Specificity Matters: Unpacking Impact Pathways of Individual Interventions within Bundled Packages Helps Interpret the Limited Impacts of a Maternal Nutrition Intervention in India

ABSTRACT Background To address gaps in coverage and quality of nutrition services, Alive & Thrive (A&T) strengthened the delivery of maternal nutrition interventions through government antenatal care (ANC) services in Uttar Pradesh, India. The impact evaluation of the A&T interventions compared intensive ANC (I-ANC) with standard ANC (S-ANC) areas and found modest impacts on micronutrient supplementation, dietary diversity, and weight-gain monitoring. Objectives This study examined intervention-specific program impact pathways (PIPs) and identified reasons for limited impacts of the A&T maternal nutrition intervention package. Methods We used mixed methods: frontline worker (FLW) surveys (n = ∼500), counseling observations (n = 407), and qualitative in-depth interviews with FLWs, supervisors, and block-level staff (n = 59). We assessed 7 PIP domains: training and materials, knowledge, supportive supervision, supply chains, data use, service delivery, and counseling. Results Exposure to training improved in both I-ANC and S-ANC areas with more job aids used in I-ANC compared with S-ANC (90% compared with 70%), but gaps remained for training content and refresher trainings. FLWs’ knowledge improvement was higher in I-ANC than S-ANC (22–36 percentage points), but knowledge of micronutrient supplement benefits and recommended foods was insufficient (<50%). Most FLWs received supervision (>90%), but supportive supervision was limited by staff vacancies and competing work priorities. Supplies of iron–folic acid and calcium supplements were low in both areas (30–50% stock-outs). Use of monitoring data during review meetings was higher in I-ANC than S-ANC (52% compared with 36%) but was constrained by time, understanding, and data quality. Service provision improved in both I-ANC and S-ANC areas, but counseling on supplement benefits and weight-gain monitoring was low (30–40%). Conclusions Systems-strengthening efforts improved maternal nutrition interventions in ANC, but gaps remained. Taking an intervention-specific perspective to the PIP analysis in this package of services was critical to understand how common and specific barriers influenced overall program impact.


Introduction
Maternal undernutrition is a longstanding and significant public health concern (1) contributing to maternal mortality, fetal growth restriction, neonatal deaths, and child undernutrition (2). Nearly 3.1 million annual child deaths and 35% of the disease burden among children <5 y are attributable to maternal undernutrition globally (2,3). These consequences are exacerbated in low-and middle-income countries (LMICs), particularly South Asia, which accounts for a third of global undernutrition (3). In India, maternal and child undernutrition is a predominant risk factor for 68% of under-5 child deaths (4), with a third of women having low BMI, more than half of pregnant and non-pregnant women having anemia, and 19% of babies born with low birth weight, of whom 4.4 million have intrauterine growth retardation (5). Improving maternal C The Author(s) 2021. Published by Oxford University Press on behalf of the American Society for Nutrition. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited. nutrition is critical for survival, health, human development, and protecting future generations (1,6).
High coverage and quality of nutrition-specific interventions through antenatal care (ANC) platforms are essential for improving maternal nutrition. A recent study in 81 LMICs estimated a reduction of 28% in maternal and neonatal deaths and 22% fewer stillbirths if evidence-based nutrition interventions are delivered to mothers already seeking care (7). Although global momentum on the importance of nutrition interventions has increased, the coverage of most interventions falls below the reach of health services through which they are delivered (8). There are concerns especially about the effectiveness and quality of maternal nutrition interventions, including counseling and micronutrient supplementation in LMICs and South Asia (9)(10)(11).
To address the high burden of maternal undernutrition (5) and critical opportunity gaps between delivery platforms and corresponding nutrition interventions (12), the Government of India prioritized provision of essential nutrition interventions through the ANC platform (13,14). Maternal nutrition interventions in India are delivered through 2 national programs: the Integrated Child Development Services (ICDS) managed by the Ministry of Women and Child Development and the Reproductive Maternal Newborn Child and Adolescent Health program under National Health Mission by the Ministry of Health and Family Welfare. Despite programmatic efforts, coverage and quality of maternal nutrition interventions remain suboptimal, and knowledge gaps exist on how to strengthen the existing ANC platform.
Renewed commitment to and insights from implementation research on interventions targeting the first 1000 days are urgent priorities for progress to be made in improving maternal and child undernutrition (15). Alive & Thrive (A&T) aimed to strengthen delivery of interpersonal counseling, micronutrient supplements, and community mobilization by integrating a package of maternal nutrition interventions within the government ANC platform in Uttar Pradesh (16). The impact evaluation of the A&T maternal nutrition interventions (2017-2019) found substantial secular improvements in service coverage but limited intervention impacts on micronutrient supplementation, diet diversity, and gestational weight gain (8-12 percentage point improvement) (16). Beyond traditional evaluation methods that conducted impact evaluations using only quantitative methods (17), additional methods to assess how or why impacts are achieved or not achieved Author disclosures: SG, TF, JE-A, and PKS were members of the program implementation team that designed and implemented the interventions studied and reported on in this manuscript. They reviewed the manuscript and provided contextual interpretation of the results, but final decisions for manuscript content lay with the primary authors from the evaluation team (SK, PHN, LMT, EAF, and PM). All other authors report no conflicts of interest. Supplemental Tables 1-5 are available from the "Supplementary data" link in the online posting of the article and from the same link in the online table of contents at https://academic.oup.com/jn/. are emerging (18). Previous studies have used a program impact pathway (PIP) analysis to examine program fidelity and identify critical steps in the implementation of behavior change communication strategies to improve infant and young child feeding in Bangladesh (19), Ethiopia (20), and Vietnam (21). To our knowledge, no study, however, has conducted a PIP analysis of multiple maternal nutrition interventions, including micronutrient supplementation, weight-gain monitoring, and diet counseling bundled into a package. In this implementationscience study, we examined intervention-specific impact pathways to identify reasons for limited impacts of the A&T overall package of interventions on maternal nutrition practices.

