Evaluation of Neonatal Intensive Care for Extremely Low Birth Weight Infants in Victoria Over Two Decades: I. Effectiveness

METHODS

The state of Victoria comprises approximately one fourth of the population of Australia. The annual number of live births of all weights in the state from 1979 to 1997 was ∼60 000, ranging between 57 767 (1979) and 66 350 (1990).² Over the whole period, there have been 3 level-III perinatal centers (high-risk obstetric referral centers with a neonatal intensive care unit [NICU]) in the state, and another NICU located within a stand-alone children’s hospital. All 4 NICUs are located in Melbourne, the capital city of Victoria. Approximately two thirds of the population of Victoria live in Melbourne.

The ELBW subjects comprised consecutive live births of birth weight 500 to 999 g born in the state of Victoria during 4 distinct eras: 1979–1980, 1985–1987, 1991–1992, and 1997. Data were collected prospectively from multiple sources including all individual hospitals through the Victorian Perinatal Data Collection Unit, the 4 NICUs, and the Newborn Emergency Transport Service (the sole retrieval service for the state) to obtain and cross-check data on the total number of live births and their survival rates.³ Controls comprised randomly selected normal birth weight (NBW, birth weight >2499 g) live births for the first, third, and fourth ELBW cohorts. The first NBW cohort was derived from births during 1981–1982 in 1 of the 3 level-III perinatal hospitals in the state (the Royal Women’s Hospital). The other 2 NBW cohorts were derived from each of the 3 level-III perinatal centers, matched with the ELBW survivors from the respective centers for the mother’s health insurance status, language spoken primarily in her country of birth (English or other), and the child’s gender. NBW controls were selected to provide a basis for comparison of outcomes with the ELBW infants, particularly for psychological test scores.

The survival rate was determined at 2 years of age (corrected for prematurity where appropriate). Survivors were traced by a variety of means, including obtaining primary and secondary contact details during the newborn period, and by maintaining contact after discharge, including sending birthday cards. Follow-up assessments were facilitated by flexible follow-up appointments, occasionally by providing assistance for families to attend and rarely by visiting some families at home (when attendance at the assessment center proved to be too difficult); the latter sometimes involved traveling to other states or countries. Survivors were assessed at 2 years of age by pediatricians and psychologists blinded to perinatal details. Over the time period of the study, the assessment staff was relatively stable; there were only 4 individual psychologists and 10 individual pediatricians involved in the assessments. As each new member joined the team, they were trained by more-experienced colleagues to ensure uniformity of data collection.

Impairments included cerebral palsy (CP), blindness (diagnosed by pediatric ophthalmologists during the first 2 years of life), deafness (requiring hearing aids), and developmental delay. Criteria for the diagnosis of CP included loss of motor function combined with abnormalities of tone and tendon reflexes.⁴ The first 3 ELBW cohorts were assessed psychologically with the Mental Developmental Index (MDI) of the original Bayley Scales of Infant Development,⁵ whereas the last group was assessed with the revised Bayley Scales.⁶ A developmental quotient (DQ) for each child was computed relative to the mean and standard deviation (SD) of the MDI for the NBW controls as (child’s MDI − mean MDI for NBW controls)/SD.

Developmental delay comprised a DQ of less than −1 SD. The mean MDI on the original Bayley Scales for the NBW controls from the first era was 105.8 and rose further over time to 114.9 for the NBW cohort from the 1991–1992 era. A target mean for MDI of 110 (SD: 16) was assumed for the second ELBW cohort to be an approximate midpoint between the first 2 NBW cohorts. Children unable to complete psychological testing because of presumed severe developmental delay were assigned a DQ of less than −3 SD. Several children assessed with alternative psychological tests to the MDI were assigned DQs based on SD scores for the respective test.

