The use of machine learning in paediatric nutrition.

PURPOSE OF REVIEW: In recent years, there has been a burgeoning interest in using machine learning methods. This has been accompanied by an expansion in the availability and ease of use of machine learning tools and an increase in the number of large, complex datasets which are suited to machine learning approaches. This review summarizes recent work in the field and sets expectations for its impact in the future. RECENT FINDINGS: Much work has focused on establishing good practices and ethical frameworks to guide the use of machine learning in research. Machine learning has an established role in identifying features in 'omics' research and is emerging as a tool to generate predictive models to identify people at risk of disease and patients at risk of complications. They have been used to identify risks for malnutrition and obesity. Machine learning techniques have also been used to develop smartphone apps to track behaviour and provide healthcare advice. SUMMARY: Machine learning techniques are reaching maturity and their impact on observational data analysis and behaviour change will come to fruition in the next 5 years. A set of standards and best practices are emerging and should be implemented by researchers and publishers.

Longitudinal pilot study of oxygen saturation indices in healthy preterm infants.

BACKGROUND: This study aimed to determine patterns of nocturnal pulse oximetry indices in moderate to late preterm infants, and investigate the relationship between oxygen desaturations, the apnoea hypopnoea index, and both corrected gestational and postnatal age. METHODS: 21 healthy infants born at 32 + 0 - 36 + 6 weeks gestation underwent serial nocturnal pulse oximetry studies and respiratory polygraphy studies until 40 weeks corrected gestational age (CGA). The main outcome measures were number of >3% oxygen desaturations/hour (ODI3), mean oxygen saturations, and number of apnoeas and hypopnoeas/hour. RESULTS: Median ODI3 increased between weeks 1 and 3 from 49.9 to 85.4/hour (p = 0.017). Mean oxygen saturations reached a corresponding nadir of 96.0% in week 3, then increased to 96.8% in week 6 (p = 0.019). Mixed effects modelling demonstrated that ODI3 and mean saturations were influenced by postnatal age but not CGA (p < 0.05). Desaturations frequently occurred without an apnoea or hypopnoea. CONCLUSION: ODI3 rises then falls during the first 8 weeks of life in moderate to late preterm infants, independently of CGA. These interesting preliminary results highlight the importance of further serial data collection to generate age-specific normal ranges, and develop a better understanding of respiratory control in preterm infants. IMPACT: The frequency of >3% oxygen desaturations (ODI3) in healthy moderate to late preterm infants rises then falls after birth, peaking in postnatal week 3. There is a corresponding nadir in mean saturations. There were significant non-linear relationships between ODI3/mean saturations and postnatal age, but not corrected gestational age. The majority of brief oxygen desaturations occurred without an apnoea or hypopnoea. Normal ranges for oxygen saturation indices are not known in this population. These results demonstrate the need for further serial data collection to generate age-specific normal ranges and inform oxygen prescribing guidelines.

Differential associations between body composition indices and neurodevelopment during early life in term-born infants: findings from the Pakistan cohort: Multi-Center Body Composition Reference Study.

OBJECTIVE: We examined associations between fat free mass (FFM) and fat mass (FM) accretion during the first 1000 days of life and neurodevelopment in term-born, low-risk infants from Karachi, Pakistan. DESIGN: Prospective, observational study nested within the larger Multi-Center Body Composition Reference Study. FFM, FM, and fat% were estimated using measured deuterium dilution method. Neurodevelopmental outcomes were assessed at 24 months on the INTER-NDA (INTERGROWTH-21st Project Neurodevelopment Assessment) (n = 132). RESULTS: Children with gross motor delays had significantly lower FFM at 18 months (8.01 ± 0.97 kg vs. 7.55 ± 0.20 kg). Children with positive and negative behavior problems had significantly higher fat% at 24 months (20.62 ± 4.30% vs. 18.23 ± 5.46%) and 20.89 ± 4.24% vs. 18.54 ± 5.38%). No associations remained significant after adjusting for covariates. Trajectory modeling showed that between 12 and 18 months, negative behavior scores changed by 13.8 points for every standard deviation change in fat accretion. CONCLUSIONS: Our findings highlight the importance of balancing neurodevelopment and metabolic risk when designing nutritional interventions for young children.

Optimising growth in very preterm infants: reviewing the evidence.