Study context and intervention description
This study used data from the impact evaluation of A&T's interventions that aimed to strengthen maternal nutrition services within the existing ANC government program in 2 districts of Uttar Pradesh (Unnao and Kanpur-Dehat) over 18 months from 2017 to 2019. The interventions included provision and counseling on iron-folic acid and calcium supplements, gestational weight-gain monitoring, interpersonal counseling on weight gain and diet during pregnancy, and community mobilization (22).
There were 2 intervention arms: 1) intensive ANC (I-ANC) areas that received the above interventions in addition to standard government services and 2) standard ANC (S-ANC) areas that received standard government services. A&T worked with 3 types of government frontline workers (FLWs): Anganwadi Workers (AWWs), Accredited Social Health Activists (ASHAs), and Auxiliary Nurse Midwives (ANMs); each of them has different tasks for providing maternal nutrition counseling to women through the course of their pregnancy (22). ANMs had a minimum of 4 contacts with women during the ANC checkup, with 1 contact each during first and second trimesters and 2 contacts during the third trimester. AWWs and ASHAs visited pregnant women at home for counseling. AWWs conducted a minimum of 3 visits (i.e., 1 visit during the second trimester and 2 visits during the third trimester), whereas ASHAs conducted a minimum of 4 visits (i.e., 1 visit each during the first and second trimesters and 2 visits during the third trimester).
Descriptions of the interventions, evaluation design, and main impacts on maternal nutrition practices have been published (16). Briefly, the impact evaluation design used a pair-matched clustered randomized design using block as the unit of randomization. Thirteen pairs were formed of 26 blocks from the 2 districts based on several demographic, infrastructure, and amenity characteristics from census 2011 data. A propensity score-matching method was used to identify pairs of intervention and control blocks. One block from each pair was randomly assigned to I-ANC (13 blocks) or S-ANC (13 blocks) through manual lottery.

Program impact pathways
The PIP was developed to map the mechanisms through which the A&T interventions were expected to achieve impact. The PIP was developed through an iterative process: review of program documents, close discussions with A&T staff, and multiple revisions based on feedback. The final intervention-specific PIP is presented in Figure 1, and details of program interventions and expectations along the PIP are in Table 1. For this article, we studied 7 domains related to improving maternal nutrition service delivery. staff. A partially mixed sequential dominant status design was used with an emphasis on complementary quantitative and qualitative methods (23,24). The quantitative data were collected first, followed by qualitative data collection, to inform quantitative findings. Data collection tools were carefully designed around the specific needs of each intervention in the bundled package.

