Each year, 2.5 million newborns die in the first 28 days of life, with 80% of these deaths occurring within the first seven days of age [1]. In resource-limited settings in low- and middle- income countries (LMICs), a majority of these deaths occur within homes largely due to preventable causes such as pneumonia, sepsis, and other illnesses [2,3]. The neonatal mortality rate in Uganda is 26 per 1000 live births [4]. In rural districts in eastern Uganda where 34 in every 1000 newborns die in their first month of life, neonates are even more vulnerable [3]. Many of these deaths could be averted by timely identification and referral to treatment [5,6].

While as many as 84% of neonatal infection-related deaths could be prevented with available interventions, identifying subtle clinical signs and symptoms of severe illness is especially challenging in neonates [5]. Thus, delays in illness identification pose a significant barrier to providing expedient care [6-8]. An analysis of 64 neonatal deaths in eastern Uganda showed that the highest contributing delays to newborn intervention were challenges in problem recognition and in the decision to seek care (50%), with a median time of 3 days from onset of illness to seeking care outside the home [3].

In order to improve recognition of neonatal illness, the World Health Organization (WHO) has developed the Integrated Management of Neonatal and Childhood Illness (IMNCI) guidelines [9-11]. Studies have shown the following seven danger signs predict severe illness in neonates: (1) difficulty feeding; (2) convulsions; (3) lethargy; (4) chest indrawing; (5) respiratory rate of 60 breaths per minute or more; (6) temperature above 37.5°C; or (7) temperature below 35.5°C [9]. Presentation of a danger sign indicates the need for immediate medical attention [9,12]. Although effective identification of these signs at the community level can intercept illness and incite care-seeking behavior capable of impacting child mortality, the tools and training needed to assess indicators of illness are lacking in low-income settings [6,13-17].

Currently, rural health care systems in Uganda and other LMICs rely on village health teams (VHTs) of volunteer community health workers (CHWs) to visit mothers at home to provide postnatal care. While women visited by CHWs are 87% less likely to lose their newborns, these volunteers are often overburdened, and limited by both availability and bandwidth [18]. Thus, CHWs are often unable to complete the three postnatal home visits recommended by the WHO during the critical first week of life. As few as 5% of Ugandan newborns receive a CHW visit in the first 48 hours of life and sick infants are often identified too late to impact survival [19]. Task-shifting neonatal assessment from CHWs to mothers is a promising strategy to reduce neonatal mortality [6,7,20]. While the aforementioned danger signs can have a sensitivity of 85% and specificity of 75% when used by a trained primary care worker to identify neonatal illness, unassisted maternal recognition of these symptoms in community settings has failed to exceed a sensitivity of 20-24% [7,9,21,22].

The opportunity to improve identification of neonatal illness in home settings was identified by a team of faculty and biomedical engineering graduate students at the Johns Hopkins University Center for Bioengineering Innovation and Design (CBID) and School of Public Health, based on over 18 years of community research in rural Bangladesh [23-25]. Engineering design teams had the opportunity to confirm these population research findings through community and rural health system immersion in Iganga, Uganda. Primary observations from ethnographic research conducted in Iganga highlighted delays in identification of neonatal illness due to a systemic reliance on overburdened CHWs. Teams also visited rural communities in Kenya and Bangladesh and observed similar patterns. The design teams identified the need for a low-cost technology to regularly, objectively, and accurately assess the seven previously mentioned IMNCI danger signs during the first week of life to expedite detection of neonatal illness in LMICs. By empowering care providers within the family unit (ie, mothers) to detect signs of illness, the team hypothesized that timely care-seeking behavior could be triggered when it otherwise might not be. Neonatal Monitoring (NeMo), a proposed solution to the need for earlier identification of neonatal illness, is a two-part system designed to empower mothers to effectively assess the seven validated IMNCI danger signs.

