To date, research on the medical care costs of preterm birth has focused nearly exclusively on inpatient care, primarily on the initial hospitalization of the infant (Zupancic, 2006). Certain studies have provided estimates of inpatient hospitalization costs exclusively for infants born with extremely low birth weights (The Victorian Infant Collaborative Study Group, 1997) or very low birth weights (VLBWs) (Rogowski, 2003), others have provided estimates exclusively by gestational age (Phibbs and Schmitt, 2006), and still others have provided estimates by both gestational age and birth weight (BW) (Gilbert et al., 2003; Schmitt et al., 2006). This literature has drawn specific attention to the high costs associated with neonatal intensive care for preterm infants. More than two-thirds of infants born extremely premature (at less than 28 weeks of gestation) and more than one-third of infants born very premature (at 28 to 31 weeks of gestation) in California who survived infancy in 1996 had respiratory distress syndrome and received mechanical ventilation in their initial hospitalization. In comparison, less than 1 percent of infants born at term received mechanical ventilation in their initial hospitalization (Gilbert et al., 2003). Ancillary costs, including respiratory care, laboratory, radiology, and pharmacy costs, contribute heavily to the costs of intensive care for low-birth-weight (LBW) infants. They comprised over 25 percent of the median $53,316 in initial hospitalization costs for those born with VLBWs (<1,500 grams) in a group of hospitals belonging to the Vermont Oxford Network in 1997 and 1998 (Rogowski, 2003).
Less research has been conducted on the medical care expenses of preterm birth beyond early hospitalization. Comprehensive data on medical care delivery and cost for the cohort of infants born between 1998 and 2000 who were covered by Intermountain Healthcare (IHC) Health Plans and who were monitored from birth through 2004 afforded the opportunity to analyze inpatient and outpatient medical care costs associated with preterm birth from birth to age 7 years, well beyond the term covered by any previous study in the United States. In making national estimates, it would have been preferable to have data from several rather than from a single health plan. Resource constraints faced by the committee precluded the incorporation of data from additional plans. As noted below, however, substantial effort was undertaken to stratify the sample and adjust estimates so as to make valid approximation of costs for the nation as a whole.
IHC is a large, nonprofit, fully integrated health care organization headquartered in Utah. In addition to providing a comprehensive set of medical care services under its Health Plans, it also provides services to patients under contract with Medicaid and Medicare, patients covered under non-IHC commercial insurance, and privately paying patients, as well as those receiving charitable care.
Of the 139,517 estimated live births in Utah between 1998 and 2000 (NCHS, 2004b), 81,931, or 59 percent of the total, were born within the IHC network of facilities. One-third of those births were covered under IHC Health Plans, 38 percent were covered by other commercial insurance, 22 percent were covered by Medicaid, and 7 percent were covered through other sources of private payment or by charity.
Finding 12-1: The medical costs of preterm birth during infancy, particularly during the neonatal period, are high and the medical needs are relatively well understood. The long-term medical, educational, productivity, and productivity costs borne by the individual, as well as by the family and society, are not well understood.
The detailed estimates of costs presented here were based strictly on the cohort covered under IHC Health Plans, as records of service utilization and cost were comprehensive for that cohort. The application of exclusion criteria based on clear coding errors for gestational age (less than 22 weeks or more than 43 weeks of gestation) or BW (<450 or >5,000 grams) or the presence of missing fields in certain instances resulted in 23,631 Health Plan infants in the cohort as of birth; 4,588 of these infants were born preterm. By the end of the 48th month of follow-up, the last possible month for all surviving infants in the cohort to be monitored, given the 2004 cutoff, 11,357 subjects (48 percent of the total) remained. The corresponding numbers (percentages) remaining at 48 months by gestational age were 36 (40 percent) among those born at less than 28 weeks of gestation, 73 (42 percent) among those born at 28 to 31 weeks of gestation, 754 (46 percent) among those born at 32 to 36 weeks of gestation, and 10,494 (48 percent) among those born at 37 weeks of gestation or later. To maintain robustness, resource utilization and cost estimates were pooled for the cohort over years 3 and 4 and for years 5 through 71.
Extrapolation of health care utilization and costs across organizations and across geographic areas is potentially confounded by differences in demographics, including the underlying health status of populations, as well as by differences in the health care delivery conventions by provider. For that reason, adjustments of charges to costs and adjustments of costs for differences across geographic areas may not be sufficient to project cost estimates from one region to the nation as a whole. Risk adjustment based on differences in population characteristics and further adjustment for organizational differences in the delivery of care may also be required (Rogowski, 1999).