Frontline worker and supervisor surveys
The FLW survey was conducted with 497 FLWs in 2017 (n = 253 in I-ANC and 245 in S-ANC) prior to the implementation of the intervention and with 459 FLWs (n = 231 in I-ANC and 248 in S-ANC) in 2019 (after 2 y of intervention) using a structured questionnaire.
FLWs were asked about receipt of training, job aid materials, and supportive supervision in the previous 6 months. Knowledge of micronutrient supplements was assessed using questions related to quantity, frequency, and benefits of supplements. Knowledge of diet diversity was assessed by asking FLWs how a pregnant woman should eat to provide good nutrition to her baby, including number of food groups and different recommended foods during pregnancy. Knowledge of weight gain was assessed based on FLWs' correct response to how much weight a woman should gain during pregnancy (10-12 kg). For service delivery, FLWs were asked about frequency of service provision and counseling content related to maternal nutrition during village health, sanitation, and nutrition days (VHSNDs) and home visits to pregnant women. Supervisor interviews (n = 103 in I-ANC and 106 in S-ANC in 2019) focused on use of data to track progress of interventions and identifying challenges in using data.

Counseling session observations
Observations of counseling sessions were used to assess the extent to which FLWs adhered to the standards of care and to evaluate their performance. A counseling observation checklist was developed based on the guidelines for the maternal service interventions, including key counseling messages related to iron-folic acid and calcium supplements, diet diversity, weight-gain monitoring, and counseling skills. In total, 405 FLWs (n = 195 in I-ANC and 210 in S-ANC) were observed during counseling sessions conducted at home visits or VHSNDs.

Qualitative in-depth interviews
Qualitative in-depth interviews were conducted with FLWs (n = 24), supervisors (n = 12), block-level staff (n = 20), and A&T program staff (n = 3) in 4 blocks belonging to I-ANC areas using a semistructured interview guide. Interviews were focused on perspectives about and experiences of program implementation, specifically, experiences with trainings and supervision, use of data during review meetings, supply chain of iron-folic acid and calcium supplements, and challenges faced in service delivery and counseling.
All study tools were developed in English and translated into Hindi. The tools were pretested locally for cultural relevance and understanding and then modified accordingly. The surveys and in-depth interviews were conducted in Hindi by fluent enumerators, supervised by field managers and researchers. Enumerators were recruited locally and trained on technical content as well as security and confidentiality issues by mixed methods (lecture, role-play, mock interview, and practice) in a classroom and field settings. Field supervisors received additional training related to quality control processes; cross-checking, editing, and coding of the questions; and security and confidentiality issues. Qualitative interviews were recorded, and field notes were taken by researchers. Enumerators conducted interviews in a separate area with only the respondent present to ensure privacy and confidentiality. Interviews took approximately 45 min.   Differences between I-ANC and S-ANC areas at baseline and endline were estimated using linear (for continuous variables) or logistic regression models (for categorical variables). A difference-in-differences (DID) analysis was conducted to estimate differences in changes over time between the 2 study groups (25). All models adjusted for clustering at the village level using a robust sandwich estimator.
Qualitative audio recordings were transcribed and independently translated back into English and compared with the original as a quality check. Interviews were transcribed by 2 local research assistants and stored on a password-protected computer to ensure privacy. The researchers randomly selected 10% of the transcripts for comparison with the audio files for quality control. We used MaxQDA and Microsoft Excel to organize interview transcripts and to manage data for identifying and summarizing inductive and deductive themes that corresponded to each PIP domain. Thematic analysis of the qualitative data was done (26). Memos were created to keep track of thought processes, categories, and themes and to develop codes. Inductive codes were created based on interview transcripts, field notes, and memos. Deductive codes were created based on themes addressed by questions in each interview guide. Inductive and deductive codes were consolidated into separate codebooks based on respondent group. Data were analyzed by 1 team member fluent in both Hindi and English, and codes and analysis were discussed and reviewed by the principal investigator throughout. Additional information on qualitative data analysis and tools has been reported previously (27).
Successes and challenges along PIP domains were identified through triangulation of quantitative (to obtain results on magnitude and direction of changes in PIP domains) and qualitative data (to identify the emerging themes on how the implementation of PIP domains occurred). For example, for the training domain, we examined the proportion of FLWs in I-ANC and S-ANC areas who were trained in the past year as well as the percentage trained on maternal nutrition topics. We further examined data from the in-depth interviews to inform the trends observed with FLWs' perceptions and feedback on training received. We analyzed specific successes and challenges for different PIP domains relevant to each individual intervention. Triangulation allows development of a comprehensive understanding of evidence-based pathways and maximizing trustworthiness through the convergence of information from different data sources (28).

