Lessons learned in implementing the Low Birthweight Infant Feeding Exploration study: A large, multi‐site observational study

Globally, early and optimal feeding practices and strategies for small and vulnerable infants are limited. We aim to share the challenges faced and implementation lessons learned from a complex, mixed methods research study on infant feeding.


| I N TRODUC T ION
Low birthweight (LBW; <2.50 kg) in newborns has multifactorial causes and can result from preterm delivery or restricted intrauterine growth. 1,24][5] LBW infants may also experience subsequent health issues such as poor growth, higher incidence of non-communicable disease in adulthood (e.g.hypertension, stroke, diabetes and hypercholesterolaemia) and long-term neurological problems. 6he vast majority (>90%) of LBW infants are born in low-or middle-income countries (LMIC), whereas most of the research designed to improve their health has been conducted in high-income settings. 7,8Further, significant variation in context, health systems and processes of care exist between high-income and LMIC that must be considered.Globally, there is limited knowledge around early feeding practices and optimal feeding strategies for LBW infants, a group of infants who are more likely to experience feeding difficulties than their normal birthweight peers.
The Low Birthweight Infant Feeding Exploration (LIFE) study aimed to investigate current feeding practices and growth patterns among moderately LBW infants (MLBW; 1.50-2.49kg) in resource-limited settings to inform potential feeding interventions.The primary study had four objectives: (1) to understand feeding practices and the standard of care underpinning them; (2) to define and document key longitudinal growth and health outcomes up to 12 months of age; (3) to examine predictors of growth faltering; and (4) to explore the beliefs, facilitators and barriers around the feeding of LBW infants. 9Here, we aim to highlight the complexities and challenges involved in conducting a large multi-country feeding study among vulnerable infants and share lessons learned during the implementation process.We believe these insights will be valuable to researchers and policymakers alike.

| M ET HODS
The LIFE study, a formative, multi-site, observational cohort study, used a mixed methods approach and was implemented in four sites (India-Karnataka, India-Odisha, Malawi and Tanzania) across three countries. 9The intention of the LIFE study was to establish the necessary foundational knowledge regarding MLBW infant feeding, growth, and health outcomes using a rigorous exploratory approach needed to design appropriate and effective interventions that would make an impact on the health and survival of these vulnerable infants.Across the four sites, we enrolled and followed up MLBW infants as well as reviewed patient charts in 12 tertiary/secondary, public/private study facilities.We followed infants after discharge via home visits or routine facility visits that aligned with the timing of follow-up visits.

| Quantitative
Using quantitative research methods, we assessed infant feeding and care by: (1) examining health facility inputs, including standard guidelines, staffing levels and equipment/ supply availability, in the study facilities prior to study initiation using a facility needs assessment; (2) assessing facility documentation of LBW infant feeding practices and growth monitoring during facility admission using chart reviews; (3) enrolling and following a small cohort of MLBW infants from birth to facility discharge using intensive and frequent observations, anthropometric measurements and maternal recall; and (4) enrolling a large cohort of MLBW infants at birth and prospectively following them for 12 months using anthropometric measurements, observations and maternal recall.
Data collector training to conduct chart reviews focused on techniques for reliable, accurate abstraction of data from medical records.Training for in-facility observations focused on effective observer techniques and consistency in data collection to ensure interobserver reliability.Observations were carried out by trained staff using a structured instrument.Observation started within 6 h after birth and occurred every 3 h for the first 7 days, twice daily for days 8-14, and then once a day for unstable infants or every 3 days for stable infants until discharge.Initial and refresher training for prospective cohort data collectors focused on survey administration techniques, use of digital application-based data collection tools (where applicable) and proper anthropometric measurement techniques.The prospective cohort was enrolled within 72 h of birth and follow-up study visits were conducted at 1, 2, 4, 6, 10, 14, 18, 26, 32, 39, 45 and 52 weeks of chronological age; 32-and 45week follow-up phone calls were only conducted in Malawi and Tanzania. 9articipants residing within a 50-km radius around the study facility were enrolled to minimise attrition and achieve high rates of follow-up.The expected visit schedule was generated using date of birth for the prospective cohort, maintained in the facility, and was used by research nurses to plan follow-up visits.Study participants were contacted a day prior to their visit to ensure availability.Additional details regarding quantitative methods can be found in the protocol paper. 9

