The stability of the diagnosis of developmental disability between ages 2 and 8 in a geographic cohort of very preterm children born in 1997
- 1Premature Infant Follow-up Program at the Royal Women's Hospital, Parkville, Victoria, Australia
- 2Department of Paediatrics, The University of Melbourne, Melbourne, Victoria, Australia
- 3Murdoch Childrens Research Institute, Melbourne, Victoria, Australia
- 4Department of Psychology, The University of Melbourne, Melbourne, Victoria, Australia
- 5Department of Obstetrics and Gynaecology, The University of Melbourne, Melbourne, Victoria, Australia
- Correspondence to Dr Gehan Roberts, Department of Obstetrics and Gynaecology, The Royal Women's Hospital, 20 Flemington Road, Parkville, VIC 3052, Australia;
Contributors All the listed authors contributed substantially to the article.
- Accepted 16 September 2009
- Published Online First 14 October 2009
Objective Studies reporting the developmental outcomes for very preterm (VPT) children often focus on development at age 2 years. The aim of this study was to assess the stability of the rates of diagnosis of developmental disability from age 2 to 8 years in a regional cohort of VPT and extremely low–birth weight (ELBW) children and compare these with term controls.
Methods VPT (22–27 completed weeks of gestation) and ELBW (birth weight 500–999 g) children and matched term controls born in Victoria, Australia, in 1997 were enrolled at birth in a regional prospective longitudinal study. Outcomes were assessed at ages 2 and 8 years.
Results Of the 283 VPT/ELBW live births, 71% survived to age 8 years and 94% were assessed. Of the 199 controls, 100% survived to age 8 years and 86.9% were assessed. At age 2 years, the rates of nil, mild, moderate and severe disabilities in the VPT/ELBW children were respectively 51.9%, 20.9%, 13.4% and 13.9%; and at age 8 years, 43.9%, 36.9%, 10.7% and 8.6%. The Cohen κ statistic revealed poor agreement between disability status at ages 2 and 8 years for VPT/ELBW children (0.20, p<0.001) but a higher level of agreement (0.37, p<0.001) for the control children. This was primarily driven by differences in classification of cognitive disability between ages 2 and 8 years.
Conclusions Developmental outcomes at age 2 years are only a moderate predictor of long-term outcome and are not a reliable end point for follow-up studies of VPT and ELBW children.
The survival rates of very preterm (VPT, 22–27 completed weeks of gestation) and extremely low–birth weight (ELBW, 500–999 g birth weight) infants continue to rise, and greater proportions of these cohorts are comprised of the least mature infants.1 Long-term follow-up studies have shown that >50% of these children have academic or behavioural difficulties at school age.2 The least mature of these children have even higher rates of impairment.3 4
Therapeutic trials in the neonatal period are increasingly using developmental outcomes to judge the effectiveness of the intervention, but these studies frequently report developmental outcomes only until age 2 years.5 Concerns have been raised in a previous study regarding the predictive validity of using developmental status in infancy to judge the efficacy of interventions.6 However, this study examined a cohort of children admitted to a single neonatal intensive care unit and did not have a prospectively recruited control group. Before recommending changes in practice based on developmental outcomes reported from intervention trials, it is important to be aware of the limitations of current practice and to re-examine the previously raised concerns using a geographically representative cohort of children compared with a matched control group. This knowledge is also important when counselling families of VPT children who tend to be followed-up up to 1–2 years of age and are no longer followed-up if they appear to be following a fairly typical developmental trajectory.
What is already known on this topic
▶. Developmental outcomes up to age 2 years are frequently used to assess the efficacy of interventions in VPT children.
▶. Concerns have been raised regarding the long-term predictive validity of diagnosing disability in infancy.
What this study adds
▶. The poor predictive validity of making a diagnosis of disability at age 2 years compared with that at age 8 years is confirmed in both preterm and term children.
▶. While neurosensory outcomes remain relatively stable between ages 2 and 8 years, there is poor agreement between the diagnosis of cognitive disabilities between ages 2 and 8 years.
