Chapter Contents
History: early neonatal care 71
The contribution of prematurity to disability in the community 72
Reporting outcomes and study design 72
Outcome in early infancy 74
Motor development 74
Developmental progress 78
Current follow-up practice 79
Predicting disability in preterm infants 79
Interventions to improve outcomes for very preterm babies 79
School-age outcomes 80
Cognitive function and school performance 80
Behaviour and psychology 81
Motor/neurology 83
Growth and medical outcomes 83
Function and quality of life in surviving ex-preterm children 83
The ex-preterm young adult 84
Pregnancies deliver prematurely because of problems with the fetus, the mother or the intrauterine environment; in each of these situations fetal health is compromised before birth and it is perhaps unsurprising that many such children develop later problems, particularly following very preterm birth before 32 weeks of gestation. This period from the middle of the second trimester into the first month of the third trimester is associated with rapid organ growth and development. In the brain, although neuronal migration is mainly complete, cortical structure evolves rapidly and there are important structural elements that develop in this critical period. The processes surrounding preterm birth may interrupt neuronal or oligodendroglial migration and cortical structural development. Furthermore, we understand little of how our clinical care and the effects of the neonatal intensive care unit (NICU) environment interact positively or negatively with these processes. It is, therefore, encouraging that the great majority of survivors develop appropriately and enjoy a good quality of life though to adulthood ( ; ; ; ), but others do develop a range of impairments that interfere with health or education and lead to problems that persist throughout the lifespan. It is these who are the focus of this chapter.
History: early neonatal care
Interest in the care of the newborn has waxed and waned through history. The invention that marked the dawn of ‘modern’ neonatal care was the incubator, first credited to a French obstetrician, Stéphane Tarnier, derived from an agricultural device to hatch hens’ eggs. The introduction into the Paris Maternité hospital in 1878 facilitated the development of the ideas published by Pierre Budin in his classic monograph entitled Le Nourissant ( The Nursling ) ( ). Although mainly concerned with inpatient care, there is one surviving photograph of his ‘graduates’.
Martin Couney, a pupil of Budin, requested a chance to exhibit the new incubators in the World Exposition in Berlin in 1896. Permission granted, he conceived the idea of a live exhibit with babies, otherwise destined to die, from a local charitable hospital. These Kinderbrutanstäldter (child hatcheries) proved a great success and were soon imitated across the world. Couney himself settled in Cooney Island, where his exhibitions became famous; his own daughter was born prematurely and incubated for 3 months ( ).
Couney’s influence led Julius Hess to establish the eight-cot premature baby station at the Sarah Morris Hospital, Chicago. This rapidly expanded and led to the first reliable reports of survival and outcome (Hess 1934). No significant deviation from normal term controls was noted in this highly selected population and no effects of birthweight, gestation, sex or maternal illness were found on outcome. Indeed, ), using graduates of Couney’s summer exhibitions, concluded: ‘prematurity does not markedly alter the normal course of mental growth. It neither retards nor accelerates’. However, thinking was soon to change.
Over the next decade the importance of the length of gestation was appreciated and increased morbidity in the smaller survivors was noted, leading to questions about care for these vulnerable babies which persist right through to today. For example, ) observed that improvements in mortality had ‘been accompanied by a rising survival of immature, malformed, birth injured and weakly babies’ and ) warned that improvements in care might lead to increasing numbers of handicapped children in the community.
Care had up to this point been crude, as exemplified by the uncertainty surrounding the use of oxygen ( ), but the 1960s heralded the start of more careful and considered interventions such as early feeding with breast milk and more appropriate respiratory interventions. There were conflicting reports as to trends in outcomes during this time. The first important systematic attempt to review trends in outcome for very-low-birthweight (VLBW) births was published by ). They indicated some of the difficulties in comparing reports, often from specialist centres reporting selected groups of infants. They summarised outcome from published reports as death, handicapped or healthy survivor and averaged over adjacent 5-year periods over 32 years. From 1960 mortality was found to have fallen progressively, but in each epoch 6–8% of total livebirths had handicaps ( Fig. 3.1 ). There was no evidence, therefore, of an excess of handicapped children in the community as a result of these improvements in care. The same trends applied equally to the subgroup <1000 g birthweight.
