Diagnosis Through Whole Genome Sequencing and Care Utilization in Children With Severe Illness

Joao M L Dias; Ravi P More; Duncan Butler; Julian Brown; Courtney E French; Helen Dolling; F Lucy Raymond; David H Rowitch; Catherine E Aiken

doi:10.1001/jamanetworkopen.2026.34372

. 2026 Sep 17;9(9):e2634372. doi: 10.1001/jamanetworkopen.2026.34372

Diagnosis Through Whole Genome Sequencing and Care Utilization in Children With Severe Illness

Joao M L Dias ¹, Ravi P More ¹, Duncan Butler ², Julian Brown ³, Courtney E French ⁴, Helen Dolling ^1,⁵, F Lucy Raymond ⁶, David H Rowitch ^1,^7,^✉, Catherine E Aiken ^8,^✉

¹Department of Paediatrics, University of Cambridge, School of Clinical Medicine, Cambridge, United Kingdom

²Prescribing Services Ltd, Norfolk, United Kingdom

³Litcham Health Centre, Norfolk, United Kingdom

⁴Children’s Rare Disease Collaborative, Boston Children’s Hospital, Boston, Massachusetts

⁵Department of Psychology, University of Cambridge, Cambridge, United Kingdom

⁶Department of Medical Genetics, University of Cambridge, School of Clinical Medicine, Cambridge, United Kingdom

⁷Departments of Pediatrics, Cedars-Sinai Guerin Children’s, Los Angeles, California

⁸Department of Obstetrics and Gynaecology, University of Cambridge, School of Clinical Medicine, Cambridge, United Kingdom

Accepted for Publication: July 24, 2026.

Published: September 17, 2026. doi:10.1001/jamanetworkopen.2026.34372

^✉

Corresponding Author: David H. Rowitch, MD, PhD, Department of Paediatrics (drh25@cam.ac.uk), and Catherine E Aiken, MB, BCh, PhD, Department of Obstetrics and Gynaecology (cema2@medschl.cam.ac.uk), University of Cambridge, School of Clinical Medicine, Cambridge CB2 0SW, United Kingdom.

Author Contributions: Prof Rowitch and Dr Aiken had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Raymond, Rowitch, Aiken.

Acquisition, analysis, or interpretation of data: Dias, More, Butler, Brown, French, Dolling, Rowitch, Aiken.

Drafting of the manuscript: Dias, More, Dolling, Rowitch, Aiken.

Critical review of the manuscript for important intellectual content: Dias, Butler, Brown, French, Dolling, Raymond, Rowitch, Aiken.

Statistical analysis: Dias, More, Brown.

Obtained funding: Raymond, Rowitch.

Administrative, technical, or material support: Dias, Butler, Dolling, Aiken.

Supervision: Rowitch, Aiken.

Conflict of Interest Disclosures: Mr Butler reported receiving grants to the institution from Prescribing Services, not to the author personally, during the conduct of the study. Dr Brown reported receiving grants to the institution from Prescribing Services, not to the author personally, during the conduct of the study. Dr Dolling reported receiving some funding from Illumina during the conduct of the study. Dr Aiken reported receiving grants from the National Institute for Health and Care Research Cambridge Biomedical Research Centre (NIHR Cambridge BRC) during the conduct of the study. No other disclosures were reported.

Funding/Support: Support for this research was funded by Rosetrees Trust and the Isaac Newton Trust (Prof Rowitch), grant GN2788 from Action Medical Research (Dr Aiken), and Wellcome Trust Discovery Award 312471/Z/24/Z and grant NIHR203312 from the NIHR Cambridge BRC (Prof Rowitch and Dr Aiken).

Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Data Sharing Statement: See Supplement 3.

Additional Contributions: We are grateful for the support of the NIHR BioResource for Translational Research in Common and Rare Diseases and in helping deliver this research study.

^✉

Corresponding author.

PMCID: PMC13586901 PMID: 42752907

Key Points

Question

Is a genetic diagnosis through whole genome sequencing associated with long-term health care utilization and cost savings in children and young people with severe illness?

Findings

In this cohort study of comprehensively linked community medical records of 270 children and young people who received whole genome sequencing for rare genetic conditions, significantly higher health care utilization, including more hospitalizations and targeted prescriptions, were found in those who received a diagnosis.

Meaning

The findings of this study suggest that a genetic diagnosis through whole genome sequencing is associated with sustained, high-intensity long-term health care utilization, which may facilitate a more precise and tailored alignment of clinical resources to the patient’s individual needs.

This cohort study assesses the association of a genetic diagnosis through whole genome sequencing with long-term health care utilization metrics among children and young people with severe illness.

Abstract

Importance

Whole genome sequencing (WGS) is increasingly used to diagnose children with severe illness, yet the long-term association of a genetic diagnosis with health care utilization and resource allocation remains poorly understood.

Objective

To assess the association of a genetic diagnosis through WGS with long-term health care utilization metrics in children and young people (hereafter children) with severe illness.

Design, Setting, and Participants

This multicenter retrospective cohort study of children aged 0 to 18 years who were severely ill who underwent WGS used data from The Next Generation of Children Project (from December 2016 to August 2020) with medical record linkage and analysis of primary care medical records conducted between January 2022 and June 2024. The primary care and hospital medical records were linked using the UK National Institute for Health and Care Research Rare Disease BioResource, Cambridge.

Exposures

Receipt of a genetic diagnosis compared with those who remained undiagnosed following WGS.

Main Outcomes and Measures

The main outcome was a comparison of 36 health care utilization parameters, including hospitalizations, primary care prescriptions, and diagnostic tests.

Results

Among the 270 children analyzed (mean [SD] age 8.65 [6.58-9.46] years; 149 males [55.2%]), those receiving a genetic diagnosis (87 [32.2%]) exhibited significantly higher overall health care utilization compared with undiagnosed peers (183 [67.8%]). This included an increase in median (IQR) hospital admissions (37 [18-66] vs 22 [10-35]) and more primary and secondary care outpatient visits each year (14 [7-26] vs 8 [4-13]), particularly for neurodevelopmental (annual treatment costs: £1280 [£507-£2529] vs £130 [£21-£354]) and seizure-related (annual treatment costs: £1277 [£320-£2271] vs £60 [£15-£143]) conditions. Children with genetic diagnoses received a median (IQR) higher volume of neurological (48 [22-88] vs 0) and gastrointestinal (5 [1-18] vs 0 [0-2])prescriptions. Median (IQR) differences specifically in neurodevelopmental (neurological prescriptions: 42 [22-64] vs 0 [0-3]) and pediatric intensive care unit (total cost prescriptions: £1045 [£244-£1914] vs £38 [£14-£220]) settings were observed. While a genetic diagnosis was associated with sustained and intensive health care utilization during the study period, it was also associated with a shift toward targeted, condition-specific medical care.

