Economic costs associated with moderate and late preterm birth: a prospective population‐based study

We sought to determine the economic costs associated with moderate and late preterm birth.


Introduction
Late and moderately preterm (LMPT) births account for approximately 6-7% of UK births and 75% of all preterm births. 1 Despite comprising a majority of preterm births, outcomes of those born late (34 +0 -36 +6 weeks of gestation) or moderately (32 +0 -33 +6 weeks of gestation) preterm remain less frequently characterised than outcomes for births at earlier gestations. Published data suggest that LMPT infants are at increased risk of adverse growth, neuropsychological, educational, and behavioural outcomes. 2- 4 The increased risk of adverse sequelae following LMPT birth is likely to translate into economic consequences for health services and other sectors of the economy, 5 the study of which has been highlighted as a priority area for research enquiry. 6 A recent structured review summarising published evidence on the economic consequences of LMPT birth, 6 found that ten studies, published between 1980 and 2011, focused on economic costs during the infant's initial hospitalisation, [7][8][9][10][11][12][13][14][15][16] whereas 13 studies reported economic costs in some form following the infant's initial hospitalisation. 5,7,9,13,[17][18][19][20][21][22][23][24][25] None of the latter studies estimated the economic costs of LMPT birth from a societal perspective, however, with 11 studies limited to an estimation of hospital costs, 5,7,9,13,[19][20][21][22][23][24][25] and with two limited to an estimation of health and social services or public sector costs. 17,18 Moreover, these economic studies were either based on patient cohorts, recruited over a decade ago, 5,7,9,13,17,[19][20][21][22][23][24] or hypothetical cohorts simulated within models. 18, 25 The aim of this study was to estimate the economic costs during the first 2 years of life associated with LMPT birth in the context of a recent prospective population-based study.
The results of our study should be considered for use within economic evaluations of preventive or treatment interventions for LMPT birth, or as inputs to studies attempting to model the economic costs of preterm birth throughout childhood.

Study population
The Late and Moderate Preterm Birth Study (LAMBS) is a population-based, prospective, cohort study of LMPT infants born in the East Midlands region of England between September 2009 and December 2010. 26 All infants born at 32 +0 -36 +6 weeks of gestation, whose mothers resided in this defined geographical region during this period, were eligible for inclusion. A control group comprised infants born at ≥37 +0 weeks of gestation to mothers resident in the same geographical area based on random sampling of the dates and times of birth of infants in this population during the previous year. Alternative methods, such as selecting the next term birth to occur at the same maternity unit after an LMPT birth, are likely to result in an over-sampling of high-risk term births, as these are more likely to be delivered in the same places and times as LMPT births. In view of the large numbers of multiple births that occur between 32 and 36 weeks of gestation, all multiples born at ≥32 +0 weeks of gestation were eligible for inclusion. Written informed consent was obtained from the parents of participating infants. The study was approved by the Derbyshire Multicentre Research Ethics Committee.

Resource use and costs
Relevant resource items were integrated into the LAMBS perinatal and follow-up data collection instruments.
Neonatal and maternal data collection forms captured a comprehensive profile of resource use during the initial hospitalisation for each infant. The total length of initial hospital stay was computed as the total number of hospital days until first discharge home or death. This incorporated any hospital stays following inter-hospital transfers that may have occurred. A clinical researcher (EB) used detailed information on interventions and feeding received by infants in the neonatal unit to map the time spent in the neonatal unit onto days by level of neonatal care (special, high dependency, or intensive), and costs were calculated using the per diem cost of the respective level of care using data from the National Health Service (NHS) Reference Costs Trusts Schedule 2010/11. 27 Non-routine investigations excluded from these per diem costs were valued using a combination of primary and secondary costs. The costs of surgical procedures were calculated by assignment to relevant Healthcare Resource Group (HRG) codes and by the application of unit costs from the NHS Reference Costs Trusts Schedule 2010/11. 27 Transfers were recorded whenever an infant was transported between hospitals for different types of care, and were valued using costs from the NHS Reference Costs Trusts Schedule 2010/11.(27) Postmortem costs were based on data from secondary sources. 28 Delivery and postnatal costs for mothers were based on the method of delivery and costs assigned using data from the NHS Reference Costs Trusts Schedule 2010/ 11. 27 As part of the LAMBS follow-up, parents completed detailed retrospective postal questionnaires about their child's resource use after initial hospital discharge at 6 months, 1 year, and 2 years, corrected for gestational age. This included use of hospital inpatient, day care, and outpatient services, community health and social care services, and medicines and drugs. We also valued adaptations to the home, provision of special equipment, and parental lost productivity attributable to the child's health status. Resource inputs were valued using a combination of primary research, based on established accounting methods, and data collated from secondary national tariff sets. 27, 29 All costs were expressed in pounds sterling and reflected values for the financial year 2010-11. All costs occurring beyond the first year after birth were discounted using the UK recommended discount rate of 3.5%. 30

