Main

Newborn screening (NBS) for rare conditions is one of the most effective public health interventions1, delivered with high uptake, quick turnaround times and at low cost. However, the increased ability to diagnose and treat rare diseases ushered in by the era of genomic and precision medicine2,3 has challenged the ability of NBS programs to keep pace, with the time to add a condition averaging 9 years in the USA4. Incorporating genomic sequencing into NBS programs offers the opportunity to substantially increase the number of conditions screened and to include conditions that do not have available biochemical markers, such as those predisposing to childhood cancers5. Further, genomic sequencing provides flexibility to add or remove conditions at low incremental cost and has the potential to provide health benefit over an individual’s lifetime through reuse of data for diagnostic, screening and research purposes6.

Genomic NBS (gNBS) raises considerable issues when implementation at the population level is considered. There is little consensus on how it should be offered in terms of timing or method of consent; what testing modality or samples to use; which conditions to screen; or how to manage downstream healthcare system impacts. Despite extensive debate over the past 10 years7, there is little evidence from prospective cohorts to guide policy and implementation8,9,10,11. Multiple studies are currently underway in diverse healthcare systems that will test the feasibility of different implementation models and provide early data on patient, family and healthcare system outcomes5. Two of these, the GUARDIAN study in the USA and the BabyDetect study in Belgium, reported interim results earlier this year in cohorts of 4,000 and 3,847 infants, respectively8,11.

The BabyScreen+ study aimed to address some of these questions through multidisciplinary evaluation of a prospective cohort of 1,000 newborns in a public healthcare system, with gNBS offered for 605 genes associated with early-onset, treatable childhood conditions alongside standard NBS (stdNBS)12,13. The study assessed a broad range of outcomes including feasibility of performing clinically accredited genome sequencing using dried blood spot (DBS) cards; screening outcomes compared to stdNBS; parental psychosocial outcomes; and acceptability.

Results

Demographics and recruitment

Study enrollment was initiated12 by 1,288 prospective parents, with 301 (23%) not completing enrollment, declining screening or withdrawing from the study. gNBS was completed for 1,000 newborns over 16 months, of whom 523 (52%) were male and 477 (48%) female, including 13 sets of twins (Fig. 1). The participant sample was underrepresented for parents under the age of 30, with two or more children, from regional areas and in Quintiles 1 and 2 of the Index of Relative Socioeconomic Advantage and Disadvantage. A relative overrepresentation of higher-educated individuals was observed (Fig. 2).

Fig. 1: Participation in BabyScreen+.
Fig. 1: Participation in BabyScreen+.
Full size image

Eligible participants initiated enrollment via the online platform Genetics Adviser, accessible via a QR code either on a study invitation card obtained from a healthcare professional (active recruitment) or on a range of other advertising material (passive recruitment). Engagement was tracked from initial login to the platform using a unique study ID. Completion of the Genetics Adviser education and decision-support module, T1 and T2 research surveys, and consent to both research participation and clinical gNBS was mandatory to complete enrollment. Completion of the T3 survey, distributed 3 months postresult, was optional. Solid line, pathway to study completion; dashed line, pathway for participants who were lost to follow-up, declined gNBS or withdrew from the study.

Fig. 2: Comparison between the birth parents of the BabyScreen+ cohort and the age- and sex-matched population of Victoria, Australia.
Fig. 2: Comparison between the birth parents of the BabyScreen+ cohort and the age- and sex-matched population of Victoria, Australia.
Full size image