Conclusions and Relevance

In this cohort study, a WGS diagnosis was associated with the integration of specialist care and the alignment of health care resources to support specific needs of children with complex disorders. These findings suggest that while longitudinal health care utilization remains intensive following a genetic diagnosis, identifying these conditions is important for accurately mapping and managing the downstream clinical resource requirements of this population.

Introduction

Children admitted to a neonatal intensive care unit (NICU) or a pediatric intensive care unit (PICU) often present with severe, life-threatening conditions affecting survival and long-term neurodevelopment.^1,2,3 Genetic disorders are recognized as leading contributors to morbidity and mortality, affecting 10% to 30% of children admitted to a NICU or a PICU.^{4,5,6,7,8,9,10,11} Whole genome sequencing (WGS) is thus increasingly common in clinical practice, particularly for suspected rare or monogenic diseases.^{6,12,13,14,15,16}

Prior health economic analyses show that WGS reduces costs in the PICU, largely due to reducing acute hospital stays and avoiding invasive procedures.^{2,3,4,8,17,18,19,20} Initiatives such as Project Baby Bear in California report short-term net health care savings of more than $14 000 per infant.²¹ Unlike other trials (eg, GEMINI,²² NICUSeq²³) or counterfactual studies (eg, Project Baby Bear,²¹ Project Baby Deer²⁴) that evaluate WGS as a clinical intervention, our observational analysis compares long-term health care utilization in children and young people (hereafter children) with an identifiable genetic diagnosis vs those without the diagnosis. We mapped baseline differences in clinical natural history and resource demands between children with monogenic diagnoses vs those with multifactorial or environmental causes for admission. In contrast to studies of short-term financial benefits, the long-term implications of a genomic diagnosis for health care utilization remain underexplored, particularly in the UK National Health Service (NHS).^25,26 We address this knowledge gap by linking pediatric WGS results to clinical outcome data provided by NHS Prescribing Services Ltd (Equality of Care Led Insights for Patient Safety & Engagement [ECLIPSE] Live), which comprises linked primary and secondary care data for more than 25 million patients.²⁷

The Next Generation of Children Project (NGC), conducted in Cambridge (from December 2016 to August 2020), was an early demonstration of utility of WGS in the intensive care setting, with more than 90% of clinicians reporting improved confidence in patient management and family communication following testing.²⁸ Initial evaluations within the NGC demonstrated an overall molecular diagnostic yield ranging from 21% to 45% via WGS.^12,28 The NHS has subsequently provided rapid genome sequencing (R14 service) for children with acute illness who meet testing criteria,^29,30 which has been shown to impact patient management and/or family reproductive counseling in nearly all diagnosed individuals.^25,31

We incorporated detailed clinical phenotypes and NHS primary care medical records alongside genomic findings from the NGC. While we report accumulated costs, we conceptualized these as a standardized metric to quantify differences in long-term utilization rather than a formal health-economic or cost-effectiveness evaluation. Rather than assessing whether a genetic diagnosis was associated with health care utilization, our objective was to map and compare longitudinal clinical utilization patterns between children who were diagnosed and undiagnosed. We aimed to delineate baseline differences in clinical natural history and resource demands between children with and without identifiable genetic etiologies. Second, we evaluated the possibility that health care utilization and other clinical characteristics might be an independent means to identify children with a higher probability of a positive genetic result from WGS.

Methods

Research Governance and Ethics

The NGC study was approved by the Cambridge South Research Ethics Committee. Written informed consent was obtained from parents for diagnostic WGS and subsequent data linkage to medical records via the UK National Institute for Health and Care Research Rare Disease BioResource. The specific linkage to primary care data was approved by the NIHR BioResource Data Access Committee. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cohort studies.

Study Design and Participant Selection

The NGC study initially enrolled children from December 2016 to August 2020 across NICU and PICU, pediatric neurology, and genetics clinics.^12,28 Enrollment required a high likelihood of an underlying monogenic condition based on clinical assessment.³² Included children presented with congenital anomalies, neurological symptoms, suspected metabolic disease, extreme intrauterine growth restriction, or unexplained critical illness. Conversely, WGS was not indicated for presentations, likely explained by nongenetic etiologies.

Demographic and Clinical Settings Data

Of the NGC participants who provided full consent for medical record linkage, the final analytical cohort comprised successfully linked survivors with complete primary identifier data for deterministic linkage, resulting in 0 missing demographic variables for the analytic cohort. Participants’ ethnicities included African, East Asian, European, Finnish-European, South Asian, and other and were ascertained by parental report. Ethnicity was collected in the study to characterize the study population and assess potential differences in representation. Due to historical consent changes, raw data for unlinked individuals were omitted, and only final statistical comparisons were reported. To address potential selection and survivor biases, we evaluated cohort representativeness by comparing the baseline demographic and clinical settings of the final analytical cohort against the broader NGC and fully consented individuals. Further details regarding selection bias, cohort representativeness, and specific methodologies are included in the eMethods in Supplement 2.

Clinical Phenotype Data

Data were obtained from electronic medical records via the NIHR BioResource. Unique Human Phenotype Ontology (HPO) terms (n = 1108) were identified, updated, and standardized via the Monarch Initiative.³³ For targeted clinical analysis, the 3 most prevalent HPO terms across the cohort (hypotonia, seizures, and neurodevelopmental and developmental delay) were selected for individual evaluation. The standardized terms were mapped to level 3 parent terms, yielding 23 unique organ-level terms for uniform phenotypic representation for consistent feature comparison.

DNA Variant Classification

Genomic data processing was performed on genetic variants. Variants were classified into pathogenic, likely pathogenic, and uncertain significance, using diagnostic criteria described previously.²⁸ Individuals with any identified pathogenic or likely pathogenic variant or variants of uncertain significance were classified as genetically diagnosed, whereas remaining individuals were designated as genetically undiagnosed.

Primary Care Records

Primary care records were accessed via the ECLIPSE Live platform.²⁷ Individual general practices, which control primary care data in the UK, permitted access records for consenting participants. Deterministic linkage was performed within a secure data environment using pseudonymized patient identifiers, specifically NHS number, date of birth, and biological sex. The resulting longitudinal dataset captured health care utilization and associated costs spanning from January 2022 to June 2024. Costs are reported in pounds sterling (to convert to US dollars, multiply by 1.36). Methodologic details for health care cost estimation are provided in the eMethods in Supplement 2. Raw data were classified into 4 broad operational categories: hospital visits and admissions, prescriptions, pathology, and conditions (eTable 1 in Supplement 1).

The primary care dataset was processed into 36 distinct features. Highly sparse features within the pathology, prescription, and condition categories were aggregated into composite groups to mitigate matrix sparsity and optimize subsequent evaluation. A comprehensive data dictionary, with explicit feature engineering and derivation methods, is presented (eTable 2 in Supplement 1 and eMethods in Supplement 2).