The NeMo system consists of a novel, wearable sensing band (the NeMo band) and a low-cost smartphone preloaded with the custom NeMo app (Figure 1). The NeMo band fastens around the neonate’s abdomen using a generic hook-and-loop fastener like Velcro (herein after referred to as velcro). It is equipped with sensors to measure temperature and respiratory rate, which are difficult to assess without appropriate training or tools. These sensors are housed in a small plastic enclosure which must be placed centrally on the neonate’s abdomen (in between the nipples and midway between the umbilical cord and the nipples) (Figure 1). Information from the respiratory and temperature sensors is transmitted to the phone for processing via a standard audio cable which plugs into the audio jack on the phone and via an identical jack embedded into the enclosure on the NeMo band. The app uses audio and visual cues in the local language to enable any mother, regardless of literacy, to assess the qualitative danger signs and place the NeMo band (Figure 2). By responding to a series of interactive audio prompts concerning her newborn’s health, the mother is asked to indicate whether or not her baby has any danger signs by selecting either a check or X on the app interface (ie, “Has your baby been refusing to breastfeed? Press check if yes, press X if no.”). The app does not require internet connectivity. By synthesizing the results from the quantitative and qualitative assessments, NeMo alerts mothers of any danger signs detected and advises users if the newborn requires medical attention.

When using the system, the mother is first taken to a home screen (Figure 2a), which introduces the navigational features of the app. The mother is then guided through individual screens which ask her to assess the presence of the four qualitative danger signs: difficulty breastfeeding (Figure 2b), chest indrawing (Figure 2c), convulsions (Figure 2d), and lethargy (Figure 2e). The app then uses audio instructions and images to walk the mother through each step of properly placing and securing the band around her baby and connecting the band to the phone with the audio cord (Figure 2f-j). Once positioned around the neonate’s abdomen, the NeMo band measures temperature and respiratory rate (Figure 2k). Following the quantitative assessment, the app displays representative images of any danger signs detected and, if indicated, produces a red phone icon to prompt the mother to call a CHW for further assessment and help (Figure 2l). The mother only needs to press the icon to initiate the phone call.

NeMo is based on a CHW-coordinated, phone-sharing framework. In this model, mothers are provided a smartphone by a CHW for the week following birth and purchase their own low cost, wearable band. After the first seven days of the newborn’s life, the phone will be returned to the CHW and the band discarded to prevent the spread of infection between newborns. By making both this technology and education to support clinical evaluation accessible in community settings, NeMo could empower mothers to assess neonatal illness at home and without support. While evidence supporting the acceptability of phone use among CHWs is abundant, evaluation of a community phone-sharing model and maternal smartphone use to enable neonatal mobile health (mHealth) interventions remains lacking [26-28]. Therefore, for any digital health intervention targeting mothers in LMICs directly, there is a need to validate acceptability and usability among target users.

While several studies have established the utility of the seven danger signs in identifying neonates in need of care, the ability of mothers to utilize a simple, digital health strategy to guide effective assessment of these signs has not been established [9,12]. Thus, this study was designed to evaluate the usability and acceptability of the NeMo smartphone app and wearable sensor among both women of child-bearing age and CHWs in the Iganga-Mayuge districts of Uganda. Further, this study seeks to evaluate women’s and CHWs’ perceptions of the phone-sharing model and their willingness to use the NeMo device to evaluate the health of their neonates.

This study was undertaken in eastern Uganda in the Iganga-Mayuge districts, over the course of two weeks, in five villages. These districts were selected for their representation of rural Uganda, offering socioeconomic diversity and corresponding literacy rates [3,29,30]. Fieldwork was supported by Uganda Development and Health Associates (UDHA), which has a regional headquarters based in Iganga. Villages in these districts were visited to capture perceptions of the usability and acceptability of the NeMo system within populations at high risk for vulnerability [29,31]. For both study participants and CHWs, a convenience sampling strategy was utilized. Village leaders randomly selected female citizens from their village within reproductive age (18-49 years old) to be interviewed. CHWs were randomly selected via their connection with UDHA.

This study was conducted following protocols approved by the Johns Hopkins Homewood Institutional Review Board (IRB00005144) and by the Institutional Review Board of Makerere University School of Public Health (MakSPH IRB) in Uganda. Written consent was obtained from all parties including study subjects and CHWs before each interview and was available in Lusoga and English.

The qualitative study consisted of semistructured interviews with all women (n=32) and volunteer CHWs (n=12). With each subject, a preliminary structured interview was conducted to collect demographic information. Subsequently, each subject participated in a verbally posed simulated scenario to assess the usability of the NeMo system. Subjects utilized the NeMo app preloaded onto an Android smartphone, along with a lookalike prototype of the NeMo device. The steps of the simulation were carried out on a NeoNatalie inflatable simulator (Laerdal Medical, Norway). A comprehensive list of the simulated tasks can be found in Table 1.