The contribution of prematurity to disability in the community
The notion that increased survival brought an additional burden of disability into the population remains even today. The concept of a ‘gain’ in a healthy survivor is set against a ‘loss’ as a child with disability. Overall, the gain in terms of healthy survivors far outweighs the increasing burden of disability in the population. Since the early 1980s there has been an explosion of interest in this area and a huge literature base has evolved. Over short periods of time it is clear that trends in disability rates are hard to demonstrate but mortality has continued to decrease as the gestational age at which birth is considered of borderline viability has fallen, generally without increasing the proportion of survivors with handicap or disability.
The increasing anxiety surrounding the provision of care for the baby of borderline viability persists even today ( ) and is now directed to babies born before 25 weeks of gestation, such is survival at later gestational ages. Survival at these low gestations is in part determined by the attitudes of caregivers, some centres preferring not to offer care at these low gestations in the belief that outcome is likely to be so poor that to attempt intensive care is considered wrong. Very few studies have attempted to define the results of different models of care. One important study compares the outcome in a health service with near-universal initiation of intensive care in the USA against selective initiation in the Netherlands; this resulted in 24.1 additional survivors and 7.2 additional cases of cerebral palsy (CP) for each 100 livebirths ( ). Thus the gain and loss account is still in favour of intensive care if CP is used as an outcome; many would argue that cognitive and behavioural impairment is an even more important adverse outcome measure, but one for which the appropriate study has not been done.
Improving survival has been recognised in the UK by the reduction in gestational age for the definition of stillbirth in 1992 from 28 to 24 weeks. Furthermore, as we will see, although most survivors are free of major handicapping conditions, many more subtle (and less easy to predict) conditions in the fields of learning, motor skills and behaviour are now apparent in survivors, closely related to the degree of immaturity at birth. For these more subtle conditions, as for serious disability such as CP, there is a close relationship to gestational age ( Fig. 3.2 ). However it must be emphasised that preterm birth only occurs in a small proportion of the population (6–13% in different countries) and that the groups most often studied with the highest prevalence of disability (very preterm or extremely preterm births) constitute less than 1% of all births. Therefore, the population-attributable risk for all of these conditions is highest in term-born children, followed by late and moderate preterm children and lowest for the very/extremely preterm groups. For example, in terms of special educational needs (SEN), the population-attributable percentage for births 24–27 weeks was 5%, for 28–32 weeks 11%, for 33–36 weeks 20% and 65% from births at full term ( ).
Reporting outcomes and study design
Data concerning the progress of neonatal intensive care graduates can be used as part of an ongoing audit of outcome or as part of a hypothesis-based research study. Such a study may explore factors that influence development or developmental trajectories (e.g. social or nutritional factors) or the predictive value of particular observations (e.g. cranial ultrasound appearances). Increasingly, randomised trials of neonatal interventions are including longer term outcomes (usually at 18–24 months) as secondary or even primary outcomes for a range of neonatal treatments and in economic appraisals of such. In an ideal world these would be collected in an agreed fashion for all children using routine systems, but this is as yet not possible.
Comparing studies in this area is fraught with methodological problems that often preclude comparison of outcome statistics between studies, despite attempts to do so (see below). It is important, therefore, that the data are presented as fully as possible to facilitate comparison year on year and between studies. Reports of population outcomes should contain as a minimum:
- •
information on the birth population
- ▪
all births or livebirths in a given geographic population
- ▪
numbers of babies admitted for neonatal intensive care
- ▪
details of transfers for care (in and ex utero)
- ▪
survivors to discharge home
- ▪
- •
information on the selection and composition of a comparator population, if applicable
- •
age at assessment (with range) and correction for prematurity if applied
- •
proportion of the population assessed and reasons for non-assessment – a diagram similar to that recommended for randomised trials in the CONSORT statement ( ) is often the clearest way of displaying this
- •
measures used to define the performance of a population (e.g. intelligence test), definitions of categories and methods used to classify outcome and the numbers with disability in each domain of outcome (e.g. CP, cognitive function, hearing or visual impairment).
This requirement applies less to targeted studies of interventions but nonetheless the discipline of identifying the sample used by reporting the population as suggested above allows much better understanding of the applicability of the results of studies to routine practice. More detailed descriptions of the requirements for such studies have been published ( ; ; for a series of reviews see ) and general reporting requirements for cohort studies should be adhered to ( ).