Statistical Analysis

Multiplicity Control

Group comparisons were performed using the Wilcoxon rank sum test. Because utilization data contained a high frequency of tied values, significance was calculated using an asymptotic normal approximation with a standard continuity correction. Effect sizes were quantified as Hodges-Lehmann median differences (95% CIs). A total of 288 statistical tests were conducted, encompassing 32 outcome features analyzed across 9 predefined clinical subgroups. Subgroups failing to meet a minimum sample size (n = 7 individuals or arm) were excluded. To strictly control the family-wise error rate, we adopted a conservative, fixed significance threshold of 2-sided P < .001. We analyzed data using R, version 4.5.0 (R Project for Statistical Computing).

Predictive Model Development and Feature Impact Analysis

A binary classification framework was developed to predict genetic diagnostic outcomes by integrating demographic, clinical phenotype, and health care utilization features into a unified feature matrix (eTable 3 in Supplement 1). Clinical phenotypes were represented as a binary matrix using the 23 organ-level HPO terms derived during data processing. Categorical features were 1-hot encoded, while numeric features were discretized into 4 quantile-based intervals prior to 1-hot encoding. Histograms of demographic and primary care features, stratified by diagnostic group, with quantile-based bins were constructed (Figure 1 and eFigure 1 in Supplement 2).

Four-panel mirrored bar charts of patient counts by diagnostic status across four measures. Four panels arranged in a two by two grid, labeled A, B, C, and D in small boxed letters near the upper left of each panel. Each panel contains horizontal bars mirrored around a central vertical line at x equals 0, with orange bars extending left for Diagnosed and light blue bars extending right for Undiagnosed. A legend centered between the top and bottom rows labels Diagnosed in orange and Undiagnosed in light blue. Panel A title Hospital admissions. Horizontal axis labeled Patient count with tick labels 40, 20, 0, 20, 40, 60. Vertical axis labeled Units with four interval categories from top to bottom: 48 to 533, 24 to less than 48, 11 to less than 24, and 1 to less than 11. Bars vary by bin, with the largest blue bar in the 11 to less than 24 bin reaching roughly 50, and the largest orange bar in the 48 to 533 bin reaching roughly 35. Panel B title Hospital admission duration. Horizontal axis labeled Patient count with tick labels 25, 0, 25, 50. Vertical axis labeled Days with four bins from top to bottom: 4 to 24, 3 to less than 4, 2 to less than 3, and 1 to less than 2. The longest blue bar appears in 1 to less than 2 at roughly 45; the longest orange bar appears in 1 to less than 2 at roughly 25. Panel C title Unique specialties. Horizontal axis labeled Patient count with tick labels 40, 20, 0, 20, 40, 60. Vertical axis labeled Units with bins from top to bottom: 9 to 26, 6 to less than 9, 3 to less than 6, and 1 to less than 3. The longest blue bar occurs in 3 to less than 6 at roughly 55; the longest orange bar occurs in 9 to 26 at roughly 40. Panel D title Total prescriptions. Horizontal axis labeled Patient count with tick labels 40, 20, 0, 20, 40, 60. Vertical axis labeled Units per year with bins from top to bottom: 19 to 65, 10 to less than 19, 4 to less than 10, and 1 to less than 4. The longest blue bar occurs in 1 to less than 4 at roughly 55; orange bars are shorter, with the largest around 20 in 19 to 65 or 10 to less than 19. — Continuous features are discretized into 4 quantile-based bins to facilitate standardized multivariable comparisons. Horizontal bar lengths represent the absolute patient count within each bin along the x-axis (split symmetrically from the central origin at x = 0 for diagnosed vs undiagnosed groups). Vertical bars represent the absolute count of individuals within each bin, grouped by diagnostic status (diagnosed vs undiagnosed). To accommodate highly skewed distributions and improve visual resolution across disparate scales, the y-axes were transformed using a pseudo-log scale and display nonoverlapping mathematical interval ranges.

Dimensionality reduction was executed using principal component analysis (PCA), followed by supervised linear discriminant analysis (LDA). PCA was conducted without centering or scaling to preserve the uniform Bernoulli variance structure and underlying matrix sparsity. Model stability was evaluated via 100 repeated random subsampling iterations (70% training and 30% validation split). To prevent data leakage, all model parameters were fitted exclusively on the training partition within each bootstrap loop. The optimal model configuration was selected based on the highest mean validation area under the receiver operating characteristic curve (AUROC).

Individual feature importance was evaluated using a leave-one-feature-out analysis conducted on the final PCA and LDA models. Comprehensive methodologic details are available in the eMethods in Supplement 2.

Results

Cohort Characteristics and Clinical Context

Of 521 NCG recruits, 430 (82.5%) provided consent for medical record linkage (Figure 2). Within the consented group, mortality was significantly associated with genetic diagnostic status: 25 of 149 diagnosed individuals (16.8%) were deceased compared with 26 of 281 undiagnosed individuals (9.3%) (χ² = 4.58; P = .032; Fisher exact, P = .03) (eTable 4 in Supplement 1). Successful linkage to ECLIPSE Live was performed for 270 of 375 living individuals (72%) (Figure 2). Unlinked individuals (n = 105) were excluded due to missing primary identifiers, withdrawn consent, patient mortality, or a change in primary care service. Postconsent exclusions were due to mortality (n = 51) and linkage failures or withdrawals (n = 109), resulting in a final analytical cohort of 270 individuals (mean [SD] age 8.65 [6.58-9.46] years; 121 females [44.8%] and 149 males [55.2%]).

Flowchart of cohort selection from 521 individuals to 270 included in analysis. Light gray rectangular nodes connected by thin gray arrows in a top to bottom sequence, with rightward arrows to exclusion boxes. Upper left main node: 521 Individuals in The Next Generation of Children Project, with a second line reading 175 With genetic diagnoses. A horizontal arrow from this node points to an upper right box reading 91 Excluded. A vertical arrow from the top node points to the next main node centered below: 430 With N I H R BioResource consent (2022), followed by two indented lines, 281 Undiagnosed and 149 Diagnosed. From this 430 node, a horizontal arrow points to a mid right exclusion box listing 55 Excluded, then separate lines reading 51 Deceased, 26 Undiagnosed, 25 Diagnosed, 2 Missing data, 1 Consultee consented, and 1 Withdrew. A vertical arrow continues downward to the next main node: 375 With demographics, clinical phenotypes (2025), followed by 365 Active and 10 Waiting for reconsent. From the 375 node, a horizontal arrow points to a lower right exclusion box reading 105 Excluded (N H S number missing, changed G P, moved abroad). A final vertical arrow leads to the bottom main node: 270 With E C L I P S E data available included in statistical analysis (genetic diagnosis, phenotypes vs 36 health care utilization metrics), followed by a final line reading 87 With genetic diagnoses. — ECLIPSE indicates Equality of Care Led Insights for Patient Safety & Engagement; GP, general practice; NHS, UK National Health Service; NIHR BioResource, UK National Institute for Health and Care Research Rare Disease BioResource.