The usability of the app interface was measured based on the subjects’ ability to navigate through the app and accurately answer the qualitative questions based on the posed scenarios. Additionally, the ability of each participant to position the device on the NeoNatalie’s chest and correctly insert the audio cord into both audio jacks was used as a measure of usability. The usability assessment was further rounded out by a series of closed- and open-ended questions, followed by a series of questions answered using a Likert scale.

Subjects were shown two alternative versions of the band and two alternative versions of the app. They were then asked to indicate their preferences based on their perceptions of ease of use, helpfulness, and safety.

This study was conducted in two phases, reflecting version I and version II of the app interface. The first phase (n=22) followed the methods detailed above and used app version I. The second phase (n=10) utilized version II of the app, which was updated based on qualitative and quantitative results from phase one. All study methods remained consistent between the two phases of the study. Women enrolled in the study were only given either version I or version II of the app.

Statistical significance was determined using a Fisher’s Exact Test to determine subjects’ association with a green check and red X and between mothers utilizing each version of the app. A Likert scale was used to gauge subjects’ opinions on the NeMo system and converted to a composite score (1=Strongly Disagree, 5=Strongly Agree). All other responses were analyzed using descriptive statistics.

Three focus groups, facilitated by 2-3 study team members and a translator, were conducted with groups of 3-5 CHWs. A total of 12 CHWs participated in focus group discussions. Each focus group lasted approximately 60 minutes, with question themes including mothers’ ability to use the system, the feasibility of a CHW-lead training and sensitization initiative, and the acceptability of a community phone-sharing model. Table 2 details key topics discussed with CHWs.

Five English-speaking study team members were trained to conduct the interviews and were provided interview guides. The majority of the interviews were conducted in Lusoga through the aid of two translators, however, some study participants and CHWs were able to understand and respond to the interview questions in English. All interviews were video recorded with participant consent.

The study cohort consisted of 32 women from 5 different villages across the Iganga-Mayuge districts. Women were screened to be of child-bearing age or older, so they ranged from 19 to 44 years of age, with an average age of 27. They had each given birth to a range of 0-6 children, averaging 2.6 children.

Of all the subjects, 30/32 (94%) either owned or had access to a cellphone. However, the majority of these were not smartphones, with only 10/32 subjects (32%) having previously used a smartphone. In acknowledgement of the importance of neonatal danger sign recognition, the Ugandan government has mandated that danger signs be taught in antenatal care education. All subjects enrolled in phase two of the study received antenatal and postnatal education. Two of the women interviewed in phase two of the study did not have children and were not included in this survey analysis. Despite the expectation of proficiency in identifying signs of illness following this training, subjects' retention was low: only 3 of the 8 mothers (37.5%) recognized fever as a danger sign. Furthermore, only 3 mothers were able to correctly identify one additional danger sign (Table 3). These results emphasize the gap in mothers’ education and knowledge surrounding the assessment of neonatal illness and highlight the potential value of an effective tool to guide identification of danger signs in the home setting.

Of all the subjects, 30/32 (94%; P<.001) correctly identified the green check to mean yes and the red X to mean no, while two women were unfamiliar with this symbology. These two women thus had the meaning of each symbol explained to them.

Of the 22 subjects that had never used a smartphone before, 18 (82%) could unlock the phone and open the app without additional guidance and 20 (91%) could do at least one or the other without additional guidance.

Subjects stuck on the homescreen in version I often attempted to press different areas on the screen (including the check mark in the NeMo logo) in an attempt to advance to the next screen. To avoid this challenge, in version II the team added more detailed audio instructions explaining the location of the button to continue to the next screen.

As a result of the poor ability of subjects to answer qualitative questions based on the simulated scenarios using version I of the app, the team evaluated what factors might be at fault. One translator suggested that subjects might be responding in reference to their own infants rather than to the posed NeoNatalie scenarios. The team also hypothesized that questions were being posed in a misleading way based on the positive and negative connotations of the check and X symbols. Originally, the app instructed subjects to press the green check if the baby had the danger sign (“Yes, my baby has this danger sign”) and the red X if the baby did not have the danger sign (“No, my baby does not have this danger sign”). However, it was suggested that subjects might not intuitively associate a green check with illness and a red X with a healthy baby. Thus, in app version II, the team rephrased the app audio, instructing subjects to press the green check if their baby was healthy and did not have the sign and the red X if their baby was sick and did have the sign.