Ethical clearance
This study was approved by the ethical review board from the International Food Policy Research Institute, Emory University, and the Suraksha Independent Ethics Committee in India. Verbal informed consent was obtained from all participants at all stages of data collection. The study was registered at clinicaltrials.gov as NCT03378141.

Sample characteristics
FLW characteristics were similar in I-ANC and S-ANC areas (Supplemental Table 1). Among 3 types of FLWs, ASHAs were younger (37 y) than AWWs and ANMs (43-44 y). Nearly half of ASHAs had lower than secondary school education, whereas most AWWs and ANMs completed secondary school or higher. Most FLWs identified as Hindus, and ∼75% belonged to disadvantaged socioeconomic groups, specifically scheduled caste/tribe.

Training and job aids
Coverage of training was high overall at endline, but key gaps were prevalent in exposure to maternal nutrition topics during trainings ( Table 2). Most AWWs and ASHAs in I-ANC and S-ANC areas received maternal nutrition training in the past year (>90%), whereas ANMs in I-ANC areas were more likely than those in S-ANC areas to receive training (79% compared with 46%) ( Table 3). Although coverage of training topics for iron-folic acid and calcium improved between baseline and endline, coverage was <50% and was similar for AWWs and ASHAs in both areas. ANMs in I-ANC areas were more likely than those in S-ANC areas to be trained on iron-folic acid and calcium (52% compared with 22%). Around 70% of AWWs and ASHAs received training on diet counseling, similar in both areas. A higher proportion of ANMs in I-ANC areas received diet training than those in S-ANC areas (60% compared with 30%). Training on weight gain was received by only about a third of FLWs, similar between areas for AWWs and ASHAs, and it was higher among ANMs in I-ANC areas compared with S-ANC areas (33% compared with 8%).
Qualitative interviews showed that FLWs in I-ANC areas had positive views of trainings and were able to better understand training content delivered through videos ( Table 2).
"They told everything very nicely. Actually, they told and showed things through video, so it was easy to understand. (ANM)" "We were benefitted from the training. Earlier, we did not have much information about so many things but after  Refresher trainings were not held regularly as planned during monthly meetings, however, as the focus was on administrative tasks. Nearly 30% of ANM positions were vacant, which limited coverage of training.
Overall, access to maternal nutrition materials was high in both areas, but use of materials was limited during contacts with pregnant women. FLWs in I-ANC areas were given a maternal nutrition calendar and flipcharts (98% for AWWs and ANMs and 92% for ASHAs) to aid them with counseling. FLWs in S-ANC areas also had access to job aids but at lower coverage (67% for AWWs, 58% for ASHAs, and 55% for ANMs) (Figure 2). More ASHAs in I-ANC areas used job aids during home visits (89% compared with 68%), and more ANMs used job aids on VHSNDs (93% compared with 72%) than their counterparts in S-ANC areas. Use of materials was lower among AWWs and similar between areas. FLWs in I-ANC areas had positive views of maternal nutrition materials and job aids, which they reported helped to provide more effective counseling.
"Initially we faced problem while making them understand but gradually they understand. We have provided them calendar and those calendars are hanging in their homes. So, if they wouldn't be able to understand orally then we explain to them through pictures of the calendar. (AWW)"

Strengthening supportive supervision for FLWs
Overall, receipt of supervision visits was high in I-ANC and S-ANC areas, but gaps were prevalent in key supportive supervision practices. Most FLWs (>90%) in both areas received supervisory visits, but only around half of FLWs reported being coached on maternal nutrition during visits and a third of AWWs and ASHAs being accompanied for home visits by their supervisors (Figure 3). ANMs in I-ANC areas were more likely to receive supervision through use of a checklist compared with those in S-ANC areas (78% compared with 62%). Qualitative interviews showed that FLWs in I-ANC areas valued supervision visits in helping them solve problems and increasing their accountability. Supportive supervision was limited, however, by vacancies among ICDS supervisory staff and supervisors being unable to conduct regular visits due to competing work priorities.
"It is good when the supervisor come for the visit, this is how we get a lot of information and can also clear our doubts, like when a supervisor comes for the visit twice a month then she check our work this is how we are also scared and do the work timely and also clear the doubts. (ASHA)"