Within the analytic cohort, 87 participants (32.2%) received a genetic diagnosis, similar to the full NGC cohort (34.0%) compared with 183 undiagnosed peers (67.8%). Among the 270 participants, 2 (0.7%) were African, 225 (83.3%) were European (83%), 1 (0.4%) was Finnish-European, 18 (6.7%) were South Asian, and 24 (8.9%) were of other ethnicities. Thirty percent of participants (n = 81) were in the Index of Multiple Deprivation (IMD) deciles 1 to 4 (lowest socioeconomic status group). Children were recruited from clinical settings including 94 (34.8%) from the NICU, 112 (41.5%) from neurodevelopmental clinics, and 64 (23.7%) from the PICU. The analytic cohort (recruited from December 2016 to August 2020) consisted of individuals from ages 5 to 22 years (analyzed in March 2025) with a slight male predominance and higher IMD than the full NCG cohort, but these differences were not statistically significant (eTables 5 and 6 in Supplement 1 and eFigure 2 in Supplement 2). While demographic testing showed no socioeconomic bias in cohort retention, our conclusions are strictly applicable to long-term survivors with continuous health care linkage.

Health Care Utilization and Cost Patterns

To evaluate disparities in health care utilization and costs between genetically diagnosed and undiagnosed individuals, comparisons were conducted across clinical outcome features and predefined analytic subgroups using a conservative multiple-testing framework with family-wise error rate control (eTable 7 in Supplement 1). Diagnosed individuals consistently demonstrated significantly higher median (IQR) health care utilization and associated costs (hospital admissions: 37 [18-66] vs 22 [10-35]; P < .001) across primary comparisons) (Figure 3). A complete inventory of P values for all 288 analyzed features is provided in eTable 8 in Supplement 1. For comparisons meeting both the significance threshold (P < .001) and the minimum sample size requirement (n ≥ 7), the magnitude of the absolute difference is formally quantified in eTable 9 in Supplement 1.

Four-panel heatmap of utilization metrics by condition with red, blue, and gray cells. Four stacked heatmap panels labeled All, Neonates, Neurodevelopmental, and Critically ill. Each panel contains a grid of small square cells with light gray background and darker gray gridlines; some cells are filled red, one cell is filled blue, and several cells are filled medium gray. Row labels appear at left. In the All panel, rows from top to bottom read Neurodevelopmental, Sensory processing, Hypotonia, Seizures, Developmental delay, and All. In the Neonates panel, a single row label at left reads All. In the Neurodevelopmental panel, a single row label reads Seizures. In the Critically ill panel, a single row label reads Seizures. A shared set of rotated column labels runs along the bottom from left to right: Hospital admission, total d; Hospital admissions; Unique specialties; Outpatient visits; A and E admissions; Total cost admissions; Unique drug groups; Total prescriptions; Allergy; Analgesia; Cardiovascular; Care and hygiene; Dental; Dermatologic; Diagnostic; Endocrine; G I regulation; Immunosuppressant; Lines, tubes, and catheters; Liver; Neurological; Nutrition; Respiratory; Short infection; Steroids; Genito urinary; Vaccination; Total cost treatment; Blood test parameters; Unique blood tests; Unique blood test parameter types; Unique conditions. Red cells appear repeatedly in the leftmost utilization columns and again in several mid and right columns, varying by row and panel. Medium gray cells appear in scattered mid columns and in a cluster near the far right in the Neonates panel. One blue cell appears in the Neonates panel near the middle columns. A legend centered below the plots contains four labeled color keys: Diagnosed in red, Undiagnosed in blue, P greater than zero point zero zero one in light gray, and Sample No. less than seven in medium gray. — Features are visually stratified based on whether they met the conservative significance threshold (P < .001 and n ≥ 7). For each condition and utilization metric, Wilcoxon rank sum tests were performed to compare diagnosed and undiagnosed individuals. The heatmaps use cells to indicate the group (diagnosed or undiagnosed) with the higher mean. A comprehensive master list of exact P values for all features shown, including those that did not meet the significance threshold, is detailed in eTable 8 in Supplement 1. A&E indicates accident and emergency department; GI, gastrointestinal.

Diagnosed individuals experienced more median (IQR) hospital admissions (37 [18-66] vs 22 [10-35]), more unique specialties (8 [4-11] vs 5 [3-8]), more primary and secondary care outpatient visits each year (14 [7-26] vs 8 [4-13]) (eFigure 3A1-A3 in Supplement 2), and higher annual admission costs (£1281 [£652-£2291] vs £795 [£377-£1296]) (Figure 4A). These trends were particularly prominent in children with hypotonia (hospital admissions: 52 [29-78] vs 20 [8-33]) (eFigure 3C1 in Supplement 2), seizures (hospital admissions: 36 [20-66] vs 20 [8-33]; outpatient visits: 13 [8-26] vs 8 [3-12]) (eFigure 3D1 and D2 in Supplement 2), and developmental delay (hospital admissions: 36 [16-67] vs 20 [8-33]) (eFigure 3E1 in Supplement 2), although individuals with seizures did not show a difference in the number of specialties. Annual median (IQR) treatment costs were also higher for diagnosed participants (£335 [£71-£1400] vs £77 [£18-£430]) (Figure 4B), especially for genetic neurodevelopmental (£1280 [£507-£2529] vs £130 [£21-£354]) (eFigure 3F3 in Supplement 2) and seizure-related (£1277 [£320-£2271] vs £60 [£15-£143]) conditions (Figure 4D). Participants with diagnosed seizures had more median (IQR) prescriptions, particularly for neurological (48 [22-88] vs 0) and gastrointestinal (5 [1-18] vs 0 [0-2]) care (eFigure 3H1 and H2 in Supplement 2). Participants with developmental delays showed a median (IQR) increase in prescriptions for care and hygiene (1 [0-16] vs 0 [0-1]) (eg, catheters, feeding tubes, and positioning aids) and nutrition (4 [0-42] vs 0 [0-1]) (eFigure 3I1 and I2 in Supplement 2). All statistically significant differences across conditions and utilization metrics are presented in Figure 4 and eFigures 2-5 in Supplement 2).