Severe neurosensory disability comprised severe CP (child unlikely ever to walk), blindness, or a DQ of less than −3 SD; moderate disability comprised moderate CP (child not walking at 2 but expected to do so eventually), deafness, or a DQ from −3 SD to less than −2 SD; and mild disability comprised mild CP (child walking at 2) or a DQ from −2 SD to less than −1 SD. The remaining children were considered to have no disability. For several children from the first 2 cohorts who were not assessed at 2 years of age but were assessed later in childhood, disabilities were classified along similar criteria, except intelligence quotient scores were used instead of the MDI. The few ELBW survivors not assessed at any age (n = 6) were assumed to be nondisabled. Utilities for survivors were assigned according to the severity of the disability: 0.4 for severe, 0.6 for moderate, 0.8 for mild, and 1 for no disability.⁷ Infants who died had a utility of 0. Utilities were multiplied for children with multiple disabilities. Utilities were summed and divided by the number of live births to calculate the quality-adjusted survival rate. The quality-adjusted survival rate will always be lower than the survival rate unless there are no disabled survivors. The mean utility for a group of survivors is obtained by the ratio of the quality-adjusted survival rate and the survival rate.

The Research and Ethics Committees at the Royal Women’s Hospital (Melbourne, Australia) approved these follow-up studies. Written, informed consent was obtained from parents of NBW infants. Follow-up was considered routine clinical care for ELBW infants.

Data were analyzed by SPSS for Windows programs⁸ and the Confidence Interval Analysis program.⁹ Differences between means were compared by the mean difference and its 95% confidence interval (CI). Differences between 2 proportions were contrasted by percent differences and 95% CIs. Differences in proportions in ordered categories were contrasted by χ² for linear trend. Differences between eras in quality-adjusted survival rates were compared by the Mann-Whitney U test. P values <.05 were considered statistically significant.

RESULTS

There were 351, 560, 429, and 233 ELBW live births in the state in the 4 eras, respectively, representing annual rates of 3.03, 3.06, 3.29, and 3.77 ELBW live births per 1000 live births in the 4 eras, respectively. There were 60, 263, and 198 NBW controls for the first, third, and fourth eras, respectively.

The survival rate to 2 years of age rose progressively between eras overall (Table 1). Not only were the differences between consecutive eras statistically significant (Table 2), they also were clinically important; the overall survival rate was only ∼1 in 4 in 1979–1980 but had increased to ∼3 in 4 by 1997. The survival rate rose progressively between eras in all birth-weight subgroups (Table 1 and Fig 1), but not all stepwise increases were statistically significant (Table 2 and Fig 1). The biggest gains in survival in the earlier eras were in heavier infants, whereas the biggest gains in survival in the later eras were in the lightest infants (Table 1 and Fig 1).

• Download figure
• Open in new tab
• Download powerpoint

Fig 1.

Survival rates and quality-adjusted survival rates in ELBW survivors in 100-g birth-weight subgroups, contrasted between eras (1979–1980: 1; 1985–1987: 2; 1991–1992: 3; 1997: 4). An asterisk indicates a statistically significant difference in both survival and quality-adjusted survival rates compared with the immediately preceding era within the 100-g birth-weight subgroup. A dagger indicates statistically significant difference in quality-adjusted survival rate compared with the immediately preceding era within the 100-g birth-weight subgroup.

The mean MDI (SD) on the original Bayley Scales for the 48 NBW controls from the 1981–1982 era able to be tested was 105.8 (16.4), significantly higher than the expected mean of 100 (mean difference: 5.8; 95% CI: 1.1, 10.6). The mean MDI (SD) on the original Bayley Scales for the 233 NBW controls from the 1991–1992 era able to be tested was 114.9 (19.5), significantly higher than the expected mean of 100 (mean difference: 14.9; 95% CI: 12.4, 17.4). In contrast, on the revised Bayley Scales for the 187 NBW controls from the 1997 era able to be tested, the mean MDI was 99.2 (15.4), not significantly different from the expected mean of 100 (mean difference: −0.8; 95% CI: −3.0, 1.5). One NBW control from the first era was severely disabled from CP and was unable to be tested on the MDI; no other NBW control had CP or was blind or deaf in any era at 2 years of age.

