Prompt: Tech for Maternity
- WOTE, Kenya – Leah Kyalo, a 23-year-old new mother, sits in her living room in Wote, Makueni County, cradling her six-week-old daughter, Oliva.
- Kyalo, who was enrolled in the program during her second week of pregnancy, receives practical information, appointment reminders, and alerts about potential warning signs.
- The program, launched in 2017, aims to reduce maternal mortality by providing crucial information to young and expectant mothers.
Text Message Program Aims to Lower Maternal Mortality in Kenya
WOTE, Kenya – Leah Kyalo, a 23-year-old new mother, sits in her living room in Wote, Makueni County, cradling her six-week-old daughter, Oliva. Kyalo says she finds reassurance in the regular text messages she receives from Jacaranda Health, a non-profit association. These messages offer advice on infant care, nutrition, and address common concerns for new mothers.
Kyalo, who was enrolled in the program during her second week of pregnancy, receives practical information, appointment reminders, and alerts about potential warning signs. She can also text questions and receive answers within minutes.She recalls using the service when experiencing vomiting after taking iron supplements, and again when she noticed blood, at which point the program urged her to seek immediate hospital care.
The program, launched in 2017, aims to reduce maternal mortality by providing crucial information to young and expectant mothers. Despite progress in healthcare, KenyaS maternal mortality rate remains high. According to the World Health Organization (WHO),Kenya recorded 530 maternal deaths per 100,000 live births in 2020. The WHO estimates this rate has increased since 2017. This is far from the United Nations’ Lasting Growth Goal of reducing the global average to less than 70 maternal deaths per 100,000 births by 2030.
Kenya’s vast and largely rural territory, coupled with inadequate road infrastructure, contributes to the problem. Factors influencing maternal mortality include limited resources in healthcare facilities,long distances to reach them,and lack of transportation.
Specialists frequently enough refer to “the three delays” in maternal care: delay in deciding to seek care, delay in reaching a healthcare facility, and delay in receiving adequate care once there. The text message program addresses the first delay, ensuring women seek timely medical attention when necessary.
Answering Mothers’ Questions
“I ask a lot of questions,” Kyalo admits. Melody Muyula, an assistance agent at Jacaranda Health in Nairobi, analyzes hundreds of messages daily. Questions arrive in English or Kiswahili, ranging from concerns about bleeding during pregnancy to how to keep a baby comfortable in hot weather, or dealing with infant diarrhea.
The platform receives between 4,000 and 15,000 messages daily from several counties in Kenya. In 2024, the platform recorded 650,000 mothers.
Jacaranda Health began as a maternity hospital in 2011, founded by Nick Pearson with the goal of providing high-quality maternal care in Nairobi’s peri-urban areas. Initially, the team launched a simple SMS reminder service for appointments and satisfaction surveys. However, a technical error led to a two-way interaction system, prompting mothers to send in their questions. Jay Patel, the technological director, recalls the rapid increase in message volume, quickly exceeding 1,000 per day, which necessitated the development of a complete assistance service.
To manage the influx, an initial AI program was integrated in 2019 to provide pre-written responses and prioritize texts based on urgency.
A more advanced version, Ulizallama, was completed in 2024 with support from the Google foundation and the International Development Research Center. Ulizallama is a large language model (LLM) in Kiswahili, capable of understanding and generating text. The system recognizes keywords in mothers’ questions and provides automated responses tailored to their stage of pregnancy or the baby’s age. Human agents intervene if the mother is unsatisfied with the AI’s response, which Jacaranda Health estimates occurs in 30% of cases. In approximately 7% of cases, the system identifies danger signs and automatically schedules a call with a healthcare professional.
Muyula recalls a case where a seven-month pregnant woman,recently enrolled in the program,had been experiencing important bleeding for three days without seeking medical attention. ”During the call, she told me innocently, ‘I thought it was normal!'” Muyula said. “We realized that this future mother did not have the knowledge of the signs of danger to monitor during pregnancy.”
Muyula emphasizes that technology alone is insufficient for handling sensitive situations. “Having worked with the system and having seen its improvement over time, I can say that the accuracy of the AI has really increased, as is its transparency, its ability to account for and its complementarity,” she said.
Data-Driven Reliability
The program’s reliability stems from its use of local health data. The thousands of messages received daily provide a rich source of information for training and refining Ulizallama. The program is available in open source for those who wish to use or adapt the model.
“I think that AI will be able to provide care to the populations who usually do not have access to it and allow health systems to serve their customers a little better,” Patel said. The program also operates in hausa, Yoruba, Xhosa, and Zulu, and has been deployed in Eswatini (2021), Ghana (2022), Nigeria, and Nepal (2024).
Studies suggest the program is effective. A 2024 study published in *BMC Pregnancy and Childbirth* found that mothers enrolled in the program were twice as likely to attend postnatal visits.Another study published in *PLOS* in 2025, involving over 6,000 pregnant women, indicated that the program improved their knowledge, preparation for childbirth, newborn care, and awareness of concerning symptoms.
Remaining Challenges
Currently, the program is available in 23 of Kenya’s 47 counties, but not yet in the northern semi-arid regions, which face the greatest challenges in accessing care. In Garissa, mandera, and Wajir, the percentage of women attending at least four antenatal visits is substantially lower than the national average. Child mortality rates are also higher in these regions.
Jacaranda health acknowledges the challenges of expanding into these areas. Operational difficulties include the nomadic nature of the population, which complicates enrollment, and the need for a robust network of healthcare facilities, which is lacking in northern Kenya.
Even in Makueni County, where Leah Kyalo resides, challenges persist, including food insecurity and a hilly terrain that isolates communities during the rainy season. Christine Muteria, head of maternity at Wote’s reference hospital, notes that cultural beliefs and superstitions also contribute to delays in seeking care.
