In March 2019, we conducted 25 semistructured in-depth interviews among mothers of newborns recruited from two county hospitals in Yunnan Province, southwestern China: the Maternal and Child Health Hospital in Nanjian County and the People’s Hospital in Jinggu County, both large hospitals where the majority of infants in the two counties were delivered. These two counties compose China’s most poverty-stricken area based on the percentage of people below the national poverty line. The study was approved by the institutional review board at the University of Georgia (STUDY00005670).

We applied the HBM and cultural competence theories to identify the preferences and barriers among the target population for receiving and practicing evidence-based parenting knowledge [22]. HBM has been used in a broad range of behavioral interventions including interventions to increase screening, reduce risky behavior, and improve adherence to medication regimen [23-26]. In its latest version, the key constructs of the model include cues to action and self-efficacy and perceived susceptibility, severity, benefits, and barriers [27]. Studies have shown that parental perception regarding the healthy development of their young children often involves perceived barriers to take action [28-30]. Behavioral change will not occur if parents’ perceived barriers outweigh perceived benefits [31].

In addition, although our target population did not have language barriers when reading ECD messages, they might have cultural beliefs that were not consistent with the evidence that supports child healthy development. Culturally competent approaches recognize and respect the cultural beliefs, norms, and values of the target population and aim to improve effectiveness of the intervention [32,33]. Therefore, in our theoretical framework, cultural beliefs and norms were included as a modifying factor (Figure 1).

The inclusion criteria were the study participants should be in good health with no severe delivery complications or clinical diagnosis of postpartum depression and have a healthy infant aged less than 1 month.

Two physicians assisted in identifying and recruiting potentially eligible participants through the hospital wards. The 25 participants were recruited in the local hospitals through criterion sampling. We reached sample saturation with 25 participants, until a point where additional interviews did not develop new theoretical insights. In other words, sampling was stopped when the data were repetitive and our research questions had been answered. To examine their eligibility for the study based on the criteria, electronic medical records were retrieved to scan for any recorded delivery complications or diagnosed depression. Informed consents were obtained from all participants who met the inclusion criteria and agreed to participate in the study.

Two trained interviewers (DZ and YZ) conducted the in-depth interviews in private rooms at the hospitals. We developed an interview guide based on the theoretical framework (Figure 1) pertaining to the design of the mobile health intervention. These topics included (1) mothers’ individual perceptions about the importance of avoiding adversities in the early years; (2) benefits and barriers when feeding and nurturing newborns according to recommendations sent over the phone; (3) family socioeconomic status, family structure, relationship, beliefs, and values related with adhering to the recommendations; (4) concerns about privacy; and (5) preferences for the frequency of receiving messages. During the interview, the Edinburgh Postnatal Depression Scale, Chinese version, was used to evaluate mothers’ mental health status [34,35]. Participants were shown sample text messages and were asked questions about the messages. The message included about 600 Chinese characters and two pictures with four sections: (1) weekly achievement, showing easily identifiable developmental milestones; (2) weekly care, providing relevant information on baby’s daily care; (3) weekly game, showing examples of age-paced games and parent-child activities including playing games, singing nursery rhymes, reading, and drawing; and (4) weekly question and answer, providing answers to common questions asked by parents. The interview was designed to begin with sociodemographic questions followed with questions according to each construct in the HBM.

Interviews were conducted on a one-on-one basis. Each interview lasts approximately 60 to 90 minutes. To aid in recruitment, ¥100 (US $15 in 2019) was provided to each participant. All interviews were tape recorded with participant permission.

Interviews were transcribed verbatim and entered into qualitative analysis software NVivo 12 (QSR International). A coding structure based on the interview guide was iteratively developed to label specific content from the interviews. The transcripts were coded separately by two members of the study team (DL and RG). After coding, the two team members met to discuss the results and resolve any disagreements in coding by consensus.

Table 1 shows the characteristics of the 25 participants. Their average age was 28 years. More participants lived in urban regions than in rural regions (56% [14/25] vs 44% [11/25]). Most participants (10/25, 40%) were ethnically Han Chinese, 32% (8/25) were Yi, and 16% (4/25) were Dai Chinese. Overall, 48% (12/25) of newborns were male. The majority of the infants’ primary caregivers (18/25, 72%) were mothers, while 12% (3/25) were grandparents. A total of 40% (10/25) of the mothers finished middle school, and 28% (7/25) obtained some college education, while 16% (4/25) finished only elementary school. Half of the participants (13/25, 52%) were unemployed, and 28% (7/25) had a full-time job; 48% (12/25) of participants reported that the main source of family income was from their husband. About half (13/25, 52%) of the participants had household income below ¥5000 (US $750) per month, whereas 52% (13/25) of them reported they were willing to spend between ¥1000 (US $150) and ¥2000 (US $300) on childcare, and 28% (7/25) reported they were willing to spend more than ¥2000 (US $300). A majority of the participants (23/25, 92%) were not receiving public financial aid. Most (23/25, 92%) were married, and 40% (10/25) lived in extended families with 7 or more family members.

