevalence of comorbidities, HCRU, and costs, increased with decreasing GA and were highest among EP infants during the first 2 years in this US analysis.

JOURNAL OF HEALTH ECONOMICS AND OUTCOMES RESEARCH to adulthood, are also predisposed to chronic respiratory diseases such as asthma and chronic obstructive pulmonary disease, 31 both of which incur long-term direct and societal costs. 21 Despite improved survival of EP infants in recent years, there remains a significant risk of developing short-and long-term complications. [32][33][34][35] Current options are limited for effectively treating neonates with complications associated with prematurity. Administered prenatally, maternal betamethasone can be used to stimulate lung maturity and magnesium sulfate can be used to reduce cerebral palsy. 36,37 Although ventilators and surfactant therapy increase survival rates in preterm infants with respiratory distress syndrome (RDS), they can contribute to pneumonia, sepsis, pneumothorax, pulmonary hemorrhage, and IVH. 38 Less-invasive surfactant therapies and oxygen titration have been developed to reduce rates of BPD and ROP, and continuous positive airway pressure oxygenation is used to reduce rates of ROP, RDS, and BPD. [39][40][41] Despite these advances, the most common complication affecting EP infants is BPD, which continues to impact respiratory function into adulthood. 7 The high costs associated with preterm birth have been documented previously, including costs of initial hospitalizations 18,25,28 and incremental costs per comorbidity, 15 costs of resource use in early childhood and through to adulthood, 23,26,27,42 societal costs for families of premature infants, 16,29,42 and annual costs relative to near-term and full-term infants. 19,23,29 All previous analyses that factored in GA and/or birthweight found that the highest costs are associated with the earliest preterm births. In a Canadian study through age 10 years, costs were higher for infants born early preterm compared with infants born moderate or late preterm. 42 A similar trend was reported in a study to age 18 years in England and Wales. 23 Studies report increased costs throughout early childhood for children delivered at less than 37 wGA, primarily due to increased resource use, including rehospitalizations and doctor visits. Through discharge to age 6 years, the odds ratio for hospitalization was 1.3-1.6 compared with children delivered at full term. 27 In an analysis of all premature infants born 1996-1997 in Quebec, Ontario, Canada, Johnston et al 42 found that costs in Canadian dollars (CAD) through age 10 years were $67 467 for early preterm infants and CAD $52 796 for moderate preterm infants vs CAD $10 010 for late preterm infants. Mangham et al 23 reported a similar trend, extending the estimate to age 18 years.
A national cost estimate in Canada from 1996 found that although EP births accounted for only 6.7% of preterm births, they accounted for 21.0% of the total national cost of preterm births. 42 Hall and Greenberg 19 reported significant cost reductions for initial hospitalization, as well as community-level costs, that would result from an increase in GA by 1-week increments.
There remains a clear unmet need for therapies and clinical practices that reduce the prevalence of complications of prematurity and their associated longer-term morbidity burden, especially in EP infants. Determination of the major drivers of costs and healthcare resource utilization (HCRU) associated with premature infants, in both the short and long term, will help inform practitioners and policy makers. This study aimed to characterize real-world patient characteristics, prevalence of comorbidities, rates of HCRU, and direct healthcare charges and societal costs among premature infants in US Medicaid programs with respect to GA and the presence of respiratory comorbidities.