Infants born before 32 weeks' postmenstrual age are at a high risk of growth failure. International guidelines have long recommended that they match the growth of an equivalent fetus, despite the challenges posed by ex utero life and comorbidities of prematurity. Several groups have recently questioned the necessity or desirability of this target, shifting attention to aiming for growth which optimises important long-term outcomes. Specifically, recent research has identified the neurodevelopmental benefits of enhanced growth during the neonatal period, but work in term infant suggests that rapid growth may promote the metabolic syndrome in later life. In this context, defining a pattern of growth which optimises outcomes is complex, controversial and contested. Even if an optimal pattern of growth can be defined, determining the nutritional requirements to achieve such growth is not straightforward, and investigations into the nutritional needs of the very preterm infant continue. Furthermore, each infant has individual nutritional needs and may encounter a number of barriers to achieving good nutrition. This article offers a narrative review of recent evidence for the competing definitions of optimal growth in this cohort. It examines recent advances in the determination of macronutrient and micronutrient intake targets along with common barriers to achieving good nutrition and growth. Finally, key implications for clinical practice are set out and a recommendation for structured multidisciplinary management of nutrition and growth is illustrated.

Changes in the growth of very preterm infants in England 2006-2018.

OBJECTIVE: To compare weight gain from birth to term equivalent age in very preterm infants in England born during two eras (2006-2011 and 2014-2018); to assess demographic and care factors influencing weight gain. METHODS: Data for infants born before 32 weeks of gestation during 2014-2018 in England were obtained (29 687 infants). Weight gain modelled using SuperImposition by Translation And Rotation (SITAR), with infants grouped by gestational week. A cohort from 2006 to 2011 was used for comparison (3288 infants). Multiple linear regression was used to assess factors influencing change in weight SD score from birth to 36 weeks postmenstrual age. RESULTS: Weight gain velocity (termed 'intensity' in SITAR models) was greater in the more recent cohort for all gestation groups born before 30 weeks of gestation. After adjustment for gestation, birth weight and other perinatal factors, care elements associated with faster weight gain included delivery in a level 3 unit (0.09 SD less weight gain deficit, 95% CI: 0.07 to 0.10) and parenteral nutrition initiation during the first day of life (0.08 SD, 95% CI: 0.06 to 0.10). Factors associated with slower weight gain included early ventilation (-0.07 SD, 95% CI: -0.08 to -0.05) and less deprived neighbourhood (-0.012 SD per Index of Multiple Deprivation decile, 95% CI: -0.015 to -0.009). CONCLUSIONS: Weight gain for extremely preterm infants was faster during 2014-2018 than during 2006-2011. Early initiation of parenteral nutrition and birth in a level 3 unit may contribute to faster weight gain.

Using machine learning to impact on long-term clinical care: principles, challenges, and practicalities.

The rise of machine learning in healthcare has significant implications for paediatrics. Long-term conditions with significant disease heterogeneity comprise large portions of the routine work performed by paediatricians. Improving outcomes through discovery of disease and treatment prediction models, alongside novel subgroup clustering of patients, are some of the areas in which machine learning holds significant promise. While artificial intelligence has percolated into routine use in our day to day lives through advertising algorithms, song or movie selections and sifting of spam emails, the ability of machine learning to utilise highly complex and dimensional data has not yet reached its full potential in healthcare. In this review article, we discuss some of the foundations of machine learning, including some of the basic algorithms. We emphasise the importance of correct utilisation of machine learning, including adequate data preparation and external validation. Using nutrition in preterm infants and paediatric inflammatory bowel disease as examples, we discuss the evidence and potential utility of machine learning in paediatrics. Finally, we review some of the future applications, alongside challenges and ethical considerations related to application of artificial intelligence. IMPACT: Machine learning is a widely used term; however, understanding of the process and application to healthcare is lacking. This article uses clinical examples to explore complex machine learning terms and algorithms. We discuss limitations and potential future applications within paediatrics and neonatal medicine.

A scoping review: urinary markers of metabolic maturation in infants with CHD and the relationship to growth.