a,b, Comparison between the study cohort and the Victorian perinatal population in 2022 in relation to age (a) and parity (b), sourced from the Australian Institute of Health and Welfare’s National Perinatal Data Collection. c–f, Comparisons between the BabyScreen+ cohort and the age- and sex-matched population of Victoria, Australia, based on Australian Census data in relation to geographic location (c), highest level of educational attainment (d), area of ancestral origin (e) and Index of Relative Socioeconomic Advantage and Disadvantage (f). Geographic location (Remoteness Area) (c) and Index of Relative Socioeconomic Advantage and Disadvantage (f) are based on the classifications used by the Australian Bureau of Statistics. These variables are from a level 2 statistical area (average population of 10,000 people). Participants were able to select more than one response when reporting their ancestry. Ancestry data (e) are presented as a proportion of the total responses. The top five ancestries from the BabyScreen+ data are presented with the addition of Aboriginal and Torres Strait Islander ancestry. Percentages are rounded to the nearest whole number and totals may not equal 100. The demographic variables were compared between BabyScreen+ participants whose newborns had genomic newborn screening (n = 987) and the relevant Victorian population datasets using chi-square tests. Significant differences were observed for age (χ2 = 150.80, P < 0.001), parity (χ2 = 69.04, P < 0.001), geographic location (χ2 = 72.46, P < 0.001), highest level of educational attainment (χ2 = 901.36, P < 0.001) and Index of Relative Socioeconomic Advantage and Disadvantage (χ2 = 37.69, P < 0.001). aThe BabyScreen+ demographic survey indicated that Oceanic ancestry refers to people from the Pacific Islands or Micronesia. The 2021 Australian Census did not provide such guidance and as such it is likely that many of the responses indicating Oceanic ancestry refer to respondents of European ancestry born in Australia. Due to the discrepancy in how this variable is ascertained, a chi-square test was not run for this variable.

Recruitment by a healthcare professional (active recruitment) resulted in the highest number of study participants completing enrollment (667 out of 998, 67%, Fig. 3), with social media advertising the most successful passive recruitment method (190 out of 998, 19%). Active recruitment had the lowest proportion of incomplete enrollments (Extended Data Table 1). A detailed breakdown of cohort demographics by recruitment method is provided in Extended Data Table 2 and Supplementary Fig. 1.

Fig. 3: Quarterly recruitment numbers by recruitment method.
Fig. 3: Quarterly recruitment numbers by recruitment method.
Full size image

Each participating family is represented as a single square, with n = 987 enrolled participants whose newborns had genomic newborn screening including 13 sets of twins. Recruitment methods were broadly categorized as ‘active’, where a study invitation card was offered to a prospective participant by a healthcare professional (midwife, obstetrician or primary care physician) in both the public and private healthcare sector. Passive methods were only utilized in the public healthcare setting and included study posters and videos in antenatal clinic waiting rooms, study invitation cards enclosed in hospital antenatal packs provided by mail, direct text messages to pregnant women, and advertisement on a designated pregnancy app used at one participating hospital. A paid social media advertising campaign on Facebook and Instagram was targeted to women between the ages of 18 and 45 located in the state of Victoria, Australia. The study team recruited eligible participants who contacted the study team directly and unsolicited. Commencement of deployment was asynchronous for different recruitment methods, with major deployment dates indicated.

Feasibility of genome sequencing from dried blood spots

DBS cards from 1,003 newborns were processed for gNBS. Reprocessing was required for 82 (8.2%) samples due to sample-related (3.2%) or process-related (5.0%) failures (Extended Data Table 3). DNA reextraction was performed for 79 (7.9%) samples, using additional punches from the existing DBS card (50), collection of additional DBS cards (7) or collection of fresh blood samples (22 samples, Extended Data Table 4). Three participants declined sample recollection.

Sequencing data generation failed target coverage requirements for 191 (19%) samples, requiring additional sequencing. Process improvements, including optimization of sample quantitation methods and sequencer loading concentrations, reduced failure rates from an initial average of 28% to less than 5% (Supplementary Fig. 2).

All reextracted and reprocessed samples were successfully reported on the second attempt. Excluding samples that required reextraction would have led to two missed high-chance results (UNC13D and GNAS, Table 1).

Table 1 High-chance results reported in BabyScreen+ and their clinical impact
Full size table

Average turnaround time (TAT) for stdNBS followed by gNBS was 24 days from sample collection (target: 28 days, minimum: 11, maximum: 68, 95% confidence interval (CI) 23.5–24.5). Average TAT for gNBS was 13 days (minimum: 6, maximum: 56, 95% CI 12.7–13.3), with 73% reported within the target of 14 days. The proportion of reports issued within target TAT increased from 65% to 81% following reduction in sequencing failure rates.

