Context. Although the increasing effectiveness of neonatal programs for extremely low birth weight (ELBW, birth weight <1000 g) infants has been established from cohort studies, there is a paucity of data on the relationship between the costs and the consequences of neonatal intensive care.
Objective. To determine the changes in the efficiency of neonatal intensive care for ELBW infants in Victoria, Australia over 2 decades.
Design. Economic evaluation (cost-effectiveness and cost-utility analyses) in a population-based cohort study of consecutive ELBW infants born during 4 distinct eras (1979–1980, 1985–1987, 1991–1992, and 1997) followed to at least 2 years of age.
Setting. The state of Victoria.
Patients. All ELBW live births of birth weight 500 to 999 g in the state in the calendar years indicated (1979–1980: n = 351; 1985–1987: n = 560; 1991–1992: n = 429; 1997: n = 233).
Main Outcome Measures. Costs were assessed primarily by the consumption of hospital resources. The consequences included survival and quality-adjusted survival rates at 2 years of age.
Results. The cost-effectiveness ratios (expressed in Australian dollars for 1997) were similar between successive eras at $5270, $3130, and $4050 per life-year gained, respectively. The cost-utility ratios were similar between successive eras at $5270, $3690, and $5850 per quality-adjusted life-year gained, respectively, and were similar to the cost-effectiveness ratios. The cost-effectiveness and cost-utility ratios were generally higher in lower birth-weight subgroups, but there were consistent gains in efficiency over time in infants of lower birth weight.
Conclusions. As there have been large increases in effectiveness from the late 1970s to the late 1990s, the efficiency of neonatal intensive care for ELBW infants in Victoria has remained relatively stable.
Neonatal intensive care for extremely low birth weight (ELBW, birth weight <1000 g) infants born in Victoria, Australia is increasingly effective; the survival rate has increased threefold, from ∼1 in 4 in the late 1970s to 3 in 4 by the late 1990s.1 Although other cohort studies have also assessed the effectiveness of neonatal programs for ELBW infants, there is a paucity of data on the efficiency, or the relationship between the costs and the consequences, of neonatal intensive care.2 There is even less information on how the relationship between costs and consequences is changing over time.
Efficiency asks whether a neonatal intensive care program is worth implementing. Because resources are finite, those spent on neonatal intensive care might be better spent on other health care programs if they can be shown to be more efficient. At the time of the landmark paper by Sinclair et al3 on the evaluation of neonatal intensive care, there had been no full economic evaluations of neonatal intensive care. Since that time, there have been several studies reporting that neonatal intensive care is relatively efficient, particularly in comparison with other intensive health care programs.4–6 Moreover, we have reported that the efficiency of neonatal intensive care for ELBW infants had apparently improved within the same geographically defined region over time.7 However, efficiency cannot improve indefinitely. Indeed, as the magnitude of improvement in mortality rates reduces and survival approaches the maximum of 100%, we hypothesized that efficiency would decline in the absence of improvements either in the quality of life of survivors or within various components on the cost side of the neonatal intensive care program. Moreover, we hypothesized that efficiency might be worse in the lightest infants, because potentially they would consume more resources and their quality of survival might be poorer. The aim of this study was to evaluate the changes in efficiency of neonatal intensive care for ELBW infants in Victoria from 1979 to 1997.