FLW knowledge
Knowledge of consuming iron-folic acid and calcium supplements during pregnancy was similarly high in both areas at endline (70-80%), but gaps were prevalent on knowledge of supplement benefits. Knowledge of consuming iron-folic acid supplements during postpartum improved more among AWWs and ASHAs in I-ANC than in S-ANC areas [DID: 35 and 39 percentage points (pp), both P < 0.001] ( Table 4). Most FLWs knew about iron-folic acid benefits of reducing the risk of anemia (∼60-70%), but knowledge of other benefits was low to moderate (30-50%) in both areas. ASHAs in I-ANC areas had higher knowledge of iron-folic acid benefits to reduce risk of low birth weight than those in S-ANC areas (DID: 33 pp, P < 0.01). Improvement in knowledge of calcium supplements during postpartum was higher in I-ANC compared with S-ANC areas for AWWs and ASHAs (DID: 35 and 30 pp, P < 0.001 and P < 0.01, respectively) and was similar for ANMs in both areas (∼75%). Although most FLWs knew about calcium benefits for growth of children's bones, knowledge of other benefits was low to moderate (30-50%). AWWs and ASHAs in I-ANC compared with S-ANC areas had higher knowledge of the benefits of calcium to ensure growth of children's bones (DID: ∼29 pp, P < 0.01 and P < 0.05, respectively) and to reduce the risk of high blood pressure/fits (DID: 22 and 36 pp, P < 0.05 and P < 0.001, respectively).
FLWs' knowledge of diet diversity was moderate in I-ANC and S-ANC areas, with ∼50-60% knowing about consuming 5 Unpacking Program Impact Pathways for Maternal Nutrition 619 or more food groups during pregnancy; knowledge of animal source foods was lower in both areas. Knowledge of diet diversity was similar for AWWs and ASHAs in both areas and higher for ANMs in I-ANC compared with S-ANC areas (DID: 36 pp, P < 0.05). For specific foods, most FLWs (>70%) in both areas knew about consuming milk and green leafy vegetables, but knowledge of animal source foods was lower. AWWs and ANMs in I-ANC areas had higher knowledge than those in S-ANC areas about consuming eggs and fish/meat daily during pregnancy (DID: 22 pp and 27 pp, respectively, both P < 0.05).
Knowledge of weight gain improved over time in both areas, with higher improvements among all FLWs in I-ANC areas compared with those in S-ANC areas (DID: 23-25 pp, all P < 0.05). Most FLWs (>70%) in I-ANC areas had correct knowledge of recommended weight gain during pregnancy (10-12 kg), but this knowledge was lower among AWWs and ASHAs (∼70%) compared with ANMs (90%).

Strengthening of iron-folic acid and calcium supplementation supply chain
Major supply chain bottlenecks related to iron-folic acid and calcium supplements were observed in both I-ANC and S-ANC areas. Interventions related to supply chain were not restricted in I-ANC areas only but implemented at the state or district level and specifically related to forecasting and procurement of supplements. Similar changes were observed over time for FLWs in both areas for supplement provision. FLWs reported providing 50-60 iron-folic acid and calcium tablets a month to pregnant women, but nearly 30% and 50% of ANMs reported stock-outs of iron-folic acid and calcium supplements in the past 3 months, respectively. A&T program monitoring data showed inadequate stock in half of I-ANC blocks for iron-folic acid and in all I-ANC blocks for calcium supplements in the month preceding endline ( Table 2).
ASHAs described how they delivered supplements at home to pregnant women who did not attend ANC during VHSNDs, which was highlighted as a positive step in improving provision of iron-folic acid and calcium. Qualitative findings also emphasized logistic issues along the supply chain, including delays due to centralization of procurement, differing methods of demand estimation between government staff, and inadequate storage facilities.
"Both of the districts had a decent amount of iron-folic acid until the end of 2018, but the shift in procurement [from district to state] did start delaying iron-folic acid procurement. (A&T program staff)" "For most of the medicines, not just for iron-folic acid and calcium, the quarterly purchase cycle was because the storehouse in Unnao was smaller and could not accommodate a 6-month or annual cycle. (A&T program staff)" "From October onwards, we started to see VHSNDs without calcium . . . in Unnao the distribution was pretty haphazard . . . most of the blocks were out of calcium by January. (A&T program staff)"