Genomic data analysis and interpretation

Analysis protocols (Fig. 4) were validated using a cohort of 108 clinical cases of critically ill infants undergoing genomic testing, with 61 known low-chance and 47 known high-chance results2. Validation data showed >97% sensitivity with 46 of the known high-chance cases correctly flagged. The high-chance case missed by the analysis had a multinucleotide variant (ACAD9, NM_014049.4, c.1376_1381delinsCCT, p.(Lys459_Ser461delinsThrCys). This was incorrectly annotated by the analysis software as two non-high impact variants, a known limitation. No cases were incorrectly classified as low-chance by the automated system; however, 32 of 61 (51%, Extended Data Table 5) of known low-chance cases required manual review of nonreportable variants.

Fig. 4: Genomic data analysis workflow for the BabyScreen+ study.
Fig. 4: Genomic data analysis workflow for the BabyScreen+ study.
Full size image

aKnown benign variants include those classified as benign or likely benign by the Victorian Clinical Genetics Services (VCGS) and/or submitted as such in national (Shariant39) and/or international databases (for example, ClinVar, (likely) benign with at least two stars). bVery common variants include those with a frequency of more than 2% in VCGS internal data and/or the Genome Aggregation Database. cKnown pathogenic variants include those classified as (likely) pathogenic by VCGS, in national databases (Shariant39) and/or international databases (ClinVar). dHigh-impact variants include nonsense, frameshift, canonical splice site, start loss, stop loss, and large deletions and duplications. eVariants in the literature include missense, in-frame and splice region variants in the Mastermind Cited Variant Reference (now known as the Indexed Variant File). fVariants too common for monoallelic disease (0.01% in Genome Aggregation Database) or biallelic disease (0.1% in Genome Aggregation Database). Mode of inheritance (MOI) is derived from the MOI of the gene in the BabyScreen+ panel in PanelApp Australia. gOnly heterozygous variants are allowed for monoallelic MOI. Biallelic MOI requires a homozygous or hemizygous zygosity, or else two (or more) variants in the same gene. VCF, variant call file.

Of the 1,000 study samples, 451 (45%) were automatically reported as low-chance. In the remaining 549 (55%) cases, 1,045 variants were manually reviewed, with 36 variants undergoing full assessment and 18 ultimately reported. Of the nonreportable variants, 1,009 were discarded following rapid manual assessment due to insufficient evidence for pathogenicity, including all copy number variants (Extended Data Table 6). The 18 variants that underwent full assessment but were not reported were excluded for a range of reasons including mismatched mechanism of disease or mode of inheritance, or association with adult-onset or mild disease (Extended Data Tables 6 and 7).

Modeling filter conditions showed an automatic low-chance report rate of 82% was achievable (Extended Data Table 8) but increased the likelihood of false negative results. Only reporting variants previously described as pathogenic in ClinVar would have resulted in the loss of one high-chance result (GNAS, Table 1).

High-chance results were issued for 16 newborns (1.6%, Table 1). No discordances with stdNBS were identified. One newborn was identified as having hypothyroidism through stdNBS. gNBS provided the precise etiology (GNAS variant) enabling cascade testing and surveillance for multisystem involvement. All high-chance results were confirmed by orthogonal testing on fresh samples, with no discordances. All variants in recessive genes were confirmed in trans.

Clinical impact of high-chance results

Clinical impact ranged from instituting preventative measures or surveillance to active management. Nine results in three genes (G6PD, MT-RNR1 and RYR1) prompted preventative measures (Table 1). These were managed by the gNBS team without involvement of specialist services. Five results (FBN1, GNAS, ENG, GJB2 and DICER1) led to initiation of surveillance measures, which included echocardiogram, blood tests, MRI scan, audiology assessment and follow-up with specialist physicians. Two results (PKHB and UNC13D) resulted in immediate treatment (Table 1). The infant with glycogen storage disorder due to PKHB variants had unrelated congenital heart disease and had been scheduled for cardiac surgery. Identification of an underlying metabolic disorder enabled appropriate multidisciplinary management of perioperative fasting to avoid episodes of hypoglycemia. The infant diagnosed with UNC13D-related hemophagocytic lymphohistiocytosis (HLH) was clinically well at the time of gNBS result disclosure, but immunological testing revealed early signs of immune dysregulation. The diagnosis enabled early commencement of therapy with immune modulators and proactive planning for bone marrow transplantation, which was performed at 4 months of age.

