Pneumococcal conjugate vaccine (PCV) was introduced under the Universal Immunization Programme (UIP) in a phased manner in India, starting in 2017 from the high burden states and districts [1,2]. The nationwide rollout of the vaccine was completed in October 2021 [3]. PCV scale-up across the country was done during the COVID-19 pandemic, amid the numerous restrictions on gatherings and travel. As is done for all new vaccine introductions, cascade trainings were conducted for the program managers and the frontline health workers/vaccinators. But as the classical classroom model could not be followed for the training, most of these were done in a hybrid or online mode. Therefore, to complement the hybrid mode and to ensure that the vaccinators have the required information about all the aspects of PCV, whenever needed, several initiatives were undertaken, including the development of animated videos and a chatbot dedicated to PCV.

The unprecedented pace of advancements in technology and the increasing amount of time people spend on the internet have led to a major change in lifestyle and community behavior change strategies. As a result, all sectors, governmental and nongovernmental, are constantly evolving and shifting to various digital platforms and solutions to meet user and customer needs. Companies have adopted and implemented information and communication technologies to enhance their relationship channels and make them constantly available in real time to assist customers [4-6].

The PCV chatbot was designed to address queries related to PCV, including dosing, scheduling, administration, cold chain management, storage, side effects, open vial policy, wastage, and eligibility [7]. It was aimed to be very easy-to-use, just like any WhatsApp group, where the auxiliary nurse midwife (ANM) can type in a few keywords related to her query, and the bot will provide a detailed response that includes those keywords. The PCV chatbot was rolled out in select geographies for use by the vaccinators and their supervisors. However, understanding end user perceptions of the PCV chatbot is essential to maximizing its effectiveness and use in the field.

This study aims to document the opportunities, challenges, mitigation strategies, and stakeholder experiences associated with the development and implementation of the PCV chatbot for PCV introduction in India.

The literature indicates that chatbots are increasingly recognized as effective digital tools to complement capacity building for vaccinators, including in the context of PCV programs [8,9]. Chatbots facilitate on-demand, scalable delivery of vaccine information and guidance, offering interactive support tailored to health care workers’ needs in both routine immunization and outbreak settings [10,11]. Scoping reviews and intervention studies show that well-designed chatbot platforms can enhance vaccine literacy, improve procedural knowledge, and support real-time decision-making for immunization staff, especially in low-resource environments where digital inequities exist [8,9,12].

Notably, chatbot interventions for vaccine confidence and uptake—across COVID-19, human papillomavirus vaccine, and routine immunizations—have demonstrated improvements in both attitudinal and behavioral outcomes, such as increased vaccination intention and actual uptake among end users [10,11,13]. For example, randomized controlled trials in Asia found that chatbot users were more likely to report increased vaccine confidence and acceptance vs nonusers, underscoring their feasibility and acceptability for health communication among frontline workers and the communities they serve [9,10,14].

However, the literature also highlights varied results depending on contextual factors such as chatbot design, conversation quality, and empathy expression by chatbots, which may sometimes negatively affect user evaluation and behavioral intention if not properly aligned with user expectations [15]. Furthermore, process evaluations stress the importance of integrating chatbots with broader digital immunization infrastructure, such as mobile health apps, electronic vaccine tracking, and learning management systems, to achieve sustainable capacity building [12,16,17].

In low- and middle-income countries, deploying chatbots as part of digital training and support strategies has enabled vaccinators to better monitor, analyze, and respond to challenges in vaccine logistics, stock management, and adverse events reporting [16,18]. These innovations empower health care personnel to harness data for targeted interventions, ultimately strengthening immunization systems and coverage [12,19].

Additional references supporting these findings include comprehensive reviews and operational frameworks by global health organizations, which advocate the integration of artificial intelligence and digital platforms to reinforce immunization workforce competencies and real-time vaccine program delivery [17,19,20]. Such evidence emphasizes both the promise and the necessity of leveraging chatbots—and related digital tools—as complementary elements in capacity building for PCV and other vaccine programs.

This study was conducted with the following objectives:

A qualitative study design was used, using semistructured interviews to understand the perspectives of key stakeholders regarding the use of a chatbot within the UIP service delivery framework, and specifically for the introduction of new vaccines. Figure 1 presents an overview of the research design. Likert scale survey statements and scoring were also used to gather user feedback on the perceived use and challenges of using the chatbot.

The study was conducted in 4 states: Arunachal Pradesh, Delhi, Karnataka, and West Bengal (Figure 2). In each state, two districts were sampled, and within each district 2 blocks were sampled for the purpose of the study. Study states, and sites were purposively sampled based on state official recommendations, John Snow India Private Limited (JSIPL) presence, and logistical and operational feasibility, ensuring geographic diversity and government approval.

