Statistics from Altmetric.com
The dilemma in premature babies
There are essentially three ways in which blindness from retinopathy of prematurity (ROP) can be prevented, including elimination of premature birth, changes in neonatal care, and improved detection and treatment of established sight threatening retinopathy. It is on the latter that the attention and efforts of the ophthalmic community have been focused, leaving the first two to the paediatricians, perinatologists, neonatologists, nurses, and others who care for these tiny babies. What Darlow et al have done in the paper in this issue of the BJO (p 1592) is direct our attention to the second possibility—that is, neonatal care may be able to be altered to decrease the incidence of potentially blinding disease.
Darlow et al document a remarkable variability in the prevalence of severe ROP among the nurseries in the neonatal intensive care units in the Australian and New Zealand Neonatal Network. Such variability is generally expected when morbidity or mortality results are compared across centres, even in an essentially population based study that this report represents (the four tertiary paediatric care referral centres are excluded from the analysis reported by Darlow et al). The variability in such reports can usually be explained by a combination of several factors, including case mix, assessment of the primary outcome measure, viability considerations by neonatologists at the various centres, and variations in clinical practice. For example, the percentage of high risk deliveries may differ dramatically among nurseries, leading to more at-risk infants in one nursery than another. Smaller nurseries may also have greater variation in the incidence of severe ROP because of sampling variability. In addition, standardisation of outcome measures in multicentre studies represents a huge challenge and can seriously impact the perceived prevalence of severe disease.
Thanks to a large enough sample and novel analytical techniques, Darlow et al have been able to take into account many of the potential confounding variables using logistic regression models. They find that variability across nurseries in the rates of severe ROP requiring treatment is not explained by case mix, sampling, or outcome assessment, but rather is probably the result of clinical practice in the individual nurseries. In their analysis, they wisely chose to highlight the improvement in rates of serious ROP that could be reached if clinical practice was altered to achieve the “best practice” rates achieved in the top 20th percentile of the nurseries, rather than focusing on the “problem” nurseries. Using this systemwide approach to changes in clinical care would lower the rate of severe ROP to 5.9% in the top 20% of nurseries from the overall rate of 9.6%. They estimate that 79 cases of severe ROP would have been prevented.
A major problem that arises when practice patterns are changed is that interventions aimed at decreasing the rate of one serious problem may increase the rate of another, equally serious problem. A very clearcut example of this was provided in the 1950s when oxygen was found to be the root cause of blinding ROP and its delivery markedly restricted.1 This led to an increase in mortality2 and spastic diplegia among premature babies,3 and it was estimated that 16 infants died from curtailed oxygenation for each case of blindness prevented.4 This observation has particular clinical relevance when evaluating possible changes in clinical care of premature babies, a population that has had an increased survival rate as a result of advances in perinatal and neonatal medicine over the past 20 years in countries with high levels of human development5; it is also a population in which we have only an evolving understanding of the developmental disabilities that are detected later in life.6–8 Since clinical care of the premature baby is directed broadly at prevention of illness, undernutrition, and infection during a time of rapid brain growth and development, it is not hard to postulate that systemwide changes in practice may differentially impact on the occurrence of cerebral palsy, visual and hearing impairments, behavioural and social problems, and learning difficulties.9 When considering changes in practice patterns, clearly the possible effect of any change must concentrate not only on survival, but also on lifelong disabilities that may occur in this vulnerable population.
Identifying and implementing practice pattern changes may decrease the prevalence of severe ROP and, with appropriate treatment, of blindness. However, most children being blinded from ROP do not live in countries with high levels of human development, but rather they are born in countries with middle levels of human development, where neonatal services are rapidly expanding and where limited resources may severely impact delivery of the highest levels of neonatal care.9 Thus, the alterations in clinical practice that may arise from examining the “best practice” nurseries in Australia and New Zealand may have limited generalisability to the nurseries that care for the largest proportion of babies at risk of blindness because of severe ROP.
Note in Proof
The dilemma in premature babies