Providers expressed that they were pregnant women’s primary source of health information, and that it was therefore their duty to educate pregnant women about maternal vaccines ( Table 1). Many providers expressed confidence that patients would accept their instructions without question, especially in rural facilities where patients are felt to have more limited knowledge. One provider in Marsabit expresses this view:

“I guess some people don’t have the understanding of the vaccines. So when they come here, they cite the myths and misconceptions about vaccines that they have been fed back in their villages.”

Providers also admitted that the relative social position of healthcare providers and patients could be a factor in high acceptance, as pregnant women are expected to defer to provider’s greater knowledge and authority in health matters. In the words of one Nairobi provider:

“Maybe sometimes they do not have that chance to say no because they look at me as their savior… and everything I tell them, they believe is right.”

However, some acknowledged that expectations of total patient deference were changing as women become more comfortable initiating requests for information. This openness to two-way exchange was more commonly expressed among urban providers, such as this Nairobi-based nurse:

“In the past, people used to be blasted by the nurses or whoever was giving the services whenever they asked questions. Those days are long gone. It is always good to ask why you are being injected.”

Some providers highlighted the limited agency of pregnant women regarding their own medical decision-making. For example, a woman might be unable to access types of care that run contrary to her husband’s beliefs ( Table 2). Conversely, husbands and male relatives who were involved in women’s antenatal care could be facilitators of maternal vaccination.

Providers cited community and religious institutions as both barriers and facilitators of vaccine acceptance among their patients. Religious institutions could provide welcome venues for vaccine education outreach, while traditional leadership could serve as gatekeepers to community acceptance. One provider in Marsabit suggested that community meetings would be an ideal venue to reach those who might otherwise not come for antenatal classes (ANC):

“ The other thing that can be done is to communicate the importance of vaccines through the chiefs during the barazas [community meetings] where even the county officials can explain it to them.”

Other providers identified certain religious denominations as barriers to vaccine acceptance, stating that the leaders of these groups instructed followers to avoid medical treatment generally or vaccines specifically. One Nairobi provider cites a traditional religious denomination as one such group:

“During the polio campaigns, the Akorino rejected it in Eastern Kenya because their religion does not allow them to take medication.”

Providers almost universally cited time constraints and provider overwork as barriers to providing adequate patient education about vaccines ( Table 3), with many suggesting that human resource allocation was the major limitation faced by their clinics in vaccine delivery. Many complained of overwork and long lines, factors seen as contributing to patient attrition from their clinics. In the words of a nurse at a private facility:

“We have had couples leaving because of the high queues.”

Educational materials for patients were often cited as absent or inadequate, a situation that was often blamed on the shifting responsibilities to county government following devolution (political decentralization) of powers from national to county governance in 2010 ( Table 3). Some providers suggested that patient brochures needed to be translated into local languages, as women at their clinics did not read English or Swahili. One provider in rural Marsabit explains:

“ I think the information should be printed out in different languages and given to the community radio stations so that it can be explained to them, because they do not understand Swahili.”

Likewise, providers expressed a desire for continuing professional education, particularly on newer vaccines. One provider links professional education to patient acceptance:

“If you do not update me properly, it will be of no use even if [a new vaccine] is brought to the facility. It will be available but I will not give it out, not because the patients are not asking for it, but because it is me who is not interested in giving it out.”

Some providers also raised concerns about shortages of physical resources, also linking these to ongoing decentralization efforts as counties accepted greater financial responsibility for vaccine delivery and administration.

While the overwhelming majority of providers are comfortable in their knowledge of vaccines currently in the national schedule and available at their facilities, few are knowledgeable about vaccines not offered at their facilities, or about upcoming vaccines ( Table 4). For example, providers at private facilities, which are likely to carry influenza vaccines, were more familiar with this vaccine than those at public facilities, where influenza vaccine is generally not available:

“It [flu vaccine] is mostly given by the private facilities. They say it is okay. I will, however, not vote for it or say much about it because I have never issued it.”

The vast majority of providers expressed the view that tetanus toxoid was beneficial to pregnant women and infants, and were open to new vaccines being introduced ( Table 5). They also believed that most healthcare providers shared their positive attitudes. Providers were less comfortable endorsing vaccines with which they were not personally familiar, especially those at public facilities. A few stated that they already received complaints from pregnant women regarding both TT2+ and childhood vaccines and that introducing more vaccines would exacerbate this problem. One Mombasa-based provider expresses doubt that pregnant women will readily accept new vaccines:

“If they are already complaining that the available vaccines are too much or painful, how about additional vaccines? That is the concern. There will be more complaints regarding additional vaccines.”

Distance to health facilities and lack of transportation was one of the major access issues perceived by providers to hinder maternal vaccine coverage ( Table 6). Other providers cited the cost of medical services at private facilities and inter-clinic mobility as sources of patient attrition during antenatal care. In the words of one Mombasa-based provider,

“Private hospitals are established to make profits. The charges may keep them away.”

Urban providers tended to express more frustration with patient mobility, as women in these settings often have access to several facilities offering the same services. This was linked to the notion that women often move during their pregnancy, leaving gaps in their medical records that are difficult to address without an integrated system. One Nairobi provider describes how patient mobility impacts coverage of TT2+ booster doses:

“ The only challenge is that the subsequent doses like when the women delivers, coming for the other doses is a challenge and you know this is cosmopolitan town so people come… Somebody would deliver and they would even go away to their homes maybe in another county so we miss them out for the subsequent doses.”