Four-panel violin plots of annual costs comparing diagnosed versus undiagnosed groups. Four violin-plot panels arranged in a two by two grid, labeled A, B, C, and D in boxed letters at the upper left of each panel. All panels use a vertical axis labeled Cost, pound sign per y, with tick labels at 0, 2000, 6000, and 14000 in panels A and B; panel C has tick labels at 0, 2000, 5000, and 10000; panel D has tick labels at 0, 1000, 3000, 6000, and 12000. Each panel contains two violins: an orange violin at left labeled Diagnosed and a light blue violin at right labeled Undiagnosed. Each violin contains three horizontal black lines marking quartiles. Panel A title: Hospital visits and admissions: total cost admissions for all participants. Under the left group: Median equals 1281, n equals 86. Under the right group: Median equals 795, n equals 169. Panel B title: Prescriptions: total cost treatment for all participants. Under the left group: Median equals 335, n equals 84. Under the right group: Median equals 77, n equals 168. Panel C title: Hospital visits and admissions: total cost admissions for participants with seizures. Under the left group: Median equals 1343, n equals 31. Under the right group: Median equals 767, n equals 103. Panel D title: Prescriptions: total cost treatment for participants with seizures. Under the left group: Median equals 1277, n equals 30. Under the right group: Median equals 60, n equals 100. Across panels, the violins extend upward into several-thousand pound ranges, with the widest sections concentrated around the middle cost ranges and narrower tails toward higher costs. — Each panel corresponds to a specific condition-utilization comparison, with violin plots contrasting diagnosed and undiagnosed groups (A, n = 255; B, n = 252; C, n = 134; D, n = 130). Horizontal lines within each violin indicate quartiles: first, 25th percentile; median, 50th percentile; and third, 75th percentile, bounding the IQRs. A pseudo-log scale to accommodate wide-ranging right-skewed cost distributions while preserving 0 values was used for the y-axis.

The health care costs associated with a diagnosis, highlighting the higher total admission and treatment costs for diagnosed individuals, are presented in Figure 4C and D and eFigure 3D and H in Supplement 2). In children recruited from NICUs, those who were diagnosed had more median (IQR) readmissions (58 [38-94] vs 16 [6-33]), specialty involvement (12 [8-14] vs 5 [2-7]), and outpatient visits (21 [15-36] vs 6 [2-12]) and higher admission costs (£2092 [£1201-£2809] vs £716 [£546-£1247]) (eFigure 4A-D in Supplement 2). Median (IQR) differences specifically in neurodevelopmental (neurological prescriptions: 42 [22-64] vs 0 [0-3]) and PICU (total cost prescriptions: £1045 [£244-£1914] vs £38 [£14-£220]) settings are shown in eFigure 5A and B in Supplement 2, and those with critical care seizures (total cost prescriptions: £2438 [£757-£4734] vs £72 [£21-£301]) are shown in eFigure 6B in Supplement 2. Diagnosed individuals consistently exhibited significantly higher health care utilization and costs, including hospital admissions, outpatient visits, and prescription costs, with the most pronounced differences observed in children with seizures, hypotonia, and developmental delay.

Predictive Model

To ascertain whether health care utilization patterns could inform the likelihood of obtaining a diagnosis via WGS, a model was trained on 70% of the cohort and validated on 30% (eFigure 7A in Supplement 2). LDA scores showed clear median (IQR) separation between diagnosed (0.62 [0.03 to 1.19]) and undiagnosed (−0.29 [−1.01 to 0.42]) individuals (P < .001), and the AUROC analysis (eFigure 7B and C in Supplement 2) resulted in an AUROC of 0.71 (sensitivity, 81%; specificity, 60%). The training AUROC was 0.78 (sensitivity, 79%; specificity, 64%), and full datasets had an AUROC of 0.76 (sensitivity, 79%; specificity, 63%).

The feature impact analysis (Figure 5) identified key clinical indicators associated with diagnostic probability (eTable 10 in Supplement 1). Key predictors were older age, higher IMD decile, management in a neurodevelopmental unit, lower accident and emergency department attendance, higher hospital admissions, higher total admissions, and higher admission costs (from £1549 to £9201). Both low (1 to 2) and high (4 to 24) inpatient days were associated with increased likelihood of diagnosis. Treatment-related predictors included higher annual prescriptions; treatment costs; and prescriptions for nutrition, neurology, gastrointestinal regulation, and hygiene and care. Low respiratory prescriptions were associated with increased likelihood, while high respiratory prescribing, low nutrition prescriptions, and management in NICU were associated with a reduced likelihood. Six HPO systems (musculoskeletal, head and neck, growth, eye, limb, and nervous system) were key predictors of a diagnosis, whereas a lower likelihood was associated with abnormalities in blood, immune, and metabolic systems. We note that the LDA model, using PCA preprocessing, prioritized clarity but may have overlooked nonlinear effects and increased false positives. In sum, feature analysis of our model suggests that the likelihood of obtaining a positive diagnosis from pediatric WGS in the context of serious illness was associated with high health care utilization, specifically frequent hospital admissions and neurological prescriptions, whereas neonatal presentation and respiratory phenotypes were associated with a lower likelihood of obtaining a genetic diagnosis.

Two-panel volcano plots of mean probability difference with 95% C I. Two stacked panels labeled A and B. Both panels use a scatter, volcano-plot layout with a horizontal axis labeled Mean probability difference and a vertical axis labeled 95% C I. Points vary in size, with larger circles indicating higher frequency; many small gray points cluster near zero on the horizontal axis. Orange circles indicate increased probability and blue-gray circles indicate decreased probability, per a legend at the right of panel A that also includes three black reference bubbles labeled 50, 100, and 150. In panel A, titled H P O term features, the horizontal axis spans approximately minus zero point zero three to plus zero point zero six two five, and the vertical axis spans zero to about zero point zero zero six. Vertical dashed lines appear near minus zero point zero one five and plus zero point zero two, and two curved dashed lines form a V shape centered at zero. Labeled orange points on the right include Abnormality of the musculoskeletal system near plus zero point zero six, Growth abnormality near plus zero point zero three five, Abnormality of head or neck near plus zero point zero five, Abnormality of the eye near plus zero point zero four five, Abnormality of limbs near plus zero point zero two, and Abnormality of the nervous system near plus zero point zero three. Labeled blue-gray points on the left include Abnormality of the immune system near minus zero point zero two five, Abnormality of metabolism or homeostasis near minus zero point zero two, and Abnormality of blood and blood-forming tissues near minus zero point zero two. Panel B, titled Demographic and primary care features, repeats the same axis labels and dashed reference lines; the horizontal axis spans roughly minus zero point zero three to plus zero point zero five. On the left, blue-gray labeled points include Nutrition near minus zero point zero two five, Neonatal near minus zero point zero two, and Respiratory near minus zero point zero two. On the right, multiple orange labeled points cluster between about plus zero point zero two and plus zero point zero five, including I M D score, Care and hygiene, Nutrition, G I regulation, Age, Neurological, Admission, d, A and E, Hospital admissions, No., Total cost admissions, y, Unique admissions, No., Total cost treatment, y, No. prescriptions, y, Neurodevelopmental, and a second Respiratory label near the far right. — Features are labeled if the absolute mean probability difference is greater than 0.02 and exceeds the width of its 95% CI. The size of each data point is proportional to each variable frequency in the cohorts. A&E indicates accident and emergency department; GI, gastrointestinal; IMD, Index of Multiple Deprivation.