The eventual follow-up rates were high for each ELBW cohort in each era (1979–1980: 100% [89 of 89]; 1985–1987: 100% [212 of 212]; 1991–1992: 98% [237 of 241]; 1997: 99% [168 of 170]), although 5 children in the 1979–1980 era and 1 child in the 1985–1987 era were not assessed until after 2 years of age. Rates of CP, deafness, and developmental delay were similar in the ELBW survivors in all eras (Table 3). There was, however, a significant fall in the rate of blindness over time. The overall rate of neurosensory disabilities was stable for ELBW survivors over time (Table 3).

Neurosensory disability rates in NBW controls were as follows: 1981–1982 (n = 49): 39 nil (80%), 8 mild (16%), 1 moderate (2%), 1 severe (2%); 1991–1992 (n = 241): 197 nil (82%), 35 mild (15%), 5 moderate (2%), 4 severe (2%); 1997 (n = 187): 156 nil (83%), 25 mild (13%), 3 moderate (2%), 3 severe (2%). These rates were significantly lower in NBW controls compared with their respective ELBW cohorts (Table 3; χ² for linear trend: first era = 18.9, P < .0001; third era = 41.3, P < .0001; fourth era = 48.5, P < .0001).

The quality-adjusted survival rate to 2 years of age rose progressively between eras overall (Tables 1 and 2 and Fig 1). As with the survival rate, the quality-adjusted survival rate rose progressively between eras in all birth-weight subgroups, but again not all stepwise increases were statistically significant (Tables 1 and 2 and Fig 1). As for survival, the biggest gains in quality-adjusted survival in the earlier eras were in heavier infants, whereas the biggest gains in quality-adjusted survival in the later eras were in the lightest infants (Tables 1 and 2 and Fig 1).

Consistent with the stable neurosensory disability rates in survivors, the mean utility per survivor for each era was also relatively stable (1979–1980: 0.765; 1985–1987: 0.800; 1991–1992: 0.840; 1997: 0.807). There was more fluctuation in mean utility rates per survivor in birth-weight subgroups, but some sample sizes were quite small (500–749 g birth weight [1979–1980: 0.620; 1985–1987: 0.863; 1991–1992: 0.799; 1997: 0.744] and 750–999 g birth weight [1979–1980: 0.780; 1985–1987: 0.839; 1991–1992: 0.853; 1997: 0.847]).

DISCUSSION

Sinclair et al¹ outlined some components of neonatal intensive care that had proven efficacy up to 1981. Since that time, there have been major advances in the perinatal care of extremely tiny or very preterm infants. An example of a known efficacious therapy since 1981 is exogenous surfactant to prevent or treat respiratory distress syndrome.¹⁰ As distinct from individual components, Kitchen et al,¹¹ before 1981, assessed the efficacy of a neonatal intensive care “package” in a clinical trial within an individual hospital. However, there are no reports of clinical trials of the efficacy or effectiveness of a complete neonatal intensive care program within the geographically defined region served by the program.

Despite the lack of randomized, controlled trials of complete neonatal intensive care programs, Sinclair et al¹ indicated that subexperimental studies such as cohort studies had provided some data on effectiveness. They reviewed the available data on the effects of neonatal intensive care on not only mortality but also morbidities such as developmental delay, CP, and retinopathy of prematurity. Despite clearly improving survival rates for very low birth weight (VLBW, birth weight <1500 g) infants, they concluded that effectiveness data for neonatal intensive care programs were not convincing up to 1981. Since that time, there have been many studies documenting improved survival rates of VLBW infants, especially those of birth weight <1000 g^12–15 or even smaller.¹⁶ In our study, not only has the survival rate for ELBW infants improved threefold, from 1 in 4 in the late 1970s to 3 in 4 in the late 1990s, but so has the quality-adjusted survival rate, from 19% in 1979–1980 to 59% in 1997, both of which are measures of the increasing effectiveness of neonatal intensive care within our region. Moreover, despite suggestions that survival rates may have peaked in the early 1990s,¹² in our region they have continued to improve in the late 1990s, particularly in the lightest infants.