A comparison of demographic and vital statistics, as well as of health care utilization and costs, between the IHC Health Plans cohort and those served outside of the plan but within the IHC network of facilities, was therefore made. A further comparison of the results with published results from other studies on initial hospitalization costs and costs outside of Utah was also performed. The comparison of demographic characteristics and vital statistics of the Health Plans cohort of births with other births in IHC facilities between 1998 and 2000 revealed—as one might expect, given the well-established inverse socioeconomic gradients in the rates of pre-
1 These estimates of those remaining in the cohort include all those still enrolled in Health Plans, even if they did not receive medical care services during a specific period. With an additional mortality adjustment, these essentially served as denominators for calculating average cost. Attrition occurred either because of mortality or because of loss of eligibility for coverage. Estimates were adjusted for mortality based on national figures, as discussed below. As long as resource utilization for those otherwise dropping coverage were similar to those remaining in the health plan, such attrition would not affect the estimates. Some attrition, however, particularly among those born extremely premature, was likely due to infants reaching a $1 million lifetime limit. While not common, the selection out of such infants introduced a downward bias into the estimates.
term birth and LBW—that those covered by IHC Health Plans were less likely to be born preterm (8.5 percent) and of LBW (6 percent) than those born in IHC facilities and covered by Medicaid (11.2 and 8.7 percent, respectively). The rates of preterm birth and LBW among births covered by other commercial insurance were nearly identical to those in the IHC Health Plan cohort. The overall rate of preterm birth in the United States in 1999-2000 was 11.6 percent (Mac-Dorman et al., 2005). The rate of infant mortality for those born extremely prematurely and for those born with VLBWs was lower in the IHC Health Plans cohort than in the cohort cared for in IHC facilities not covered by Health Plans; but summary statistics (mean, median, interquartile ranges, and box plots) revealed that the rate of neonatal intensive care unit utilization, the average length of stay, and the average cost of treatment for the initial hospitalization by gestational age were not significantly different between those who were covered by IHC Health Plans and those who were covered by Medicaid or other commercial insurance in IHC facilities. With the incorporation of adjustments for mortality, in other words, the IHC Health Plan cohort yielded reliable estimates of average service utilization across plans by gestational age and BW.
The levels of health care provision under the umbrella of a single organization and for a population with demographics different from those for the nation as a whole (in Utah, the population is younger and has lower proportions of ethnic and racial minorities than the population of the nation as a whole) could still deviate from those for the general population. Because studies of resource utilization associated with preterm birth have been specific to particular organizations or geographic regions, no "gold standard" against which a definitive assessment can be made exists. Table 12-2 provides a summary of the results of several recent studies, including the results for the IHC Health Plans cohort, on the length of stay for initial or early hospitalizations by gestational age and BW. Data from those studies were reconfigured to make more direct comparisons with the IHC Health Plans cohort by gestational age and BW.
The first three columns of Table 12-2 summarize the length-of-stay data for recent population-based studies in the state of California. The samples in each of those studies essentially drew from the same statewide database on hospital discharges linked to vital records. Because of the absence of reporting of charges by the Kaiser Permanente Medical Care System, the lengths of stay for hospitalizations under that system were largely excluded from the statistics in the CA1 and CA2 columns of Table 12-2. The major difference between the hospital stay figures from the study of Gilbert et al. (2003) reported in column CA1 and those from the study of Phibbs and Schmitt (2006) reported in column CA2 is that the former included only survivors of infancy, whereas the latter included all infants. Given that the length of stay for survivors is longer than that for nonsurvivors, it is counterintuitive for the average length of stay reported in the study of Phibbs and Schmitt (2006) to be uniformly longer by gestational age than that reported in the study by Gilbert et al. (2003). Nevertheless, the gradients in the lengths of stay by gestational age are quite similar between the two studies, and the difference was not statistically significant. The average length of stay and the gradients exhibited by gestational age and BW from the IHC Health Plan cohort reported in the final column are very similar, in this context, to those reported for the California population-based data in columns CA1 to CA3 of Table 12-2.
It should be noted that the IHC data provided in the table summarize the inpatient lengths of stay for all admissions in the first month after birth and therefore exclude transfers from initial hospitalizations that transpire after the first month. The California studies included only the lengths of stay for initial hospitalizations but included all transfers connected to that hospitaliza-tion, even if they occurred after the first month. Therefore, among the extremely premature and very premature infants, who have the longest initial hospitalizations and who have the highest rates of transfer after the first month, the California data would be expected to exhibit longer average lengths of stay than the IHC data. Among those born moderately preterm (32 to 36 weeks of gestation), who are more likely to be discharged from their initial hospitalization but then readmitted for other problems within the first month after birth, the IHC data would be expected to reveal relatively longer average lengths of stay. These are the precise patterns exhibited in Table 12-2.