Study Design
Premature infants were identified retrospectively from approximately 26.6 million Medicaid beneficiaries from 6 US state Medicaid databases that were available from 1997 to 2018 (Iowa, Kansas, New Jersey, Mississippi, Missouri, and Wisconsin) and followed for up to 2 years of corrected age (CA). Patients were identified by GA using the International Classification of Diseases, Ninth/Tenth Revision, Clinical Modification (ICD-9-CM/ICD-10-CM) codes (Supplementary Table S1). Inclusion criteria included premature infants (<37 wGA) who were born during the data availability period; infants admitted to the intensive care unit (ICU) or neonatal ICU (NICU) within 1 day after birth, where the first hospitalization with ICU/NICU admission is defined as the index (birth) hospitalization; and infants with continuous eligibility from birth to 2 years CA or before death, where CA was defined as: Infants were excluded if they died before reaching 36 weeks postmenstrual age, or if they were diagnosed with congenital heart disease, diaphragmatic hernia, or other major congenital malformations from birth to death, 1 year CA, or end of continuous eligibility, whichever occurred first.
The database contained data related to patient demographic characteristics, enrollment history, and complete medical and pharmaceutical claims from Medicaid records for all eligible patients (including Medicare/Medicaid crossovers), identified by codes listed in Supplementary Tables S2-S9. The first NICU/ICU admission was defined as the index hospitalization. Bronchopulmonary dysplasia was defined as at least 1 diagnosis code of BPD in the ICD-9-CM/ICD-10-CM, during the period from birth to death or end of continuous eligibility (Supplementary Table S7). Chronic lung disease, determined for the EP cohort only, was defined as infants with at least 1 of the following events in 2 consecutive quarters during 1 year CA, excluding the index hospitalization: use of home oxygen therapy or respiratory aids, use of respiratory medications, or presence of respiratory symptoms (cough and wheeze) ( Supplementary Tables S5 and S6).
Overall HCRU, direct charges, and societal costs were calculated from claims records. Subsequently, the numbers of events per personyear (PPY) and counts and percentages of patients with any use in each utilization category were estimated. Healthcare charge adjustments were made by the health plan and were not part of this analysis. The charges given in the claims data were used and adjusted to 2018 USD using the US Medical Services Component of the Consumer Price Index.
For direct charges, the majority of claims were capitation claims, which are fixed advance payments adjusted for each patient based on age and health status, paid monthly by the health plan directly to service providers irrespective of actual services received (eg, the number of visits a patient had). Because the amount paid for a capitation claim does not reflect the costs of the actual services provided, healthcare charges were analyzed individually for each claim.
Because this was a retrospective cohort analysis of existing deidentified patient data and did not involve experiments on humans or the use of human tissues/samples, patient consent and approval from an institutional review board were exempt, per Title 45 CFR § 46.101. 43 Analysis Group implemented all safeguards that were required for the protection of the individuals whose data were included in the study. All methods were carried out in accordance with relevant guidelines and regulations.

Statistical Analyses
Descriptive statistics, including counts and percentages for categorical variables, and mean and SD for continuous variables, were generated. Unadjusted and adjusted generalized linear models with a negative binomial distribution and a log link were used to compare the number of HCRU events and length of stay between GA cohorts up to 2 years CA or death, and models with a Tweedie distribution and a log link were used to compare charges and costs between cohorts. Unadjusted and adjusted logistic regression models were used to compare the presence of HCRU events between cohorts. Adjusted models comparing GA cohorts accounted for demographic characteristics (sex, state of birth, length of index hospitalization, cesarean vs vaginal delivery, birth year, and multiple birth) and comorbidities (BPD, IVH, ROP, periventricular leukomalacia [PVL], necrotizing enterocolitis, bacterial sepsis, and spontaneous intestinal perforation).
Incidence rates of rehospitalizations, emergency department (ED) visits, and outpatient visits after the index hospitalization due to respiratory causes and all causes were calculated by GA up to 2 years CA or death.
Societal costs were estimated for each individual patient by multiplying the average hourly wage (calculated from state-specific average hourly wages from the 2017 US Census data, inflated to 2018 USD) by the hours of work loss incurred for 1 parent: work loss was assumed to be 2 hours for each outpatient visit, 4 hours for each ED visit, and 8 hours for each day in the hospital.
Data were reported by GA cohort: EP (≤28 wGA), very premature (VP) (>28 to <32 wGA), and moderate to late premature (M-LP) (≥32 to <37 wGA) infants. Note that although the World Health Organization definition of "extremely preterm" denotes infants less than 28 wGA, infants born at 28 weeks cannot be separated out owing to a limitation of the ICD-9-CM code, specifically, the ICD-9-CM code 765.24, which includes infants born at 27 to 28 completed gestation weeks; thus, the definition of EP for this study includes infants born up to 28 weeks 6 days GA.

