DA Systematic Review of Costs Associated with Preterm Birth

Publication Details

John A. F. Zupancic1

The costs of prematurity have been the subject of close scrutiny for more than 2 decades. During that time, there have been dramatic advances in therapeutic interventions, improvements in mortality, and significant inflation in medical care costs. Although the neonatal intensive care phase is most prominent in discussions of these costs, the sequelae of prematurity constitute a chronic disease with significant long-term implications. This appendix has two parts: the first examines the literature on costs associated with the initial hospitalization for prematurity, whereas the second examines longer-term economic factors.

To evaluate the quality and applicability of cost-of-illness studies, it is important to have agreed-upon definitions of certain technical terms. There are, for example, several different categories of cost that might be important to stakeholders. Direct medical costs arise from resources directly consumed by patients. In the context of this review of costs of initial hospitalization, these include costs for the hospital bed and support services (“accommodation”), ancillary services (pharmacy, radiology, laboratory, and respiratory care services), and professional fees. In contrast to these direct costs, overhead costs are those expenses that are incurred in running the institution but that are not directly attributable to a particular patient. These include overhead expenditures for housekeeping, administration, and the physical plant. Costs to patients include nonmedical direct costs (such as the cost of transportation to the hospital, meals, and child care for other siblings). Finally, there are costs related to lost productivity, also referred to as indirect costs. In the short run, these are the lost contributions of labor by the parents; in the long run, they also include reduced work alternatives for the patient.

The particular choice of which costs to include is dependent on the perspective of the stakeholder who will be using the information. For example, hospital administrators will be most interested in the direct medical costs that affect hospital financial operations and less interested in wage losses by parents. However, in valuing the financial burden of prematurity to society, it is vitally important to understand the full extent of costs, regardless of the individuals to whom they accrue. In the absence of such information, policy makers may make decisions that appear to be fiscally desirable but in fact simply shift costs to another, unmeasured facet of care.

Studies that measure costs on their own without reference to the effectiveness of an intervention are known as “cost-of-illness studies.” When costs are combined into a single metric with a measure of effectiveness, such as life years saved, the study is referred to as a “cost-effectiveness analysis.” When patient preferences for particular outcomes are used as the measure of effectiveness, as they are in quality-adjusted life years, the study is a known as a “cost-utility analysis.” Both cost-effectiveness and cost-utility analyses express costs and effects as a ratio measure. In contrast, cost-benefit analyses measure effects in monetary terms and subtract this value from the cost of the intervention to determine value for money.

METHODS

Population Description

Although much attention has been focused on infants at the borderline of viability, the large volume of children at higher gestational ages may drive the overall financial impact of prematurity. Therefore, broad inclusion criteria were maintained for the review and encompass children in the following categories: (a) premature, defined as birth at less than 37 weeks of gestational age, and (b) low birth weight, defined as a birth weight of less than 2,500 grams.

Although the assessment of efficiency is a critical component of technology assessment, the focus of this appendix is on the absolute financial burden of prematurity and low birth weight rather than on the value for money of treatment modalities. Therefore, only costing studies are included and cost-effectiveness, cost-utility, and cost-benefit studies are excluded, except when the population measurement of pertinent costs was a central component. In some cases cost descriptions of cohorts alongside trials of efficacy of therapeutic interventions are included, but these are restricted to control groups to optimize external validity.

Two factors have significant impacts on the generalizability of the findings from economic studies. First, the therapies offered in the field of neonatal intensive care and their resulting outcomes have changed substantially in the past 2 decades. In the late 1980s, there was a dramatic decrease in the rate of mortality because of prematurity with the introduction of surfactant replacement therapy, a therapy subsequently shown repeatedly to have important cost implications. Because the underlying structure of care has changed, this problem is difficult to overcome. For the review of short-run costs, therefore only studies reporting on cohorts after 1990 are included to optimize the external validity of the results and avoid bias against lower birth weight in incremental cost comparisons. Since follow-up care of infants following discharge has changed to a lesser degree, reports of long-run costs are included if the cohort was born after 1980. In a few cases, cohorts were assembled with patients from before and after 1980; these studies are included if >50 percent of the cohort met the inclusion date criterion.

Similarly, substantial effort has been made to assess the costs of neonatal intensive care in other countries. In some cases, as in the United Kingdom, Europe, and Canada, health care delivery and financing may have certain similarities that allow extrapolation of the results to the U.S. population. However, studies from middle-income and developing countries were excluded on the grounds that both outcomes and health care delivery cost structures are too different to interpret the applicability of the results in the U.S. context.

Data Sources and Search Strategies

The following databases were searched for candidate articles: Medline (1990 to 2005), electronic abstracts from the annual meetings of the Society for Pediatric Research (1998 to 2005), Econlit (1990 to 2005), and the Proquest dissertation and theses full-text index (1990 to 2005). The Science citation index was searched for articles citing any of the first authors of included studies. The bibliographies of other systematic or informal reviews were scanned (1–7). The websites of organizations known to be involved in similar research were also scanned; these included the websites of the March of Dimes, the Alan Guttmacher Institute, and the Centers for Disease Control and Prevention. The search strategies were intentionally broad to ensure optimal sensitivity. The intersection of the exploded MeSH headings “infant, newborn” and “costs and cost analysis” was used for Medline searches. For other databases, the keywords “econom*” OR “cost*” AND “infant” were used. The titles and abstracts of the articles retrieved with these strategies were manually screened to determine whether they met the inclusion criteria described above.