Table 2 shows participant health status and lifestyle. All participants self-reported having good health. Most participants had a low risk of postpartum depression with a depression scale score below 13, but 6 people (24%) had a relatively high risk of depression, although no clinical diagnosis of depression was reported. A total of 76% (19/25) of participants reported light to moderate workload for doing housework, and 64% (16/25) of participants would spend more time on childcare than on other things, while 20% (5/25) reported that they would spend more time on family business, housework, or farm work. Most participants (16/25, 64%) had no difficulty in breastfeeding, while 28% (7/25) did not start breastfeeding, and one participant reported having no breastfeeding plan because she had hepatitis B. A majority of the participants (18/25, 72%) planned to breastfeed for 6 months to 1 year, and 12% (3/25) planned to breastfeed for less than 6 months.

This study focused on assessing feasibility of an ECD intervention using mobile phones to meet the needs of mothers of newborns based on their beliefs, traits, and preferences. By interviewing 25 mothers, we enhanced our understanding of the necessary elements for an ECD intervention to be acceptable to low-income communities and determined the factors that may affect the usability of the intervention. From the demographic information, the samples in our study were distributed evenly in rural and urban areas and covered a variety of races. The majority of mothers we interviewed reported under ¥5000 (US $750) estimated household income per month. Overall, the samples in our study fitted our targeted ECD intervention, which was mothers of newborn children from the poverty-stricken area.

The results indicated that mothers living in high-poverty regions of China had strong perceived demand for understanding the risk of poor childhood development. They had limited knowledge on ECD and needed evidence-based information. They were concerned that children may experience health problems such as malnutrition and poor cognitive development. From this perspective, we believe that the general guidance on how to deal with common disease and malnutrition as well as anticipatory guidance on development should be easily accessed in the mHealth program to increase the motivation of use. Furthermore, they realized that parent-child interaction played a vital role in ECD. The participants expressed their interests to learn how to read books and play games with their child. Therefore, education information on those aspects should be maintained in the mobile phone–aided intervention.

In terms of likelihood of action, the participants believed that learning parenting knowledge would help them better understand the child’s needs. However, perceived barriers such as having limited time and resources, different opinions in childcare among family members, and distrust of village doctors exist, which may affect whether participants follow the evidence. Therefore, it might help to expand the mobile health program to incorporate training for local doctors and community health workers, in order to build a trustworthy relationship between parents and local health professionals [36]. In addition, many participants lived with their parents, who may influence their decision making, and several participants reported that their husband or mother-in-law made decisions on childcare. Although some indicated that they would prefer to follow the evidence-based information, they were still concerned that their parents or parents-in-law would insist on the traditional beliefs. A potential approach to address this barrier is to create a shared account among family members so that everyone in a household can learn ECD knowledge. It is worth noting that some participants had a high risk of postpartum depression based on the cutoff point of our screening tool, and messages/sections on postpartum mental health support should be incorporated in the ECD intervention.

As for the feasibility and ideas on the ECD app content, we found most participants had access to the internet and smartphones, thus the mobile health program was feasible for this population. However, the participants had relatively low breastfeeding rate/attempts and low health literacy, and thus multimedia (eg, video, graphics, and text) and plain language should be applied to make the content more understandable. Visuals help make the information more persuasive and memorable [37], which is particularly important for people with low health literacy [38]. Plain language with a concise and organized writing style helps participants better understand the messages. Moreover, since the participants indicated that they didn’t have much time due to childcare responsibilities, text message that took more than 5 minutes to read might be lengthy for them. Prior studies have proven that media-assisted interventions, such as video episodes, can increase the reach and reduce the cost and time for parent education; these are also parents’ strongest preferences [39,40]. Therefore, instead of using text, we plan to create 1-minute videos in the new designed app conveying parenting knowledge and early childhood interventions to decrease learning time and make learning more convenient and attractive for the parents.

We did not specifically interview mothers whose children are at older ages, and their needs for parenting education could be different with those moms who have a newborn child. However, we did interview several moms who already have an elder child, so we were able to retrieve some information on this end.

Our study proposed that mHealth should be considered as one of the techniques in interventions on ECD due to its feasibility, low cost, and potential to cover a wide range of the population. This was the first study to ever understand the beliefs and barriers for a mobile phone–based intervention to improve ECD in a high-poverty region in southwestern China. We conclude that even in poverty-stricken areas, both the technology and the health literacy meet the requirement for the implementation of mHealth programs. Furthermore, this study brought out some potential ideas to make it more feasible when applied on the ground. The results of the study will aid in the development of potentially effective early childhood programs in underserved regions in LMICs.