Towards stronger antenatal care: Understanding predictors of late presentation to antenatal services and implications for obstetric risk management in Rwanda

Background Early antenatal care (ANC) reduces maternal and neonatal morbidity and mortality through identification of pregnancy-related complications, yet 44% of Rwandan women present to ANC after 16 weeks gestational age (GA). The objective of this study was to identify factors associated with delayed initiation of ANC and describe differences in the obstetric risks identified at the first ANC visit (ANC-1) between women presenting early and late to care. Methods This secondary data analysis included 10,231 women presenting for ANC-1 across 18 health centers in Rwanda (May 2017-December 2018). Multivariable logistic regression models were constructed using backwards elimination to identify predictors of presentation to ANC at ≥16 and ≥24 weeks GA. Logistic regression was used to examine differences in obstetric risk factors identified at ANC-1 between women presenting before and after 16- and 24-weeks GA. Results Sixty-one percent of women presented to ANC at ≥16 weeks and 24.7% at ≥24 weeks GA, with a mean (SD) GA at presentation of 18.9 (6.9) weeks. Younger age (16 weeks: OR = 1.36, 95% CI: 1.06, 1.75; 24 weeks: OR = 1.33, 95% CI: 0.95, 1.85), higher parity (16 weeks: 1–4 births, OR = 1.55, 95% CI: 1.39, 1.72; five or more births, OR = 2.57, 95% CI: 2.17, 3.04; 24 weeks: 1–4 births, OR = 1.93, 95% CI: 1.78, 2.09; five or more births, OR = 3.20, 95% CI: 2.66, 3.85), lower educational attainment (16 weeks: primary, OR = 0.75, 95% CI: 0.65, 0.86; secondary, OR = 0.60, 95% CI: 0.47,0.76; university, OR = 0.48, 95% CI: 0.33, 0.70; 24 weeks: primary, OR = 0.64, 95% CI: 0.53, 0.77; secondary, OR = 0.43, 95% CI: 0.29, 0.63; university, OR = 0.12, 95% CI: 0.04, 0.32) and contributing to household income (16 weeks: OR = 1.78, 95% CI: 1.40, 2.25; 24 weeks: OR = 1.91, 95% CI: 1.42, 2.55) were associated with delayed ANC-1 (≥16 and ≥24 weeks GA). History of a spontaneous abortion (16 weeks: OR = 0.74, 95% CI: 0.66, 0.84; 24 weeks: OR = 0.70, 95% CI: 0.58, 0.84), pregnancy testing (16 weeks: OR = 0.48, 95% CI: 0.33, 0.71; 24 weeks: OR = 0.41, 95% CI: 0.27, 0.61; 24 weeks) and residing in the same district (16 weeks: OR = 1.55, 95% CI: 1.08, 2.22; 24 weeks: OR = 1.73, 95% CI: 1.04, 2.87) or catchment area (16 weeks: OR = 1.53, 95% CI: 1.05, 2.23; 24 weeks: OR = 1.84, 95% CI: 1.28, 2.66; 24 weeks) as the health facility were protective against delayed ANC-1. Women with a prior preterm (OR, 0.71, 95% CI, 0.53, 0.95) or low birthweight delivery (OR, 0.72, 95% CI, 0.55, 0.95) were less likely to initiate ANC after 16 weeks. Women with no obstetric history were more likely to present after 16 weeks GA (OR, 1.18, 95% CI, 1.06, 1.32). Conclusion This study identified multiple predictors of delayed ANC-1. Focusing existing Community Health Worker outreach efforts on the populations at greatest risk of delaying care and expanding access to home pregnancy testing may improve early care attendance. While women presenting late to care were less likely to present without an identified obstetric risk factor, lower than expected rates were identified in the study population overall. Health centers may benefit from provider training and standardized screening protocols to improve identification of obstetric risk factors at ANC-1. This secondary data analysis used data obtained in a cluster randomized controlled trial on prenatal care and birth outcomes, conducted by The Preterm Birth Initiative–Rwanda ({“type”:”clinical-trial”,”attrs”:{“text”:”NCT03154177″,”term_id”:”NCT03154177″}}NCT03154177) [35]. The present analysis was restricted to health centers randomized to the control group, which included 18 facilities in 5 districts (Bugesera, Rubavu, Nyamasheke, Nyarugenge, and Burera). These health centers were selected for inclusion in the trial based on their location in one of the five districts, their monthly ANC volume, and the presence of at least two ANC providers at the facility. Women ≥15 years of age who presented to one of the participating health centers between May 2017 and December 2018 for ANC services were invited to enroll in the study. Only participants with completed ANC records were included in the final analysis. There were no significant differences in the characteristics of those with incomplete ANC records, as determined by linear models of key sociodemographic variables. Trained data collectors employed by the research team were embedded at each of the 18 health centers. After introducing the study to each woman presenting for ANC, data collectors obtained consent for participants to be included in the study. Upon enrollment, data collectors administered an initial survey to all participants (S1 File). Data collected included age, educational attainment, occupation, contribution to household income, level of partner communication, proximity of the health center to their home and tobacco and alcohol use. Food security over the past month was assessed with a two-question series recommended by the American Association of Pediatrics [36]. Women were also asked whether they had received a pregnancy test and/or whether a community health worker (CHW) had recommended that they visit a health center to confirm their pregnancy. Multiparous participants were asked to report any previous preterm births, low birth weight infants, fresh stillbirths, neonatal deaths (first 28 days of life) and repeated miscarriages. These participants were also asked to report the number of ANC appointments that they attended during their most recent past pregnancy. Data from participants’ first ANC visit (ANC-1) were abstracted from existing national collection tools, including health center registers and patient files. All health centers participated in data strengthening training prior to the start of the study to improve accuracy and completeness of these existing data collection tools. Information abstracted from participants’ antenatal registers included gravidity, parity, and GA at ANC-1. Obstetric risk factors included the presence of anemia, proteinuria, hypertension (≥140/90), multiple births, middle upper arm circumference (MUAC) <21cm, and HIV positive status (either positive test or known positive status documented in the chart). Syphilis or malaria identified at ANC-1 were also recorded. If no obstetric risk factors were identified in the chart, data collectors recorded “none.” Additional history collected from participants’ ANC-1 files included a documented history of diabetes and/or chronic hypertension. In Rwanda, anemia, proteinuria, hypertension, multiple births, MUAC <21cm, diabetes, and syphilis are universally screened for at ANC-1 and noted as either present or absent in the maternity register using tick boxes. The primary outcomes in this analysis were late and very late presentation to ANC. Late presentation was defined as ≥16 weeks GA in accordance with the Rwandan Ministry of Health’s prenatal care performance indicators, and very late was defined as ≥24 weeks GA [37]. Univariate logistic regression models were constructed to identify variables significantly associated with late presentation to ANC at both GA cut-offs for all 10,231 participants, both nulliparous and multiparous. Variables assessed included age, educational attainment, employment, contribution to household income, food insecurity, health center proximity, prior spontaneous abortion, gravidity/parity, pregnancy testing, CHW recommended health center visit, partner communication and the health center proximity. Variables that were significantly associated with delayed presentation to ANC at the αcrit = 0.20 level in univariate analyses were retained for multivariable model building [38, 39]. For covariates that were identified to be collinear (variance inflation factor >2.5) the variable more strongly associated with delayed ANC was retained. Final multivariable logistic regression models were constructed using manual backwards elimination. A full model including all candidate predictors was constructed, and the predictor with the highest p-value greater than αcrit = 0.20 was removed. The model was refit and this process was repeated until all variables maintained in the model had a p-value less than αcrit, with the exception of age which was considered by the investigators to be a potential confounder. Cluster-robust standard errors were used to account for the clustering effects of health centers. Odds ratios and 95% confidence intervals are reported. For the 7,380 multiparous participants, obstetrics history predictors of late (≥16 weeks GA) and very late (≥24 weeks GA) presentation to ANC were assessed using logistic regression models. Obstetric history variables included a history of a preterm delivery, low birthweight infant, previous fresh stillbirth, 28-day mortality of a neonate, and repeated miscarriages. Self-reported ANC attendance in the most recent prior pregnancy was also assessed. The secondary outcomes of interest were the obstetric risk factors identified at ANC-1. Logistic regression models were used to identify associations between late and very late presentation to ANC and the types of obstetric risk factors identified at a woman’s first ANC visit. Risk factors assessed included anemia, proteinuria, hypertension, multiple births, smoking, alcohol use, HIV positive status, and MUAC <21. Diabetes, syphilis and malaria were not reliably recorded in the ANC-1 records, and thus were excluded from the final analysis. Additional logistic regression models were also used to assess for associations between parity and the identification of pregnancy-related risk factors at ANC-1. To assess whether parity was a moderator of the relationship between GA at ANC-1 and each of the obstetric risk factors identified at ANC-1, logistic regression models with interaction terms were used. All analyses were conducted in R (version 3.6.1). This study was approved by the Rwanda National Ethics Committee (No.0034/RNEC/2017), and the University of California, San Francisco Institutional Review Board (16–21177). Written consent was obtained from all participants prior to administering the enrollment survey and reviewing patient health records. All consent forms were translated into Kinyarwanda. Participants provided consent by reading and signing the consent form. For participants who were illiterate, a member of the study team verbally read the consent form in the presence of a witness and both the consented participant and witness signed the consent. The Rwandan National Ethics Committee and the University of California, San Francisco Institutional Review Board waived parental consent requirements for pregnant minors.