Cost Projections

Cost estimates were inflated to 2005 constant dollars by using both the medical component of the consumer price index (8) and the hospital producer price index (9). Currencies were converted to U.S. dollars by using purchasing power parities (10). Per-patient estimates of cost were projected to the U.S. birth cohort by multiplying by the number of infants in a given gestational age or birth weight category, according to the 2002 natality file of the National Center for Vital Statistics (11). State-level population cost estimates were projected to the U.S. birth cohort by dividing by the ratio of the state’s birth cohort to the total number of births nationally for 2002 (the most recent year for which such statistics were available).

YIELD OF SEARCH STRATEGY

The manual screening of titles and abstracts from the initial literature searches yielded 170 articles. Of these, the majority were eliminated because they were review articles or economic evaluations of interventional studies, were based on data from middle-income or developing countries, were studies at the neonatal intensive care unit level not stratified by birth weight or gestational age, or had cohort dates prior to 1990 for short-run cost reports or 1980 for long-run cost reports. Descriptions and methodological assessment of the short-run (12–27) and the long-run (15, 27–41) cost studies retrieved are shown in Tables D-1 and D-2, respectively.

TABLE D-1A Studies of Costs During Initial Hospitalization.

Table

TABLE D-1A Studies of Costs During Initial Hospitalization.

TABLE D-2A Studies of Costs Following Discharge from Initial Hospitalization.

Table

TABLE D-2A Studies of Costs Following Discharge from Initial Hospitalization.

METHODOLOGICAL LIMITATIONS OF STUDIES RETRIEVED

Studies of Initial Hospitalization

The methodological quality of the short-run cost studies since 1990 was variable. Only five of the reports involved the collection of data alongside prospective research projects (12, 16, 23, 24, 27), whereas the remainder were retrospective analyses of administrative and clinical data collected for other purposes. In some cases, particularly those analyses involving statelevel data in California, there was a preexisting process for checking the integrity of data through reabstraction (18, 25). However, the accuracy of the smaller secondary data sets was not reported in the publications.

TABLE D-1B Studies of Costs During Initial Hospitalization.

Table

TABLE D-1B Studies of Costs During Initial Hospitalization.

TABLE D-2B Studies of Costs Following Discharge from Initial Hospitalization.

Table

TABLE D-2B Studies of Costs Following Discharge from Initial Hospitalization.

The inconsistent level of regionalization in neonatal care delivery makes the nature of the cohort critically important for cost analyses of prematu-rity. Six of the reports (12, 16, 18, 21, 25, 27) involved cohorts that were geographically based at the state or regional level. Others used claims data from nationally representative insurance organizations (13, 15) or data from a large number of self-selected hospitals in a neonatal quality improvement network (23, 24). Although the latter data do not guarantee freedom from selection bias, the large number of patients and hospitals involved makes such bias less likely than the likelihood of selection bias with the two small single-center studies (20, 26). In most cases, the inclusion criteria were adequately described; however, three studies defined gestational-age exclusively by diagnosis-related group (DRG) codes (13, 15, 21) rather than by direct observation of patient information. Because there is some discretion in the use of some DRG codes (specifically, “normal preterm” and “preterm with problems”), generalization of the findings to a larger population is more difficult.

Because of the very low incidence of extreme prematurity and the marked variability and skew of cost data, it is also important that studies have adequate sample sizes. As these studies were targeted toward the estimation of costs rather than inference, it is difficult to define an appropriate sample size a priori. The U.S. state-level geographic studies (18, 21, 25), insurance-claims studies (13, 15), and neonatal-network studies (23, 24) all based cost estimates on at least 1,000 premature infants. In contrast, the hospital-specific studies (19, 20, 26) and non-U.S. studies (12, 16, 27) examined as few as 105 premature infants, a sample size that is unlikely to yield useful confidence intervals, especially when the data are projected to national levels.

Earlier cost-of-illness studies in neonatalogy often relied on hospital charges, as these were easily available in administrative databases. However, the use of charges as a measure of economic cost may distort cost estimates because charges reflect a hospital’s decisions regarding markups and internal cost transfers. Information on actual costs may be derived from proprietary cost accounting systems or from conversion factors known as cost-to-charge ratios, which are mandated for the reporting of cost information to federal agencies. Two studies in the current review relied only on billed charges (15, 21), one relied on paid insurance claims (13), two relied on older costing studies corrected for inflation (12, 16), and two referred to costs but did not specify the means of derivation (20, 27). The remaining six studies used Medicare cost to charge ratios and are likely to be more comparable to each other.

Given the issues of variance in cost data and the number of comparisons made between groups of infants in the costing studies, it is particularly surprising that they relied almost entirely on point estimates without any quantification of statistical or parameter uncertainty. Statistical uncertainty was reported in only three studies (20, 25, 27), and sensitivity analyses were performed in three studies (12, 16, 27).

Because almost any medical care involves a net expenditure of resources, it is important that the costs of prematurity be viewed as incremental on the medical costs for neonatal care had the premature infant been delivered later in gestation. Five of the studies did not include a cohort of infants delivered at full term (12, 16, 19, 23, 24).

In regionalized systems of neonatal care, infants are often transferred to other hospitals for higher levels of care or to less intensive care settings for a period of further maturation following the resolution of acute cardiorespiratory instability. Failure to include posttransfer costs would lead to underestimates of the cost of care, whereas failure to include transferred infants would affect the external validity of the estimates and possibly overestimate treatment costs, as retrotransferred infants are more stable. Three studies explicitly noted that transferred infants were excluded (19– 21), two explicitly included them (15, 25), and the remainder did not specify the inclusion or exclusion of transferred infants.

Several authors also excluded nonsurviving infants from cost estimates (18, 21). One article that did measure the costs for nonsurviving infants observed that they represented 9 percent of total costs (27). Because nonsurviving infants have shorter lengths of stay, they are likely to be less costly (24, 26), even with increased per diem resource utilization due to their higher illness acuity. Including only the costs accruing to survivors will yield underestimate, as resources must also be invested in nonsurvivors to yield a surviving infant.