The states were selected so as to ensure that the research is not restricted to any one geographical spread in the country. Factors contributing to the state selection also included approvals from the respective state governments to conduct the research in their state, logistical and operational feasibility. For the districts, recommendations from the state officials and the ease of travel to the said district were considered during selection. Two districts in each of the 4 states were selected based on the said criteria. Then, from each district, 2 blocks were selected based on the recommendations of the district officials and operational feasibility. For the FGDs and the short survey, the ANMs who were available for the interaction on the day of the data collection in their respective area were contacted. Thus, this was a purposive sampling for the purpose of the study (Figure 3).

Participant demographics such as age or gender were not recorded, as the study emphasized cadre-specific perspectives (eg, ANMs as frontline vaccinators or state expanded program on immunization officers/district immunization officers/medical officers [MOs] in-charge as immunization in-charges or officials). ANMs were included based on availability at selected facilities, ensuring diverse field experiences across states.

The respondents included members from the teams that conceptualized the development and deployment of the PCV chatbot, the IT agency that developed the PCV chatbot, and a selection of targeted end users and their supervisors at the block, district, and state levels. The detailed sample size and the theme of the respective enquiry for each category of the respondents are mentioned in Table 1.

Separate tools were developed for each cadre of the respondents in the study. All the tools developed for the KIIs had sets of open-ended questions covering different themes pertaining to the rationale, design, functionality, content, operability, benefits, and challenges of the PCV chatbot.

In addition to these open-ended tools, a Likert scale–based short quantitative tool was developed for the ANMs, which concentrated on the agreement/satisfaction level of the end users toward the PCV chatbot.

As the ANMs in the field are more comfortable responding and interacting in their local language, the ANM tools were translated into the regional languages for all the 4 states (ie, Hindi for Delhi and Arunachal Pradesh, Bengali for West Bengal, and Kannada for Karnataka).

As this research involves primary data collection and interactions with human subjects, ethical clearance for carrying out this research was obtained from the Convergent Institutional Review Board (IRB; IRB certificate number Convergent IRB/2024-25/015).

Ethical approval for this study was obtained from the Convergent IRB agency, with certificate number Convergent IRB/2024-25/015 on August 28, 2024.

In addition, a separate permission to conduct the research was obtained from each of the highest state officials.

Consent procedures were in line with those laid out by the ethical certification agency. Study participants were requested to provide informed written and signed consent prior to any data collection. No personal details, such as names, mobile numbers, or addresses of any of the participants, were collected during the research.

It was ensured that any personal identifier information related to the participants was protected, and confidentiality was maintained.

Results are structured thematically (1-6), corresponding to study objectives: development opportunities/challenges (Objective 1), and user experiences/benefits (Objective 2).

The sample size included 36 KIIs (program managers: n=8; SIOs: n=4; DIOs: n=8; MOICs: n=16) and 16 FGDs (ANMs: n=74).

Demographics were not collected (because emphasis was on role-based sampling), and thus purposive sampling was done by cadre/availability across 4 states.

This qualitative study met its 2 objectives through thematic analysis of 36 KIIs and 16 FGDs (74 ANMs) across 4 Indian states. Objective 1 documented PCV chatbot development amid COVID-19 constraints: hybrid trainings revealed ANM knowledge gaps, prompting a WhatsApp-based FAQ tool translated to 13 languages, though deployment hit technical glitches, government priorities, and scope limits (Themes 1-3). Objective 2 captured experiences: 92% (68/74) ANMs were unfamiliar, with high ratings for completeness/accuracy (Figures S2-S4 in Multimedia Appendix 1), ease (Figure S5 in Multimedia Appendix 1), and preference (61/74, 82%; Figure S9 in Multimedia Appendix 1), valuing self-sufficiency over handbooks/Google, but noting internet/digital literacy barriers (Themes 4-6). Overall, chatbots show promise as scalable UIP complements.

Globally, as well as in India, there are a number of chatbots being used in the health care system, designed for different end users. Some of these chatbots have been evaluated for their effectiveness, usability, ease of use, reach, and accuracy.