Table 12-2 also reports the median lengths of stay for 6,797 VLBW infants who were born in 1997 and 1998 and who received intensive care in 29 of 34 hospitals located in several states across the United States belonging to the Vermont Oxford Network (Horbar et al., 2003; Rogowski, 2003). Estimates were made before the participation by these hospitals in a neonatal intensive care quality improvement initiative (NIC/Q) beginning in 2000. The 47-day estimate reflects only the portion of initial hospitalizations that transpired within a participating hospital and therefore excludes that portion of the initial length of stay for 1,364 infants before transfer into a study hospital as well as that portion of the initial stay after 1,264 infants transferred out of participating hospitals. Given these exclusion criteria, which generated slightly shorter stays relative to those for the IHC cohort, the median length of stay of 51 days for the IHC cohort is very similar to that of the NIC/Q hospitals.
The STAT column of Table 12-2 is taken from a study of infants who survived their first year of life and provides the average length of stay over that entire year (Medstat, 2004). The sample included infants born between 2000 and 2002 covered under the health plans of certain large employers in the United States. The selection of infants was made on the basis of specific International Classification of Diseases (Ninth Revision) and diagnosis-related group codes and not on the basis of a reported gestational age or BW. Despite such differences, the average length of stay for the IHC cohort, when the length of stay was recalculated for survivors in the entire first year, was 16.6 days, which is nearly identical to the 16.8 days from the Medstat analysis.
Table 12-2 also summarizes data on the length of stay for the initial hospitalization from a population-based cohort study conducted in the United Kingdom (Petrou et al., 2003). The infants in the sample were born between 1970 and 1993 and were monitored for 5 years. Given the vast improvements in neonatal intensive care technology that have taken place since 1970 and that have resulted in dramatic increases in the rates of infant survival, it is not surprising that the average length of stay for the most premature infants in this sample was substantially less than that for the more recent periods in the United States reported in the other columns of Table 12-2. The longer average length of stay among those born at 32 to 36 weeks of gestation and among normal term infants in the United Kingdom than in the United States reflect the well-established differences in practice patterns between the two countries (Profit et al., 2006).
The comparative data on the length of stay for initial or early hospitalization provided in Table 12-2 suggest that the IHC Health Plans cohort data on health care utilization associated with preterm birth, when adjusted for gestational age and mortality, is reflective of that in other large population-based samples in the United States. When the data are also adjusted for geographic differences in prices, the IHC data were considered sufficiently reliable for the approximation of the rates of medical care utilization and the costs associated with preterm birth for the United States as a whole. Although the intensity and the quality of care per inpatient stay or physician visit cannot be directly observed from the data, they could still create differences in costs between the IHC cohort and the national population. Price-adjusted cost comparisons of initial and early hospital stays by BW between the IHC cohort and those reported in other analyses, however, provided additional support for the conclusion that such differences may not be conse quential. An analysis of VLBW infants (birth weights of <1,500 grams) receiving neonatal intensive care in 1997 and 1998 in 29 hospitals belonging to the Vermont Oxford Network, for example, estimated the median cost of the initial hospitalization to be $53,316 in constant 1998 dollars (Rogowski, 2003). Although that study did not include the costs associated with the care received before some transfers to or after some transfers out of the sample hospitals, the comparative figure for first-month hospitalizations among such infants in the IHC cohort was $56,433, which is remarkably similar to that for the Vermont Oxford Network, given the geographic differences and the differences in the costing methodologies applied in the two studies. Geographic and inflation adjustments applied to initial hospitalization costs reported in one California study (Schmitt et al., 2006) demonstrated patterns similar to those reported in Table 12-2 (column CA3) for length of stay.
The algorithm developed by IHC Health Plans for the computation of cost is constructed such that allowed charges are reflective of cost at a detailed level of service provision. Allowed charges for services provided to the IHC cohort between 1998 and 2004 were tabulated and adjusted to 1998 constant dollars on the basis of a separate inflation algorithm developed within IHC. Adjustment of IHC costs to the United States as a whole was based on geographic adjustment factors that were separately constructed for inpatient and outpatient care. The geographic adjustment factor for inpatient care was based on the 1998 Medicare Prospective Payment System wage and capital wage indices for each metropolitan statistical area (MSA) and for rural Utah. To yield a single inpatient geographic adjustment factor, these area indices were weighted by the population distribution in each Utah MSA and rural Utah, according to intercensus population estimates for 1998; and the capital and wage components were weighted according to their relative contributions to hospital care costs. A parallel method was applied by using Medicare geographic adjustment factors for physician work, practice, and malpractice expenses in Utah to generate an outpatient geographic adjustment factor. An adjustment for national cost inflation between 1998 and 2005 was then made separately for inpatient care and outpatient care on the basis of Medicare Prospective Payment System price adjustments and physician practice expense price adjustments, respectively.