Testing in first-degree relatives resulted in 20 additional diagnoses (12 parents and 8 siblings). None of the affected relatives were previously suspected of having a genetic condition, although clinical findings and family histories consistent with the diagnoses were present in 4 parents (FBN1, GNAS, ENG and DICER1). Two participants required reanalysis of gNBS data for diagnostic purposes within the study period. One was a newborn with a low-chance result who had a diagnosis of moderate bilateral sensorineural hearing loss from newborn hearing screening. Diagnostic reanalysis did not identify a genetic cause for hearing loss. The second newborn with a high-chance result for G6PD deficiency had diagnostic reanalysis following admission to neonatal intensive care with multiorgan failure. A reanalysis report was issued within 24 h of the request, with no additional genetic diagnosis identified.

Psychosocial outcomes and attitudes toward screening

Out of 1,012 parents who consented to gNBS, 998 (99%) completed a survey (Fig. 1). We conducted 48 interviews with 46 birth parents and 3 partners. This included 22 pretest gNBS acceptors, 2 gNBS decliners, 17 low-chance results recipients and 8 high-chance results recipients.

Most survey respondents (80%) indicated they consented ‘immediately’ using Genetics Adviser, with the decision perceived as either ‘easy’ or ‘very easy’ (Supplementary Table 1). Only 8% found decision-making difficult. Interviewees valued Genetics Adviser, with education content generally used to reaffirm decisions, provide guidance on what to consider, understand the impact of their decision or facilitate discussions with partners.

Interviewees also described how they considered clinical, psychosocial and practical factors in gNBS decisions. They weighed up benefits of screening, the types of conditions being screened, potential barriers and their ability to navigate results.

Decisions to have gNBS were most strongly influenced by a desire “To know what to expect for my baby’s future” (77% survey respondents) (Fig. 5). The main influence for declining gNBS (10 surveys completed) was concern about the result having negative impact on parents (80%) (Supplementary Table 2).

Fig. 5: Factors influencing participants’ decision to have genomic newborn screening.
Fig. 5: Factors influencing participants’ decision to have genomic newborn screening.
Full size image

Participants whose newborns had genomic newborn screening (n = 987) were asked to rate the extent to which their decision was influenced by each statement at enrollment.

Older age (for ages 30–34, odds ratio (OR) 2.41, 95% CI 1.35–4.33; ages 35 and over, OR 2.58, 95% CI 1.38–4.81), English as the main language spoken at home (OR 1.9, 95% CI 1.02–3.54) and prior experience with a genetic test (OR 1.8, 95% CI 1.09–3.10) were factors that were associated with participants choosing to have gNBS (Extended Data Table 9). At enrollment, the median trait anxiety score was 32.63 (interquartile range 28.42–38.94). Most survey respondents scored under the cutoff for probable clinical state anxiety at enrollment (80%) and at the T3 survey (85%, n = 501). Postresult return, decision regret was very low (median 0, interquartile range 0–10, n = 506). Parents of infants who received a high-chance result were asked to complete an adapted version of the Genomics Outcomes Scale (GOS). Eight participants (50%) responded to the GOS scale, with a mean empowerment score of 26.3 (95% CI 23.98–28.52) out of 30. Interviewees who received a low-chance gNBS result reported positive impacts such as reassurance. Interviewees who received a high-chance result valued results due to their clinical utility. Prompt genetic counseling and access to high-quality information facilitated adaptation.

Most respondents (80%) would choose to have gNBS for a future baby, and 92% would recommend it to a family member (Supplementary Table 3). All but one respondent thought that gNBS should be available to all parents, and 97% thought it should be publicly funded.

Discussion

The BabyScreen+ study provided screening for disease-causing variants in 605 genes associated with severe, early-onset, treatable childhood conditions in a prospective cohort of 1,000 newborns. We found 1.6% of newborns had an increased chance of a screened condition, leading to a range of interventions from preventative measures and surveillance through to bone marrow transplant. Only one of these diagnoses was identified by stdNBS, highlighting the ability of gNBS to identify a much broader range of actionable rare disorders. Parental attitudes toward gNBS were positive with minimal decisional regret.