The aim of the current study was to compare the stability of the rates of diagnosis of developmental disability in a regional cohort of VPT/ELBW children at age 2 years with rates at age 8 years. The following hypotheses were tested: (1) that rates of disability would not be stable over time and fewer children would meet the classification criteria for moderate-to-severe disability at age 8 years than at age 2 years and (2) that change in disability classification would be primarily driven by an overdiagnosis of cognitive disabilities at age 2 years compared with age 8 years. A secondary aim was to compare these rates of change with those of a term control group.
Of 283 consecutive live births between 1 January and 31 December 1997, in the state of Victoria, Australia, with a gestational age range of 22–27 completed weeks or birth weight of 500–999 g, 201 survivors up to 8 years of age comprise the VPT/ELBW cohort born, a survival rate of 71%. The normal–birth weight cohort comprised 199 randomly selected children with a gestational age >36 weeks or birth weight >2499 g, all of whom survived to 8 years of age. The controls were born on the expected date of birth of VPT/ELBW children and recruited before their discharge. They were matched for sex, the mother's country of birth (English speaking or not) and health insurance status (private health insurance or not). All the children were enrolled during the newborn period in a prospective longitudinal study of growth and development throughout childhood; results at 2 years of age have been reported.7 Written informed consent was obtained from the parents of the control children. The research and ethics committees of the Royal Women's Hospital, Melbourne, approved these follow-up studies.
The survivors were assessed at 2 and 8 years of age by paediatricians and psychologists blinded to the perinatal details, predominantly in specialised follow-up clinics, although a few were tested elsewhere if they could not attend the clinics. Age was corrected for prematurity to be consistent with assessments of earlier cohorts and to avoid bias in the cognitive test scores.8 Impairments that were diagnosed at follow-up included cerebral palsy (CP), blindness, deafness and cognitive impairment (index scores <−1 SD relative to the mean for the controls). CP was diagnosed by the paediatrician on the basis of loss of motor function, combined with abnormalities of tone or tendon reflexes, blindness comprised visual acuity worse than 6/60 in the better eye and deafness comprised hearing loss that required amplification or worse.
At age 2 years, cognitive ability was assessed using the mental developmental index (MDI) of the Bayley Scales of Infant Development—Second Edition (BSID-II)9; and at age 8 years, with the full-scale intelligence quotient (IQ) of the Wechsler Intelligence Scale for Children—Fourth Edition (WISC-IV).10 The index scores on both scales have a mean of 100 and an SD of 15; however, the VPT/ELBW children were compared with the term controls in assigning the disability criteria to study the VPT/ELBW children in the context of their typically developing peers from the same geographic area. Standardised scores for developmental quotient (DQ) and IQ were calculated relative to the mean (SD) of the controls for the MDI and full-scale IQ, respectively. Some children with significant disabilities were not able to be assessed with the BSID-II or WISC-IV and were assigned scores of −4 SD. A small number of children were not able to complete all the subtests of the WISC-IV, primarily because of neurosensory impairments. For the children who were unable to complete the visual-motor subscales of the WISC-IV because of CP-related motor impairment or visual impairment, their verbal comprehension index was used as an estimate of IQ. For the children who were unable to complete the language-based subscales of the WISC-IV because of significant hearing impairment, the perceptual reasoning index score was used as an estimate of IQ. Two children were unable to complete any of the WISC-IV subtests because of severe disability. These children were assigned an IQ standard score of −4 SD. Two children did not complete all the subscales of the WISC-IV because of a lack of compliance, and their IQ score was calculated based on the completed subscales. Analyses were carried out including and excluding the children with incomplete assessments.
Severe disability was defined as severe CP (child is expected never to walk), blindness or a cognitive score <−3 SD; moderate disability, as moderate CP (not walking at 2 years old but expected to walk or walking with considerable difficulty, with or without appliances at age 8 years), deafness or a cognitive score from −3 to <−2 SD; and mild disability, as mild CP (walking with minimal limitation) or a cognitive score from −2 to <−1 SD.11
Descriptive analyses were used to show survival and follow-up rates of the VPT/ELBW and full-term control groups at age 8 years. The rates of overall disability and specific disabilities at age 8 years were compared between the VPT/ELBW and control groups using χ2 analyses or the Fisher exact test for the categorical outcomes. Mean differences in IQ were compared between groups using unpaired t tests.