| Qualitative
Qualitative study components included focus group discussions (FGDs) and in-depth interviews (IDIs) of clinicians, mothers and family members of LBW infants, and key stakeholders (e.g.those with expertise/knowledge in supply chain and donor human milk banks) to understand beliefs, facilitators and barriers around the feeding of LBW infants.Each respondent was interviewed at home or in the health facility by trained qualitative researchers using a semi-structured interview guide.Purposive sampling was used.Additional details regarding qualitative methods can be found in the protocol paper. 9

| Gathering lessons learned
Over the course of the 4-year LIFE study consortium (2018-2022), comprising 14 institutions and more than 50 researchers, we gathered insights about study conduct and collaboration/partnership through feedback sessions, publication committee meetings and study progress monitoring.In this manuscript, we share lessons learned (including successes and challenges) related to study setting, implementation, partnership strategies and opportunities for improvement.

| R E SU LTS
Across the LIFE study, operationalisation of the study required consideration of contextual realities, engagement of facility leadership and staff, modification of study protocols, and solution-oriented approach to address challenges.Primary study results were recently published and additional manuscripts are forthcoming 10 ; here, we focus on data and experiences that affected study implementation.

| Context: facility assessment
Care for MLBW infants occurred in the postnatal award, sick newborn care units and neonatal intensive care units (NICU).The size of each hospital's NICU ranged from 9 to 126 beds.Across the 12 facilities, the majority of care was provided by nurses with paediatrician and neonatologist support.Daytime nurse-to-infant ratios in the NICU varied from 1:3 up to 1:22; physician-to-infant ratios spanned 1:3 to 1:35.
We observed significant variability across the 12 study facilities with respect to infant feeding guidelines, supplies and support.All 12 facilities provided lactation teaching for new mothers; however, topics covered and the extent of the teaching varied across facilities.Some facilities had standard reminders related to breastfeeding, whereas others had less structured and less frequent lactation teaching.
Six of the 12 facilities (50%) had written policies/procedures regarding the preparation of expressed breastmilk for infant feeding as well as the cleaning, drying and storage of implements used for milk expression, feed preparation and infant feeding.About half of the facilities surveyed had a designated location for mothers to express milk.A few facilities without a dedicated location noted that mothers express milk while sitting on their beds in the postpartum wards.Most maternal milk expression was expected to be done by hand.Breast pumps were available at two facilities (both in India); facilities did not provide pump parts, so women generally needed to bring their own.In Malawi and India, babies in the postnatal or Kangaroo Mother Care ward were encouraged to feed on demand.However, across all facilities, maternal-infant separation was common, particularly if infants were admitted to a special newborn care unit (SNCU) or NICU.In Malawi, NICUadmitted babies were fed every 2 h when mothers came to breastfeed, both day and night.In India, feeding intervals varied for infants in the various hospital NICUs/SNCU.Some hospitals conducted feedings every 2 h, others every 3 h.In Tanzania, the feeding schedule varied between every 2 h and every 3 h.
Of the 12 study facilities, only one (in India) had a Donor Human Milk (DHM) bank.Term formula was available in six of 12 facilities; preterm formula was available at two facilities.Only powdered formula was available; premixed formula was not an option at any of the facilities.

| Documentation of care: chart review
From March to October 2019, 603 infant charts were retrospectively examined and data were extracted using a structured survey capturing information on maternal demographics, infant feeding, growth, morbidity and documentation by research nurses.Overall, key infant and maternal demographics were well documented in the charts; however, anthropometric parameters, such as head circumference and length at birth and daily weights, were not well documented (Table 1).
Detailed documentation, critical for clinical decisionmaking and patient hand-off between clinicians and facilities, was poor, particularly concerning feeding.The time to initiate feeding, a key indicator for the WHO, was also poorly documented across all sites.Even among infants who were admitted to the NICU/SNCU for closer monitoring, feeding information was not well documented.

| Current care: in-facility observational cohort
In the in-facility cohort conducted from August 2019 to April 2020, we enrolled 148 MLBW infants for in-facility observation and followed them during their hospital stay in the postnatal unit or NICU.We completed 1076 contacts/observations across the 148 infants.Infants were typically in the facility for 3.1 days (median, IQR: 1.5, 5.7) (Figure 1).To be discharged, infants had to meet certain criteria set by each facility concerning their feeding, stability and weight.Of the 140 infants assessed at discharge, 97.1% were fed only breastmilk (69.3% direct, 7.1% expressed only, 20.7% both direct and expressed).However, 18.8% of infants were discharged with weights below the site-specified criteria: 20.6% (7/34) in India-Karnataka and 5.7% (2/35) in India-Odisha were <1.80 kg; 40.0%(16/40) in Tanzania <2.00 kg; and 3.5% (1/29) in Malawi <1.50 kg.