The currency date was not provided in seven of the articles; when this information was necessary for calculations, the most recent year of the cohort was assumed.

Studies of Costs Following Discharge from the Initial Hospitalization

The longer-run costing studies were similarly methodologically variable. Several of these limitations overlapped with those of the short-run studies.

Most of the studies were secondary analyses of data collected for other purposes. One study was a randomized trial with prospective data collection (28), six used linked governmental vital statistics and administrative or survey databases (27, 29, 31, 33, 34, 38), four used commercial or governmental claims databases (15, 36, 37, 41), three used clinical data (35, 39, 40), and two collected primary data through patient interviews (27, 30, 32). For preexisting data sets, validation was generally not described.

The majority of long-run studies were from broadly representative cohorts. These geographically based cohorts were at the regional (27, 33–35, 39, 40), state (36, 38), and national (29, 31) levels. Only six smaller studies used convenience or hospital-specific samples (15, 28, 30, 32, 37, 41).

There were, however, other problems with generalizability. Six of the 16 long-run studies were from countries other than the United States (27, 33–35, 39, 40). Of note, these were all geographical cohorts and represented the only studies in the review with assessment of direct medical costs beyond 2 years of age. Generalizability was also compromised by the broad time range of the studies. As noted above, the literature search was restricted only to studies conducted after 1980, given that postdischarge care is not as technologically sophisticated and therefore not as rapidly changing. This decision also allowed the inclusion of longer periods of follow-up. The cohort dates of these studies, however, ranged from 1980 to 2001; and the degree of comparability or applicability of earlier studies to the current context is unclear. Most studies did specify the birth weight and gestational age inclusion criteria for the samples, which allowed projection of the results for health policy. Studies that relied on claims databases used DRG designations, which must be applied with a degree of discretion, thus limiting external validity (15, 41).

In contrast to the studies of initial neonatal hospitalization, the sample sizes of most of these studies were adequate, possibly because of the greater reliance on administrative data. Four of the 16 studies had sample sizes of >20,000 infants (15, 31, 33, 34). Although these were heavily weighted toward full-term controls, the sizes of the subsamples of infants born preterm still comprised several hundred to several thousand infants. Only five studies had sample sizes of <150 infants (27, 30, 32, 35, 40). Although the confidence intervals around the cost estimates are consequently broad, most of these studies involved direct patient interviews or questionnaires and yielded unique information that would otherwise not have been obtainable.

Ten of the studies stated that they used costs derived from cost-to-charge ratios or other cost-accounting systems. None of the authors conducted primary costing research, and direct validation of preexisting cost-accounting systems was not presented; however, in at least some of the studies, these appear to be well-established and accepted systems (33, 34). The validity of patient-reported costs is equally difficult to ascertain. None of the studies presenting data from patient interviews, diaries, or questionnaires provided validation of these sources (27, 30, 32).

Uncertainty was again assessed inconsistently. Seven of the studies presented no information on sampling or parameter uncertainty, either through descriptive statistics or sensitivity analysis (15, 31, 35, 36, 39–41).

The assessment of costs for premature infants over the longer term is a more complex exercise, requiring the assembly of data from multiple sources. In some studies, this led to a large number of assumptions regarding the applicability of data, the choice of unit costs, and natural history (29, 31). Although these assumptions were well explained, it is very difficult to know the extent to which they are valid.

Unlike the assessment of initial hospital costs, the study of postdischarge resource utilization ideally tracks costs over several years. Most individuals have a positive rate of time preference; that is, they prefer to receive benefits sooner rather than later and to defer negative factors such as costs. To take account of this preference, economists apply a discount rate to reduce the weight of future investments in a stream of costs over time. This is variably set at 0 to 5 percent, with a 3 percent rate generally being considered reasonable for application in the base case. Only 5 of the 16 studies in the present review applied a discount rate (28, 29, 35, 39, 40).

CONTENT LIMITATIONS OF RETRIEVED STUDIES

Studies of Initial Hospitalization

Cost Categories

Cost categories in the studies of initial hospitalization costs for prematurity were variable. Only one study included both direct nonmedical costs and productivity losses (27). Only three studies included professional fees as a component of direct medical costs (13, 15, 26). The inclusion of overhead costs could be inferred in most cases but was usually not specified.

Only one author disaggregated costs into their component parts, in two studies. Rogowski showed that for infants with birth weights of <1,500 grams, accommodation costs represented most of the total cost of care (72 percent). Of the remaining ancillary costs, 22 percent was for respiratory therapy, 24 percent was for laboratory work, 7 percent was for radiology, and 16 percent for pharmacy (24).

Maternal Costs

This review was not specifically targeted to maternal costs, but these were measured in several of the studies reviewed (13, 15, 18, 20, 21). This is an important potential confounder, as the marginal contribution of maternal care may introduce bias in the comparison of costs for various gestational ages. For example, for infants born at from 25 to 29 weeks gestation, maternal costs were approximately 8 percent of the neonatal costs, for infants born at from 30 to 34 weeks gestation they were 32 percent, for infants born at from 37 to 38 weeks gestation they were 192 percent (18). Corresponding U.S. population projections were as follows: for infants born at 25 to 29 weeks gestation, $100 million; for infants born at 30 to 34 weeks gestation, $429.5 million; and for infants born at 37 to 38 weeks gestation $1.94 billion. Adams et al., found that maternal costs were $7,451 for those with full-term infants, $10,626 for those with healthy preterm infants, and $11,508 for those with extreme preterm infants (13). Chollet et al. found that maternal costs were $7,850 for the full-term DRG, $13,017 for the healthy preterm DRG, and $14,815 for the extreme preterm DRG (15). Thus, maternal costs vary inversely with birth weight, in parallel with neonatal costs, but to a lesser degree.