Statistical analysis
Maternal and neonatal characteristics and resource use items were summarised by gestational age at birth metagroup (LMPT versus term) and subgroup (moderately preterm, late preterm, term). Resource use values were time period adjusted to account for differential lengths of follow-up between children. Differences between groups were analysed using Student's t-tests for continuous variables and v 2 test for categorical variables. Mean (standard error, SE) costs by cost category and mean (SE) total costs were estimated by gestational age at birth. Cost comparisons for LMPT versus term infants were carried out using Student's t-tests, and differences in mean costs and their respective confidence intervals were estimated. Non-parametric bootstrap estimates, 31 based on 1000 bias-corrected replications, were also calculated for these differences in mean costs and their respective confidence intervals calculated. The bootstrap method does not rely on parametric assumptions concerning the underlying distribution of data, hence its usefulness for generating confidence intervals for skewed data. 31 Total costs were estimated from a hospital service perspective for the initial hospitalisation. Total costs were also estimated from: (1) the NHS and personal social services (PSS) perspective recommended by the National Institute for Health and Care Excellence (NICE) in England and Wales for evaluative purposes; 30 and (2) a broader societal perspective, for the entire period of follow-up.
Regression modelling was used to estimate the relationship between gestational age at birth and economic costs. The models were estimated using both ordinary least squares (OLS) and generalised linear modelling (GLM). 32 For the GLM models, a gamma distribution and log link function for costs was selected on the basis of its Akaike information criterion (AIC) statistic, a test commonly used to assess model fit for models of this type. 33 The first set of regression models were estimated with total hospitalisation costs, from birth to initial hospital discharge (or death), representing the dependent variable in the analyses. The first multivariable model (model 1) explored the impact of infant and birth characteristics on initial hospitalisation costs, and covariates (categories presented in results tables) considered in these regression analyses included: gestational age at birth status; place of delivery; mode of delivery; baby survival status; multiplicity; gender; small for gestational age; 34 presence of a congenital anomaly; and first-born status. The second model (model 2) additionally included the following maternal sociodemographic and lifestyle characteristics as covariates: age; body mass index (BMI); ethnicity; highest educational qualification; marital/civil status; social class based on maternal occupational status; home ownership; pre-pregnancy EQ-5D utility score; 35 chronic health problems; socio-economic deprivation (a composite indicator was derived for each postcode sector using the 2010 Index of Multiple Deprivation, IMD, which uses census-derived indicators of income, education, employment, environment, health, and housing at a smallarea level), 36 recreational drug use during pregnancy; and smoking during pregnancy. The third model (model 3) built on model 2 and included two additional variables (previous premature birth and maternal drinking during pregnancy), for which there was a high degree (60.1%) of missing data.
Two further regression models (models 4 and 5) were also estimated with total societal costs, from birth to 24 months (or death), representing the dependent variable in the analyses. The first of these models (model 4) contained all of the covariates included in model 1 except for the baby survival status variable, whereas the second (model 5) included all of the covariates included in model 2 except for baby survival status and maternal BMI. These covariates were excluded from models 4 and 5 as they precluded the estimation of GLM models and we wanted to present results for comparable models using alternative estimators.
Two sets of analyses were conducted: the first included cases with cost data available at all time points (complete cases), whereas the second used the inverse probability weighting method to adjust for the presence of censored data. 37,38 All estimates were additionally recalculated following the weighting of the random sample of term births, to adjust for the over-sampling of term multiple births. All analyses were estimated using STATA 11 (Stata-Corp, College Station, TX, USA).