To examine the stability of the diagnosis of developmental disabilities at age 2 years with the rates at age 8 years, the Cohen κ was used. Cohen κ is a measure of agreement for categorical data.12 Values of 0.0–0.4 can be thought of as representing slight to fair agreement; 0.4–0.6, moderate agreement; and 0.6–1.0, substantial to perfect agreement.13 The ordinal disability category at age 2 years was subtracted from the ordinal disability category at age 8 years to generate a new ordinal variable that measured change in disability categorisation over time. Systematic change in disability classification was measured using the Wilcoxon signed-rank test. Each of the components of the disability diagnosis (CP, deafness, blindness and DQ–IQ differences) were analysed individually to assess which contributed to changes in diagnosis. The difference between IQ score at age 8 years and DQ score at age 2 years was analysed by linear regression to determine the effects of the matching sociodemographic variables (mother's health insurance status and country of birth, and the child's sex) on changes in cognitive function over time within the VPT/ELBW group.
Table 1 outlines the characteristics of the VPT/ELBW and control groups. Most VPT/ELBW children and all of the controls survived to age 8 years; there were no deaths between ages 2 and 8 years. Follow-up rates were higher for the VPT/ELBW children than for the controls. All preterm/ELBW children seen at age 8 years were also seen at age 2 years; however, three term controls seen at age 8 years did not have a complete assessment at age 2 years.
At age 8 years, the VPT/ELBW group had a statistically significantly higher rate of CP and overall disability compared with the control group as shown in table 2. The mean IQ for the VPT/ELBW group was significantly lower than the term controls. The mean IQ score for the VPT/ELBW children after exclusion of those who did not complete the full WISC-IV 94.4 (SD 14.2) was very similar to that of the entire VPT/ELBW group.
Disability categorisation for the VPT/ELBW children at age 2 years had a poor level of agreement with categorisation at age 8 years, as seen in table 3 (κ=0.20, p<0.001). Reanalysis of the data from only the VPT (n=144) or only ELBW (n=160) children had little effect on the poor level of agreement (κ=0.21, p<0.001 and κ = 0.19, p<0.001, respectively). The overall rate of children with disability increased between 2 and 8 years of age, although fewer children met the criteria for moderate to severe disability at age 8 years. In contrast, there was a higher level of agreement for disability classification over time for the control group (κ=0.37, p<0.001, data not shown).
Table 4 shows that only 46.0% of the VPT/ELBW children remained at the same disability classification at age 8 years compared with those at age 2 years. There was no systematic change in disability classification, with 27.0% of the VPT/ELBW children classified as less disabled at age 8 years and 25.9% classified as more disabled (p=0.47). In contrast, 80.9% of the control group remained at the same disability classification at both ages, again with no systematic change (p=0.31).
Most of the change for the VPT/ELBW cohort occurred within the classification of cognitive disability, with poor agreement between ages 2 and 8 years, as shown in fig 1 (κ=0.11, p=0.02). The degree of agreement in cognitive abilities for the control group was much higher (fig 1; κ=0.37, p<0.001), The disability classifications for the other domains were relatively stable for the VPT/ELBW children. One of the four children classified as blind at 2 years of age was found not to have a severe visual defect at age 8 years. One child who did not need hearing aids at age 2 years did have hearing aids at age 8 years. The agreement in CP classification was moderate (κ=0.56, p<0.001).
Within the VPT/ELBW group, sociodemographic variables contributed to the variation in cognitive outcomes over time. Female sex had a negative effect of 0.58 SD on cognitive change between 2 and 8 years of age (p=0.002); that is, females, on average, deteriorated relative to the males over time. Children whose mothers were not born in an English-speaking country had, on average, positive change over time (0.58 SD, p= 0.013). Private health insurance status did not make a significant difference (−0.30 SD, p=0.15). These variables explained 8.9% of the variance in the change in cognitive outcomes between 2 and 8 years of age.