BACKGROUND: Growth failure in infants born with CHD is a persistent problem, even in those provided with adequate nutrition. OBJECTIVE: To summarise the published data describing the change in urinary metabolites during metabolic maturation in infants with CHD and identify pathways amenable to therapeutic intervention. DESIGN: Scoping review. ELIGIBILITY CRITERIA: Studies using qualitative or quantitative methods to describe urinary metabolites pre- and post-cardiac surgery and the relationship with growth in infants with CHD. SOURCES OF EVIDENCE: NICE Healthcare Databases website was used as a tool for multiple searches. RESULTS: 347 records were identified, of which 37 were duplicates. Following the removal of duplicate records, 310 record abstracts and titles were screened for inclusion. The full texts of eight articles were reviewed for eligibility, of which only two related to infants with CHD. The studies included in the scoping review described urinary metabolites in 42 infants. A content analysis identified two overarching themes of metabolic variation predictive of neurodevelopmental abnormalities associated with anaerobic metabolism and metabolic signature associated with the impact on gut microbiota, inflammation, energy, and lipid digestion. CONCLUSION: The results of this scoping review suggest that there are considerable gaps in our knowledge relating to metabolic maturation of infants with CHD, especially with respect to growth. Surgery is a key early life feature for CHD infants and has an impact on the developing biochemical phenotype with implications for metabolic pathways involved in immunomodulation, energy, gut microbial, and lipid metabolism. These early life fingerprints may predict those individuals at risk for neurodevelopmental abnormalities.

A Scoping Review: Urinary Markers of Metabolic Maturation in Preterm Infants and Future Interventions to Improve Growth.

Background: Growth failure in infants born preterm is a significant issue, increasing the risk of poorer neurodevelopmental outcomes and metabolic syndrome later in life. During the first 1000 days of life biological systems mature rapidly involving developmental programming, cellular senescence, and metabolic maturation, regulating normal growth and development. However, little is known about metabolic maturation in infants born preterm and the relationship with growth. Objective: To examine the available evidence on urinary markers of metabolic maturation and their relationship with growth in infants born preterm. Eligibility criteria: Studies including in this scoping review using qualitative or quantitative methods to describe urinary markers of metabolic maturation and the relationship with growth in infants born preterm. Results: After a screening process 15 titles were included in this review, from 1998-2021 drawing from China (n = 1), Italy (n = 3), Germany (n = 3), Greece (n = 1), Japan (n = 2), Norway (n = 1), Portugal (n = 1), Spain (n = 2) and USA (n = 1). The included studies examined urinary metabolites in 1131 infants. A content analysis identified 4 overarching themes relating to; (i) metabolic maturation relative to gestational age, (ii) metabolic signature and changes in urinary metabolites over time, (iii) nutrition and (iv) growth. Conclusion: The results of this scoping review suggest there are considerable gaps in our knowledge relating to factors associated with metabolic instability, what constitutes normal maturation of preterm infants, and how the development of reference phenome age z scores for metabolites of interest could improve nutritional and growth outcomes.

A multimodal quality improvement approach to promote normothermia in very preterm infants.

AIM: To achieve the National Neonatal Audit Programme (NNAP) standard of 90% normothermia among preterm infants born under 30 weeks of gestation. METHODS: Project SHIP (Stopping Hypothermia In Premmies) was a quality improvement programme to improve admission normothermia. Phase 1 of the project implemented low-fidelity simulations during 2011-2016. In Phase 2 (2017), a multimodal approach to quality improvement was used, including in situ simulations, videos of simulated scenarios, an allocated team member for thermal care, a clear protocol for thermal care, a coordinating 'lollipop man' role and monthly performance feedback. Additionally, continuous temperature monitoring using servo-control during stabilisation was introduced during Phase 2. Phase 3 (2018-2019) focused on embedding practice and maintaining performance. RESULTS: Phase 1 initiatives resulted in improvement of normothermia rates from 58% to 75%. However, the results plateaued. During Phase 2, the hypothermia rate fell from 16% to 3%. During Phase 3, this improvement in the hypothermia rate was sustained, achieving the standard of 90% normothermia in 2018 and falling just short in 2019 due to an increased hyperthermia rate. CONCLUSION: A multimodal quality improvement approach achieved sustained improvement in normothermia. Continuous temperature monitoring during stabilisation allows resuscitating teams to plan interventions to treat hypothermia and hyperthermia.

Total body water in full-term and preterm newborns: systematic review and meta-analysis.