The articles that do include maternal costs list only the cost for the delivery admission; other costs are often associated with premature delivery (closer monitoring, repeat admissions, etc.) that may not be captured and would inflate the costs of the delivery itself. In an exception to this pattern, Schmitt et al. showed that, for pregnancies that resulted in births of infants less than 1,500 grams, prenatal hospitalizations accounted for $6.4 million of a total of $49.6 million of maternal hospital expenditures in California (25). Infertility treatments were also not captured in the studies reviewed.

Birth Weight and Gestational Age Cohorts

The majority of studies reported only on very low birth weight infants (in some cases with full-term controls). Although the per-infant cost is significantly lower for moderately preterm infants, the much higher incidence of moderate prematurity results in significant population costs. Schmitt et al. showed that infants with birth weights of 1,500 to 2,500 grams had total projected costs of $2.46 billion nationally (25). St John et al. projected that the costs for infants born at 30 to 34 weeks of gestation were $2.29 billion (26).

No studies determined the incremental contribution of low birth weight after controlling for gestational age.

Financial Burden on Minorities

Although certain ethnic and racial groups experience an increased burden of prematurity, the cost studies reviewed here did not specifically examine the issue of which populations are associated with the initial hospitalization costs.

Financial Burden on Payers

The studies reviewed did not provide specific disaggregations of the sources of payment for care for prematurity. Two of the studies relied on claims databases from large third-party payers. Chollet et al. calculated that prematurity accounts for $4.7 billion to employer-sponsored health insurance plans, beyond the costs for normal-term deliveries (15). Older studies suggested that the newborn period is a major source of uncompensated care (42), accounting for as much as 19 percent of the direct delivery and uncompensated care dollars allocated to maternity and neonatal care (43). The March of Dimes estimates that 11 percent of newborns covered by employer health plans are born premature and that $7.4 billion (in 2002 dollars) in hospital charges was billed to private insurers (44).

Studies of Costs Following Discharge from the Birth Hospitalization

Cost Categories

Thirteen of 16 studies included direct medical costs (15, 27, 28, 31–37, 39–41). However, three of these included only rehospitalization costs and omitted the costs of outpatient physician visits or pharmaceuticals (31, 33, 34). Several of the studies failed to disaggregate the costs of rehospitalization from outpatient care (35, 37) or the costs of initial hospitalization from those of postdischarge care (31).

Educational costs were assessed in only four studies (27, 29, 31, 38). Two of these dealt with preschool costs (27, 31), one dealt with kindergar-ten costs (38), and two dealt with educational costs beyond kindergarten (29, 31). None followed an incidence cohort of infants longitudinally through school or assessed costs throughout the school spectrum by a consistent methodology.

Parental out-of-pocket expenditures were assessed in five studies (14, 17, 22, 27, 30). All of these included travel costs. Only one, non-U.S. study examined parental wage losses (27). None of the studies retrieved examined the lost productivity of the child.

Only one study attempted to ascertain costs in the direct medical, educational, and parental out-of-pocket expense categories (27).

Time Horizon

Only four of the studies tracked costs beyond 5 years of age. Despite the availability of cohorts reporting clinical follow-up to young adulthood (45, 46), there is almost no information on the cost implications of prematurity beyond early childhood.

Birth Weight and Gestational Age Cohorts

In contrast to the studies of initial hospitalization, most studies of longer-term costs included infants across the birth weight and gestational age spectra.

Comorbid and Resulting Conditions

The studies retrieved did not make any attempt to attribute postdischarge costs to particular neonatal or postnatal conditions other than birth weight or gestational age. Thus, limited information is available on the proportion of rehospitalization or outpatient costs that are associated with chronic lung disease, retinopathy of prematurity, or cerebral palsy. There are, however, estimates of the costs of certain of these (47–52) conditions separately, without the prematurity denominator afforded by the studies reviewed here. Studies of premature infants with chronic lung disease, for example, show that the costs for infants with admissions for respiratory syncytial virus infection are significantly higher; however, the proportion of costs for the entire birth cohort cannot be derived directly (51, 52). Similarly, although costs are available for cerebral palsy, most children with this condition were not born prematurely; the extent to which the costs for cerebral palsy drive the longer-term costs for premature infants is not known.

CHARACTERIZATION AND ESTIMATES OF COSTS OF INITIAL HOSPITALIZATION

Direct Medical Costs

Despite the heterogeneity of methods, designs, and sources, estimates of the per-patient cost are moderately consistent. The second column of Table D-3 shows the per-patient costs derived from the studies, expressed in 2005 U.S. dollars. For infants with birth weights of 500 to 999 grams, cost estimates ranged from $67,027 to $221,450, with four of six estimates in the range $84,847 to $126,380 per infant.

TABLE D-3. Direct Medical Costs During Initial Hospitalization.

TABLE D-3

Direct Medical Costs During Initial Hospitalization.

There are several potential sources of variability in cost estimates. First, the studies were performed with cohorts spanning a 15-year period. The characteristics of the adjusters for differences over time are therefore significant. The results are reported with both the medical component of the consumer price index (CPI), which relies on charges and may therefore overestimate charges, and the hospital producer price index (PPI), which used the actual net costs of resources, purchased by both public and private sources. As shown in Table D-3, use of these indices had a minimal effect on the results, suggesting that the source of variation lies elsewhere. A second potential source of variability in estimates is in the geographic diversity of the cohorts. International comparisons are facilitated to some extent by the Organisation for Economic Co-operation and Development purchasing power parity, although this accounts only for differences in price and not differences in practice style. Within the United States, neonatal care is fairly homogeneous, so it is likely that price differences will outweigh geographic practice style differences. An adjustment for geographic differences was not made because of the difficulty in defining the location of care for most of the cohorts, which were often either at the state level or at institutions in several states. Finally, variability arises because of the authors’ choices of which types of costs to include. Unfortunately, most authors did not disaggregate costs into components, thus making an assessment of the contributions of these choices to the cost estimates difficult to quantify. Table D-3 presents neonatal costs separately from maternal costs, showing a greater degree of variability for the former.