We have demonstrated feasibility of delivering clinically accredited gNBS within a public healthcare system using a scalable model designed to be minimally disruptive to the healthcare system and to families. The model was informed by prior studies5,14,15,16, as well as public and professional consultation including focus groups17,18, key informant interviews19 and discrete choice experiments with over 2,000 members of the Australian public20. Our model includes education and consent using online tools during pregnancy; use of existing DBS collection pathways; and laboratory processes that integrate with stdNBS and balance automation and manual review of genomic data to minimize false positive results. The use of clinically accredited genome sequencing and analysis facilitates reuse of the data for diagnostic purposes and for further age-appropriate screening21.

While prior studies have mostly elected to offer gNBS in the newborn period using trained personnel, concerns have been raised about the scalability and appropriateness of these models, given the complexity of information required for gNBS consent and the potential impact on consent for stdNBS5,22. The stated preference, of both parents and healthcare providers, for information provision and consent during pregnancy18,19,23 led us to implement a model where gNBS was introduced during pregnancy using multimodal approaches. An extensively evaluated online digital platform24,25 provided education and decision support, including case vignettes and value clarification exercises. The resultant cohort is diverse, with 80% of parents reporting being able to make an ‘easy’ decision to have gNBS. However, the temporal separation of consent from sample collection in this model raises challenges such as the need to ensure ongoing consent and to develop laboratory protocols for accurate sample identification. The overall model requires further evaluation at scale to ensure it supports equitable access, particularly to families disadvantaged by socioeconomic or cultural and linguistic factors. Further consideration also needs to be given about how information provision and consent will integrate with other screening tests offered in pregnancy, notably reproductive carrier screening where there is considerable overlap in the conditions screened13,26,27, potentially creating confusion for healthcare providers and potential parents alike. This issue is highlighted by the high-chance result for UNC13D-related HLH in our cohort where the parents had completed expanded reproductive carrier screening for 200 genes. This test included two more frequent genetic causes of familial HLH but excluded UNC13D.

Similar to other studies, we found using DBS cards for gNBS was feasible, with 3.2% of samples requiring reprocessing to obtain results due to sample-related failures28. Taking additional punches from the original DBS card addressed most reprocessing requirements, negating the need for sample recollection. Opting not to reprocess would have resulted in two missed high-chance results, including that of HLH (UNC13D).

The average time to report of 13 days compares favorably with stdNBS and is faster than the time to report from other studies, which ranged from 32.5 days11 to 64 days8. While we elected to perform gNBS after stdNBS to minimize disruption, we envisage the two occurring in parallel. Achieving clinically meaningful turnaround times for gNBS is important as many of the conditions are immediately actionable, as demonstrated in this cohort by the newborn diagnosed with glycogen storage disease who had forthcoming cardiac surgery for unrelated reasons. Knowledge of the underlying condition allowed appropriate planning of perioperative care under the guidance of a metabolic physician to avoid fasting hypoglycemia.

A high degree of automation will be required to scale gNBS to public health programs but careful calibration is needed to minimize false positive and negative results5. We adopted an integrated approach, incorporating stdNBS results and multidisciplinary review before reporting. Adherence to predefined variant lists can lead to false negatives8 and would have likely led to at least one missed result in our cohort as well. Similarly, automation can miss complexities, such as common in cis variants in genes with recessive inheritance11, increasing false positives with significant impact on families and the workforce when considered at scale (Extended Data Table 7). Our analysis approach mitigates these common pitfalls; our rate of high-chance results (1.6%) is comparable with that of other studies8,11 despite hundreds more genes being analyzed13 (Extended Data Table 10). While the rate of manual variant review required in our setup currently cannot be considered scalable, we have shown that a small number of simple changes can substantially increase automation without major compromises to accuracy (Extended Data Table 8). Continual refinement of automation pipelines and the integration of data generated by pilot gNBS studies will optimize this balancing act moving forwards29.

We identified a wide range of conditions, with G6PD deficiency the commonest finding. Just over half of the high-chance results were managed without involvement of specialist services. The remainder required input from multiple specialists and prompt access to other investigations. In addition, 20 relatives received a molecular diagnosis through cascade testing. As gNBS scales, there will be a need to systematically develop dedicated downstream pathways that integrate results, investigations and referrals to a broad range of pediatric and adult services with adequate psychosocial support for families. Equitable access to these pathways, including for regional and remote communities, will be a key consideration for public health programs.