Strategic use of data to track the progress of interventions and identify needed improvements
Data use was higher in I-ANC compared with S-ANC areas, but key gaps were prevalent in the strategic use of data during monthly review meetings in both areas. Use of data was higher in I-ANC than S-ANC areas during monthly meetings (66% compared with 50% among AWWs and 58% compared with 42% among supervisors), during cluster review meetings (52% compared with 36% among AWWs), and for decision-making purposes (23% compared with 15% among supervisors) ( Table 5). Block-level staff described using Health Management Information Systems data to monitor iron-folic acid and calcium stocks and estimate demand. Supervisors in I-ANC areas reported using supportive supervision checklist data to provide feedback to FLWs on maternal nutrition counseling, including gaps and areas of improvement. Collaboration between government staff from different departments to review data was also highlighted as a positive step. "We will sometimes look at the trends in ANC registration in the first trimester from the past 2-3 years to see how many pregnant women are in the village. That way we can estimate how much iron-folic acid and calcium we need in the block. Challenges in using data were reported by ∼10-25% FLWs and 5-15% supervisors in both areas, with key challenges including difficulty in understanding data, poor data quality, and lack of time. Qualitative data highlighted additional constraints in data use, including server issues, staff vacancies, lack of technology, and reporting burden. In addition, monthly review meetings mostly focused on administrative tasks, which limited time to discuss data.

FLW service delivery
Most service provision improved in both areas over time, but critical gaps remained in I-ANC and S-ANC areas on fully using service delivery platforms to provide iron-folic acid and calcium supplements, conduct food demonstration, and weigh pregnant women. Higher support to ANMs and mobilization by ASHAs for ANC checkups were observed in I-ANC compared with S-ANC areas (34 pp) ( Table 6). During monthly VHSNDs, ∼60% of AWWs and ASHAs and 75% of ANMs provided iron-folic acid supplements for pregnant women at endline, but calcium provision was much lower. All ASHAs and ANMs conducted monthly home visits to pregnant women, but less than a third provided iron-folic acid and calcium supplements during home visits.
Although ∼60% of AWWs and ASHAs provided nutrition counseling during home visits, less than a third conducted food demonstrations to help women understand which foods they should consume during pregnancy. These findings were similar between I-ANC and S-ANC areas.

FLW counseling
Despite improvements over time, critical gaps were prevalent for specific counseling on supplement benefits (∼30%), diet diversity (50-60%), and weight-gain monitoring (30-40%) in I-ANC and S-ANC areas. Findings from counseling observations showed that most FLWs advised women to take iron-folic acid regularly during pregnancy (60-75% among AWWs and ASHAs and >80% among ANMs) ( Table 7), but the percentage of FLWs who advised women to take calcium was lower (33-46% among AWWs and ASHAs and 53-63% among ANMs). Several gaps remained in both areas in counseling on how to consume iron-folic acid (40-57% among AWWs and ASHAs), benefits of iron-folic acid (26-46%), and side effects and how to manage them (12-43%). Discussions on benefits and side effects of calcium were even lower and similar among FLWs in both areas. AWWs and ASHAs in I-ANC areas were more likely to advise women on how to take calcium compared with those in S-ANC areas (47% compared with 26%). Reported counseling on iron-folic acid and calcium showed similar patterns as observations, with some higher counseling in I-ANC compared with S-ANC areas for consuming supplements during pregnancy, how to consume, and benefits of consumption (Supplemental Tables 2-3).
Counseling observations for diet diversity showed ∼50-60% of FLWs counseled on the importance of a diverse diet, and 70-80% counseled on consuming at least 5 food groups in a day, similar between I-ANC and S-ANC areas ( Table 7). In addition, ∼60% of AWWs and ASHAs in both areas enquired about foods pregnant women consumed in the previous 24 h during counseling sessions. Reported counseling on diet diversity showed similar patterns as observations, with some improvements in I-ANC compared with S-ANC areas for counseling on animal source foods during pregnancy (Supplemental Table 4).
Counseling observations for weight monitoring showed gaps at endline and were nondifferential between study areas (Table 7). Less than a quarter of FLWs checked whether weight gain was adequate using measurements recorded in the Mother and Child Protection card; only around a third of FLWs explained optimal gestational weight gain to women. Counseling on the importance of weight gain during pregnancy was 37-55% among AWWs and ASHAs and 50-71% among ANMs. Reported counseling on weight-gain monitoring showed patterns similar to the observations, with some improvements in I-ANC compared with S-ANC areas for counseling on weight-gain monitoring during VHSNDs and recommended weight gain (Supplemental Table 5).