Because it is difficult to fully address variability in prices, an attempt was also made to describe resource utilization independent of its price. Table D-3 shows the length of stay (LOS) and the number of days of assisted ventilation for the studies reviewed. Although not all studies reported these measures, the variability in the direct markers of resource utilization appears to be somewhat less than that for cost. The LOSs for infants with birth weights of 500 to 999 grams ranged from 64 to 106 days in six studies, with four studies showing LOSs in the narrow range of 73 to 82 days.

Studies that reported both birth weight- and gestational age–specific costs show that these are consistent. Gilbert et al. give a per-patient cost of $202,700 for a survivor born at 25 weeks gestation and a per-patient cost of $224,400 for a survivor born with a birth weight of 500 to 750 grams (18). For survivors born at 30 weeks gestation, the per-patient cost is $46,400, whereas for a survivor with a birth weight of 1,250 to 1,499 grams, the per-patient cost is $51,900. These authors excluded infants whose birth weights and gestational ages did not match; the birth weight-specific costs may therefore be less biased, as the gestational age on vital statistics records is more often likely to be inaccurate. Across all studies, under the assumption that DRG Code 387 for extreme prematurity corresponds to gestational ages of <28 weeks, costs ranged from $70,451 to $100,389, similar to those for the corresponding birth weight range of 500 to 999 grams.

When the costs were projected to the U.S. population, the range for infants born at <37 weeks gestation was $4.62 billion to $13.38 billion, with the latter estimate derived from a study based on charges (15). For infants with birth weights of 500 to 999 grams, the costs ranged from $1.53 billion to $5.06 billion, with the range for the U.S. population being $2.72 billion to $5.06 billion.

Full-term controls had per-patient costs of $734 to $4,303. If the stud ies that used charges or failed to specify this were excluded, the range for the U.S. population was $1,545 to $1,900. Corresponding population projections were $2.37 billion to $6.58 billion.

The previously noted inverse relationship between costs and gestational age or birth weight was observed in all studies. The marked right skew of the cost data was also noted in those studies that reported means and medians (23–25, 27).

Direct Nonmedical Costs and Lost Productivity

Table D-4 describes studies that assessed parental out-of-pocket expenditures and lost productivity. In the single study that examined both non-medical direct costs and productivity losses, parental mean costs before discharge for extremely low birth weight infant were €2,755, or 4 percent of total costs (27). These were composed of 64 percent travel costs, 30 percent lost earnings, and 6 percent accommodation. An earlier interview study of travel costs for mothers of 109 low-birth-weight infants in Britain showed that 36 percent of families traveled more than 21 miles, that 88 percent of families visited the infants daily, and that the median total expenditure was between £101 and £200 (in 1990 United Kingdom pounds sterling), with a maximal expenditure of £1,000 (22). More recently, Gennaro observed the families of 224 infants born at less than 37 weeks of gestational age discharged from a single center between 1990 and 1994 (30). That study did not detail the productivity losses during the initial hospitalization. The average mother in that study used 4 weeks of maternity leave during the initial hospitalization. Out-of-pocket expenses averaged $433, with the largest category being transportation.

TABLE D-4 Parental Out of Pocket Expenses and Lost Productivity During Initial Hospitalization.

Table

TABLE D-4 Parental Out of Pocket Expenses and Lost Productivity During Initial Hospitalization.

CHARACTERIZATION AND ESTIMATES OF COSTS FOLLOWING DISCHARGE FROM THE BIRTH HOSPITALIZATION

Direct Medical Costs

Estimates of direct medical resource utilization and costs following discharge from the birth hospitalization are given in Table D-5. There was a greater degree of heterogeneity compared with that in the studies of initial hospitalization, in large degree because of the various time horizons and costs included. For this reason, it is problematic to provide a single estimate or range of estimates for these costs. As for short-run costs, costs and resource utilization were inversely related to gestational age. In addition, in studies with longer follow-ups, the majority of the costs accrued in the first year of life. In studies that measured both outpatient and rehospitalization costs, the latter accounted for the majority of costs in the first year of life, with a range of 54 to 60 percent of the costs in the non-U.S. studies and 70 to 86 percent of the costs in the U.S. cohorts.

TABLE D-5. Direct Medical Expenses Following Discharge from the Initial Hospitalization.

TABLE D-5

Direct Medical Expenses Following Discharge from the Initial Hospitalization.

Petrou examined factors associated with higher rehospitalization costs (33). These included maternal conditions, such as hospitalization, age >35 or <20 years, or smoking; perinatal factors, such as instrumental delivery or delivery complications; demographic factors, such as lower socioeconomic status; and prematurity.

Educational Costs

Details of studies assessing educational costs are given in Table D-6. As noted above, the assessment of educational costs was inconsistent with respect to both the study methodology and the age at the time of the assessment. After controlling for medical, economic, and social factors, Chaikind and Corman found a 49 percent relative risk for special education placement in low-birth-weight infants (29). This translated to a net present value of $1,240 (in 1988 U.S. dollars) per low birth weight infant, or $370 million per year for special education costs alone. Roth et al. found that infants with birth weights of <1,000 grams had kindergarten costs 60 percent higher than those for normal birth weight controls (38). However, the authors noted that low levels of maternal education and poverty accounted for more than three quarters of the total excess costs of kindergarten, whereas low birth weight accounted for only 1 percent of the total excess costs because of the relative numbers of children with these predisposing conditions.