Because a convenience sample of providers was interviewed, it is possible that selection bias towards those with higher vaccine knowledge or more positive vaccine attitudes occurred. Moreover, many providers were recruited and interviewed from their homes due to an ongoing healthcare workers’ strike during the data collection period. Among public sector workers, the strike may also have influenced some responses, particularly those pertaining to working conditions. Private facility workers may also have been experiencing overwork and higher patient volumes due to closures of public facilities. In addition, more HCPs from private facilities not participating in the strike were recruited for interviews, resulting in oversampling from private facilities.

To maintain high coverage of TT2+ and to ensure widespread acceptance of future maternal vaccines, providers must be empowered to use effective, pro-vaccine messaging in their interactions with pregnant women. Communication strategies for vaccine promotion can be incorporated into current professional education efforts. Print materials for patient education may also facilitate these interactions, and political decentralization provides opportunities and challenges for county-level government to publish regionally tailored resources for patients and providers. Finally, cultural and religious influencers should be leveraged to facilitate the introduction of new maternal vaccines.

This analysis draws from a large, multi-component, mixed methods study of determinants of maternal vaccine acceptance, which includes data from pregnant women, healthcare providers, and key informants in Kenya. This paper considers only qualitative data from healthcare providers; however, triangulation with pregnant woman and key informant data was used to validate results.

Semi-structured interviews were conducted at 15 facilities in four diverse geographic areas. An interview guide ( Supplementary File 1) was developed using a grounded theory approach drawing on observational data gathered in antenatal clinics (ANCs), literature review, and themes emerging from interviews of pregnant women. This approach was chosen to limit investigator bias and to account for a lack of qualitative vaccine acceptance literature relevant to the Kenyan context. Interviews covered topics including resources for vaccine delivery, experiences with patient education, knowledge of vaccines, and strategies for demand creation. This study was approved by the Institutional Review Boards at Emory University [IRB00089673], the Centers for Disease Control and Prevention (CDC) [covered under reliance agreement with Emory], and the Kenya Medical Research Institute (KEMRI).

Kenya was chosen as the site of the in-depth phase of the study due to its large birth cohort, relatively high vaccine coverage, and longstanding collaboration with the CDC through KEMRI. Interviews were conducted in four geographically diverse locations in country: Marsabit, Nairobi, Mombasa, and Siaya. These sites were purposively selected to create a sample representing a range of urban and rural locations ( Table 7). A convenience sample of healthcare providers representing 37 public and private facilities was recruited at each site until data saturation was reached; all providers who consented to participate were interviewed. Inclusion criteria were current employment as a healthcare provider, proficiency in English or Swahili, and ability to grant informed consent. Healthcare providers were defined as physicians, nurses, community health workers, and clinical officers (non-physician clinicians licensed by the Kenyan government who undergo additional training beyond that of nurses ¹⁵ ).

Development of the qualitative study protocol and data collection instruments was overseen by an anthropologist trained in qualitative methodology [FAM] in partnership with the principal field investigator [NAO], a Kenyan expert in maternal and child health. Meetings between senior members of the Kenya-based data collection team and the US-based analysis team occurred on a weekly basis during the study to ensure the validity and cultural relevancy of all study materials, results, and conclusions. All data collection activities were performed by Kenyan team members who received training in interview protocol, note taking, consenting, data management, and transcription. One member of the US-based analysis team [SWN] was fluent in Swahili and was able to confirm local terms that remained in the translated interviews.

An interview guide was developed using an iterative, team-based approach incorporating findings from the World Health Organization’s Strategic Working Group of Experts on Immunization, field notes from observations performed in antenatal clinics, and themes emerging from pregnant women interviews. The guide was revised after field testing with three interviews based on input from the Kenya-based data collection team. The final guide covered the following topics: (a) proportion of patients estimated to have received or refused maternal vaccines; (b) gestational age at which vaccines are given; (c) barriers or reasons for refusal cited by patients; (d) perceived ability and methods used to address these barriers/refusals; (e) comfort discussing vaccine recommendations with patients; (f) knowledge and experiences of vaccine preventable diseases; (g) existing patient education efforts; (h) introduction of future maternal vaccine recommendations into the current schedule; and (i) knowledge and attitudes surrounding tetanus toxoid (TT2+) and influenza vaccine effectiveness and safety.

Interviews were conducted in private rooms by teams of two members of the data collection team [WNW, SNW, ROA]. Written informed consent, covering interview participation, transfer of data to the US for analysis, and presentation of de-identified portions of interview transcripts, was obtained prior to the start of each interview. Face-to-face interviews were conducted in English or Swahili at participants’ places of employment, and in some cases, at their homes (to accommodate public sector workers who were on strike) between January and August 2017. Interviews were audio recorded, transcribed verbatim, translated where necessary by the data collection team, and assessed for accuracy by a team lead before analysis. No identifying information was included in written transcripts, which were stored on a secure server along with copies of the audio recordings. Original audio recordings were deleted immediately following transcription.

Principles of thematic analysis ¹⁶ informed our coding methodology. Codes were created both deductively and inductively through an iterative, team-based process based on interview guide questions and emergent themes captured by reading a sample of 12 transcripts from all sites. After finalization of the preliminary codebook, three transcripts representing different sites were selected and coded by the data analysis team [IB, SWN, CAA]. Inter-coder reliability was assessed using NVivo 11 Pro (QSR International Pty Ltd, Victoria, Australia), after which inconsistencies in coding were resolved and the codebook was revised to accommodate further emergent themes. After several iterations of this process, the codebook was finalized with inter-coder reliability at kappa >0.8 on 10% of transcripts representing all sites ( Supplementary File 2). All interview transcripts were coded using NVivo 11 Pro by a team of four research assistants [IB, SWN, CAA, VLF]. Thematic analysis of the coded transcripts was performed by the coding team [IB, SWN, CAA, VLF] and senior investigators [PMF, FAM].