The ELBW subjects comprised consecutive live births of birth weight 500 to 999 g born in the state of Victoria during 4 distinct eras: the calendar years 1979–1980, 1985–1987, 1991–1992, and 1997. The methods for obtaining data on the outcomes, in terms of survival and quality-adjusted survival relative to normal birth weight (birth weight >2499 g) controls, are described in the previous article.1
The methods for the economic evaluation of intensive care followed those of Drummond et al,8 as applied previously in newborn infants.4,5,7 Data on the consumption of hospital resources were collected prospectively for each live birth in each of the 4 level-III neonatal intensive care units in the state of Victoria. The number of patient-days of assisted ventilation (AV, any of intermittent positive-pressure ventilation, high-frequency ventilation, or continuous positive-airways pressure via an endotracheal tube, or nasal continuous positive-airways pressure delivered by various means) and stay in a tertiary hospital while not receiving AV were recorded to the next whole day. Patient-days in a tertiary hospital not receiving AV were converted into equivalent patient-days of AV, assuming they were worth one third of a patient-day of AV. Some ELBW infants born in the state were not cared for within a level-III center; all but 1 of these infants died over the 4 eras, and all those who died did so on the first day of life. The duration of hospitalization for the 1 survivor not cared for in a level-III center was obtained. For survivors in 1991–1992 and 1997, patient-days in step-down nurseries were recorded and also assumed to be worth one third of a patient-day of AV. For survivors in 1985–1987 and 1979–1980, such episodes were rare, and the data were not recorded. For all survivors, patient-days of rehospitalization in early childhood were recorded, to 5 years of age for the first 3 cohorts and to 2 years of age for the 1997 cohort. These patient-days were also assumed to be worth one third of a patient-day of AV. Patient-days of AV and equivalent patient-days of AV were summed and divided by the number of live births to obtain the mean duration of both.
Life-years gained per live birth and quality-adjusted life-years gained per live birth were calculated by multiplying the survival rate and the quality-adjusted survival rate, respectively, by the life expectancy; a life expectancy of 70 years was assumed except for multiply, severely disabled children, whose life expectancy was assumed to be 40 years. A discount rate of 3% was applied as recommended,8 which effectively reduces the life expectancy from 70 to 29.1 years and from 40 to 23.1 years, respectively.
Cost-effectiveness ratios were calculated by dividing the change in costs per live birth (costs) by the change in survival rates (consequences) between successive eras and were expressed as costs per life-year gained. Cost-utility ratios were calculated by dividing the change in costs per live birth (costs) by the change in quality-adjusted survival rates (consequences) between successive eras and were expressed as costs per quality-adjusted life-year gained. Both ratios were calculated for successive eras, both overall and in 250-g and 100-g birth-weight subgroups. Relative inefficiency would show as a higher ratio, and relative efficiency would be indicated by a lower ratio, or even a negative ratio, given progressive increments in survival or quality-adjusted survival rates. Costs were expressed as patient-days of AV and equivalent patient-days of AV predominantly. Dollar costs were calculated by the cost of 1 patient-day of AV of $1630 (Australian dollars in 1997), increased from the previous estimate of $1470 (Australian dollars in 1992) to allow for an increase caused by inflation of 11.2% since 1992 (see the Australian Bureau of Statistics web site at www.abs.gov.au). The dollar cost of 1 patient-day of AV was derived primarily from the proportion of the operating budget of one of the neonatal intensive care units consumed by neonatal intensive care and was similar to estimates calculated more exhaustively in other Australian studies.9
The robustness of the cost-effectiveness and cost-utility analyses to changes in certain assumptions underlying the calculations was determined in a sensitivity analysis.4 The following values were varied independently: 1) discount rate from 0% to 5% (baseline: 3%); 2) life expectancy from 50 to 90 years (baseline: 70 years) or from 20 to 60 years for multiply, severely disabled survivors (baseline: 40 years); 3) patient-days in hospital not receiving AV from the equivalent of one sixth to two thirds of a patient-day of AV (baseline: one third); 4) utility values from 0.9, 0.8, and 0.7 for mild, moderate, and severe disabilities to 0.7, 0.4, and 0.1, respectively (baseline: 0.8, 0.6, and 0.4, respectively); 5) utility value 0 for children not seen or fully assessed at ≥2 years of age years of age (baseline: 1); and 6) additional costs for surfactant (available only in the last 2 eras), each dose assumed to be equal to one third of a day of AV.