Descriptive statistics
Of 1376 eligible infants born at 32-36 weeks of gestation, 1146 (83%) were recruited; 1258 infants born at ≥37 weeks of gestation (79% of all eligible) acted as controls. The corresponding number of mothers who gave birth to these infants was 1041 for the LMPT group and 1120 for the term group.
The characteristics of the infants are presented by gestational age group (moderately preterm, late preterm, LMPT, and term) in Table 1. Multiples comprised 19% of moderate preterm, 18% of late preterm, and 2% of term births. Four per cent of moderately preterm and 2% of late preterm infants had a congenital anomaly, compared with 1% for the term group. There were significant differences in the perinatal characteristics of the LMPT infants compared with term infants across all variables. Similarly, comparisons of demographic characteristics of the mothers revealed significant differences in all characteristics, with the exception of marital/civil status and smoking and alcohol consumption during pregnancy (Table S2).

Resource use
Compared with term infants, the median duration of initial stay in the neonatal unit (NNU) and hospital was substantially longer for LMPT infants ( Figure S1). LMPT infants also had significantly higher levels of resource use across all other resource categories, except hospital transfers, compared with term infants (Table S1).
Resource use values between initial hospital discharge and 24 months are presented in Table S3 for the comparator groups, by period of follow-up. For the period between initial hospital discharge and 6 months, the proportion of infants using health services was significantly higher for LMPT infants than for term infants for all categories of hospital services and all categories of community based services, except for GP visits (P < 0.05). There were also significant differences in the proportion of infants prescribed medications or for whom adaptions were made to the home between the study groups. The proportion of infants attending hospital day care and outpatient hospital departments was significantly higher for LMPT infants than for term infants between 6 and 12 months of age. The proportion of infants using health services was also significantly higher for LMPT infants than term infants for all categories of hospital and community based services between 12 and 24 months of age.

Economic costs
Birth to initial hospital discharge The hospital costs up to initial hospital discharge or death are summarised in Table 2 for moderate preterm, late preterm, and term infants. Mean costs were significantly  Initial hospital discharge to 24 months Economic costs between initial hospital discharge and 24 months of age or death (whichever was earliest) are presented in Table S4 for moderately preterm, late preterm, combined LMPT, and term infants, by cost category and follow-up period. For the period between initial hospital discharge and 6 months, all categories of hospital costs were significantly higher for LMPT infants. Mean (SE) NHS and PSS costs were £1351 (£172) and £1073 (£54) for moderately preterm infants and late preterm infants, respectively, compared with £773 (£43) for infants born at term. The corresponding median (IQR) NHS and PSS costs values were £767 (£944) and £694 (£727) for moderately preterm infants and late preterm infants, respectively, compared with £579 (£488) for infants born at term. Mean societal costs were significantly higher for the combined LMPT group than for the term group over this period (£1165 versus £810; P < 0.001). For the period between 6 and 12 months, all hospital services, except visits to an accident and emergency department, and all community based services, except contact with a health visitor and visits to walk-in centres, generated significant cost differences between the comparator groups. For the period between 12 and 24 months, all costs remained significantly higher amongst LMPT infants.
Birth to 24 months of age Economic costs from birth to 24 months of age or death are presented by cost category in Table 3 for moderately preterm (n = 85), late preterm (n = 509), combined LMPT (n = 594), and term infants (n = 716), for whom cost data were available for all follow-up periods. Mean neonatal care costs, hospital care costs, NHS and PSS costs, and societal costs were significantly higher for LMPT infants than for term infants. There were no significant differences in mean community care costs, medication costs, parental lost earnings, special equipment costs, and adaption costs between

<0.001
*All estimates were additionally recalculated following weighting of the random sample of term births to adjust for the over-sampling of term multiple births. **These are comparisons of term versus combined late and moderate preterm groups. ***Comparisons of term versus combined late and moderate preterm groups carried out using Student t-tests.