The results for the average medical care cost by gestational age and by year of life are presented in Tables 12-3 through 12-5 separately for inpatient care (Table 12-3), outpatient care (Table 12-4), and total care (Table 12-5). Birth-year inpatient costs constitute the large majority of incremental medical care costs associated with preterm birth, and the familiar steep inverse gradient for first-year inpatient medical care costs is evident in the first column of Table 12-3. Higher incremental inpatient costs are evident beyond birth through age 4 years for those born at less than 32 weeks of gestation and through age 7 years for those born at less than 28 weeks of gestation. Although early inpatient costs are disproportionate relative to subsequent costs, postin-fancy medical care costs associated with preterm birth are not insignificant. Incremental costs for outpatient care exceed those for inpatient care beginning in the second year of life, and incremental outpatient care costs continue for children born preterm through age 4 years, regardless of gestational age.
The differences in mean outpatient costs exhibited in Table 12-4 appear to be driven largely by differences in the upper tails of the cost distribution rather than by the median by age 2 (Table 12-6). This upper tail dominance appears to continue with age, particularly for those under 32 weeks of gestation, but sample size limitations because of restriction of the analysis to the upper 5 percent of the sample, coupled with attrition in the later years of the IHC cohort database, precluded a definitive assessment.
The average cost estimates in Tables 12-3 through 12-5 were multiplied by national cohort estimates at each age to make national cost estimates. The size of the birth cohort by gestational age was based on vital statistics for the 2003 birth cohort (Martin et al., 2005), and infant mortality was based on linked birth-death records from 2001 and 2002 (MacDorman et al., 2005). These data indicated that 91, 71, and 56 percent of the cases of infant morality occur in the neonatal period for those born at less than 28, 28 to 31, and 32 to 36 weeks of gestation, respectively. Adjustments were therefore made for average costs in the first year of life for infant mortality, such that all cases of infant mortality were assumed to take place at the end of the first month of life for those born at less than 28 weeks of gestation and at the end of the neonatal period for those born at 28 to 31 and 32 to 36 weeks of gestation. Normal survival was assumed beyond infancy. Costs beyond the first year of life were discounted back to the year of birth at a 3 percent rate. The results for aggregate national costs and the cost per case by gestational age and care category are provided in Table 12-7.
Although the literature on initial hospitalization has often suggested a roughly equal distribution of total inpatient costs between the three groups of infants born preterm—that is, those born extremely preterm (less than 28 weeks of gestation), those born very preterm (28 to 31 weeks of gestation), and those born moderately preterm (32 to 36 weeks of gestation)—the inclusion of a longer period of follow-up in this analysis demonstrates that the overall contribution to inpatient costs among those born extremely preterm is even larger. Given the relatively small numbers of infants born at less than 28 weeks of gestation, this translates into an even steeper cost-concentration curve, with the 6 percent of infants born at less than 28 weeks of gestation accounting for nearly 38 percent of total medical costs (Table 12-7). The cost per preterm infant increases nearly exponentially with each categorical decrease in gestational age. The concentration of neonatal intensive care among those born at the lowest gestational ages, coupled with its very high cost, is a primary driver behind this gradient.
Although the average cost reflects the generally higher cost of preterm birth with progressive decreases in gestational age, the variance of the cost is substantially higher among pre-term infants than among term infants. The ratio of the 75th cost percentile to the 25th cost percentile for early inpatient care, for example, is about 3 to 1 for each of the gestational age categories, similar to that found by Phibbs and Schmitt (2006). Although the distribution of cost is affected by mortality, particularly among those born extremely preterm, the high variance in cost is not driven by mortality (Phibbs and Schmitt, 2006). The association of birth defects with preterm birth (Rasmussen et al., 2001), coupled with the high cost of repair of several of those defects in infancy (Waitzman et al., 1996), may explain part of the variance in cost. More research that formally accounts for comorbidities to explain the variance in cost associated with preterm birth is required.
Finding 12-2: The variance in the costs associated with preterm birth is large, even within gestational age groups. Sufficient knowledge about the factors that explain this variance is not available.
Was this article helpful?