Prospective parents and the general public express positive views toward gNBS, while health professionals are typically more cautious5,30,31,32,33,34,35. Many concerns have been raised about the potential psychosocial risks of sequencing newborns, including effects on parent–child bonding, perceived child vulnerability and self and partner blame36. In this cohort, parents held positive views following gNBS, generally found the decision easy to make and supported future public funding. We found no evidence of adverse psychosocial outcomes such as increased anxiety or decision regret, consistent with smaller previous cohorts14,37,38. Rather, participants reported feeling empowered. Even those with a high-chance result reported adapting to the information through access to prompt high-quality information. The views of parents with direct experience of gNBS form an important part of assessing the acceptability of gNBS and complement the findings of the focus groups17,18 and discrete choice experiments20 conducted with the Australian public as part of this study.

The limitations of this study included the relatively small cohort size and recruitment over an 18-month period. This means that the laboratory and clinical systems were not tested at the scale that would be required to deliver a national screening program. For the population of Australia this would require 300,000 samples to be processed per annum and over 4,500 high-chance results to be returned in a variety of settings, necessitating substantial investment in infrastructure, workforce development and national consistency. In addition, there was an overrepresentation of highly educated parents and targeted efforts would be required to ensure equity of access both to screening and to downstream care pathways. Other areas that remain underexplored include the potential for the data to be reused for clinical and screening purposes throughout an individual’s lifetime. While we were able to demonstrate the benefits of data reuse for diagnostic purposes in two infants within a relatively short period of time, large-scale implementation of a data reuse model would include further considerations of consent models, infrastructure requirements and integration with electronic medical records for example.

In conclusion we have demonstrated feasibility and acceptability of gNBS in a public healthcare system, using a model that integrates with the existing stdNBS program and delivers clinically accredited results with rapid turnaround times. While we provide comprehensive multidisciplinary evaluation, much larger longitudinal studies are now required to demonstrate scalability to population level and assess other outcome measures such as equity, cost-effectiveness and long-term impacts on families and healthcare systems.

Methods

Ethical approval

The BabyScreen+ study obtained ethics approval from the Royal Children’s Hospital Melbourne Human Research Ethics Committee (main BabyScreen+ protocol: HREC/91500/RCHM-2023; key informant interviews: HREC/90929/RCHM-2022; and focus groups and discrete choice experiments: HREC/91392/RCHM-2022).

Study setting

The previously published study protocol12 is summarized here. BabyScreen+ was designed to evaluate the acceptability and feasibility of gNBS in a prospective cohort of 1,000 newborns from the state of Victoria, Australia, using clinically accredited genome sequencing. The study is governed and administered by the Murdoch Children’s Research Institute (MCRI), Melbourne, Australia. All laboratory procedures, data analysis and reporting were performed by VCGS, a wholly-owned not-for-profit subsidiary of MCRI, in Melbourne, Australia. VCGS is clinically accredited (NATA/RCPA) to ISO15189;2012 to carry out genetic and genomic testing. In the state of Victoria, stdNBS is a publicly funded program and parents provide explicit written consent for stdNBS in the postpartum period. VCGS is responsible for the delivery of stdNBS for the state, and DBS cards were accessed with permission from participants and the Victorian State Government.

Recruitment

During the third trimester of pregnancy gNBS was offered, free of charge, in both private and public healthcare settings. Recruiting healthcare professionals attended an education session and were provided with study cards, posters and videos. The study was also advertised via social media, text message and a pregnancy app. Birth parents could enroll if aged 16 or over; planning to give birth in Victoria, Australia and intending to participate in stdNBS. Enrollment was ideally in pregnancy but was available up until 2 weeks after birth. Study cards and advertising materials contained a QR code connecting potential participants to the online platform Genetics Adviser, which provided education and decision support24,25. Participants provided informed consent to research and to gNBS separately. Contact with a genetic counselor was available at any stage.