If cognitive impairment had been computed on the basis of a mean of 100 and SD of 15 for both the MDI and the full-scale IQ on the WISC-IV, the rates of cognitive impairment would have been greatly underestimated in both groups of children at 8 years of age compared with that of those at 2 years of age (controls, 6.0% at 8 years of age compared with 16.8% at 2 years of age; Wilcoxon rank sum test Z=3.8, p<0.001; VPT/ELBW group, 24.2% at 8 years of age compared with 45.7% at 2 years of age; Wilcoxon rank sum test Z=6.0, p<0.001).
The primary study hypothesis was confirmed: the rates of disability for the VPT/ELBW children were not stable between the ages of 2 and 8 years. As hypothesised, fewer children met the criteria for moderate and severe disabilities at age 8 years compared with those at age 2 years. An unexpected finding was that fewer children also met the criteria for no disability at age 8 years, so that there was no systematic change in disability classification. The rates of disability for the control children were more stable between the ages of 2 and 8 years. The poor agreement between ages 2 and 8 years in the VPT/ELBW children was primarily driven by differences in classification of cognitive disability, with the number of children with moderate or severe cognitive disability reducing between ages 2 and 8 years. This is not surprising given that cognitive impairment is the most common reason for disability at either age. On the other hand, there was much better agreement in the classification of CP, deafness and blindness over time.
Of interest was the shift in classification for the VPT/ELBW group from no disability at age 2 years to mild disability at age 8 years. This was largely because of movement within the cognitive classification. Age-appropriate cognitive development at age 2 years does not imply that cognitive impairments will not emerge later. However, this movement in cognitive disability classification was not restricted to nil or mild disability, and a large proportion of the children who were classified with a moderate to severe cognitive disability at age 2 years did not retain this classification at age 8 years. These findings indicate that early classification of cognitive disability is not a reliable predictor of later cognitive functioning in the VPT/ELBW children, supporting previous reports.6 Sociodemographic variables only explained a small proportion of the variance in cognitive outcomes between ages 2 and 8 years.
This discrepancy could be at least partly related to the different cognitive measures used at the two time points. The BSID-II was designed to detect developmental delay. Trajectories of development vary greatly in early childhood, and a significant proportion of children who exhibit early developmental delay will catch up to their peers, and a proportion of young children who display accelerated development will not maintain this advantage. Cognitive outcome on the BSID-II is heavily dependent on language skills, and this may have led to increased identification of disability at age 2 years in the VPT/ELBW cohort. The VLBW children have been previously shown to have positive changes in expressive vocabulary between ages 3 and 8 years.14 In contrast, the WISC-IV was specifically designed to assess general intellectual functioning and, by age 8 years, is considered a reasonable predictor of later IQ.15 The WISC-IV has a heavy emphasis on reasoning ability and working memory, both elements of executive functioning that has been reported to be a specific area of vulnerability in VPT/ELBW children.16 This may account for the increase in the rates of children identified with mild disabilities at age 8 years.
It is important to note that although the mean IQ score of the VPT/ELBW children was within the mean range according to test norms, it was 0.8 SD below the mean IQ score of the control group, which is a clinically significant difference. Instead of classifying disability according to test norms, in this study, children were classified according to the mean of the matched full-term control group in the attempt to minimise the problems associated with (1) contrasting values from two different measures, (2) relying on norms obtained from a different country and (3) the Flynn effect, whereby general ability scores rise over time such that test norms overestimate true ability. Examining the proportions of children who have IQ scores <−1SD below the test mean provides validation of this method: the proportion of control children with scores in this range is greatly underestimated using comparison with test norms, 6.8% compared with an expected value of 15.9% for a normal distribution. Although this approach increases the rate of VPT/ELBW children with a disability, it is a better reflection of the challenges these vulnerable children face compared with typically developing peers. In the EPICure cohort of children born at <26 weeks of gestation and assessed at 30 months and 6 years of age, moderate cognitive disability was reported in 21% of children when based on test norms, but this proportion rose to 41% when based on the mean of classroom controls.17
Hack et al6 have shown in a single hospital cohort of ELBW children that Bayley MDI scores at age 20 months were poorly predictive of cognitive function at age 8 years, based on the Kaufman Assessment Battery for Children. Although 39% of the ELBW children had MDI scores that were in the moderate-to-severe range (<−2 SD below the mean) at age 2 years, this dropped to 16% at age 8 years. It is important to note that these scores were according to test norms and not compared with the controls.