To investigate the possible effect of multiple, severe, long-term disabilities on costs beyond hospital discharge, yearly costs for children with multiple, severe disabilities cared for at home were assumed to be $27 800 (Australian dollars in 1997) for food, shelter, and clothing and an opportunity cost for parental care based on the estimate by Boyle et al4 of $8462 (Canadian dollars in 1978), adjusting for currency differences and inflation. This is equivalent to an additional cost of 17 days of AV per year. These yearly costs were multiplied by the life expectancy of 40 years discounted by 3%, summed for all multiple, severely disabled children, and added to the costs for the respective cohorts, and the cost-effectiveness and cost-utility ratios were recalculated in the sensitivity analysis, assuming baseline values for the other variables. Institutional care was not an option for survivors with multiple, severe disabilities.
Data were analyzed by Microsoft Excel, carrying up to 14 decimal places during all calculations but rounded down for final presentation. Cost-effectiveness and cost-utility ratios were determined for all ELBW infants initially and then within 250-g and 100-g birth-weight subgroups to assess whether there was any systematic variation in either ratio with birth weight. Dollar costs were all expressed to the nearest $10.
The number of live births, the survival and quality-adjusted survival rates, the life-years and quality-adjusted life-years gained, and the consumption of resources are all shown in Table 1, both overall and in 250-g birth-weight subgroups. Except for the first era, the majority of the resources consumed per live birth were days of AV, as distinct from non-AV hospital time converted to equivalent days of AV. The increments in equivalent days of AV in each era attributable to hospital readmissions after discharge home were small (1979–1980: 1.1; 1985–1987: 1.2; 1991–1992: 1.9; 1997: 1.3).
The relationships between the consumption of equivalent days of AV and the survival rates for each era for the whole birth-weight range of 500 to 999 g and 250-g and 100-g birth-weight subgroups are shown in Fig 1 (A and B, respectively). For most birth-weight subgroups, there was an approximately linear relationship between the consumption of hospital resources and survival rate over successive eras, indicating that efficiency was relatively stable over time. Moreover, the relationship was approximately linear through the origin, particularly for the overall group of birth weight 500 to 999 g. The consumption of hospital resources per survivor was higher in lower birth-weight subgroups, ie, the line connecting successive eras was further to the left in lighter infants. There was more variation in the relationship as birth weight was split into smaller subgroups. A cost-effectiveness ratio between any 2 eras, expressed as the increment in equivalent days of AV per additional survivor, can be determined from the slope between the 2 data points for the respective eras. For the lightest infants, the slopes were not as steep, because there were bigger gains in survival in the later eras compared with earlier eras. In other words, efficiency was better in later eras in lighter infants.
For the whole birth-weight range of 500 to 999 g, the cost-effectiveness and cost-utility ratios were all positive, indicating that there were net costs per life-year gained or quality-adjusted life-year gained; the results are expressed in units of hospital resources in Table 2 and in dollars in Fig 2. The cost-effectiveness ratios (Australian dollars in 1997) for all birth weights of 500 to 999 g were similar between successive eras at $5270, $3130, and $4050 per life-year gained, respectively. The cost-utility ratios were similar between successive eras at $5270, $3690, and $5850 per quality-adjusted life-year gained, respectively, and were similar to the cost-effectiveness ratios. Ratios with costs expressed as days of AV alone were mostly lower than when costs were expressed as equivalent days of AV (Table 2). The cost-utility ratios were mostly slightly higher than the corresponding cost-effectiveness ratios.
Within the 250-g birth-weight subgroups, those of birth weight 500 to 749 g had cost-effectiveness and cost-utility ratios that were more positive (less efficient) than for those of birth weight 750 to 999 g for the first 2 comparisons but not for the last comparison (Table 2 and Fig 2). The most consistent gains in efficiency over time were in those of birth weight 500 to 749 g. Within the 100-g birth-weight subgroups, there was more variation in all ratios, but trends between eras and within lower birth-weight groups were similar to those described for the 250-g birth-weight subgroups (Fig 2). However, it was not possible to identify a birth weight below which efficiency was markedly worse on any ratio. In the latest comparison (1997 vs 1991–1992), the birth-weight subgroup of 700 to 799 g was more inefficient on both ratios than either the 500- to 599-g and 600- to 699-g birth-weight subgroups. Some ratios were negative within the 100-g birth-weight subgroups, indicating a net reduction in resources for the gain in survival.