With improving survival rates after the introduction of modern neonatal intensive care, morbidities such as CP initially increased in VLBW survivors in geographically defined cohorts.^17,18 More recently, however, there are suggestions that CP rates might be decreasing in the early 1990s within discrete populations.¹⁹ However, we have not observed this in our regional ELBW cohort, with the rate of CP in survivors remaining ∼10% in each era. As predicted by Bhushan et al,²⁰ as the survival rate has improved, the absolute number of survivors per year with CP has increased. The rate of blindness in survivors fell significantly over time in our regional cohort, which is pleasing, particularly because there has been a second “epidemic” of blindness from retinopathy of prematurity with modern intensive care.²¹ Deafness, developmental delay, and overall neurosensory disability rates were not significantly different over time in survivors in our cohorts.

Contemporaneous NBW controls were essential in evaluating our ELBW cohorts for several reasons. First, the original Bayley Scales,⁵ published in 1969, were clearly outdated for our cohorts, with the NBW controls scoring significantly higher than the test mean of 100. We would have underestimated the rate of developmental delay in both the ELBW and NBW cohorts if we had relied on the test mean of 100 rather than using the mean from the NBW controls. The revised Bayley Scales,⁶ on the other hand, were appropriate for the 1997 ELBW and NBW cohorts. Second, NBW controls also have disabilities, which are overlooked if only outcomes for ELBW children are reported. Because NBW children far outnumber ELBW children, it must be remembered that most disabled children in any region will be NBW rather than ELBW. For example, in Victoria in 1997, the number of NBW live births exceeded 58 000,² of whom >9000 would be disabled if the rate of 16% observed in our NBW cohort applied to the whole region. In contrast, there were 82 disabled ELBW children, <1% of the expected number of NBW disabled children.

It is difficult to synthesize the numerous reports of outcomes of ELBW infants because of the many variations between the studies’ designs, of which birth-weight selection is but 1 variable. We have presented data not only overall from 500–999 g birth weight but also in 250-g birth-weight and 100-g birth-weight subgroups to facilitate comparisons with other studies.

We are not the only group to report increasing survival rates of ELBW infants within the same geographically defined region with or after the introduction of modern intensive care. For example, survival rates to hospital discharge of ELBW infants in 1 region in Ontario, Canada increased from 11% in the mid-1960s to 22% in the mid-1970s,²² and in those of birth weight 501-1000 g, survival to hospital discharge from 46% in 1977–1980 to 48% in 1981–1984.²³ In Alberta, Canada, survival rates to 1 year of age in ELBW infants were 13% in 1978–1979,²⁴ 44% in 1988–1989,²⁴ and 59% in 1990.²⁵ In Western Australia, survival to at least 1 year of age in ELBW infants was 35% in 1980–1983 and 43% in 1984–1987.¹³ In North Carolina, survival to 28 days of age increased over time in infants of birth weight 500-1000 g as follows: 20% in 1968–1974; 28% in 1975–1979; 51% in 1980–1984; 62% in 1985–1989; and 69% in 1990–1994.²⁶ Within the limitations of different birth-weight selection, years of birth, and ages at determining outcomes, these mortality rates are similar to those in our study.

A few of these regional studies have also reported on changes in neurosensory outcome over time. Boyle et al²² reported that quality-adjusted life-years gained per live birth increased from 5.5 in 1964–1969, before intensive care, to 9.1 in 1973–1977, after intensive care. Later, from the same region, Saigal et al²³ reported a significant drop in functional disability in survivors of birth weight 501-1000 g, from 50% in 1977–1980 to 27% in 1981–1984, although the survivors did not have formal psychological tests. Robertson et al²⁴ described stable impairment rates in ELBW survivors from 1978 to 1979 and 1988 to 1989 and again in 1990.²⁵ Disability rates were also stable in ELBW survivors in Western Australia from 1980 to 1983 and 1984 to 1987.¹³ Disability rates in survivors were stable over time in our study, but because there were large gains in survival, the quality-adjusted survival rate improved substantially, supporting the progressive effectiveness of neonatal intensive care within our region.

Neonatal intensive care has been increasingly effective for ELBW infants born in Victoria from the late 1970s to the late 1990s. However, clearly, the rates of disabilities in ELBW children are too high relative to NBW controls and, because mortality rates have fallen, have become the greatest challenge to improving the outcomes of neonatal intensive care for ELBW infants.