In 1994, because of a lack of consistency between studies in definitions for long-term outcomes, guidelines for the categorisation of severely impaired outcome following prematurity were developed in the UK ( ). This was the outcome that was felt to be best predictive of later ongoing disability and ascertainment at 2 years was chosen as the age at which serious CP could be identified with some confidence, bearing in mind the difficulty in interpreting very early assessments. More recently this has been revised to include categories of severe and moderate impairment in both neurodevelopmental and somatic domains ( Table 3.1 ) ( ). This simple functional classification should serve as a minimum basis for reporting outcomes for preterm populations and may be easily incorporated in routine clinical follow-up. Alongside this the prevalence of CP (see below) should be reported.
CRITERIA FOR DOMAIN | SEVERE NEURODEVELOPMENTAL DISABILITY | MODERATE NEURODEVELOPMENTAL DISABILITY |
---|---|---|
Any one of the below | Any one of the below | |
Motor | Cerebral palsy with GMFCS level 3, 4 or 5 | Cerebral palsy with GMFCS level 2 |
Cognitive function | Score < −3 sd below norm (DQ <55) | Score −2 sd to −3 sd below norm (DQ 55–70) |
Hearing | No useful hearing even with aids (profound >90 dB HL) | Hearing loss corrected with aids (usually moderate 40–70 dB HL) or Some hearing loss not corrected by aids (usually severe 70–90 dB HL) |
Speech and language | No meaningful words/signs or Unable to comprehend cued command (i.e. commands only understood in a familiar situation or with visual cues, e.g. gestures) | Some, but fewer than five, words or signs or Unable to comprehend uncued command but able to comprehend a cued command |
Vision | Blind or Can only perceive light or light-reflecting objects | Seems to have moderately reduced vision but better than severe visual impairment or Blind in one eye with good vision in the contralateral eye |
Other disabilities (included as additional impairments to SND or NDI) | ||
Respiratory | Requires continued respiratory support or oxygen | Limited exercise tolerance |
Gastrointestinal | Requires TPN, NG or PEG feeding | On special diet or has stoma |
Renal | Requires dialysis or awaiting organ transplant | Renal impairment requiring treatment or special diet |
There is less consensus regarding the definition of disability at later ages. Formal definitions of impairment, disability and handicap are still very useful and should underpin any scheme ( Ch. 2 , Table 2.7). However, these definitions have recently been revised to include broader categories of impairment, activity limitations and participation restrictions within society in the International Classification of Function, Disability and Health (ICF), which, although more relevant to society and individuals, makes simple categorisation for outcome studies much less clear.
In pragmatic terms it is usual to continue to categorise impairment across the domains shown in Table 3.1 , but with the addition of behaviour (using a relevant behavioural scale) and the inclusion of the Manual Abilities Classification System ( ) alongside the Gross Motor Function Classification System ( ). Categorisation on continuous outcomes using standard deviation bands (−1 to −2 sd ; −2 to −3 sd ; < −3 sd ) to denote, respectively, mild, moderate and severely impaired outcomes is conventional. The use of standardised outcome measures (e.g. cognitive, neurological, motor or behavioural domains) differs widely and, to a large extent, depends upon the hypothesis under investigation. Where there are clear internationally accepted definitions (e.g. Diagnostic and Statistical Manual of Mental Disorders, 4th ed. (DSM-IV)), these should be reported alongside results on measures specific to the individual study. However the research hypothesis will determine the choice and selection of investigation required and the selection of such tests has been discussed in reviews ( ; ).
Research evaluations in childhood should still, where possible, be focused on a population base, as described above, and data reported in a standardised fashion. The importance of recruiting contemporary comparison groups is stressed as secular upward drifts in IQ scores are well described ( ) and the results of categorisation of children’s performance will depend on which reference norms are selected with older measures ( ). More recently, changes in test structure and standardisation, for example in the new third edition of the Bayley Scales (Bayley-III: ), have made this more important.
The importance of attempting to examine or at least obtain classifying information on all children in a population as far as possible has been stressed. There is some evidence that the non-responders may comprise an excess of severely disabled children ( ; ; ), although there is a paucity of data from other studies to support this. Increasingly the issue of research governance and the necessity to respect parents’ privacy means that fewer attempts to contact parents for follow-up investigations are justifiable and research ethics committees are mindful to limit these. Difficulties with contacting parents after some time means that the taking of consent for later research evaluations should be done at the point of original entry into the study and contact maintained with newsletters, birthday cards or similar, as this should help to minimise loss to follow-up.