Regression models for economic costs
Birth to initial hospital discharge Table S6 shows the results of three generalised linear regression models (models 1-3) for costs of total hospitalisation to initial discharge or death (whichever was earliest). The first model explores the impact of infant and birth characteristics on cost, whereas the second and third models also include maternal sociodemographic and lifestyle characteristics, to explore whether they additionally influence hospitalisation costs. Model 1 shows that, compared with term-born infants, the mean (SE) cost ratios for moderate preterm and late preterm birth were 5.02 (0.22) and 1.98 (0.07), respectively. Model 1 also shows that place of birth (centre 2), mode of delivery (assisted/instrumental delivery, caesarean section during labour, and caesarean section not in labour), presence of a congenital anomaly, and whether the infant was firstborn were each associated with elevated costs. The addition of maternal sociodemographic and maternal lifestyle variables in models 2 and 3 had little significant impact on these findings. Table S7 contains the corresponding results for the OLS regressions. Model 1 shows that, even after controlling for infant and birth characteristics, moderate preterm and late preterm birth increased initial hospitalisation costs by an average of £8340 (SE £610; P < 0.0001) and £2224 (SE £282; P < 0.0001), respectively, in comparison with birth at full term. The same patterns of statistical significance are present as in the corresponding GLM model (Table S6).
Similarly, the addition of maternal sociodemographic and maternal lifestyle variables in models 2 and 3 had no significant impact on these findings. Table S8 shows the results of two generalised linear regression models on total societal cost from birth to 24 months or death (whichever was earliest). The results are presented for infants with complete data. The first model (model 4) explores the impact of the reduced set of infant and birth characteristics on total societal costs, whereas the second model (model 5) also includes the reduced set of maternal sociodemographic and lifestyle characteristics to explore whether they additionally influence societal costs. Model 4 shows that, compared with term-born infants, the mean (SE) cost ratios for moderate preterm and late preterm birth were 2.72 (0.20) and 1.55 (0.07), respectively. Model 4 also shows that place of birth (centre 2), mode of delivery (assisted/instrumental delivery, caesarean section during labour, and caesarean section not in labour), presence of a congenital anomaly, and whether the infant was firstborn were each associated with elevated costs. The addition of maternal sociodemographic and maternal lifestyle variables in model 5 had no significant impact on these findings. Table S9 contains the corresponding results for the OLS regressions. Model 4 shows that, even after controlling for infant and birth characteristics, moderate preterm and late preterm birth increased total societal cost from birth to 24 months by an average of £7730 (SE £886; P < 0.0001) and £2079 (SE £370; P < 0.0001), respectively, in comparison with birth at full term. The same patterns of statistical significance are present as in the corresponding GLM model (Table S8). Table S10 shows the results of two generalised linear regression models on total societal cost from birth to 24 months of age or death (whichever was earliest), using the inverse probability weighting method to account for the presence of censored data. 37,38 These analyses replicated the model specifications applied to the complete case analyses (Table S8). The same patterns of statistical significance present in model 4 for the complete case analysis are found in model 4 for the inverse probability weighted data. The addition of maternal sociodemographic and maternal lifestyle variables in model 5 resulted in one additional statistically significant effect. An unknown maternal ethnicity status resulted in a mean cost ratio of 0.73 compared with the white reference group. Table S11 contains the corresponding results for the OLS regressions. Model 4 shows that, even after controlling for infant and birth characteristics, moderate preterm and late preterm birth increased total societal cost from birth to 24 months by an average of £7607 (SE £875; P < 0.0001) and £1859 (SE £323; P < 0.0001), respectively, in comparison with birth at full term. The same patterns of statistical significance are present as in the corresponding GLM model (Table S10). The addition of maternal sociodemographic and maternal lifestyle variables in model 5 resulted in maternal age displaying a statistically significant effect, with each additional year of maternal age increasing costs by an average of £84 (SE £42; P < 0.05).