Discussion

While several prior studies have indicated benefits and cost-savings of WGS in the short-term for children with severe illness, this cohort study investigated longer-term implications of a WGS diagnosis. We generated new insights by combining primary care records and genomic data of children with serious illness.

Genetic diagnosis via WGS was not associated with a reduction in overall long-term health care costs but rather was associated with more precise allocation of clinical resources tailored to individual patients. Individuals who received a genetic diagnosis used significantly more health care resources than undiagnosed peers including increased hospitalizations, outpatient visits, and condition-specific prescriptions for neurological, gastrointestinal, and nutritional management.

While associated HPO terms were equivalent between diagnosed and undiagnosed groups, children who received a diagnosis had enhanced use of specialist pathways and personalized treatment. The most pronounced differences in utilization patterns were observed among children initially diagnosed in intensive care settings (NICU or PICU), underscoring the clinical complexity of this population.

Such findings are important for informing clinical services. Our team’s previous work suggests that early detection could help prioritize faster management in the acute illness phase.^12,28 We now show that diagnosed children can often access more specialized care after the acute phase. While this might not lower immediate costs, it optimizes long-term care for chronic neurodevelopmental disorders. Importantly, our results support the continued application of current NHS WGS eligibility criteria, which prioritize children with a high clinical suspicion of monogenic disorders, rather than expanding WGS inclusion to broader, nonspecific clinical categories.

None in our cohort were recipients of gene therapies, such as antisense oligonucleotides, attracting high annual costs. Our findings of higher health care utilization and cost are thus independent of considerations around advanced therapies, which is an important future consideration.

Health care utilization patterns were closely tied to phenotype-diagnosis associations. The leave-one-feature-out analysis showed that older age, higher socioeconomic status, and some aspects of prescribing and admissions were associated with a higher diagnostic yield. Notably, neonates in the undiagnosed group exhibited significantly higher immunosuppressant use. In the absence of a genetic diagnosis to inform genotype-directed therapy, this likely reflects a reliance on empiric immunomodulation for undifferentiated inflammatory states. It may also reflect complex acquired neonatal morbidities, such as chronic lung disease, underscoring the distinct therapeutic challenges inherent in managing neonates with unknown underlying etiologies.

Limitations

This study had several limitations. We acknowledge that this study characterizes the inherent clinical complexity of the cohorts rather than any causal effect of a diagnostic intervention. A limitation of this observational design is the inability to decouple whether the diagnostic process itself was associated with increased costs via specialist interventions or whether these expenditures reflect the intrinsic baseline needs of managing genetic disease. Furthermore, these metrics do not clarify whether higher utilization reflects more targeted clinical management or is a by-product of frequent, sustained contact with the health care system. Consequently, these findings indicate high clinical complexity and systemic engagement rather than serving as direct proxies for care quality or therapeutic optimization.

Sample size constraints precluded a 3-way data split with an independent hold-out test set. Because such splits can compromise statistical power in smaller cohorts,^34,35 we instead used a rigorous internal validation, in which the validation partition informed principal component selection within a nested loop. These results represent a stable internal estimate. However, subsequent evaluation in independent, multicenter external cohorts will be required to confirm the model’s accuracy and account for recruitment-era drift. Furthermore, while our models quantified clinical trajectories, they did not capture the lived experiences of navigating increased health care utilization. Qualitative research within this cohort has highlighted complex parental support needs in the postdiagnostic period,³⁶ as well as varying parent perceptions of how genomic results may alter medical care and support availability.³⁷ Future research should integrate these qualitative experiences with utilization metrics to comprehensively evaluate the broader impacts of genomic diagnosis.

Further limitations include a relatively small analytical dataset. We were unable to include participants who did not consent for linkage studies, who had died, and who were not registered with a general practice with available linked data. The attrition between consent and final linkage introduced a potential source of selection bias. Because diagnostic status and clinical acuity are often correlated with early mortality, this survivor-only cohort may underrepresent the highest-intensity health care utilization associated with the most severe genetic presentations (eTable 11 in Supplement 1). Consequently, our findings may provide a conservative estimate of the true utilization gap. Conversely, an early diagnosis may facilitate transition to palliative care, which substantially reduces subsequent health care costs; thus analysis may also underestimate the total economic impact of early diagnosis. While statistical comparisons confirmed that the final cohort remained representative of the original study population in terms of baseline demographics (eTables 5 and 6 in Supplement 1 and eFigure 2 in Supplement 2), the potential for survivor bias must be considered.

Furthermore, we acknowledge the potential for confounding by phenotypic severity. Children with severe or multisystemic presentations are more likely to both receive a pathogenic diagnosis and require intensive, sustained medical follow-up. Since this study did not use propensity-based matching or standardized clinical severity scores, the observed utilization gap reflects a combination of the diagnostic state and the intrinsic severity of the underlying condition, characterizing the aggregate burden rather than isolating the independent economic impact of the diagnosis itself. In addition, the up-front costs of WGS were not available for inclusion. However, this aligns with using accumulated cost as a standardized proxy for systemic health care utilization rather than conducting a formal cost-effectiveness evaluation. Additionally, while our models adjusted for baseline socioeconomic status using IMD scores, our dataset lacked granular family-level covariates or geographic parameters. Unmeasured confounders, such as regional health care accessibility and parental health status, could independently modulate health care utilization.

Conclusions

The findings of this cohort study show feasibility of linkage of genetic data to community or general practice medical records and mapping of health care utilization needs at scale for children with diagnosed rare genetic conditions. Children with conditions diagnosed via WGS had significantly higher costs for care over time than those without an identified genetic diagnosis. While these findings do not measure the effect of WGS as a clinical intervention, they provide a longitudinal characterization of health care utilization patterns and highlight the profound clinical complexity and sustained resource requirements inherent in monogenic disorders. The findings suggest that a WGS diagnosis, rather than reducing overall costs, may facilitate integration of specialist care, helping align health care resources with individual needs. Future work should validate these findings in different populations. Long-term strategies should focus on predicting care intensity and treatment needs, with cost evaluations that include primary, secondary, and social care costs.