Outcome in early infancy
Motor development
Schemes for neurological examination of the newborn, such as those of Amiel-Tison, Dubowitz and Prechtl, are described in Chapter 40 part 1 . Neonatologists must be aware of the different patterns of development over the first year and, where supported by evidence, be prepared to embrace a range of interventions in the nursery and after discharge to optimise a child’s development.
Despite the reliance on ultrasound-detected brain lesions for prognosis, clinical assessment of the infant at discharge is of at least equal importance to evaluating the results of cranial imaging. ) have described a simple functional examination to be performed at around 40 weeks’ postmenstrual age. At this age, the very preterm baby is quite different from a newborn term infant, particularly if born at a very low gestational age at birth. Ex-preterm babies are more visually active and demonstrate less flexion and more extensor activity. When the examination is optimal and the results of cranial ultrasound show no abnormality, the risk of abnormality at 12 months of age is negligible ( ). The positive predictive value is less useful as many babies will eventually lose their non-optimal signs early in the first year.
The very preterm baby is hypotonic and weak, with a poorly calcified skeleton. These impart particular vulnerability to external influences on the development of neurological tone and to skeletal deformations. Nursing postures may lead to changes in head and chest shape if they are not varied and prone positioning will encourage extensor posturing and external hip rotation with shortening of the hip adductors ( ). This may encourage hip dislocation if the child develops spasticity but may easily be avoided using simple postural management ( ). The recent interest in nursing positioning will facilitate normal postural development for very preterm infants. Long-term ventilation and chronic lung disease are associated with similar positional issues; neck retraction and truncal extension are features of airway-shortening manoeuvres. Irritability associated with these deformities produces a picture that may be attributed to neurological injury and handling or feeding children with fixed postural deformity is difficult and may lead to problems in maternal attachment.
These influences may all modify the trajectory of development. ) have described a group of ex-preterm infants with discrepancies between active and passive muscle tone that are most obvious in the extensor muscles of the trunk and may also be asymmetric ( ). This usually transient dystonia may be a feature of the developing ex-preterm infant independent of positional changes. de Groot and colleagues argue that these tonal abnormalities lead to delay in unsupported sitting and rotation towards the end of the first year. These have implications for transition between postures and also lead to delay in fine motor development ( ) and behavioural changes, through impairment of ability to optimise gross motor and hand function in an otherwise normal child because of truncal fixation.
Transient dystonia was described in preterm children by ) as infants with an excess of extensor hypertonicity in the trunk and legs, increased hip adductor tone and delayed supporting reactions. Such changes tend to resolve at around the child’s first birthday or over the second year and be initially diagnosed as CP. Other more recent studies have evaluated motor development over the first 18 months. ) studied a geographically based cohort of babies <2000 g birthweight. They categorised 29% as dystonic, 10% as hypotonic and 8% as suspected CP. The prevalence of dystonia was 35% in babies <1000 g, 35% for those of 1000–1499 g and 21% in babies 1500–1999 g. The peak prevalence of dystonia was at 7 months of age corrected for prematurity. Of children at 18 months of age with suspected CP, when examined at 7 months, five had been labelled as suspect CP, eight dystonia and one hypotonia. In another study the prevalence of dystonia was 36% of 260 VLBW infants ( ). In a further study, although truncal hyperextension had disappeared by 24–26 months of age, associated abnormalities of arm and hand function appeared to persist, perhaps as precursors of later motor abnormalities ( ).
Four studies have indicated long-term outcomes for dystonic infants. In ) original study they were found to have an excess of educational difficulties; in a study of 50 children there was a trend towards more neuromotor difficulties ( ); and in a further study dystonic infants had lower cognitive scores and higher disability grades at early school age ( ).
Cerebral palsy
The definition of CP has been updated to include wider aspects of function ( ):
Cerebral palsy describes a group of permanent disorders of the development of movement and posture, causing activity limitation, that are attributed to non-progressive disturbances that occurred in the developing fetal or infant brain. The motor disorders of cerebral palsy are often accompanied by disturbances of sensation, perception, cognition, communication, and behaviour, by epilepsy, and by secondary musculoskeletal problems.