| Health and growth outcomes during infancy: prospective observational cohort
From April 2019 through March 2020, we screened 2152 infants within 72 h of birth for study inclusion.A cohort of 1114 moderately LBW infants were enrolled at birth and followed through 12 months of age.Given that enrolment was occurring at the time of birth and not during the antenatal period, we estimated gestational age based on the best obstetric estimate, prioritising early (first or second trimester) ultrasound or last menstrual period.In this cohort, only 4% had dating based on ultrasound, 65% on last menstrual period and 15% on fundal height.The use of Dubowitz scoring to assess gestational age at the time of birth was acceptable to providers and had high agreement during training between trainers and trainees. 11However, after implementation of the Dubowitz exam and best obstetric estimate algorithm, only 1.4% of gestational age determination was based on the Dubowitz score.This experience highlighted key challenges in enrolling a cohort of low birthweight infants with limitations in accurate and standardised estimation practices and documentation of gestational age during pregnancy.
With a strong community engagement strategy and dedicated staff to follow families at home or in the facility, follow-up rates in the study were high, with only 10.1% (n = 1002) lost to follow-up (Table 2).Study visit completion for all 11 visits was 76% despite enrolment and follow-up occurring during the COVID-19 pandemic.During the initial stages of the pandemic, we stopped enrollment in all sites for a period of time and established safety protocols to protect data collectors, clinicians, study participants and their families.

| DISCUS SION
Since 2019, the LIFE study has been completed with high rates of follow-up and has advanced our understanding of  75-100% missing 51-75% missing 26-50% missing 0-25% missing MLBW infant feeding and care. 9,10With our results in mind, we share four key lessons learned during the implementation of this multi-country study around working with vulnerable populations, pursuing multi-country studies while accounting for disruptions, ensuring that the evidence-base for trials/interventions is solid, and considering which methods are needed to solicit the strongest evidence for impact.

| Lesson #1: Enrolment and follow-up of vulnerable populations requires additional effort from researchers and the community
In the LIFE study, we focused on enrolling MLBW infants and their mothers, who require additional care compared with their normal birthweight counterparts.As expected and despite no intervention being introduced, there was some reluctance among parents to involve their children in a research study due to the potential risks around privacy, confidentiality and repeated assessments focused on growth and development.We were able to overcome this challenge successfully by putting a series of practical measures in place.First, we focused on engaging community and health facility authorities to obtain approval for study conduct and buy-in for longer term engagement.Through a series of stakeholder meetings, research teams shared study aims, objectives and relevance to this vulnerable population, local facilities and communities.In the hospital postnatal wards and SNCUs/NICUs, postgraduates and other staff were provided with an overview training on the LIFE study.Study staff counselled and engaged the family members with the local health official authorisation throughout the study to sustain their participation.For example, in Tanzania, study staff worked closely through longstanding partnerships with health facility staff and their current outreach programmes, including partnering with social workers who were familiar with the community.This raised study awareness as well as increased acceptance of study participation.Secondly, we focused on hiring study staff (i.e. research nurses) with experience working with and physically handling these vulnerable infants, familiarity with the health facilities, and comfort with the conduct of research.
Thirdly, across the 12 facilities, our study staff tried to efficiently use the ongoing systems of the hospital and work closely with hospital staff and community to identify potential study participants.The Malawi study site worked with community members via a community advisory board (CAB).Community educators at the study site had regular meetings with CAB members prior to and during conduct of research.These CAB members were key in advising the team on culturally appropriate study procedures.CAB members were also critical in providing accurate information about the study to the communities.
Finally, we instituted a safety net to identify and refer to the vulnerable infants and their mothers who experienced severe illness or displayed danger signs requiring referral.