Finally, while most women using version I of the app were able to correctly place the device, the NeMo app’s audio cues were updated to further assist subjects with device placement. Greater detail was given about anatomical structures on the newborn, specifying the position of the device relative to the nipples and umbilical cord in version II.

In addition to assessing the usability of the band and the phone app, subjects were asked questions regarding the amount they would be willing to pay for the device. A total of 30/32 subjects (94%) were willing to make a one-time payment to purchase a NeMo band usable for the seven days following birth. The amount they were willing to pay varied between 1,000 to 100,000 Ugandan shillings (UGX), or approximately $0.27 to $27.00 United States dollars (USD). The median price was 10,000 UGX, or approximately $2.70, and the mean was 19,000 UGX, or $5.14. Figure 8 illustrates the distribution of prices subjects indicated they would pay for the NeMo band. While most subjects indicated they would be willing to pay for the device, participants were less receptive to the prospect of making a security deposit to borrow the smartphone. In total, 20/27 subjects (74%) indicated they would be comfortable making a security deposit they could collect after the device was returned.

Three focus groups conducted with a total of twelve CHWs highlighted critical logistical considerations for the successful implementation of the NeMo system in Uganda. Through these conversations, it was confirmed that CHWs are expected to visit mothers three times during the first week of life. Additionally, CHW assessments of critical quantitative danger signs, including respiratory rate and temperature, are most often only qualitative in nature and are conducted by visually observing the baby’s breathing as well as touching the baby to determine temperature. Regarding environmental factors, CHWs were optimistic about mothers’ abilities to charge the phones, especially considering that most families own at least one phone and solar power is becoming increasingly more available in village homes.

While early identification of neonatal illness can impact neonatal mortality rates and reduce the burden of treatment, identifying subtle clinical signs and symptoms of possible severe illness is especially challenging in neonates. Thus, delays in illness identification pose a significant barrier to providing expedient care. Implementation of IMNCI and danger sign recognition have been shown to improve inequities in neonatal mortality and care-seeking behavior; however, the success of such interventions may be limited by the consistency of CHW follow-up and training [15,16,32]. Further, the current lack of sensitivity in maternal screening limits the feasibility of at-home danger sign screening [7]. The proposed NeMo system has the potential to empower mothers in community settings to assess their neonates’ health from home and without support. Further validation of mothers’ use during the newborn’s first week of life, behavior change catalyzed by triage information, and integration into the existing health care system must still be conducted.

When asked about their prior training in antenatal care, all subjects indicated they had received some antenatal education. However, the observed discrepancy between training and recollection of danger signs emphasizes the value the NeMo device might add as an educational tool and assessment aid.

Although many subjects had never used a smartphone before the study, the majority were able to unlock the phone and navigate through the NeMo phone app without external guidance. Improvements made in the app between version I and version II pointed to the importance of precise verbal instruction. Small changes to how questions were framed and how instructions were articulated, without any change to the visual appearance of the app, significantly increased subjects’ ability to correctly answer qualitative questions based on a posed scenario (Figure 4). In version II of the app, no subjects required prompting or further guidance from the study team to navigate the app, despite their unfamiliarity with smartphones prior to the study (Figure 3). This indicates sufficient verbal instruction and intuitive app design can overcome the barrier of introducing unfamiliar technology. Further, these results suggest lack of prior smartphone experience may not be a barrier to successful use of the NeMo system.

Many participants expressed their trust in the device’s evaluation, with agreement to the statement, “I would trust this device to assess the health of my baby,” receiving a score of 4.66 on the 5-point Likert scale. Women also indicated willingness to engage with the proposed system of phone-sharing. This acceptance suggests NeMo and other such frameworks may be acceptable in low-income, eastern Ugandan communities. While 29/32 subjects (91%) thought the device was easy to use, perceptions around learnability were not as conclusive. In addition to these results, unanimous desire for CHW support indicates the value of engaging CHWs in training programs and device management in order to successfully shift neonatal assessment to the home setting.