William Little is credited with the observation that cerebral spasticity and paralysis were caused by ‘difficulties arising around the time of birth’, including preterm birth (and asphyxia), hence the epithet ‘Little’s disease’ applied to diplegia. The specific association between a broad category of bilateral spasticity, which he termed spastic diplegia, and prematurity was first formally described in a paper entitled ‘Zum Cerablen Diplegien’ published in 1896 by Sigmund Freud, during his work as a paediatric neurologist with Charcot in Paris ( ). Thus the increased risk of CP following preterm birth has been known for some time and, indeed, CP is often considered to be the most common and important disability following preterm birth. In fact this is not the case as developmental delay and cognitive impairments are by far the commonest disabilities. Nonetheless CP remains one of the most important outcomes and in the early 1980s spastic diplegia was used as a marker of the success of perinatal services ( ).
The risk of CP is inversely related to gestational age at birth and increased in the presence of fetal growth restriction, at least among moderately preterm children ( ), although the use of fetal growth standards instead of birthweight standards indicates a consistent relationship between prevalence of CP and size at birth across the gestational range ( Fig. 3.3 ) ( ). As many older studies are birthweight-based, the results must be treated with some caution because of the varying proportion of babies born after fetal growth restriction ( Ch. 10 ). Thus the evaluation of secular trends in CP must be related to either birthweight or gestation because of the close links. The evolution of birthweight-specific trends in the prevalence of CP over the time period of introduction of neonatal intensive care in Merseyside, UK, has been described ( ). As birthweight-specific survival increased there was an initial rise in the prevalence of CP within that birthweight band followed by a levelling out. This may relate to the increasing survival of children with CP who would otherwise have died, up to an equilibrium point.
Population trends in the prevalence of CP as a group or in birthweight-specific groupings are difficult to interpret, although towards the late 1990s the rate of CP appeared to be decreasing in Sweden ( ) and in the Surveillance for Cerebral Palsy in Europe (SCPE) collaboration of CP registers there was a fall in CP prevalence for babies 28–32 weeks from 1981 to 1995 but little change for babies <28 weeks ( Fig. 3.4 ) ( ). Furthermore, data from the 4 Child Register of CP in four UK counties seem to indicate a falling off in CP prevalence for births of less than 28 weeks of gestation through to 2001–2003, demonstrating the value of longitudinal data ( Fig. 3.5 ). These findings are consistent with low prevalence in three recent reports of extremely preterm children from Victoria, Australia (2005: 9.8% 22–27 weeks’ gestation: ), from Edmonton, Canada (2000–2003: 1.9% <27 weeks: ) and Cleveland, OH, USA (2000–02: 5% <1000 g birthweight: ).
The distribution of impairment in CP is determined by the pattern of brain injury. Diplegia, in which the impairment is more severe caudally, is considered to be the result of damage to the internal capsule as part of periventricular leukomalacia. This usually symmetrical lesion is commonly observed on magnetic resonance imaging (MRI) scans at follow-up even if no neonatal ultrasound evidence of injury was observed ( ). Diplegia is often associated with relatively mild disability and there is often less severe impairment in other domains compared with other patterns of CP. In the EPICure cohort of babies of 25 weeks’ gestational age or less, only 44% (12/27) of children with diplegia had severe disability and thus were likely to have persisting disability at later ages ( ).
Spastic hemiplegia, where the distribution of impairment is the reverse of diplegia and where there is usually significant asymmetry, results from lesions involving the cortex, such as cerebral venous infarction accompanying germinal matrix and intraventricular haemorrhage. Spastic quadriplegia, where there is four-limb involvement equally distributed in upper and lower limbs, occurs in variable proportions in reports. The distribution of disability in children with diplegia and hemiplegia is not always confined to the legs or to one side of the body and it is a subjective decision as to how to categorise the impairments. This accounts for many of the differences found between different observers. However, the impairment in quadriplegia is more extensive than in the other two and the disability is frequently severe (11/12 in the EPICure cohort). Other types of CP are found less frequently in very preterm populations.
In the discussions of classification of CP the above categories provide a sensible aetiologically relevant categorisation. However, over the years the classification has changed to reflect the prevailing fashion ( ). Most recently the SCPE group has published a classification and an algorithm, which should be used in future classifications based on three groupings ( Table 3.2 ).