Stability in disability rates over time has been reported previously in the EPICure study.17 Severe disabilities at 30 months of age (using criteria very similar to those in the current study) were predictive of severe disabilities at 6 years of age: of the children who met the criteria for severe disability at 30 months of age, 86% continued to do so at age 6 years, compared with 35% in the current study. Less severe disabilities, however, were poorly predictive of disability status at age 6 years. Possible explanations for the greater stability of more severe disabilities over time in the EPICure study compared with that in the current study may be the older initial assessment age (30 vs 24 months), follow-up at a closer time point to the initial assessment (6 vs 8 years) and the more immature cohort of preterm children, with correspondingly higher disability rates, in the EPICure cohort.
In contrast to the poor stability of cognitive outcomes over time, the other neurosensory outcomes (deafness, blindness and CP) had a much higher level of stability between ages 2 and 8 years. As shown in the current study, these disabilities can be diagnosed with a high degree of confidence at age 2 years when reporting outcomes of intervention trials and providing feedback regarding long-term outcomes to families. The stability of these outcomes over time has been previously demonstrated in an earlier VPT/ELBW cohort from the same geographic area.11 18
The current study has several strengths. The VPT/ELBW cohort was a regional cohort, free of referral bias. The follow-up rates were high. The controls were recruited prospectively and were matched for variables that are known to influence long-term outcomes.19 The study weaknesses include the lower retention rate at age 8 years of the controls, although the controls seen at age 8 years were representative of the total control population in their cognitive development at age 2 years (data not shown). The disability criteria used in the current study do not include additional areas of difficulty that have been shown to occur commonly in VPT/ELBW children, including behaviour, attention and learning difficulties.2 Examining only neurosensory deficits and cognitive development underestimates the true frequency of difficulties in this high-risk population.
In conclusion, developmental outcomes at age 2 years are not a reliable end point for assessing long-term outcomes of children who are born VPT or with ELBW. Outcomes at this age should not be used to make final judgements about the effectiveness of therapeutic interventions or to provide definitive feedback to families regarding long-term prognoses. The diagnosis of a disability should be made with caution during infancy and toddlerhood. Developmental assessment at age 2 years may be most appropriately thought of as a “snapshot” of current status that is best used to direct intervention services appropriately, and follow-up of these vulnerable children should be continued until at least school age.
Participants Convenor Lex W Doyle, MD, FRACP; Collaborators (in alphabetical order): Peter Anderson, PhD; Catherine Callanan, RN; Elizabeth Carse, FRACP; Margaret P Charlton, MEd Psych; Noni Davis, FRACP; Cinzia de Luca, BSc; Julianne Duff, FRACP; Marie Hayes, RN; Esther Hutchinson, BSc (Hons); Elaine Kelly, MA; Marion McDonald, RN; Gillian Opie, FRACP; Gehan Roberts, MPH, PhD, FRACP; Linh Ung, BSc (Hons); Andrew Watkins, FRACP, and Heather Woods, RN.
Funding This study was funded by the Victorian Government, who had no involvement in the design, the collection, the analysis and the interpretation of data; in the writing of the report and in the decision to submit the paper for publication.
Competing interests None.
Ethics approval This study was conducted with the approval of the Royal Women's Hospital Research and Ethics Committee.
Provenance and peer review Not commissioned; externally peer reviewed.