One challenge facing implementation of the NeMo system is creating trust in the device output sufficient enough to trigger behavior change and promote care-seeking. While care-seeking is a multifactorial and complex decision, preliminary questions were asked to gauge the subjects’ initial response to the NeMo system. In order to assess the ability to incite care-seeking, the team asked subjects how they would respond if the device said the baby was sick when they believed it was healthy. As 29/32 (91%) women indicated they would seek care from a CHW or hospital, this suggests that the NeMo device may be able to trigger a behavior change to reduce time to care for sick neonates. One potential risk posed by the device is mothers ignoring their own intuition and relying too heavily on NeMo. The results of this study indicated 24/32 subjects (75%) would take their baby to a facility if they thought it was sick, even if the device indicated it was healthy. This majority suggests most mothers retain autonomy in their decision making.

The ability of the NeMo device to improve care-seeking for neonates will rely heavily on the accessibility of the technology in the home setting for LMICs. In order to successfully transition assessment from CHWs to mothers, the smartphone and device must be available at a low enough cost to enable use in low-income households. In the eastern Ugandan districts where the study was conducted, subjects indicated a median price point of 10,000 UGX ($2.70) for the sensor. To improve accessibility of the smartphone, the study team developed a community phone-sharing framework in which CHWs would manage phones for all mothers in their village to borrow in the week following birth. Interviews with CHWs revealed concern about coordinating a phone-sharing network if the CHW is receiving payment directly, and some CHWs worried about the phones being lost or broken. However, focus groups also emphasized the ability of community sensitization to educate mothers on the value of the device so as to minimize these risks and increase likelihood of purchasing the device.

Despite the promising acceptability and usability results in this study and potential for improved maternal recognition to reduce neonatal mortality, the study revealed several barriers to implementing and scaling the NeMo system. One of these issues is the phone-sharing framework, which attempts to minimize the barriers to technology in rural, low-income regions by distributing phones on a community-by-community basis. However, in order to initiate the implementation of this phone-sharing system, government or private foundations will need to provide funding and support. Without their aid, the likelihood of the populations studied herein accessing this technology is low until smartphone penetration increases substantially. Although implementation of the NeMo system could significantly increase the frequency of neonatal screening, the device distribution system proposed herein would continue to require CHW interaction and support. Incorporation of distribution and training in existing antenatal care programs could enable this contact for many women. As access to antenatal care is limited in many areas, alternative intervention strategies centered around the opportunity for training and distribution around the time of birth may be employed to overcome barriers to CHW access. Further, while this study demonstrates promising results regarding perceived intent to act on a recommendation provided by the device, past researchers have found that barriers, including money and transportation, often bar mothers from following through with referrals to health facilities [3,33]. Even after a mother acts on NeMo’s recommendation to seek care, infrastructural or other barriers may prevent sick newborns from receiving the necessary treatments to improve neonatal mortality. This suggests initial intervention in a region with more robust infrastructure may foster a greater impact.

While the study was conducted in multiple subcounties and villages within the Iganga-Mayuge districts with a notably wide range of socioeconomic backgrounds, the limited sample size prevents greater generalizability of the study results. When posed hypothetical situations of neonatal illness during the simulated scenarios, a few subjects were noted to have responded about the condition of their own child instead of following the study member’s prompts, leading to incorrect responses to the qualitative questions posed by the app. Due to the use of a convenience sampling methodology, there was potential for positive bias and interviewer bias. Steps, including the use of structured questionnaires and predetermined simulated use procedures, were implemented to minimize this bias. Trained local interpreters were used to minimize interviewer bias. Because two villages overlapped between implementation of version I and version II of the app, there exists the possibility of study contamination between these populations; however, as the NeMo system never remained with the women beyond the interview and study members gave no indication of actions or questions being completed correctly, the risk of contamination was considered to be very low.

The results of this study indicate that women are receptive to and capable of using NeMo to screen their newborns for neonatal danger signs. This study suggests that the NeMo system could be an acceptable, easy to use resource capable of expediting identification of neonatal illness to decrease delays in care-seeking behavior. While additional studies are required to assess the NeMo system’s ability to safely and effectively decrease delays in accessing neonatal care, this system, if further developed, could be an approach to decrease preventable neonatal deaths by empowering and educating mothers to detect danger signs in their newborn.