BACKGROUND: Total body water (TBW) is one component of fat-free mass and changes in TBW are influenced by fluid shifts (especially during transition to postnatal life), electrolyte balance and nutritional status. Normal values for term-born neonates and preterm infants at birth have not been defined in large cohorts, limiting investigation into its monitoring and use in clinical practice. OBJECTIVE: To systematically review the evidence base for percentage of TBW in term-born infants, quantify the effect of prematurity on TBW at birth, and describe normal progression of TBW over time in preterm infants. METHODS: Systematic review of Medline, Web of Science Core Collection and EBSCO-CINAHL (January 1946 to January 2020). Included articles used dilutional methods to assess TBW. RESULTS: Searches identified 2349 articles of which 22 included data suitable for analysis. Mean TBW in term-born newborns was 73.8% (95% CI 72.47% to 75.06%, 15 studies, 433 infants). Meta-regression showed that TBW was higher in preterm infants (up to 90% at 26 weeks gestation, dropping to 75% at 36 weeks corrected gestation) and was negatively correlated with gestation at birth, falling 1.44% per week (95% CI 0.63% to 2.24%, 9 studies, 179 infants). Analysis of TBW over time during the ex utero growth of preterm infants was not possible due to paucity of data. CONCLUSION: This review defines the normal TBW percentage in term-born infants and confirms and quantifies previous findings that preterm infants have a higher TBW percentage.

Growth failure is rare in a contemporary cohort of paediatric inflammatory bowel disease patients.

AIM: We assessed growth in a paediatric inflammatory bowel disease (PIBD) cohort. METHODS: Paediatric inflammatory bowel disease patients were eligible if they were diagnosed at Southampton Children's Hospital from 2011 to 2018. Weight and height standard deviation scores (SDS) were retrieved. Mean SDS values, SDS change and anti-TNF status were analysed at diagnosis and during follow-up. RESULTS: Four hundred and ninety patients were included, 313 with Crohn's disease (CD). CD patients presented with mean height SDS -0.13, -0.1 at 1-year, -0.11 at 2-years and -0.03 at 5 years, reflecting preserved linear growth. There was no significant height-SDS change from diagnosis to 5-year follow-up, +0.12, 95%-CI: 0.48 to -0.24. Mean weight-SDS at diagnosis was -0.39, driven by CD patients (-0.65). Mean weight-SDS approached 0 after 1 year and remained at the 50th centile throughout follow-up. Growth in ulcerative colitis was maintained. In multivariable regression males had worse height growth from diagnosis to transition (P = .036). Anti-TNF treatment (P = .013) and surgical resection (P = .005) were also associated with poorer linear growth. Patients treated with anti-TNF therapy had lower height-SDS compared to those never treated with anti-TNF at 1 year (-0.2 vs -0.01, P = .22), 2-years (-0.27 vs -0.01, P = .07) and 5 years (-0.21 vs 0.25, P = .051). CONCLUSION: Height was generally maintained in Crohn's disease, and impaired linear growth was rare in this cohort.

Generating longitudinal growth charts from preterm infants fed to current recommendations.

OBJECTIVE: To use repeated measurements of weight, length and head circumference to generate growth centile charts reflecting real-world growth of a population of very preterm infants with a well-described nutritional intake close to current recommendations. DESIGN: Infants born before 30 weeks gestational age (GA) were recruited. Infants received nutrition according to an integrated care pathway, with nutrient intake recorded daily, weight recorded twice-weekly and length and head circumference weekly. The LMS method was used to construct growth centile charts between 24 and 36 weeks corrected GA for each parameter. SETTING: A single tertiary neonatal unit in England. PATIENTS: 212 infants (124 male) (median GA at birth: 27.3 weeks, median birth weight: 900 g). RESULTS: Median daily energy, protein, carbohydrate and fat intake were within 3% of published recommendations. The total number of measurements recorded was 5944 (3431 for weight, 1227 for length and 1286 for head circumference). Centile charts were formed for each parameter. Data for male and female infants demonstrated similar patterns of growth and were pooled for LMS analysis. A web application was created and published (bit.ly/sotongrowth) to allow infants to be plotted on these charts with changes in SD score of measurements reported and graphically illustrated. CONCLUSIONS: These charts reflect growth in a real-world cohort of preterm infants whose nutrient intakes are close to current recommendations. This work demonstrates the feasibility of forming growth charts from serial measurements of growing preterm infants fed according to current recommendations which will aid clinicians in setting a benchmark for achievable early growth.