DNA extraction and sequencing

Following completion of stdNBS, DNA was extracted from four 3 mm DBS punches using the Mag-Bind DNA Blood and Tissue kit (Omega Biotek). PCR-free genome sequencing libraries were created using the PCR-free DNA prep kit (Illumina) and sequenced using a 2 × 150 bases paired end read configuration to an average depth of ×30 on a NovaSeq X Plus instrument (Illumina). Sequencing library quantitation was initially performed using the Qubit dsDNA High Sensitivity kit (Thermo Fisher Scientific), and was transitioned to real-time PCR-based library quantitation using the KAPA Library Quantification kit for Illumina Platforms (Roche) on a QuantStudio 7 Pro instrument (Thermo Fisher Scientific) as per standard protocol, following troubleshooting of sequencing data issues.

Genomic data analysis and interpretation

Details of the gene selection process are previously published13. Data analysis and interpretation were performed using Illumina Emedgene variant analysis software version 35.0, with custom in-house filter configuration (Supplementary Table 4) designed to identify variants for potential reporting in 605 genes associated with early-onset treatable childhood conditions. Variants passed by the filters were flagged for manual assessment by an analyst. Manual assessment was performed in two steps: a rapid assessment (<10 min per variant) was undertaken to exclude variants that were erroneously flagged by the filter configuration (for example conflicting ClinVar classifications and insufficient literature support, Extended Data Table 6). Variants that could not be rapidly excluded underwent full variant curation and classification. Variants were considered for reporting if they were classified as likely pathogenic or pathogenic based on the American College of Medical Genetics and Genomics and the Association for Molecular Pathology guidelines40 and were consistent with the relevant mode of inheritance. Carrier status, adult-onset or mild forms of conditions, and variants of uncertain significance were not reported.

Return of results and clinical management

Results were designated either low-chance, where no reportable variants were identified, or high-chance. For low-chance results, participants were informed via email and text message of result availability on Genetics Adviser. Genetic counseling was available on request. For high-chance results, parents were contacted by a genetic counselor to discuss the result, and to arrange a clinical genetics appointment, followed by additional testing (including confirmation and segregation of reported variants), and referral to specialist services as needed.

Participant surveys

Participating birth parents provided pregnancy and demographic information (T1) and completed a research survey (T2) as part of enrollment, with a further optional survey 3 months postresult (T3). All participant survey and clinical data was captured in a Research Electronic Data Capture (REDCap version 12.5.16) database. Pregnancy and demographic variables were compared to age- and sex-matched Australian Government Census data, or the Australian Institute of Health and Welfare’s National Perinatal Data Collection. For participants who consented to gNBS, study-specific multiple-choice survey items were used to measure the difficulty of decision-making; reasons for choosing to have gNBS; acceptability; attitudes toward gNBS and preferences for having gNBS for future children. Trait anxiety was measured at enrollment and 3 months postresult using the short form 6-item State-Trait Anxiety Inventory41. The Decision Regret Scale42 was administered 3 months postresult. For participants who chose not to have gNBS, reasons for declining were identified at enrollment using a study-specific matrix survey comprising 11 options rated on a 5-point Likert scale. Item logic was used in the data collection platform to administer the GOS in the T3 survey to participants with an infant who received a high-chance result. The GOS is a patient reported outcome measure for clinical genetics services43. The GOS consists of 6 items with 5-point Likert scale responses options ranging from strongly disagree to strongly agree.

Participant interviews

A random sample of participants who accepted gNBS were invited to take part in interviews before receiving their results and/or approximately 3 months after receiving a low-chance result. These participants were purposively sampled to ensure that participants were spread across age groups, study sites and ancestry. We also invited all participants who declined gNBS or received a high-chance result.

Semistructured interviews were conducted via Zoom or by phone and were audio recorded for transcription. An external transcription company transcribed all audio recordings verbatim, the researcher conducting the interviews then removed identifying details from the transcripts and assigned pseudonyms to all interviewees. Written notes were used for two interviews due to technical issues associated with those recordings.

Analysis

Statistical analysis of participant survey, recruitment and laboratory data were done using Stata version 18.0 and Microsoft Excel version 16.0. Chi-square tests were run to examine the differences between the proportion of BabyScreen+ participants and the relevant Victorian datasets across five demographic variables (age, parity, geographical location, education and socioeconomic advantage and disadvantage). We explored the differences between acceptor and decliner participants using binary logistic regression. TAT for samples that required recollection were measured from receipt of the fresh sample. For all qualitative data, NVivo version 12.0 was used to facilitate content analysis.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.