Patient Characteristics
From 1997 to 2018, a total of 25 573 premature infants (46.1% female) were included in the study; 4462 (17.4%) were EP (≤28 wGA), 2904 (11.4%) were VP, and 18 207 (71.2%) were M-LP (Figure 1). Characteristics and demographics during the index hospitalization are shown in Table 1. Sex distributions were similar between GA cohorts. Among race categories, Black infants comprised 32.5%, 28.3%, and 20.1% of EP, VP, and M-LP infants, respectively, while White infants comprised 38.0%, 42.2%, and 46.0% of EP, VP, and M-LP infants, respectively. Index hospitalization length of stay (LOS) increased with decreasing GA: EP infants had a mean LOS of 83.24 days compared with 48.82 days among infants with VP and 20.02 days among infants with M-LP. Follow-up duration was similar across the 3 cohorts.

Comorbidities
The prevalence of comorbidities during the birth hospitalization increased with decreasing GA (  Infants born at less than 37 wGA were identified using ICD-9-CM/ICD-10-CM codes from the US Medicaid claims database (see Supplementary Table S1). b Admission to ICUs/NICUs were identified using hospital standard revenue codes (see Supplementary Table S2). Patients with an index hospitalization greater than 365 days were excluded. c Congenital heart disease, diaphragmatic hernia, and congenital malformation were identified using ICD-9-CM and ICD-10-CM codes (see Supplementary Table  S3) during the period from birth to death, 1 year CA, or end of continuous eligibility, whichever happened first. Infants with a diagnosis of patent ductus arteriosus or patent foramen ovale were not excluded.

JOURNAL OF HEALTH ECONOMICS AND OUTCOMES RESEARCH
In the VP and M-LP cohorts, RDS was also the most prevalent comorbidity; however, a smaller percentage of infants were affected (82.1% and 45.3% of infants, respectively, compared with 91.5% in the EP group). In these cohorts, PDA was present in 19.1% and 4.5% of infants, BPD in 14.4% and 2.4%, IVH in 14.3% and 3.3%, ROP in 10.2% and 2.1%, and PVL in 1.8% and 0.2%, respectively.

Healthcare Resource Utilization
The proportions of infants with rehospitalizations, ED visits, and outpatient visits from index hospitalization to 2 years CA or death were higher among EP infants than infants born at more than 28 wGA.  (Figure 2). Of the total number of rehospitalizations and visits, respiratory causes were responsible for 70.7% of rehospitalizations, 70.6% of ED visits, and 69.2% of outpatient visits, respectively.
Length of stay for all rehospitalizations was higher for infants with lower GA. The mean ± SD LOS for rehospitalizations for any cause was 22
Compared with the EP cohort, all-cost healthcare charges for the EP cohort were 5.4 and 2.3 times higher than for the M-LP and VP cohorts, respectively. Findings were similar for respiratory-related healthcare charges; unadjusted ratios of healthcare charges were 10.3 and 3.5 times higher, respectively. There were no differences between cohorts in healthcare charge cost ratios when demographic characteristics and comorbidities were considered (adjusted ratios of 1.2 for both comparisons); however, for respiratory-related healthcare charges, the adjusted ratios were 1.6 and 1.4 times higher than the EP cohort compared with the ML-P and VP cohorts, respectively.