Main findings
A recent review of published evidence on the economic consequences of LMPT birth found sparse evidence on this topic. 6 No recently published studies have, to our knowledge, disentangled the economic costs associated with LMPT birth, by period of follow-up and cost category, in a UK setting. Our main findings are that, during the initial hospitalisation, LMPT infants have higher rates of resource use compared with term infants across all hospital service resource categories, and that mean costs are significantly higher for LMPT infants than for term infants across all categories of hospital services. Mean (SE) hospital costs to initial discharge or death were estimated at £11 629 (£605) and £4528 (£336) for moderately preterm infants and late preterm infants, respectively, compared with £1864 (£36) for infants born at term. Our study also generated mean (SE) total societal costs estimates over the first 2 years of life of £12 037 (£1114) and £5823 (£1232) for moderately preterm infants and late preterm infants, respectively, compared with £2056 (£132) for children born at term.

Strengths and limitations
The strengths of this study lie in the fact that it was based on a large geographically determined prospective population cohort, included a term comparison group, and captured a comprehensive profile of resource use between birth and 2 years of age, based on previously validated instruments. 39,40 Furthermore, the rigorous costing methodology applied followed national guidance for health economic evaluation purposes. 30,41 A number of caveats must be borne in mind. First, because the study was based in the East Midlands of England, it may not be representative of other populations of LMPT infants, nor indeed of the breadth of organisational structures for perinatal care across industrialised nations. Nevertheless, as we were able to collect information about the numbers of infants that were not recruited, together with denominators for 'all births', it has still been possible to create reasonable underpinning epidemiological estimates for the population, which should be suitable for generalisation. Moreover, we are unaware of any evidence to suggest that alternative organisational structures would have significantly affected the cost differences between the comparator groups. Second, our study only considered hospitalisation costs during the period between birth and initial hospital discharge, whereas a broader societal perspective for economic costs over this initial time horizon might also appropriately consider costs borne by parents and informal carers. A recent structured review of the economic costs associated with preterm birth highlighted the importance of non-healthcare costs associated with the initial period of hospitalisation, such as parental travel to neonatal units and costs associated with lost parental productivity, and it is likely that these categories of costs are relevant to some families and carers of children born LMPT. 42 Third, the study covered a time horizon of birth to 2 years of age, whereas a longer time horizon could have captured the economic consequences of potential longerterm sequelae, such as physical, neuropsychological, and behavioural difficulties. 2- 4 The economic consequences of LMPT birth are likely to continue in infants with long-term adverse sequelae, and it is possible that the need for special educational support in this group is substantial but not yet fully recognised.

Interpretation
The comparability of our cost estimates related to initial hospitalisation with other studies is constrained both by a paucity of broader evidence and by a number of methodological factors, including variability in cohort dates, the way perinatal and neonatal care is organised, and differences in accounting procedures. In contrast, a significant body of literature exists on the initial hospitalisation costs associated with very or extreme preterm birth. For very preterm infants, estimates of mean costs associated with the initial hospitalisation have varied between $29 679 (2007 prices) and $91 343 (price date not stated). 10,16 For extremely preterm infants, estimates range between $11 397 (2007 prices) and $195 254 (price date not stated). 7,10,16 Although these cost estimates are generally greater in magnitude than our own, disentangling the relative contributions of gestational age, in itself, and the methodological factors underpinning alternative cost study designs, is beyond the scope of this study.
Similarly, the comparability of our cost estimates over the first 2 years of life across studies is complicated, particularly by differences in follow-up periods, types of costs included, and accounting procedures. There are no directly comparable results for extremely or very preterm infants. The incremental societal cost of LMPT birth from birth to 2 years of age was estimated at £7583 (£874) and £1963 (£337), respectively, compared with birth at full term. In 2011, there were 6916 and 35 565 infants born moderately preterm and late preterm, respectively, in England and Wales. 43 The application of our incremental societal cost estimates to these epidemiological data translates to an annual national economic burden of LMPT birth of approximately £122 million.