Supplement 1.

eTable 1. Primary Care Features Used in the Analysis, Including Outpatient Visits, Prescriptions, Pathology, and Conditions

eTable 2. Data Dictionary for ECLIPSE-Provided Variables and Their Transformation for Use in This Study

eTable 3. Case-Level Feature Matrix Used for Predictive Modeling

eTable 4. Attrition Analysis and Differential Mortality by Diagnostic Status

eTable 5. Distribution of Age, Gender, Ethnicity, IMD Score, Clinical Setting, and Genetic Diagnosis

eTable 6. Assessment of Cohort Representativeness Based on Consent and Linkage Status

eTable 7. Distribution of Cases Across the Full Cohort and Within Each Hospital Unit

eTable 8. Statistical Significance of Differences in Health Care Utilization Between Diagnosed and Undiagnosed Cases

eTable 9. Effect Size Estimates for Statistically Significant Health Care Utilization Differences

eTable 10. Numerical Values of Feature Impacts on Diagnostic Probability

eTable 11. Comparison of Diagnostic Yield by Clinical Acuity Setting

jamanetwopen-e2634372-s001.xlsx^{(211.5KB, xlsx)}

Supplement 2.

eMethods

eFigure 1. Histogram of Demographic and Primary Care Features Across 270 Cases, Stratified by Diagnostic Group

eFigure 2. Assessment of Cohort Representativeness and Selection Bias Across Age and Socioeconomic Deprivation Metrics

eFigure 3. Violin Plot Visualization of Health Care Utilization: Hospital Visits and Admissions, Prescriptions, and Unique Conditions

eFigure 4. Violin Plot Visualization of Health Care Utilization: Hospital Visits and Admissions and Prescriptions

eFigure 5. Violin Plot Visualization of Health Care Utilization: Neurodevelopmental Prescriptions

eFigure 6. Violin Plot Visualization of Health Care Utilization: Critically Ill Prescriptions

eFigure 7. Predictive Modeling of Genetic Diagnosis

jamanetwopen-e2634372-s002.pdf^{(2MB, pdf)}

Supplement 3.