The cost-effectiveness and cost-utility ratios were quite robust to variations in most of the assumptions relevant to the study, some of which are shown as percent changes from baseline in Fig 3 for cost-utility. The percent changes from baseline in the cost-effectiveness ratios in the sensitivity analyses are not shown, because they were virtually identical to those for the same variables affecting the cost-utility ratios. Variation in the discount rate resulted in more change in the cost-effectiveness and cost-utility ratios than variation in most other variables. The effects of varying life expectancy, fractions of days of AV for nonventilated time in the hospital, the costs of long-term care for multiple, severely disabled children, and of changes in utility values (relevant to cost-utility only) were mostly small. If those who were not seen were assumed to have utility of 0, the change from baseline ranged between 0.4% and 7.2%. If additional costs for surfactant were added (relevant only to 1991–1992 and 1997), the change from baseline was an additional 2.2% for 1991–1992 vs 1985–1987 and an additional 2.1% for 1997 vs 1991–1992. If costs for readmission to the hospital after the primary hospitalization were ignored, the change from baseline ranged between −7% and +5%.
Neonatal intensive care is relatively efficient compared with other health care programs, both intensive and nonintensive.6,10 However, the main purpose of the present study was not to compare neonatal intensive care with other health care programs via a “league table”; others have warned that such tables should be interpreted cautiously.11 Our intent was to be sure that neonatal intensive care was not becoming too inefficient within our own region, particularly in the lightest infants who were surviving in increasing numbers in later eras.
We were concerned that increased consumption of resources might outstrip additional gains in survival and quality-adjusted survival, but this was not observed. Indeed the relationship between increasing consumption of resources and increasing survival and quality-adjusted survival was approximately linear overall, through the origin, as illustrated in Fig 1. This means that regardless of which 2 time points are used for comparison, even if 1 is the theoretical comparison of no treatment with no survivors (the origin in Fig 1), the cost-effectiveness and cost-utility ratios would be approximately the same, as are the incremental ratios between consecutive eras. Moreover, although the lightest infants consumed more resources, there was no clear demarcation where it seemed that the incremental costs per additional life-year (cost-effectiveness) or additional quality-adjusted life-year (cost-utility) were so high as to consider denying intensive care for the tiniest infants on grounds of inefficiency alone. Furthermore, in the most recent era, care of lighter infants was relatively more efficient, ie, their cost-effectiveness and cost-utility ratios were lower than in earlier eras.
Zupancic et al,2 recognizing the higher readmission rates of ELBW infants after discharge, emphasized the need to consider hospital resource consumption after the primary hospitalization in any economic evaluation. That we had rehospitalization data only to 2 years of age for the 1997 cohort (because they had not yet reached 5 years of age) rather than to 5 years of age (as for the other 3 cohorts) was relatively unimportant, because the exclusion of rehospitalization data beyond the primary hospitalization had little effect on the cost-effectiveness and cost-utility ratios. Days of AV, in contrast with nonventilated days converted to equivalent days of AV, represented the bulk of the costs of improving outcomes in our study, particularly in more-recent times. The overall cost-effectiveness and cost-utility ratios for the latest era expressed in equivalent days of AV were only ∼7% higher than when expressed solely in days of AV.