| Lesson #2: Exploratory studies are essential for proper planning of targeted large-scale interventions
In the original development of this consortium and fundraising process, the co-investigators intended to conduct a large-scale randomised controlled trial around infant feeding, focused on donor human milk and provision of fortification, when needed.In 2018, it quickly became apparent that the data on guidelines, standard practices and interventions for these MLBW infants across and within countries were limited.At the time of study initiation, the most recent guidelines from the World Health Organization (WHO) focused on LBW infant feeding were from 2011, with more than 70% of those guidelines based on poor or very poor quality evidence. 12Although new guidelines were released in 2022, focused on care for preterm and LBW infants, guidance is still limited, particularly concerning feeding and nutrition for this vulnerable group. 13cross study sites, there was a lack of consistency in policies and standard protocols for LBW infants, particularly by location of care (i.e. in and outside of the SNCU/NICU).In India, guidelines formulated by national professional societies, such as the National Neonatology Forum of India, were available but not uniformly followed.In Tanzania, specific policies and protocols existed for LBW infants but varied by facility.Regardless of the policies formally in place, they were not always followed due to resource gaps and overcrowding.In the in-facility observational cohort, nearly one in five infants was discharged with a weight lower than facility guidance.
To address this lack of information, guidelines and concrete evidence, we developed the LIFE study to answer critical questions around care of MLBW infants and identify potential intervention points.Using the mixed methods approach, we were able to address the existing gaps and identification of problems with different stakeholder perspectives.For example, we held discussions with mothers, grandmothers and traditional leaders to learn about barriers and perceptions of LBW feeding practices in the community.The chart review component of our study demonstrated incomplete documentation of the management of the newborns even though essential newborn care was provided.Being able to promote and support optimal feeding may be more difficult without the proper documentation and handover between clinicians.In addition, clinical decision-making concerning feeding support, length of stay, discharge and risk management would improve with better documentation.Potential explanations for these gaps included heavy workloads and verbal rather than documented transfer of information or documentation elsewhere.As noted by English and colleagues, health worker shortages and heavy workload can undermine high quality and compassionate care. 14ulti-site trials must take into account these variations in adherence to protocols and procedures when examining outcomes; if not considered, variation may distort study outcomes.Through this formative work, we were able to identify some early points of intervention that would have been missed if we had dived right into a formal randomised controlled trial.The qualitative component of the study provided an explanation for various practices and the key influencers in the community and facility setting.

| Lesson #3: A mixed methods approach adds value to the study rigour through triangulation of data and ability to tell the whole story
Often, research studies involve only one type of methodology, with either quantitative or qualitative approaches prioritised.Here, we used a mixture of research strategies, including direct observations, chart abstraction, selfreported surveys, interviews in groups or individually, anthropometric assessments, feeding diaries and reporting.Our work and that of other researchers has noted that while more complex and requiring additional expertise, the use of mixed methods has proven beneficial and has added meaning to our findings. 15,16or the most part, we have noted similar results and key findings across the study methods.This has allowed us to cohesively fit all pieces of the moderately low birthweight infant feeding experience into a single set of outputs.The variety of methods complemented each other; the quantitative data provided an overview of infant care, while the direct observation or interviews provided nuance.7,18 4) co-presentation of research findings locally and globally.Although these practices may seem obvious, there are still plenty of reports of parachute science, 19 poor collaborative practices and limited engagement with data collectors and clinicians at the frontlines of care.As previously mentioned, we completely modified the original study questions from a randomised trial of a feeding intervention to a broad landscape of moderately low birthweight infant feeding.That decision to change the research approach and design emerged in a 2-day meeting held in Boston, USA, where investigators from all partners/institutions attended and presented their opinions.Those collaborative meetings for study design and development were only achieved through mutual respect, trust, partnership and transparency, aligned with the four principles recently highlighted by Hodson and colleagues on equitable partnerships. 20econdly, through protocol development and training curricula, we were able to ensure that consistent approaches were being used across the study.In July 2019, we conducted standardised training across all partners focused on ethics and data collection with didactic sessions including role play.Additionally, a specialised training was focused on Dubowitz scoring for gestational age assessment at birth. 11inally, through the creation of a Publications Committee with an amalgamation of ICMJE and institutional rules, we were able to create a venue for Principal Investigators and each partner to put forth ideas for manuscript development, engage on manuscript writing, agree on authorship and writing timelines, and hold each other accountable to study goals and outputs.The Publications Committee met every 6-8 weeks with representation from the Principal Investigators in India, Malawi, Tanzania and the USA.Using a Publications Committee charter, which outlined agreedupon procedures, we had a standardised approach for idea submission, authorship order assessment and dispute resolution.We resolved any concerns or disagreements around authorship through discussion aiming to reach consensus.We intentionally aimed to provide opportunities for early career researchers to lead papers and partner with more experienced authors.Over the past 18 months in analysis, data interpretation and results findings, our consortium has been dedicated to highlighting each partner's strengths, contributions and impact on the LIFE study findings.Through broad dissemination of this exploratory work, we believe the science around vulnerable infant feeding can be strengthened.