Discussion
To address the challenges of maternal undernutrition, A&T implemented a bundled package of maternal nutrition interventions, but the impact evaluation only found limited impacts on key outcomes related to consumption of micronutrient supplements, diet diversity, and gestational weight gain. Using mixed methods and a theory-driven PIP analysis, this study identified successes and challenges along PIP domains for a bundled package of maternal nutrition: micronutrient supplementation, diet diversity counseling, and weight-gain monitoring. We observed secular trends in several domains along the PIP in both intervention areas and found greater improvements in some domains in I-ANC compared with S-ANC areas. Across interventions, common areas of successes Across the 3 types of FLWs, ANMs in I-ANC compared with S-ANC areas showed the greatest improvements in exposure to maternal nutrition training content. ASHAs and ANMs in I-ANC compared with S-ANC areas were more likely to use maternal nutrition materials, whereas use of materials was similar among AWWs in both areas. Knowledge of and counseling on iron-folic acid and calcium supplements were low to moderate among AWWs and moderate to high among ASHAs and ANMs, with some improvements in I-ANC compared with S-ANC areas. For diet diversity, knowledge and counseling were moderate for AWWs, ASHAs, and ANMs with some higher knowledge in I-ANC compared with S-ANC areas among AWWs and ANMs. Knowledge of and counseling on weight gain were moderate and higher in I-ANC compared with S-ANC areas for all FLWs. We also identified key successes and challenges for each specific intervention. For micronutrient supplementation, successes included FLW knowledge of and counseling on taking iron-folic acid and calcium during pregnancy. Key challenges were limited exposure to training content on iron-folic acid and calcium, low knowledge of supplement benefits, supply chain stock-outs, gaps in provision of iron-folic acid and calcium, and counseling on supplement benefits and side effects. Intervention-specific successes for diet diversity counseling included training content on diet counseling, FLW knowledge of diet diversity and importance of dark green leafy vegetables, and counseling on the importance of diet. Challenges for diet diversity counseling included low knowledge of animal source foods, food demonstrations during home visits, and low counseling on specific food groups. For weight-gain monitoring, most FLWs had correct knowledge of recommended weight gain during pregnancy, but critical challenges included training on weight-gain monitoring, weighing of women during VHSNDs, and counseling on recommended weight gain and weight monitoring.
Across maternal nutrition interventions, exposure to training was high overall, but gaps occurred in training content. Monthly meetings were available as a platform for refresher training but not fully used, with irregular meetings and overemphasis on administrative tasks, reducing time for technical content discussion. Previous literature has highlighted the importance of regular continuous trainings to improve FLW knowledge  and skills and deliver quality services (11,29); strengthening monthly meetings as an active platform for refresher trainings to exchange experiences and fill in knowledge gaps is critical. Ensuring that training manuals are comprehensive and include context-appropriate information for problem solving is important for improving content (19). Although FLWs had positive perceptions of job aids, only around twothirds of them used materials. Ensuring that materials are concise and easily understandable by FLWs is important so that use of materials is not obstructed due to lack of time (20). Improving knowledge of FLWs to deliver services is a critical next step following training. Overall knowledge was moderate to high but with gaps in detailed knowledge of supplement benefits and recommended foods during pregnancy. Tailored frequent trainings are important to equip FLWs with the knowledge to provide accurate counseling to mothers.
Although supportive supervision was found to be a key intervention element influencing FLWs' performance in LMICs (29), high reporting burden and focus on administrative tasks among supervisors and block-level staff obstructed supportive supervision practices and effective use of data in our study. Potential improvements include providing constructive feedback, coaching, and problem solving during visits instead of only monitoring and finding faults (20). A change in system culture is needed to shift the focus from administrative functions toward training and coaching support for improvements in FLW performance. Literature on strategic use of data to track intervention progress and identify areas of improvement is limited, and this topic is important to study in future research, as well-resourced nutrition data and information systems are essential for effective implementation (15,27). Setting up monthly or quarterly review systems at the district and block levels can be used to track performance, supplement stock, and identify bottlenecks. The study findings showed the potential of monthly meeting platforms but also highlighted challenges of regularity in their functioning.