The PCV chatbot was developed and deployed in the pre–generative artificial intelligence period, when such interventions and ideas were not very prevalent. Thus, it was developed so that it was very easy to use, just like any WhatsApp group, where the ANM can type in a few keywords related to her query, and the bot will provide a detailed response that includes those keywords. The evaluation of one such chatbot—the ASHABot—designed to address the information needs of community health workers (CHWs) in India, found that the chatbot provided a private and trustworthy channel for CHWs to ask rudimentary and sensitive questions they hesitated to ask their supervisors. The study suggests that chatbots should be seen as supplemental resources within the community health care system, rather than a replacement for CHWs’ supervisor support [22]. Our study findings highlight a dual perspective on chatbot use among ANMs: on one hand, strong endorsement for enhancing self-sufficiency—82% (61/74) preferred querying the bot before supervisors, citing busy MOICs and instant access (Figure S9 in Multimedia Appendix 1). On the other hand, apprehensions about overreliance, diminishing peer/human interactions essential for nuanced field support, potentially fostering isolation in hierarchical UIP settings. This tension underscores the value of hybrid models that harness chatbots for digital autonomy while preserving essential human interactions and relational training within the UIP framework.

The recommendations collated from the study are summarized below under specific areas.

While all respondents were largely satisfied with the content, there were several recommendations on enhancing the existing content and adding new content.

In all states, following the discussions on monitoring and accountability, outlined below are the recommendations made by different levels of respondents.

Update regularly as per the government guidelines: all respondents recommended that it would be mandatory for the chatbots to be updated as per the new guidelines and protocols that are introduced by the government for the same vaccine. If this is not done, the chatbot will become more of a liability than an aid.

This study has several limitations that should be considered when interpreting the findings. First, the purposive sampling strategy, particularly for ANMs based on availability at health facilities during data collection, constitutes convenience sampling, which limits generalizability beyond the selected sites in Arunachal Pradesh, Delhi, Karnataka, and West Bengal. While states were chosen for geographic diversity and logistical feasibility, findings may not fully represent all UIP contexts, such as high-burden states or union territories.

Second, no demographic data (eg, age, gender, and years of experience) were collected from participants, as the focus was on cadre-specific perspectives. This assumes relative uniformity within roles (eg, state expanded program on immunization officers/district immunization officers/MOs in-charge as MBBS physicians, ANMs as frontline vaccinators), which may overlook variations in digital literacy or tenure affecting chatbot perceptions, particularly among older ANMs noted in qualitative responses.

Third, data collection occurred in September 2024, over 3 years post–PCV rollout (2021), relying on retrospective recall. This introduced potential recall bias, evidenced by 92% (68/74) ANM unfamiliarity with the chatbot, largely due to staff transfers common in public health systems. Real-time experiences during deployment may have differed, though field monitoring notes from program managers provided contemporaneous corroboration.

Fourth, the mixed methods sequence—Likert surveys preceding FGDs with the same ANMs—may have introduced response bias, as quantitative priming could influence qualitative discussions toward positive framing. While this order minimized groupthink in FGDs, reverse sequencing (FGDs first) might have yielded more unprimed views.

Fifth, the study focused exclusively on the PCV chatbot, limiting insights into multivaccine or routine immunization applications. Stakeholder suggestions for broader UIP integration remain untested empirically.

Additional methodological constraints include the hybrid nature of PCV trainings influencing baseline knowledge gaps, and technical demonstrations during FGDs potentially inflating perceived ease-of-use ratings (Figures S5-S9 in Multimedia Appendix 1). Finally, as an implementation study by chatbot developers, researchers’ familiarity with stakeholders risked social desirability bias, though ethical protocols (eg, anonymity and external IRB) and triangulation mitigated this.

Despite these limitations, the large sample (n=110), multistate design, and convergence of qualitative themes with Likert data enhance credibility and transferability to similar low- and middle-income country vaccine introductions. Future prospective, multivaccine trials could address these gaps.

This study affirms WhatsApp-based chatbots like PCV chatbot as viable, low-cost complements to cascade trainings for new vaccine introductions in India’s UIP, particularly during disruptions such as COVID-19, by enabling on-demand, multilingual access that 82% (61/74) of ANMs preferred over supervisor calls for routine doubts. Beyond reinforcing knowledge gaps and self-sufficiency, findings reveal untapped potential for community advocacy, positioning chatbots as equity tools for remote caregivers hesitant about AEFI or schedules.

Broader implications extend to UIP resilience: integrating predictive, voice-enabled, offline-capable bots for multivaccines could standardize field support nationwide, reducing dropout rates in hard-to-reach areas where connectivity lags. National policies should prioritize user-centric design—audio-visual content, real-time updates—drawing from successes like ASHABot, potentially lifting coverage by 10%-20% per Asian trials while preserving peer networks.

For low- and middle-income countries, this model offers a scalable blueprint: government-approved FAQs via familiar platforms bypass digital divides better than apps, fostering ANM autonomy without overreliance. Future innovations—artificial intelligence analytics on queries, caregiver portals—could transform immunization from reactive to predictive systems, advancing Sustainable Development Goals 3 targets amid climate/geopolitical shocks.