TABLE D-6 Educational Expenditures Following Discharge from Initial Hospitalization.

Table

TABLE D-6 Educational Expenditures Following Discharge from Initial Hospitalization.

Early intervention costs may also be significant. A recent study of statelevel data in Massachusetts assessed costs from birth to age 3 years for early intervention services by gestational age. Infants born at less than 28 weeks of gestation had early intervention costs of $7,182 (in 2003 U.S. dollars), whereas the early intervention costs were $613 for those delivered at 40 weeks of gestation or more (53).

Direct Nonmedical Costs and Lost Productivity

Table D-7 shows the costs associated with postdischarge parental out-of-pocket expenses and lost productivity. Tommiska et al. documented firstyear parental wage losses of €5,990 for infants with birth weights of <1,000 grams, whereas the loss was €880 for controls. These costs increased to €8,175 in the second year for the parents of children born with low birth weights (27). Similar data are not available for the United States.

TABLE D-7 Out-of-Pocket Expenses and Lost Productivity Following Discharge from Initial Hospitalization.

Table

TABLE D-7 Out-of-Pocket Expenses and Lost Productivity Following Discharge from Initial Hospitalization.

SUMMARY OF COST STUDIES FOR PREMATURITY

Several themes emerged from this systematic review of cost-of-illness studies:

  1. Most studies had significant methodological limitations. These included a reliance on administrative data sets without adequate checks on data validity, small sample sizes and possible selection bias, a lack of quantification of uncertainty through the use of descriptive statistics or sensitivity analysis, the use of charges or inadequately described costing, the lack of a control group, and a failure to discount appropriately for longer studies.
  2. All cost estimates in the literature omitted at least some potentially important components of costs and are therefore likely to underestimate true resource use. These omissions included professional fees, hospital costs for transferred infants, costs for nonsurviving infants, and out-of-pocket costs for parents. There is almost no information on the lost earnings of parents and no information on the productivity losses of the infants themselves. Maternal costs for the delivery hospitalization and the costs of antenatal admissions were omitted by most studies. The analysis of infants and mothers separately risks both an underestimation of the costs of prematurity and the possibility of missed shifting of costs between the two groups. Very few studies have addressed educational costs, and these do not provide adequate information across the spectrum of school age and disability.
  3. Most authors did not disaggregate costs into their components, making it problematic to target the sources and predictors of high costs.
  4. Studies confirmed the previously noted inverse relationship of maternal, neonatal, and postdischarge costs with birth weight and gestational age.
  5. Although the per-patient cost for moderately preterm infants is lower than that for extremely preterm infants, the larger numbers of these infants results in population costs of similar magnitudes in both groups. This has important implications for the setting of policy, as interventions that ameliorate moderate prematurity may be as cost-effective as the more difficult interventions that would eliminate extreme prematurity.
  6. The time horizon for economic follow-up is inadequate. Very few studies have followed infants beyond 5 years, and none have documented the implications of prematurity in young adulthood, despite the availability of appropriate cohorts.
  7. Most of the studies with longer time horizons or more comprehensive methods of cost ascertainment have been from countries other than the United States. The appropriateness of generalizing these cost estimates to the United States is uncertain.

RECOMMENDATIONS FOR POLICY AND FURTHER RESEARCH

The findings summarized above have direct implications for policy and research:

  1. Because policy makers are likely to accept the dollar estimates of the costs of illness without adequately assessing the quality of the underlying studies, it is important that the peer review process identify the more significant methodological limitations, such as the use of unadjusted charges or the lack of assessment of uncertainty.
  2. In light of the prominence of the economic implications on health policy decision making for this population, it is essential that a U.S. study with a societal perspective be performed. This should include a comprehensive assessment of productivity losses and parental out-of-pocket expenditures in the U.S. context.
  3. Studies with longer time horizons should be undertaken and should preferably use an incidence approach rather than a prevalence approach to costing. At a minimum, these should extend well into school age. The implications of prematurity as young adults enter the work force should be examined, as should lifetime labor force implications.
  4. Because of the interactions between socioeconomic status and outcome, costing studies should include specific analyses of the increased burden on particular racial and ethnic populations.
  5. The focus of short-run costing studies should be shifted toward a perinatal perspective, in which the unit of analysis is the maternal-infant dyad. This is now feasible with improved data linkage.
  6. Attention should be directed toward the economic implications of moderate prematurity, in addition to extreme prematurity.
  7. A longitudinal analysis of the economic implications of prematurity to the educational system should be undertaken.
  8. A comprehensive longitudinal study of the long-term economic implications of prematurity should be undertaken.

