How to reduce participant attrition in a longitudinal study

Studying maternal and newborn health delivery

The Maternal and Newborn Health (MNH) E-Cohort study collected longitudinal data on maternal and newborn health quality in near-real time. The longitudinal data focused on antenatal care, delivery, and postnatal care for newborns and mothers. This data complemented administrative data and routine household surveys that comprise the majority of data on health service quality. Through this study, we aimed to further understand care processes and inform future health service improvements.

This study was conducted in Kenya, Ethiopia, South Africa and India. KEMRI Wellcome Trust and Laterite led the study in Kenya. Together, we conducted the study over the course of a year, from mid-2023 to mid-2024. This study recruited 1,006 women, with 500 from Kiambu and 506 from Kitui counties.

The different study modules

We divided the study into five modules. Modules 1 and 5 were conducted through in-person interviews, while modules 2 to 4 were conducted by phone. Laterite led the modules that involved phone surveys.

• Module 1 – In-person interviews with pregnant women having their first antenatal care visit at the health facilities. This focused on obtaining respondents’ health and obstetric history, content of care, and health systems rating of the health facility visited.
• Module 2 – Monthly phone interviews with pregnant women until delivery, stillbirth or miscarriage. The aim of this was to understand the care women received during pregnancy as well as the new consultations, medications, and costs of care.
• Module 3 – A phone interview after the woman has given birth or experienced a miscarriage or stillbirth. This focused on understanding additional health care women received prior to delivery, during labor and delivery, and following delivery; maternal and newborn complications, and post-partum health. Questions on pregnancy loss were also asked to the relevant respondents.
• Module 4 – A phone interview to understand new mothers’ post-natal care. The focus was to understand the health of the mothers and newborns, and the care the woman and baby received since delivery, including user experience, costs, and new medications/vaccines/supplements.
• Module 5 – In-person interview conducted at the woman’s nearest health facility with the aim of understanding additional care the woman and baby received after birth. In addition, a health assessment for both the mother and baby were done.

Reducing participant attrition

The biggest challenge with longitudinal studies is attrition. Attrition refers to the loss of study participants from a sample. Another challenge we faced was tracking the large number of study participants and knowing which module to administer and when.

We employed several strategies to tackle these two challenges. One strategy we employed to reduce attrition was allocating one enumerator to each respondent throughout the study. Having one enumerator per participant allowed for good rapport between the respondent and enumerator. It also saved time during interviews, since enumerators didn’t have to repeatedly introduce themselves or the study. Participants also felt more comfortable, making the survey smoother and more efficient.

We also ensured that interviews were conducted at the participants’ preferred schedule. Respondents would suggest a preferred time or day of the week that they were comfortable being called and enumerators would record and implement this during the next interview.

Finally, we provided reimbursements to respondents for participating in the interviews in the form of airtime. Airtime is prepaid credit or time purchased for making phone calls, sending text messages, or using mobile data on a cell phone. Initially, we faced an issue with participants assuming they didn’t receive their airtime. After we implemented an intervention to ensure participants were notified when their airtime was sent, the attrition rate significantly decreased.

Improving participant tracking

To address the second challenge, we developed a comprehensive dashboard to keep track of every participant. This central dashboard consolidated information from the individual dashboards we created for each module. It contained information on the pregnancy status of each respondent, their delivery date, their assigned enumerator and the last module administered to them. Importantly, it highlighted whenever a module was due to be administered to a respondent. This allowed us to send out weekly lists to enumerators with information on the respondents and the particular module to be completed. The centralization of data helped monitor the surveys and ensured that we administered the right module at the right time.

Outcome

The strategies we employed allowed us to interview a large number of sampled participants, contributing to the richness and robustness of data collected. This helped the team to thoroughly investigate the quality of detection and management of pregnancy risk factors during antenatal care in Kenya.

There were 1,006 pregnant women enrolled in this study, and 839 women completed all four study modules. Fifty-four women experienced miscarriages or stillbirths and concluded their participation in the study after module 3. These women are not included in attrition calculations, as they completed all studies relevant to their personal circumstances. Therefore, the total number of respondents who had completed all study modules relevant to their circumstances was 893. The attrition rate from the start of the study until the end of module 4 was 11%.

Some common reasons for attrition were mobile phone connectivity issues and respondents’ unwillingness to continue participating in the study. While our strategies were successful, additional measures could be implemented in future studies to further reduce attrition, particularly by addressing issues related to mobile connectivity and providing additional support for participants in challenging circumstances. This support could be through providing counselling services and social welfare checks on women who seem to be undergoing emotional distress and also those who have experienced stillbirth or miscarriage.