Neonatal intensive care has changed in complexity since the 1970s, and dollar costs per unit of time also may have changed. Also, currencies fluctuate widely between countries over relatively short periods of time. For example, in the late 1970s $1 Australian was worth approximately $1.10 US, whereas today it is worth only half as much relative to the US dollar. In the only other full economic evaluation of neonatal intensive care by researchers other than our group,4 the cost-effectiveness and cost-utility ratios for neonatal intensive care for ELBW infants were higher than in our first economic evaluation.5 However, the explanation for the difference may reside not so much in relative efficiencies or inefficiencies but more in the difficulties of comparing studies between countries and eras. It is for these reasons that we have expressed costs primarily in units of ventilated time in hospital and not only in dollars. This will facilitate comparisons with neonatal intensive care programs either in other countries or in different eras.
We have reported outcomes in various subgroups of birth weight, which increases the complexity of the data. However, not only does this allow contrasts between birth-weight subgroups in an attempt to identify more inefficient birth-weight subgroups, it also facilitates comparison with others who undertake similar calculations but who may have birth-weight subgroups that match some but not all of the subgroups in our study.
In common with all economic evaluations, certain assumptions were necessary, and wide margins of uncertainty were allowed for the sensitivity analyses. Variations in the discount rate were most important, with variations in life expectancy, the nonventilated time assumed equivalent to ventilated time (including rehospitalization time) utilities, and costs for long-term care for multiply, severely disabled survivors all being less important. Additional costs for surfactant, relevant only in 1991–1992 and 1997, also had a negligible effect. Allowances for children not fully assessed had little effect, because the follow-up rate was so high for all cohorts.
As there have been large increases in effectiveness from the late 1970s to the late 1990s, the efficiency of neonatal intensive care for ELBW infants in Victoria has remained relatively stable. However, neonatal intensive care for ELBW infants must be continually reevaluated economically, particularly as survival rates approach their maximum.
This work was supported in part by a grant from Health and Community Services (Victoria, Australia) and the National Health and Medical Research Council (Canberra, Australia).
- Received February 18, 2003.
- Accepted May 15, 2003.
- Address correspondence to Lex W. Doyle, MD, FRACP, Department of Obstetrics and Gynaecology, Royal Women’s Hospital, 132 Grattan St, Carlton, Victoria 3053, Australia. E-mail:
The following were the participants in the Victorian Infant Collaborative Study Group. Convenor: Lex W. Doyle, MD, FRACP (Royal Women’s Hospital and University of Melbourne); Collaborators (in alphabetical order): Ellen Bowman, FRACP (Royal Women’s Hospital and Newborn Emergency Transport Service), Catherine Callanan, RN (Royal Women’s Hospital), Elizabeth Carse, FRACP (Monash Medical Centre), Dan Casalaz, FRACP (Mercy Hospital for Women), Margaret P. Charlton, MEd Psych (Monash Medical Centre), Noni Davis, FRACP (Royal Women’s Hospital), Geoffrey Ford, FRACP (Royal Women’s Hospital), Simon Fraser, FRACP (Mercy Hospital for Women), Jane Halliday, PhD (Victorian Perinatal Data Collection Unit), Marie Hayes, RN (Monash Medical Centre), Elaine Kelly, MA (Royal Women’s Hospital and Mercy Hospital for Women), Anne Rickards, PhD (Royal Women’s Hospital), Michael Stewart, FRACP (Royal Women’s Hospital and Royal Children’s Hospital), Andrew Watkins, FRACP (Mercy Hospital for Women), Heather Woods, RN (Mercy Hospital for Women), and Victor Yu, MD, FRACP (Monash Medical Centre) (all affiliations are located in Melbourne, Australia).
- ↵Doyle LW, the Victorian Infant Collaborative Study Group. Evaluation of neonatal intensive care for extremely low birth weight infants in Victoria over two decades. 1: Effectiveness. Pediatrics.2004;3 :505– 509
- ↵Drummond MF, O’Brien B, Stoddart GL, Torrance GW. Methods for the Economic Evaluation of Health Care Programmes. Oxford, United Kingdom: Oxford University Press; 1997
- Copyright © 2004 by the American Academy of Pediatrics