REFERENCES

  1. Mugford M. The cost of neonatal care: reviewing the evidence. Soz Praventivmed. 1995;40(6):361–368. [PubMed: 8578874]
  2. Petrou S. Economic consequences of preterm birth and low birthweight. Br J Obstet Gynaecol. 2003 Apr;110(Suppl 20):17–23. [PubMed: 12763106]
  3. Petrou S, Davidson LL. Economic issues in the follow-up of neonates. Semin Neonatol. 2000 May;5(2):159–169. [PubMed: 10859710]
  4. Petrou S, Sach T, Davidson L. The long-term costs of preterm birth and low birth weight: results of a systematic review. Child Care Health Dev. 2001 Mar;27(2):97–115. [PubMed: 11251610]
  5. Richardson DK, Zupancic JA, Escobar GJ, Ogino M, Pursley DM, Mugford M. A critical review of cost reduction in neonatal intensive care. I. The structure of costs. J Perinatol. 2001 Mar;21(2):107–115. [PubMed: 11324356]
  6. Richardson DK, Zupancic JA, Escobar GJ, Ogino M, Pursley DM, Mugford M. A critical review of cost reduction in neonatal intensive care. II. Strategies for reduction. J Perinatol. 2001 Mar;21(2):121–127. [PubMed: 11324358]
  7. Zupancic JA, Richardson DK, Lee K, McCormick MC. Economics of prematurity in the era of managed care. Clin Perinatol. 2000 Jun;27(2):483–497. [PubMed: 10863661]
  8. U.S. Bureau of Labor Statistics. Consumer Price Index—All Urban Consumers—Medical Care (Series ID CUUR0000SAM). [accessed August 1, 2005]. [Online]. Available:www​.bls.gov/data.
  9. United States Bureau of Labor Statistics. Producer Price Index Industry Data—Hospitals (ID PCU622). 2005. [accessed September 20, 2005]. [Online]. Available:www​.bls.gov/data.
  10. Organization for Economic Cooperation and Development. Purchasing Power Parities for OECD Countries since 1980. 2005. [accessed August 1, 2005]. [Online]. Available:www​.oecd.org.
  11. Martin JA, Hamilton BE, Sutton PD, Ventura SJ, Menacker F, ML M. Births: Final Data for 2002. National Vital Stat Rep. 2003;52(10):1–113. [PubMed: 14717305]
  12. The Victorian Infant Collaborative Study Group. Economic outcome for intensive care of infants of birthweight 500-999 g born in Victoria in the post surfactant era. J Paediatr Child Health. 1997 Jun;33(3):202–208. [PubMed: 9259293]
  13. Adams EK, Nishimura B, Merritt RK, Melvin C. Costs of poor birth outcomes among privately insured. J Health Care Finance. 2003 Spring;29(3):11–27. [PubMed: 12635991]
  14. Brazier L, Harper K, Marrington S. Hospital visiting costs. An exploratory study into travelling expenses incurred by parents with babies in a regional neonatal unit. J Neonat Nurs. 1995;1(2):29–31.
  15. Chollet DJ, Newman JF Jr, Sumner AT. The cost of poor birth outcomes in employersponsored health plans. Med Care. 1996 Dec;34(12):1219–1234. [PubMed: 8962588]
  16. Doyle LW. Evaluation of neonatal intensive care for extremely low birth weight infants in Victoria over two decades. II. Efficiency. Pediatrics. 2004 Mar;113(3 Pt 1):510–514. [PubMed: 14993542]
  17. Giacoia GP, Rutledge D, West K. Factors affecting visitation of sick newborns. Clin Pediatr (Phila). 1985 May;24(5):259–262. [PubMed: 3987165]
  18. Gilbert WM, Nesbitt TS, Danielsen B. The cost of prematurity: quantification by gestational age and birth weight. Obstet Gynecol. 2003 Sep;102(3):488–492. [PubMed: 12962929]
  19. Kilpatrick SJ, Schlueter MA, Piecuch R, Leonard CH, Rogido M, Sola A. Outcome of infants born at 24-26 weeks’ gestation. I. Survival and cost. Obstet Gynecol. 1997 Nov;90(5):803–808. [PubMed: 9351768]
  20. Luke B, Bigger HR, Leurgans S, Sietsema D. The cost of prematurity: a case-control study of twins vs singletons. Am J Public Health. 1996 Jun;86(6):809–814. [PMC free article: PMC1380399] [PubMed: 8659654]
  21. Marbella AM, Chetty VK, Layde PM. Neonatal hospital lengths of stay, readmissions, and charges. Pediatrics. 1998 Jan;101(1 Pt 1):32–36. [PubMed: 9417147]
  22. McLoughlin A, Hillier VF, Robinson MJ. Parental costs of neonatal visiting. Arch Dis Child. 1993 May;68(5 Spec No):597–599. [PMC free article: PMC1029312] [PubMed: 8323364]
  23. Rogowski J. Measuring the cost of neonatal and perinatal care. Pediatrics. 1999 Jan;103(1 Suppl E):329–335. [PubMed: 9917475]
  24. Rogowski J. Using economic information in a quality improvement collaborative. Pediatrics. 2003 Apr;111(4 Pt 2):e411–e418. [PubMed: 12671160]
  25. Schmitt SK, Sneed L, Phibbs CS. The costs of newborn care in California: a populationbased study. Pediatrics. 2006;117:154–160. [PMC free article: PMC8720276] [PubMed: 16396873]
  26. St John EB, Nelson KG, Cliver SP, Bishnoi RR, Goldenberg RL. Cost of neonatal care according to gestational age at birth and survival status. Am J Obstet Gynecol. 2000 Jan;182(1 Pt 1):170–175. [PubMed: 10649175]
  27. Tommiska V, Tuominen R, Fellman V. Economic costs of care in extremely low birthweight infants during the first 2 years of life. Pediatr Crit Care Med. (2003) Apr;4(2):157–163. [PubMed: 12749645]
  28. Broyles RS, Tyson JE, Heyne ET, Heyne RJ, Hickman JF, Swint M, et al. Comprehensive follow-up care and life-threatening illnesses among high-risk infants: a randomized controlled trial. JAMA. 2000 Oct 25;284(16):2070–2076. [PubMed: 11042755]