“There is a fear of certain women to come to the hospital, a mixture of superstitions and culture. They don’t want to bring their pregnancy bad luck,” she said. Despite these challenges, Muteria notes that the situation has improved, and the program has helped reassure mothers about medical follow-up. In Makueni County, the maternal mortality rate has
Text Message Program Aims to Lower Maternal Mortality in Kenya: Your Questions Answered
In Kenya, a groundbreaking program using text messages is making strides in reducing maternal mortality rates by providing crucial information and support to expectant mothers. Let’s delve into the details of this innovative initiative with a Q&A-style guide to help you understand its impact.
What is this Text Message Program All About?
This program, run by the non-profit Jacaranda Health, leverages SMS technology to provide vital information, appointment reminders, and alerts to pregnant women and new mothers in Kenya. Through regular messages and a two-way communication system, the program aims to reduce maternal mortality by addressing the challenges related to accessing timely healthcare.
How Does the Text Message program Work?
The program uses a large language model (LLM) called Ulizallama,which is developed in Kiswahili,to understand and respond to questions from mothers. Here’s a breakdown:
- regular Messages: Offer advice on infant care,nutrition,and address common concerns for new mothers,as well as pregnancy advice.
- two-Way Communication: Mothers can text questions and receive answers within minutes.
- AI-Powered Responses: Ulizallama recognizes keywords and provides automated responses tailored to the mother’s stage of pregnancy or baby’s age.
- Human Intervention: Human agents intervene if the AI’s response is unsatisfactory or in case of medical urgency.
Why is Maternal Mortality a Significant Problem in Kenya?
Maternal mortality remains high in Kenya. In 2020, Kenya recorded 530 maternal deaths per 100,000 live births, which is significantly higher than the United Nations’ goal of less than 70 deaths per 100,000 births by 2030. Several factors contribute to this, including:
- Geographic Challenges: Vast, largely rural areas and inadequate road infrastructure.
- Healthcare Access: Limited resources, long distances to reach facilities, and lack of transportation.
- “Three Delays”: Delays in deciding to seek care, reaching a healthcare facility, and receiving adequate care once there.
How Does the Program Address the “Three Delays”?
The text message program primarily addresses the first delay – the delay in deciding to seek care. By providing information about warning signs and encouraging timely medical attention, it helps women make informed decisions and seek care when needed. the reminders also address the other delays.
What Services Does the Text Message Program Offer?
the program aims to address issues like:
- Practical information.
- Appointment reminders.
- Alerts about potential warning signs.
- A platform to text questions and receive answers.
What are the Key Features of Ulizallama?
Ulizallama is a large language model (LLM) designed to understand and generate text in Kiswahili. Key features include:
- Keyword Recognition: identifies keywords in mothers’ questions.
- Automated Responses: Provides tailored responses based on the stage of pregnancy or baby’s age.
- Human Oversight: Human agents can intervene if the automated response is unsatisfactory.
- Danger Sign Alerts the system identifies danger signs and automatically schedules a call with a healthcare professional in approximately 7% of the cases.
How Accurate is the AI in Providing Healthcare Information?
The accuracy of the AI, Ulizallama, has improved over time.The Jacaranda Health team ensures that human agents intervene when the AI’s response is unsatisfactory, which happens approximately 30% of the time. This human oversight guarantees the quality and relevance of the information provided.
What Types of Questions Do Mothers Ask Through the program?
Mothers ask a wide range of questions. They frequently enough include concerns about:
- Bleeding during pregnancy.
- Keeping babies cozy in hot weather.
- Dealing with infant diarrhea.
- Infant care and nutrition.
- Common pregnancy concerns.
What is the Program’s Data-Driven Reliability?
The program heavily relies on local health data,and draws information from daily incoming messages,which provides a rich source for training and refining Ulizallama.
Is the Program Available in Other Languages and Countries?
In addition to Kiswahili and English, the program operates in:
- Hausa
- Yoruba
- Xhosa
- Zulu
The program has also been deployed in various countries, including:
- Eswatini (2021)
- Ghana (2022)
- Nigeria
- Nepal (2024)
Has This Program Been Proven Effective?
Yes, studies suggest that the program is effective. For example:
- A 2024 study in *BMC Pregnancy and Childbirth* found that enrolled mothers were twice as likely to attend postnatal visits.
- A 2025 study in *PLOS*, involving over 6,000 pregnant women, showed the program improved knowledge, preparation for childbirth, newborn care, and awareness of concerning symptoms.
What are the Challenges to Expanding the Program?
While the program has made significant strides, there are still challenges:
- geographic Limitations: Currently available in 23 of Kenya’s 47 counties, but not yet in the northern semi-arid regions.
- Operational Difficulties: Nomadic populations complicate enrollment.
- Infrastructure Gaps: The need a robust network of healthcare facilities, which is lacking in northern Kenya
- Cultural and Societal Factors Cultural beliefs and superstition in seeking medical care can be a factor in delayed care.
- Food Insecurity: Food insecurity and the hilly terrain isolate communities during the rainy season.
How Does the Program Impact Mothers’ Awareness of Medical Follow-up?
The program has helped reassure mothers about medical follow-up, improving their awareness.while the situation has improved, challenges persist, including superstitions, cultural beliefs, and difficulties during the rainy season.
what is the Role of Human Intervention?
human agents play a critical role. They intervene when the AI’s automated response is unsatisfactory, which occurs in about 30% of cases. They ensure the quality of the provided information and intervene in medical emergencies.
can this program be used elsewhere?
Ulizallama is available in open source for those who wish to use or adapt the model.