Data Sharing Statement

jamanetwopen-e2634372-s003.pdf^{(15KB, pdf)}

References

1.Arias AV, Lintner-Rivera M, Shafi NI, et al. ; Pediatric Acute Lung Injury and Sepsis Investigators (PALISI) Network on behalf of the PALISI Global Health Subgroup . A research definition and framework for acute paediatric critical illness across resource-variable settings: a modified Delphi consensus. Lancet Glob Health. 2024;12(2):e331-e340. doi: 10.1016/S2214-109X(23)00537-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
2.van Hasselt TJ, Kanthimathinathan HK, Kothari T, et al. Impact of prematurity on long-stay paediatric intensive care unit admissions in England 2008-2018. BMC Pediatr. 2023;23(1):421. doi: 10.1186/s12887-023-04254-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
3.Long D, Minogue J, Charles K, et al. Neurodevelopmental outcome and quality of life in children admitted to the paediatric intensive care unit: a single-centre Australian cohort study. Aust Crit Care. 2024;37(6):903-911. doi: 10.1016/j.aucc.2024.05.001 [DOI] [PubMed] [Google Scholar]
4.Rodriguez KM, Vaught J, Salz L, et al. Rapid whole-genome sequencing and clinical management in the PICU: a multicenter cohort, 2016-2023. Pediatr Crit Care Med. 2024;25(8):699-709. doi: 10.1097/PCC.0000000000003522 [DOI] [PMC free article] [PubMed] [Google Scholar]
5.Hill M, Hammond J, Lewis C, Mellis R, Clement E, Chitty LS. Delivering genome sequencing for rapid genetic diagnosis in critically ill children: parent and professional views, experiences and challenges. Eur J Hum Genet. 2020;28(11):1529-1540. doi: 10.1038/s41431-020-0667-z [DOI] [PMC free article] [PubMed] [Google Scholar]
6.Kansal R. Rapid whole-genome sequencing in critically ill infants and children with suspected, undiagnosed genetic diseases: evolution to a first-tier clinical laboratory test in the era of precision medicine. Children (Basel). 2025;12(4):429. doi: 10.3390/children12040429 [DOI] [PMC free article] [PubMed] [Google Scholar]
7.Sanford EF, Clark MM, Farnaes L, et al. ; RCIGM Investigators . Rapid whole genome sequencing has clinical utility in children in the PICU. Pediatr Crit Care Med. 2019;20(11):1007-1020. doi: 10.1097/PCC.0000000000002056 [DOI] [PMC free article] [PubMed] [Google Scholar]
8.Kingsmore SF, Nofsinger R, Ellsworth K. Rapid genomic sequencing for genetic disease diagnosis and therapy in intensive care units: a review. NPJ Genom Med. 2024;9(1):17. doi: 10.1038/s41525-024-00404-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
9.Petrikin JE, Cakici JA, Clark MM, et al. The NSIGHT1-randomized controlled trial: rapid whole-genome sequencing for accelerated etiologic diagnosis in critically ill infants. NPJ Genom Med. 2018;3(1):6. doi: 10.1038/s41525-018-0045-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
10.Stark Z, Scott RH. Genomic newborn screening for rare diseases. Nat Rev Genet. 2023;24(11):755-766. doi: 10.1038/s41576-023-00621-w [DOI] [PubMed] [Google Scholar]
11.Wojcik MH, Schwartz TS, Thiele KE, et al. Infant mortality: the contribution of genetic disorders. J Perinatol. 2019;39(12):1611-1619. doi: 10.1038/s41372-019-0451-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
12.French CE, Delon I, Dolling H, et al. ; NIHR BioResource—Rare Disease; Next Generation Children Project . Whole genome sequencing reveals that genetic conditions are frequent in intensively ill children. Intensive Care Med. 2019;45(5):627-636. doi: 10.1007/s00134-019-05552-x [DOI] [PMC free article] [PubMed] [Google Scholar]
13.Auber B, Schmidt G, Du C, von Hardenberg S. Diagnostic genomic sequencing in critically ill children. Med Genet. 2023;35(2):105-112. doi: 10.1515/medgen-2023-2015 [DOI] [PMC free article] [PubMed] [Google Scholar]
14.Farnaes L, Hildreth A, Sweeney NM, et al. Rapid whole-genome sequencing decreases infant morbidity and cost of hospitalization. NPJ Genom Med. 2018;3:10. doi: 10.1038/s41525-018-0049-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
15.Morton SU, Christodoulou J, Costain G, et al. Multicenter consensus approach to evaluation of neonatal hypotonia in the genomic era: a review. JAMA Neurol. 2022;79(4):405-413. doi: 10.1001/jamaneurol.2022.0067 [DOI] [PMC free article] [PubMed] [Google Scholar]
16.Gano D, Boardman JP, Agarwal S, et al. ; Newborn Brain Society Guidelines and Publications Committee . Neonatal neurocritical care considerations for prenatally identified neurological disorders. Pediatr Res. Published online January 7, 2026. doi: 10.1038/s41390-025-04691-w [DOI] [PubMed] [Google Scholar]
17.van Hasselt TJ, Gale C, Battersby C, Davis PJ, Draper E, Seaton SE; United Kingdom Neonatal Collaborative and the Paediatric Critical Care Society Study Group (PCCS-SG) . Paediatric intensive care admissions of preterm children born <32 weeks gestation: a national retrospective cohort study using data linkage. Arch Dis Child Fetal Neonatal Ed. 2024;109(3):265-271. doi: 10.1136/archdischild-2023-325970 [DOI] [PMC free article] [PubMed] [Google Scholar]
18.Choi J, Park E, Choi AY, Son MH, Cho J. Incidence and mortality trends in critically ill children: a Korean population-based study. J Korean Med Sci. 2023;38(23):e178. doi:10.3346/jkms.2023.38.e178 [DOI] [PMC free article] [PubMed]
19.Schults JA, Hall L, Charles KR, et al. Hospital-acquired complications in critically ill children and PICU length of stay, duration of respiratory support, and economics: propensity score matching in a single-center cohort, 2015-2020. Pediatr Crit Care Med. 2025;26(3):e304-e314. doi: 10.1097/PCC.0000000000003668 [DOI] [PubMed] [Google Scholar]
20.van Hasselt TJ. Preterm Birth and Paediatric Intensive Care: Using National Data to Examine Critical Illness in Early Childhood. University of Leicester; 2024. [Google Scholar]
21.Dimmock D, Caylor S, Waldman B, et al. Project Baby Bear: rapid precision care incorporating rWGS in 5 California children’s hospitals demonstrates improved clinical outcomes and reduced costs of care. Am J Hum Genet. 2021;108(7):1231-1238. doi: 10.1016/j.ajhg.2021.05.008 [DOI] [PMC free article] [PubMed] [Google Scholar]
22.Maron JL, Kingsmore S, Gelb BD, et al. Rapid whole-genomic sequencing and a targeted neonatal gene panel in infants with a suspected genetic disorder. JAMA. 2023;330(2):161-169. doi: 10.1001/jama.2023.9350 [DOI] [PMC free article] [PubMed] [Google Scholar]
23.Krantz ID, Medne L, Weatherly JM, et al. ; NICUSeq Study Group . Effect of whole-genome sequencing on the clinical management of acutely ill infants with suspected genetic disease: a randomized clinical trial. JAMA Pediatr. 2021;175(12):1218-1226. doi: 10.1001/jamapediatrics.2021.3496 [DOI] [PMC free article] [PubMed] [Google Scholar]
24.Michigan: Project Baby Deer. Radys Children’s Institute for Genomic Medicine. Accessed September 10, 2025. https://radygenomics.org/case-studies/michigan-project-baby-deer/
25.McDermott H, Sherlaw-Sturrock C, Baptista J, Hartles-Spencer L, Naik S. Rapid exome sequencing in critically ill children impacts acute and long-term management of patients and their families: a retrospective regional evaluation. Eur J Med Genet. 2022;65(9):104571. doi: 10.1016/j.ejmg.2022.104571 [DOI] [PubMed] [Google Scholar]
26.van Hasselt TJ, Newman S, Kanthimathinathan HK, et al. ; United Kingdom Neonatal Collaborative and the Paediatric Critical Care Society Study Group (PCCS-SG) . Transition from neonatal to paediatric intensive care of very preterm-born children: a cohort study of children born between 2013 and 2018 in England and Wales. Arch Dis Child Fetal Neonatal Ed. 2025;110(4):369-376. doi: 10.1136/archdischild-2024-327457 [DOI] [PMC free article] [PubMed] [Google Scholar]
27.ECLIPSE. Prescribing Services Ltd. Accessed December 3, 2025. https://www.eclipselive.org
28.French CE, Dolling H, Mégy K, et al. ; Next Generation Children’s Project Consortium . Refinements and considerations for trio whole-genome sequence analysis when investigating Mendelian diseases presenting in early childhood. HGG Adv. 2022;3(3):100113. doi: 10.1016/j.xhgg.2022.100113 [DOI] [PMC free article] [PubMed] [Google Scholar]
29.Redman DM. R14: Acutely unwell children with a likely monogenic disorder (rapid genome testing). National Health Service England, National Genomics Education Programme/GeNotes. Updated August 15, 2025. Accessed November 18, 2025. https://www.genomicseducation.hee.nhs.uk/genotes/knowledge-hub/r14-acutely-unwell-children-with-a-likely-monogenic-disorder/
30.Turnbull C, Scott RH, Thomas E, et al. ; 100 000 Genomes Project . The 100 000 Genomes Project: bringing whole genome sequencing to the NHS. BMJ. 2018;361:k1687. doi: 10.1136/bmj.k1687 [DOI] [PubMed] [Google Scholar]
31.Wu B, Kang W, Wang Y, et al. Application of full-spectrum rapid clinical genome sequencing improves diagnostic rate and clinical outcomes in critically ill infants in the China Neonatal Genomes Project. Crit Care Med. 2021;49(10):1674-1683. doi: 10.1097/CCM.0000000000005052 [DOI] [PubMed] [Google Scholar]
32.National Genomic Test Directory. National Health Service England. Updated July 16, 2026. Accessed March 4, 2026. https://www.england.nhs.uk/publication/national-genomic-test-directories/
33.Putman TE, Schaper K, Matentzoglu N, et al. The Monarch Initiative in 2024: an analytic platform integrating phenotypes, genes and diseases across species. Nucleic Acids Res. 2024;52(D1):D938-D949. doi: 10.1093/nar/gkad1082 [DOI] [PMC free article] [PubMed] [Google Scholar]
34.Steyerberg EW, Harrell FE Jr, Borsboom GJJM, Eijkemans MJC, Vergouwe Y, Habbema JDF. Internal validation of predictive models: efficiency of some procedures for logistic regression analysis. J Clin Epidemiol. 2001;54(8):774-781. doi: 10.1016/S0895-4356(01)00341-9 [DOI] [PubMed] [Google Scholar]
35.Altman DG, Vergouwe Y, Royston P, Moons KGM. Prognosis and prognostic research: validating a prognostic model. BMJ. 2009;338:b605. doi: 10.1136/bmj.b605 [DOI] [PubMed] [Google Scholar]
36.Dolling H, Rowitch S, Bromham M, et al. Fathers’ and mothers’ support needs and support experiences after rapid genome sequencing. Eur J Hum Genet. 2026;34(2):260-269. doi: 10.1038/s41431-025-01987-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
37.Dolling H. Psychosocial Impact of Paediatric Early Rapid Genomic Testing and Diagnosis: A Mixed Methods Study of Parental Adjustment, Adaptation, Risk and Resilience. Apollo–University of Cambridge; 2025. [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials