Study design
This study utilizes data collected from a prospective pilot cohort study, known as Mobile Solutions for Women, Adolescents, and Children’s Health: Neonate (Mobile WACh NEO Pilot). The Mobile WACh NEO Pilot study, designed to support maternal and infant outcomes by promoting facility delivery, infant survival and family planning uptake, enrolled 800 pregnant women seeking antenatal care services from two public facilities in Kenya: Mathare North Health Centre (Nairobi County, peri-urban) or Rachuonyo Sub-County Hospital (Homa Bay County, rural) from December 2017 to January 2019. These two public facilities offer a wide range of services for antenatal, birth center and postpartum care as well as have a high volume of daily antenatal appointments (> 10 new mothers per day) and serve a large community of low-income women and babies at high risk of neonatal morbidity and mortality. The Nairobi site serves women living in a large urban slum, and the Western Kenya site serves women living in a low-income rural area.
Mobile WACh NEO Pilot participants received pre-programmed SMS messages from enrollment during pregnancy until 14 weeks postpartum. Mobile WACh NEO is designed for maximal impact on neonatal morbidity. Thus, this intervention was implemented at the time when women and their infants are most at risk for experiencing morbidity and mortality, i.e., the last ANC visit (30–36 weeks) to the first postpartum visit (often delayed to approximately 14 weeks postpartum) in order to support and augment perinatal care in this critical period. Content and frequency of pre-programmed messages were dependent upon the woman’s pregnancy status and were delivered in the participant’s preferred language and time of day. Message content was not altered based on depression status. These messages targeted specific actionable health outcomes and encouraged engagement with the nurse. Participants could communicate with the nurse via SMS at any time free of charge. Study nurses managed the bidirectional SMS communication and used national guidelines and local practice standards for the care of pregnant/postpartum women and their infants when responding to participants’ questions. Prior to the start of the study, nurses were trained in these study procedures by two OBGYN physicians (Drs. Kinuthia and Unger) and the study nurse coordinator (Brenda Wandika). These physicians and nurse were available at all times for consultation. In addition, a twice monthly review of all messages was performed by Dr. Unger, and a message review implemented with the team of nurses. Two interviews, one at the time of enrollment and one post-intervention, were conducted with participants to collect demographic, health outcomes, and study-related information.
mHealth intervention
Mobile WACh NEO is a two-way SMS communication intervention designed to engage women with a health care worker at their local clinic with the aim of improving maternal and neonatal outcomes. Messages are personalized, behavioral theory based and action oriented specific to the time point in pregnancy or postpartum. Participants are encouraged to reply to all messages they receive and initiate their own spontaneous messages throughout the study. A study nurse managed SMS communication with participants.
Schematic 1 SMS messaging content and frequency
Period
|
Pregnancy
(Enrollment – Delivery)
|
Early neonatal period
(Delivery- 4 weeks infant age)
|
Postnatal period
(4 weeks – 12 weeks infant age)
|
SMS frequency
|
Weekly
|
Daily × 1 week ➔ Every other day × 3 weeks
|
Twice a week
|
Topics
|
Birth preparation
|
Infant and maternal health evaluations
|
Infant health and family planning
|
Study population
Pregnant women seeking ANC services from the two sites were recruited to participate in the Mobile WACh NEO Pilot intervention. This source population encompassed both rural and peri-urban areas, an ethnically diverse population, and areas with generally low socioeconomic status and high neonatal mortality. Pregnant women were eligible if they had daily access to a mobile phone, were ≥ 14 years of age, and were between 30 and 36 weeks gestation. If a woman was not sufficiently literate but had access to a partner or family member whom she would be comfortable having read her messages, she was eligible for the study. Pregnant women were recruited by community health workers who introduced the study to potential participants, answered questions, and invited women to participate. We introduced the study to all women attending ANC visits at the two clinics between December 2017 and May 2018, and using a convenience sample, we recruited those who agreed to participation and met inclusion criteria. Women were recruited and enrolled on the same day at these two facilities. It was emphasized that participation was completely voluntary and would not in any way affect their antenatal, postnatal, or infant care services. Women who were referred and willing to participate were given a screening questionnaire in order to assess eligibility. Oral consent was obtained for participation in screening. Eligible women who agreed to participate and receive SMS messages provided written informed consent and were entered into the Mobile WACh system along with their preferences for SMS message delivery. Eligible women who chose not to participate were asked their reasons for non-participation, with responses recorded in the screening questionnaire. Women who screened positive for depressive symptoms were referred to available mental health resources, which was the same referral process utilized by the ANC clinics for women not included in the study and typically involved a social worker. Researchers were blinded to a woman’s depression status until after the study concluded; thus, referrals were made within the standard process of the clinical site without interference from the study itself.
Data collection
Women were followed during pregnancy and for 14 weeks postpartum. Participants were administered a standardized questionnaire at enrollment and one follow-up visit (at 14 weeks postpartum) using a tablet-based system (Open Data Kit, ODK) [17]. Exit surveys were conducted either in-person or via telephone, but data on depressive symptoms was not gathered during the phone exit surveys. We collected patient information including questions pertaining to demographics, medical history, experience with SMS and technology, and depression. The Abuse Assessment Screen (AAS) was used to evaluate maternal experience of violence based on participant reports of experiencing physical abuse during the current pregnancy [18]. The AAS has been used as a measure of intimate partner violence in previous studies but is not specific to abuse inflicted by a sexual partner [19]. Undesired pregnancy was defined as the mother reporting she did not want to have a/another baby at the time of becoming pregnant with the current pregnancy. History of miscarriage was a binary variable consisting of women who had reported at least one spontaneous abortion prior to current pregnancy. SMS communication was collected continuously in the Mobile WACh platform throughout the entirety of the study period.
Infant and delivery information
Preterm births were defined as delivery prior to 37 weeks estimated gestational age. The type of delivery was classified as either a vaginal delivery, planned Cesarean section (C-section), or unplanned/emergency C-section, based on self-report. Infant morbidity was defined as a mother affirming her child had been to any clinic/hospital for any illness after delivery but before follow-up at 14 weeks or her child had been admitted to the hospital after delivery. Infant mortality was based on maternal report of infant death.
Depression status
The prevalence of depressive symptoms was assessed by dichotomizing self-reported participant Edinburgh Postnatal Depression Scale (EPDS) scores into: < 10, categorized as “no depression” and ≥ 10, categorized as “depression”. For the purposes of this study, we stratified women in this cohort into four patterns of depressive symptoms: 1) antenatal depression (EPDS score ≥ 10 at enrollment), 2) postpartum depression (EPDS score ≥ 10 at follow-up), 3) persistent perinatal depression (EPDS score ≥ 10 at both time-points) and 4) any perinatal depression (EPDS score ≥ 10 at any time-point) [3, 20]. Therefore, women with postpartum depression could represent new-onset postpartum depression (incident depression) or continuation of depression from the antepartum period [20], while any perinatal depression could represent women with antenatal depression only, postpartum depression only, or persistent perinatal depression. Any perinatal depression is meant to account for the fact that depressive symptoms during the perinatal period are dynamic and have historically not been well-characterized longitudinally but rather only in the antenatal or postpartum periods [3].
SMS message data
All SMS messages sent to and received from participants were recorded in the Mobile WACh messaging platform. The sender of the message was classified as the system, the participant, or the nurse. For the purposes of this analysis, only messages that originated from the participants were used to assess level of interaction with the two-way messaging system. Character counts of messages were documented based on the original text messages sent. Participant SMS messaging by participants was categorized into ever having sent ≥1 SMS during the study period vs. never having sent an SMS. Messages whose character count was ≥10 characters were classified as long SMS messages, in an effort to capture conversational messages rather than messages that simply acknowledged receipt of system message. If a participant sought advice from the study nurse, this initiative was considered a nurse consult in this study.
Statistical analyses
Statistical analysis was performed using the program R studio version 1.2.5001 (Boston, 2019). All tests were considered statistically significant at an alpha level of 0.05.
Descriptive analyses for baseline sociodemographic and obstetric characteristics of women who completed a second EPDS at follow-up were conducted.
Correlates of depression were identified using univariable and multivariable generalized estimating equation (GEE) with Poisson link clustered by facility, with exchangeable correlation structure and robust standard errors. Poisson regression was used to generate effect estimates that could be interpreted as relative risk despite common outcomes. GEE was used to account for similarities by site. Multivariable models were run to control for potential confounders, determined a priori based on literature review: adolescent age (< 20), marital status, education level, employment status, HIV status, monthly income, number of living children, primigravid status, history of miscarriage, experience of abuse during pregnancy, pregnancy desire, and previous family planning use. The main analysis identified correlates of any perinatal depression and secondary analyses were conducted on correlates of each pattern of depression. For the continuous variable of monthly income, results were interpreted as the relative risk of having the depression comparing mothers with 1000 KSh (~ 10 USD) difference in monthly household income. The relative risk for the variable “living children” was interpreted as relative risk of depression associated with one additional living child.
Association of antenatal depression as a predictor of infant outcomes of preterm birth, infant morbidity, and infant mortality was analyzed using GEE with Poisson link, clustered by facility, with exchangeable correlation structure and robust standard errors. Univariable and multivariable regressions were run. Confounders in multivariable analyses were chosen a priori from literature review and included adolescence, marital status, education level, employment status, abuse during pregnancy, HIV status, monthly income, distance from clinic, type of delivery, location of delivery, complications during pregnancy, and complications during delivery.
We evaluated association of depression with several measures of participant engagement with the SMS platform: a binary outcome of any SMS message sent over the study period; the total number of SMS messages sent; and the total number of long SMS messages (≥10 characters). Each measure of engagement was compared between women with and without depression using univariable and multivariable GEE with Poisson link, clustering by facility, exchangeable correlation structure and robust standard errors. Only antenatal and persistent perinatal depression were included as depression patterns so that depressive symptoms preceded the outcomes of interest. All women with persistent perinatal depression were included in the antenatal depression cohort given the definitions described above.
Chi-squared tests were used to assess differences in study nurse consultation between women with and without depression. Again, only antenatal and persistent perinatal depression were included as these depression patterns preceded the outcomes of interest. Fisher’s exact tests were used when sample sizes were too small to get accurate results from Chi-squared tests.