Spastic cerebral palsy |
Characterised by at least two of:
|
Ataxic cerebral palsy |
Characterised by both of:
|
Dyskinesic cerebral palsy |
Characterised by both of:
Dyskinesic cerebral palsy may be either:
|
The timing of diagnosis in follow-up reports is important in determining the prevalence of CP; the predictive value of a label of CP increases as the child grows up. There is some consensus that the diagnosis of CP producing significant disability is usually accurate by the age of 2 years ( ), but that, before that, abnormal patterns of motor development and dystonia or developmental delay may confuse the unwary. In the National Collaborative Perinatal Project CP was overdiagnosed at 12 months compared with assessment at 7 years, such that half of the children with the diagnosis at 1 year were free of neurological signs 6 years later ( ). The most common types of CP to resolve were categorised as mild and were of the monoparetic, dyskinetic or diplegic types; resolution was more frequent in black infants. However a significant proportion of the children in whom signs had resolved (13% white and 25% blacks) had an IQ ≤70 at 7 years, emphasising the importance of careful early examination. Despite the recommendation to wait until 2 years, even at this age there is overdiagnosis ( ) and some less severe impairment may not be detectable until early school age ( ).
Careful sequential neonatal cerebral ultrasound scanning may identify the majority of children who go on to develop severe CP ( ) ( Ch. 40.5 ) and early sequential neuroassessment in infancy is mandatory if CP is to be identified and managed appropriately. The role of other specific investigations is less clear; for example, somatosensory evoked potentials ( ) and Prechtl’s assessment of general movements ( ) are good predictors of outcome but have not found a place in routine practice as they are time-consuming and require great skill to do.
Screening vision and hearing
Screening for retinopathy of prematurity (ROP) is addressed in Chapter 33 , but this cannot form the basis for screening for visual or ocular impairments. Squints and refractive errors are frequently found at follow-up in infancy and later childhood, thus each assessment must include adequate examination of visual activity and ocular movements. Refractive errors and subtle disorders of vision such as poor stereognosis and contrast sensitivity are found frequently at longer term follow-up ( ; ; ; ), and should be sought in assessments at school age. In a teenage follow-up of the East Midlands cohort of children who had been screened but not treated for ROP, ophthalmic morbidity was found in 67.8% of children born ≤1000 g, 51.6% of those 1001–1250 g, 44.3% of those 1251–1500 g and 49.5% of those 1501–1750 g compared with 19.5% of their comparison group ( ).
Sensorineural hearing loss is more prevalent in preterm populations, although the aetiology is still obscure and may well be multifactorial ( ). Found in 1–4% of VLBW children, this represents a 10-fold increase over unselected populations ( ; ; ; ). Because of this, most services have developed targeted neonatal screening policies in which prematurity (either as VLBW or birth ≤32 weeks) is included in the screened population alongside children with a family history of deafness, orofacial anomalies and perinatal central nervous system infection. More recently there has been a move towards universal neonatal hearing screening, predicated on the thesis that early intervention for children with severe or profound hearing loss may improve later language and speech development ( ). Most targeted screening involves automated or manual evoked responses (auditory or brainstem response (ABR)) that will identify children with significant hearing loss. Universal screening in the USA and most European countries involves testing with automated ABR technology, whereas currently in the UK a ‘two-step’ approach has been adopted using initial screening with otoacoustic emissions and automated ABR for failures. Universal screening has a sensitivity of between 80% and 90% and a specificity of above 90% ( ) and the median age of detection of severe hearing loss may be as low as 2 months, facilitating early intervention.
Developmental progress
Central to the definition of disability in preterm populations is an estimate of the child’s developmental level. Correction of age for prematurity is usually applied up to 2 years, but may be extended to 3 years for extremely preterm children where the correction still makes around 10% difference depending on gestational age ( Fig. 3.6 ). This correction is controversial, as some prefer to assess against chronological age-related norms, claiming that this more correctly identifies those with developmental problems, but as increasingly immature children survive this has fallen out of practice. Correcting for prematurity once the child is in the education system, where he or she is compared with peers, is generally deemed unnecessary.
Clinic assessment using one of the widely available developmental screening tools (e.g. Denver Developmental Screening Test or Schedule of Growing Skills) may indicate normal or very abnormal progress but the diagnostic utility and predictive value are not known; in particular, screening tests are used to identify children for more detailed assessment and are likely to overidentify children with developmental problems.