| CONCLUSION
Research focused on vulnerable newborns is critical given the scale of the challenge (with more than 20 million infants each year), their excessive burden of morbidity and mortality.Fortunately, there is recent interest and investment from donors and policymakers.The associated poor outcomes for these infants and their families must be mitigated in effective and efficient ways utilising the health system and community engagement to improve their well-being.Multi-site partnerships in global health research, which require active and equal engagement, are instrumental in avoiding duplication and building a stronger, generalisable evidence base.We believe that the experience of the LIFE study and consortium, made up of 14 international partner organisations and more than 50 individuals, has provided helpful insights for consortium development and study implementation that should inform future research design, planning and execution.

AU T HOR C ON T R I BU T ION S
Study co-PIs (KEAS, TM, LV, KM, CRS) developed the concept for the paper.SSV, SS and KM drafted the first version of the paper; all co-authors reviewed the first and second drafts of the paper and provided critical input.All co-authors reviewed and agreed with the final version of the paper.

AC K NO W L E D GE M E N T S
We would like to thank clinical leadership and staff at all study facilities for their partnership, support and contribution to this work; the mothers and infants for allowing us to have a glimpse into their experiences and sharing key moments of their lives; and all data collectors and study staff for conducting study activities.

F U N DI NG I N FOR M AT ION
This publication is based on research funded in part by the Bill & Melinda Gates Foundation (BMGF).The findings and conclusions contained within are those of the authors and do not necessarily reflect positions or policies of the Bill & Melinda Gates Foundation (INV-007326).The funder reviewed the overall study design but did not have any input on the decision to publish.

E T H IC S S TAT E M E N T
This protocol was approved by 11 ethics committees in four countries: India, Malawi, Tanzania and the USA; more information can be found in the published protocol.These approvals apply to mother-infant dyads recruited from all 12 study facilities.The study is registered with clini caltr ials.gov(NCT04002908) and CTRI/2019/02/017475 (Clinical Trial Registry of India-http://ctri.nic.in).Maternal or surrogate written consent was required to participate in the study.

PAT I E N T A N D PU BL IC I N VOLV E M E N T
As part of the study design, the LIFE team involved clinicians, researchers and community stakeholders familiar with the respective settings and populations.Study tools were piloted with patients and community members to ensure that research questions and indicators were culturally appropriate, acceptable and relevant to the study population.

T A B L E 1
Heatmap of incomplete documentation in charts of moderately low birthweight infants in Tanzania, Malawi, India-Karnataka and India-Odisha.

F I G U R E 1
Length of facility stay for 148 moderately low birthweight infants in India, Malawi and Tanzania.T A B L E 2 Prospective cohort enrolment and follow-up rates for moderately low birthweight infants in Tanzania, Malawi, India-Karnataka and India-Odisha.

4. 4 |
Lesson #4: A multi-site consortium can build evidence in a more efficient, generalisable way by avoiding duplication and building on others' expertise There is an oft-quoted African proverb: 'If you want to go fast, go alone; if you want to go far, go together'.Collaborative research involves time, commitment and resources to promote equitable partnership, collaboration and development of generalisable learning for the world.A few key examples of this consortium's strategies in partnership and joint learning focused on: (1) collaborative meetings for study design and development; (2) standardised training; (3) a publications committee; and ( All authors completed the ICMJE conflict of interest form and were funded by the Bill & Melinda Gates Foundation for this work as part of the LIFE study.Co-authors (KEAS, LV, DET, CRS, KM) receive additional funding from BMGF beyond the LIFE study and have received funding from BMGF for maternal and newborn health and nutrition work at large.All other authors have declared no conflicts of interest.DATA AVA I L A BI L I T Y S TAT E M E N T Data related to the study and manuscript are available in a public, open access repository.Deidentified individual participant data are available through the Harvard Dataverse platform under the BetterBirth Dataverse website.This can be found at: https://datav erse.harvard.edu/datav erse/Bette rBirt hData.