DISCUSSION
In this study, we evaluated the effect of GA on the prevalence of comorbidities, HCRU, and direct healthcare charges and societal costs associated with preterm birth using birth claims and follow-up claims up to 2 years CA from a large public insurance system in the United States (Medicaid). We found considerable increases in the number of   comorbidities, rehospitalizations, ED visits, outpatient visits, direct charges, and societal costs associated with EP infants vs those born at more than 28 wGA. Respiratory comorbidities were a significant driver of the cost increases. The underdeveloped respiratory system is a major contributor to morbidity and mortality in premature infants. The findings from our analysis underscore the costs associated with respiratory comorbidities such as BPD and the need for improvement in respiratory care to minimize the risk for development of these complications. Although protocols and standard of care vary between healthcare systems in the United States, some quality improvement networks (ie, collaborations within states) have developed more specific guidelines for respiratory support in the EP population. Standardizing care has long been shown to improve outcomes and therefore would be a valuable practice to start toward decreasing the incidence of BPD and its associated long-term consequences.
From our analysis of 25 573 infants born prematurely in 6 US state Medicaid databases from 1997 to 2018, 17.4% were EP (<28 wGA). Of note, the proportion of infants who were Black increased with lower GA, whereas the proportion of White infants increased with higher GA. The proportion of EP infants overall in our study was higher than the previous reported national estimate of 6%. 3 The higher proportion of EP births we found in our data may be related to the low socioeconomic status of Medicaid insurance recipients in the 6 states analyzed. Socioeconomic factors of preterm birth in the United States have been documented 44 ; in the present analysis, 5 of the 6 states included are ranked in the lower half of all US states by median household income (2015). 45 Birth hospitalization LOS was found to approximately double with each 4-week decrease in GA in our study. This is in agreement with a 2006-2007 study of preterm birth in Canada, in which LOS was 83.1 days for infants born at less than 28 wGA, 42.6 days for infants 28 to 31 wGA, and 21.2 days for infants 32 to 33 wGA. 22 Similarly, in a UK study of preterm births 24 to 31 wGA, LOS ranged from 123 days for infants 24 wGA to 66 days for infants 28 wGA. 46 Mean follow-up times were 8% (19 days) longer for infants born at 28 or more wGA, which may be due to increased mortality for EP infants, but these data were not available for confirmation.
In our study, RDS was the most common comorbidity across all GA cohorts, affecting 57.5% of infants; twice as many infants in the EP cohort were affected (91.5%) than in the M-LP cohort (45.3%). This is consistent with recent reviews on complications of prematurity that find respiratory complications among the most common comorbidity in preterm infants. 7,30,31,44 In a retrospective study of 4292 preterm births at 24 to 36 wGA at the Medical University of South Carolina, RDS was the most common comorbidity, affecting 19.8% of infants. 15 Respiratory distress syndrome was reported to increase with decreasing GA in a study of 147 224 singleton preterm births, affecting 58.5% of infants born at 28 wGA compared with 82.3% of infants born at 25 wGA. 18 Bronchopulmonary dysplasia, the next most frequently occurring comorbidity across all GA cohorts in the present study, affected 50.6% of EP infants; the prevalence was 3.5 times higher than in the VP cohort (14.4%) and more than 20 times higher in the EP cohort than in the M-LP cohort (2.4%). Invasive respiratory support is a major contributor to the development of BPD 47 ; thus, its prevalence is intrinsically linked with overall respiratory complications. In EPIPAGE-2, a national, prospective, population-based cohort study conducted in all maternity and neonatal units in France, severe BPD was reported in 25.6% of infants born at 23 to 26 wGA, 4.6% of infants 27 to 31 wGA, and 0% of infants 32 or more wGA. 34 The trend of increased prevalence of PDA, IVH, ROP, and PVL with decreasing GA, observed to varying degrees in this study, is also in agreement with several recent studies. 15,34,48 In the present study, all-cause rehospitalizations were reported in just under half of EP infants (44.8%), compared with a quarter of ML-P infants. For EP infants with BPD, this percentage rose to 51.6%. These findings are higher overall than reported in previous studies, although the trends by GA are consistent. The rates of rehospitalization found by Reed 49.6% for infants with BPD at initial discharge. 50 Our study found that in the first 2 years of life, total healthcare charges for EP infants were significantly higher than for infants born at more than 28 wGA; total all-cause healthcare charges in USD, inflated to 2018 values, were 5.4 times higher in the EP cohort ($843 499) than in the M-LP cohort ($158 430), and 2.3 times higher than in the VP cohort ($379 581). Excluding birth hospitalization, charges associated with the treatment of respiratory comorbidities accounted for 52.7% of the all-cause costs for the EP cohort, vs 43.5% for the M-LP cohort. Healthcare charges have been found to increase with decreasing GA in numerous previous studies. 10,[14][15][16][17][19][20][21][22][23]27,28,50,51 A Canadian study by Bérard et al 14 found that the mean total healthcare costs in the first 2 years of life are almost double for late preterm births compared with full-term births (CAD $2568 preterm vs CAD $1285 full-term; cost ratio, 1.99 [95% confidence interval, 1.90-2.09]); however, data did not address lower GA cohorts. Increasing healthcare costs are also associated with increasing rates of comorbidities as GA decreases. 15,24,48 Black et al 15 modeled the cost of individual comorbidities of preterm births and calculated the increase with each week reduction in GA. Healthcare charges associated with treatment of BPD were estimated at USD $23 132 for an infant born at 24 wGA compared with USD $9356 for an infant born at 36 wGA.
We found that all-cause societal costs increased 1.8-fold for EP infants compared with those born M-LP. However, there is a paucity of reported data comparing societal costs of preterm birth by GA. The Neonatal Adequate Care for Quality of Life study of infants up to age 18 months compared very low birthweight infants without prematurity-related morbidities with full-term infants to assess productivity loss costs during the initial hospitalization. 16 These costs were estimated to be more than 5 times higher for very low birthweight vs full-term infants, and total societal costs of initial hospitalization from birth to discharge were more than 12 times higher. 16