Most services now evaluate development with a formal assessment using one of the available test instruments. Any scale with which the assessor is familiar may be used in clinical practice but the commonest ones in current use are the Bayley Scales of Infant and Toddler Development (Bayley-III) and the Griffiths Scales (second edition). Both have had recent standardisations which enhance the value of the results but the Bayley-III was restructured and rescaled in a new manner and this has resulted in relatively high scores (+7 points compared with the second edition – the BSID-II), which makes scores difficult to interpret. Most observers have found that fewer children are identified as low scorers using these scales and recommend that all new preterm cohort studies recruit contemporary term-born comparison children ( ), which will add greatly to the logistics and cost of doing such studies. The predictive value of individual developmental score results into childhood is relatively weak, although preterm children appear to track better than normal term-born children. However, in population studies the predictive value of proportions in different outcome categories is much better, hence these scores are useful in categorising outcomes for research purposes. The Bayley-III scales have no values for very poorly performing children (scores < −3 sd ) and nominal scores must then be allocated to children who perform below the lower level of the test range; this must be acknowledged in reports as even allocating a score of 50 may produce very different results from allocating a score of −4 sd (40), as is often done.
Using the BSID-II, 30% of the EPICure cohort had scores <70 and 11% <55, representing −2 and −3 sd respectively. Scores were lower in boys but did not vary with gestational age or plurality ( ). Mean scores were MDI (Mental Development Index) 82 ( sd 15) and PDI (Psychomotor Development Index) 83 ( sd 16) for those tested but 9.5% of the children assessed failed to complete the BSID-II. Evaluating mean scores of various studies is fraught with difficulty, mainly because exclusions can seriously modify results. It is important to include the whole population in any assessment where possible or to describe why children fail to complete the test.
A formal developmental assessment is expensive and time-consuming for service or research purposes, particularly where a categorisation around one figure is required. Increasingly, 18–24-month assessments are being used as outcome measures for randomised trials and their inclusion adds very significantly to the cost of the project. We have recently validated an adapted parent report questionnaire ( ) against the BSID-II at 2 years in very preterm children ( ; ). This has high correlation ( r = 0.68) and good diagnostic utility (sensitivity 81%; specificity 81%) for a BSID-II cognitive score <70 and is likely to be as accurate as repeating the developmental assessment with another observer. This has proved a useful tool in the routine follow-up of preterm populations and has been used in multicentre trials, where follow-up is important to the evaluation of the safety and efficacy of an intervention but there are cost constraints. Although the potential cost of overreferrals may be a cause for concern for clinical use of such parental questionnaires, we have shown that very preterm infants with false-positive screens are at greater risk for neurodevelopmental and behavioural problems than those with true negative results and further assessment may be warranted ( ).
There are no particular patterns of developmental impairments associated with prematurity, rather as a group scores are globally depressed; this reflects the perhaps surprising lack of specific learning impairments at later ages. Behavioural difficulties are particularly difficult to evaluate in infancy but the knowledge that attention-deficit hyperactivity disorder (ADHD) is a persistent and constant finding in later studies encourages assessment in infancy. Sadly there are no well-respected measures which help to identify children at risk, although direct observation in the clinic does seem to identify a significant proportion of children with inattention and distractibility ( ).
Current follow-up practice
Our present practice in University College Hospital is to review children in the multidisciplinary neonatal follow-up clinic until 2 years corrected for prematurity, if they fall into one of three distinct high-risk groups, namely babies of 28 weeks’ gestation or less, babies with bronchopulmonary dysplasia and babies who have demonstrated neonatal encephalopathy. Review is by a neonatologist, respiratory nurse, dietician and speech and language therapist, neurologist and a developmental specialist. Children are seen approximately 3-monthly over the first year and then at 2 years. Annual review visits for those with bronchopulmonary dysplasia then are offered until 5 years. A formal neurological assessment ( ; ) is done at each visit together with a Bayley-III assessment. Video analysis of general movements ( ) is carried out at 12 weeks postterm. Babies with identified developmental or neurological issues are referred to their local neurodisability team for assessment and intervention.
It should be stressed that this is an environment where we have good community-based family support from health visitors, primary care physicians and community paediatric staff. There is also good support from and easy access to advice from dietetics and locality-based physiotherapy and other family support teams for children with special needs. In other health systems alternative arrangements may be necessary.
Predicting disability in preterm infants
The major determinants of later disability are in essence similar to those predicting perinatal death, often described as part of the continuum of reproductive casualty ( ). Perinatal factors increasing risk of disability include male sex ( ; ), gestational age and birthweight.
Using simple data such as these one can make broad statements about disability-free survival, which can help inform decisions around the time of birth ( Fig. 3.7 ) ( ). A more specific prognosis for serious disability may be derived from brain imaging using ultrasound, MRI or neurosensory assessments ( Ch. 40.5 ) and used to counsel parents on an individual risk basis.