Conclusion
Compared with birth at full term, LMPT birth is associated with significant additional costs during the period of the initial hospitalisation and throughout the first 2 years of life. Clinical decision-makers and budgetary and service planners should recognise the overall economic impact of LMPT birth in their service planning, as well as the potential contribution of clinical and sociodemographic factors to future hospital and broader societal costs. The results of this study support an increasing evidence base that suggests that delivering babies LMPT is neither risk-free nor without significant cost, economic as well as medical, and there is potential to reduce iatrogenic early delivery. In addition, economic variables should be incorporated into future longitudinal studies of moderate and late preterm infants with the view to estimating the economic costs of LMPT birth over the longer term. The results of our study should also be considered for use as inputs to studies attempting to model the economic costs of preterm birth throughout childhood. Moreover, our results should act as an input into economic evaluations of preventive and treatment interventions for LMPT birth that aim to ensure the efficient allocation of finite resources in this area.

Disclosure of interests
Full disclosure of interests available to view online as supporting information.

Contribution to authorship
KAK carried out the bulk of the analyses, reviewed and revised the article, and approved the final article as submitted. SP designed the economic study and its data collection instruments, coordinated and supervised data collection, critically reviewed the article, and approved the final version for publication. MD carried out some of the cost analyses, reviewed and revised the article, and approved the final version. SJJ conceptualised and designed the study, reviewed and revised the article, and approved the final version. BM conceptualised and designed the study, reviewed and revised the article, and approved the final version. ESD conceptualised and designed the study, reviewed and revised the article, and approved the final version. LKS conceptualised and designed the study, reviewed and revised the article, and approved the final version. SES conceptualised and designed the study, reviewed and revised the article, and approved the final version. NM conceptualised and designed the study, reviewed and revised the article, and approved the final version. JD conceptualised and designed the study, reviewed and revised the article, and approved the final version. DJF conceptualised and designed the study, reviewed and revised the article, and approved the final version. EMB conceptualised, designed, and provided overall oversight of the LAMBS study, and its ancillary studies, reviewed and revised the article, and approved the final version.

Details of ethics approval
This is a secondary analysis of a pre-existing data set with non-identifiable data, so no ethics approval was required.

Supporting Information
Additional Supporting Information may be found in the online version of this article: Figure S1. Duration of initial hospitalisation by gestational age group. Table S1. Resource use by birth status (all births), from birth to hospital discharge. Table S2. Summary descriptive statistics for mothers by gestational age at birth (all births). Table S3. Resource use by birth status (all births) from initial hospital discharge to 24 months. Table S4. Economic costs of late and moderate preterm and term infants from initial hospital discharge to 24 months. Table S5. Economic costs of late and moderate preterm and term infants from initial hospital discharge to 24 months. Table S6. Relationship between gestational age at birth and total birth with initial hospital discharge costs; generalised linear model (gamma distribution with log link). Table S7. Relationship between gestational age at birth and total birth with initial hospital discharge costs; ordinary least squares model. Table S8. Relationship between gestational age at birth and total birth with costs for 24 months; generalised linear model (gamma distribution with log link), complete cases. Table S9. Relationship between gestational age at birth and total birth with societal costs over 24 months; ordinary least squares model (complete cases).
Table S10. Relationship between gestational age at birth and total birth with costs over 24 months; generalised linear model (gamma distribution with log link), inverse probability weighted.
Table S11. Relationship between gestational age at birth and total birth to 24 months societal costs; Ordinary Least Squares Model (inverse probability weighted). &