  29. Chaikand S, Corman H. The impact of low birthweight on special education costs. J Health Econ. (1991) Oct;10(3):291–311. [PubMed: 10170854]
  30. Gennaro S. Leave and employment in families of preterm low birthweight infants. Image J Nurs Sch. 1996 Fall;28(3):193–198. [PubMed: 8854539]
  31. Lewit EM, Baker LS, Corman H, Shiono PH. The direct cost of low birth weight. Future Child. 1995 Spring;5(1):35–56. [PubMed: 7633866]
  32. McCormick MC, Bernbaum JC, Eisenberg JM, Kustra SL, Finnegan E. Costs incurred by parents of very low birth weight infants after the initial neonatal hospitalization. Pediatrics. 1991 Sep;88(3):533–541. [PubMed: 1652734]
  33. Petrou S. The economic consequences of preterm birth during the first 10 years of life. Br J Obstet Gynaecol. 2005 Mar;112(Suppl 1):10–15. [PubMed: 15715587]
  34. Petrou S, Mehta Z, Hockley C, Cook-Mozaffari P, Henderson J, Goldacre M. The impact of preterm birth on hospital inpatient admissions and costs during the first 5 years of life. Pediatrics. 2003 Dec;112(6 Pt 1):1290–1297. [PubMed: 14654599]
  35. Pharoah PO, Stevenson RC, Cooke RW, Sandu B. Costs and benefits of neonatal intensive care. Arch Dis Child. 1988 Jul;63(7 Spec No):715–718. [PMC free article: PMC1590135] [PubMed: 2458075]
  36. Rogowski J. Cost-effectiveness of care for very low birth weight infants. Pediatrics. 1998 Jul;102(1 Pt 1):35–43. [PubMed: 9651411]
  37. Rolnick SJ, Jackson JM, O'Connor P, DeFor T. Impact of birthweight on healthcare charges within a managed care organization. Am J Manag Care. 2000 Dec;6(12):1289–1296. [PubMed: 11151806]
  38. Roth J, Figlio DN, Chen Y, Ariet M, Carter RL, Resnick MB, et al. Maternal and infant factors associated with excess kindergarten costs. Pediatrics. 2004 Sep;114(3):720–728. [PubMed: 15342845]
  39. Stevenson RC, McCabe CJ, Pharoah PO, Cooke RW. Cost of care for a geographically determined population of low birthweight infants to age 8-9 years. I. Children without disability. Arch Dis Child Fetal Neonat Ed. 1996 Mar;74(2):F114–F117. [PMC free article: PMC2528521] [PubMed: 8777657]
  40. Stevenson RC, Pharoah PO, Stevenson CJ, McCabe CJ, Cooke RW. Cost of care for a geographically determined population of low birthweight infants to age 8-9 years. II. Children with disability. Arch Dis Child Fetal Neonat Ed. 1996 Mar;74(2):F118–F121. [PMC free article: PMC2528526] [PubMed: 8777658]
  41. Thomson Medstat. The Cost of Prematurity to U.S. Employers. March of Dimes; 2004. http://www​.marchofdimes​.com/prematurity/15341_15349.asp.
  42. Imershein AW, Turner C, Wells JG, Pearman A. Covering the costs of care in neonatal intensive care units. Pediatrics. 1992 Jan;89(1):56–61. [PubMed: 1728023]
  43. Long SH, Marquis MS, Harrison ER. The costs and financing of perinatal care in the United States. Am J Public Health. 1994 Sep;84(9):1473–1478. [PMC free article: PMC1615185] [PubMed: 8092374]
  44. March of Dimes. Impact on Business. March of Dimes; 2005. http://www​.marchofdimes​.com/prematurity/15341_15349.asp.
  45. Hack M, Youngstrom EA, Cartar L, Schluchter M, Taylor HG, Flannery D, et al. Behavioral outcomes and evidence of psychopathology among very low birth weight infants at age 20 years. Pediatrics. 2004 Oct;114(4):932–940. [PubMed: 15466087]
  46. Saigal S, Pinelli J, Hoult L, Kim MM, Boyle M. Psychopathology and social competencies of adolescents who were extremely low birth weight. Pediatrics. 2003 May;111(5 Pt 1):969–975. [PubMed: 12728073]
  47. Darmstadt GL, Bhutta ZA, Cousens S, Adam T, Walker N, de Bernis L. Evidence-based, cost-effective interventions: how many newborn babies can we save? Lancet. 2005 Mar 12-18;365(9463):977–988. [PubMed: 15767001]
  48. Leader S, Yang H, DeVincenzo J, Jacobson P, Marcin JP, Murray DL. Time and out-of-pocket costs associated with respiratory syncytial virus hospitalization of infants. Value Health. 2003 Mar-Apr;6(2):100–106. [PubMed: 12641860]
  49. Stang P, Brandenburg N, Carter B. The economic burden of respiratory syncytial virusassociated bronchiolitis hospitalizations. Arch Pediatr Adolesc Med. 2001 Jan;155(1):95–96. [PubMed: 11177073]
  50. Waitzman NJ, Romano PS, Scheffler RM. Estimates of the economic costs of birth defects. Inquiry. 1994 Summer;31(2):188–205. [PubMed: 8021024]
  51. Greenough A, Alexander J, Burgess S, Bytham J, Chetcuti PA, Hagan J, et al. Health care utilisation of prematurely born, preschool children related to hospitalisation for RSV infection. Arch Dis Child. 2004 Jul;89(7):673–678. [PMC free article: PMC1720002] [PubMed: 15210503]
  52. Greenough A, Cox S, Alexander J, Lenney W, Turnbull F, Burgess S, et al. Health care utilisation of infants with chronic lung disease, related to hospitalisation for RSV infection. Arch Dis Child. 2001 Dec;85(6):463–468. [PMC free article: PMC1719001] [PubMed: 11719328]
  53. Clemens K, Avadi F, Wilber N, Barfield WD. The cost of prematurity: birth to age three early intervention costs [abstract] Pediatr Res. in press.

Footnotes

1

John A. F. Zupancic MD, ScD, Department of Pediatrics, Harvard Medical School, Boston, Massachusetts.