Limitations
Although our findings are in line with previous studies on the health burden and healthcare costs of preterm birth, there are limitations to the current study, including limitations related to the nature of data collected from medical insurance claims. The use of ICD-9-CM and ICD-10-CM diagnosis codes for medical claims approximate actual diagnoses and may not capture all relevant medical diagnoses. Medical services obtained outside the Medicaid system are not captured. Interstate differences in the coding of Medicaid patient data require a non-uniform treatment of the study data. For example, dates of birth of infants from New Jersey and Missouri were not available, due to requirements for de-identification of data; in these instances, the date of birth was approximated using the date of the earliest claim in the birth year. For premature infants identified by ICD-9-CM and ICD-10-CM diagnosis codes, GA was conservatively approximated using the upper bound of the GA range. Because the Medicaid system is a public insurance with strict maximum income eligibility requirements, our results may not be representative of data from private insurers or a broader population.
Limitations of our analytic methods include a variable follow-up period across infants, due to a small number of infants who died be-
JOURNAL OF HEALTH ECONOMICS AND OUTCOMES RESEARCH fore 2 years CA. Additionally, there was a high prevalence of capitation claims throughout the dataset; thus, actual healthcare costs could not be estimated without strong assumptions. Estimates of work time lost (due to ED and outpatient visits) are reflective of the geographic regions analyzed, and could be considered comparatively low, resulting in underestimation of total societal costs. Lastly, premature infants included in this study were identified retrospectively from 1997 to 2018, and it is possible that clinical care changes over that period may have affected the results.

CONCLUSIONS
This retrospective cohort analysis of US Medicaid data demonstrated that presence of comorbidities, HCRU, direct hospital charges, and societal costs from birth to 2 years CA were higher among EP infants than for premature infants born at more than 28 wGA and increased with decreasing GA. A significant portion of healthcare costs is associated with respiratory complications of preterm birth, which increase in prevalence with decreasing GA. Our findings highlight a need for strategies to improve clinical outcomes for EP infants to reduce the burden of comorbidities of prematurity on healthcare systems and families. Funding: This study was funded by Shire, a Takeda company. Although the sponsor had input on the study design, agreed to submit the article for publication, and reviewed the final version of the article for accuracy, the interpretation of the data was made by the authors independently.

Data Availability:
The data that support the findings of this study are available from Medicaid records from individual states, but restrictions apply to the availability of these data due to applicable privacy laws and data use agreements.
The data were used pursuant to data use agreements between Analysis Group, Inc. and the individual states. Requests for data access should be directed to Alexandra Miller, Data Governance, Analysis Group, Inc. (Alexandra.Miller@analysisgroup. com). The data may be available with permission directly from individual states, subject to the state's requirements for data access to Medicaid records.