Availability of supplements and a well-functioning supply chain are essential for micronutrient supplementation interventions. A study from Kenya examined the integration of ironfolic acid and calcium supplements into primary health care and found substantial supply chain bottlenecks, particularly for calcium supplements (30). Delays were prevalent in demand estimation of supplements, which caused stock-outs in some areas (30). Our study reported similar issues with the supply chain for both iron-folic acid and calcium supplements, especially the latter. One such bottleneck in the supply chain identified through our study was differing methods of stock estimation at different levels of staff.
Counseling on diet during pregnancy is a critical intervention to improve dietary intakes for pregnant women, which are suboptimal in LMICs, particularly India (31, 32). Recent literature has highlighted several constraints for women to adopt nutritious diets, including resource constraints leading to limited affordability and accessibility to nutritious foods, dependence on husbands to procure food, and food preferences (33,34). Context-appropriate diet counseling is important to help women understand which locally available foods they should consume to provide adequate nutrition for themselves and their children. Counseling of husbands can also help to improve family support and build awareness of the importance of ensuring nutrition during pregnancy. Our study results indicated moderate levels of counseling on overall diet diversity and certain food groups, including green leafy vegetables, with gaps for counseling on other foods, particularly animal source foods. Closing these gaps is essential to improve quality of counseling and its effectiveness.
Although WHO (2016) (35) recommendations on maintaining healthy weight gain within a specified range are tailored to categories of prepregnancy BMI, current recommendations on calculating weight gain, counseling on weight gain, recordkeeping, and use of data are not specified by national protocols (14). Existing protocols do not specify weight-gain counseling based on BMI but recommend an average weight gain of 10-12 kg. In this study, gaps were prevalent in exposure to topics of weight gain during trainings and knowledge of optimal weight gain, leading to low provision of weight monitoring and weight-gain advice.
The secular trends observed could be explained by the parallel implementation of a national nutrition agenda and the systems-strengthening efforts adopted by A&T. The Indian government nationally prioritized the integration of maternal nutrition services into trainings for FLWs and community-based events focused on maternal nutrition. A&T's maternal nutrition interventions included technical assistance on supply chain and anemia management and sharing of behavior change materials across all districts in Uttar Pradesh; these activities may have led to some spillover effects.
In this study, data on training, knowledge, and service delivery were self-reported by FLWs, with potential for social desirability and recall bias. To objectively measure FLWs' performance, we included direct observations of counseling to reduce potential bias from reported measures. The qualitative data collection was limited to 4 intervention blocks and may not reflect issues pertaining to nonintervention blocks or the larger study area. Finally, subjective measures used in qualitative interviews, including asking FLWs about their perceptions of supervision received, may have had response bias due to sociocultural norms that discourage criticism of supervisors. We reduced this bias by ensuring privacy during interviews and reassuring participants about response confidentiality.
To our knowledge, our study is one of the first to conduct an intervention-specific PIP analysis to examine the integration of a multifaceted maternal nutrition intervention into existing government systems, offering a unique contribution to the current literature. Many of the identified gaps in FLW knowledge, service delivery, counseling, and supply chains-and the reasons for the gaps-are likely experienced in similar programs in other contexts. The method used allowed going beyond overall impact-focused analyses (which have mostly explained successes rather than limited impacts of interventions) and can be used in analogous implementation research elsewhere. Systems-strengthening efforts within complex programs must focus on the specific demands of each intervention, which can be elucidated by conducting an intervention-specific PIP analysis to understand the areas of successes and challenges unique to different interventions.