Objectives: To determine population attributable risks (PARs) estimates for factors associated with non-use of postnatal care (PNC) in Nigeria. Design, setting and participants: The most recent Nigeria Demographic and Health Survey (NDHS, 2013) was examined. The study consisted of 20 467 mothers aged 15-49 years. Non-use of PNC services was examined against a set of demographic, health knowledge and social structure factors, using multilevel regression analysis. PARs estimates were obtained for each factor associated with non-use of PNC in the final multivariate logistic regression model. Main outcome: PNC services. Results: Non-use of PNC services was attributed to 68% (95% CI 56% to 76%) of mothers who delivered at home, 61% (95% CI 55% to 75%) of those who delivered with the help of non-health professionals and 37% (95% CI 31% to 45%) of those who lacked knowledge of delivery complications in the study population. Multiple variable analyses revealed that non-use of PNC services among mothers was significantly associated with rural residence, household poverty, no or low levels of mothers’ formal education, small perceived size of neonate, poor knowledge of delivery-related complications, and limited or no access to the mass media. Conclusions: PAR estimates for factors associated with non-use of PNC in Nigeria highlight the need for community-based interventions regarding maternal education and services that focus on mothers who delivered their babies at home. Our study also recommends financial support from the Nigerian government for mothers from low socioeconomic settings, so as to minimise the inequitable access to pregnancy and delivery healthcare services with trained healthcare personnel.
Data from the 2013 NDHS data set were used for this study. The 2013 NDHS household survey was conducted by the National Population Commission (NPC) in conjunction with ICF International. The household survey information on demographic and health issues such as maternal and child health, childhood mortality and education were gathered by interviewing eligible women and men of reproductive age, aged 15–49 and 15–59 years, respectively. Three questionnaires (household, women’s and men’s questionnaires) were used to record all information gathered. Sampling procedures used in the NDHS have earlier been published in detail elsewhere.15 A total of 38 948 women were successfully interviewed, yielding a response rate of 97.6%. More than 50% (20 467) of these women had the most recent birth within 5 years prior to the survey interview, and were used for our study analyses. The analysis was restricted to births that occurred within the previous 5 years because only those births had detailed information on the use of perinatal health services, and to limit the potential for differential recall of events from mothers who had delivered at very different durations prior to the survey date. The outcome variable for this study was non-use of PNC services. This takes a binary form, such that PNC will be regarded as a ‘case’ (1= if healthcare service was not received during the first 6 weeks after delivery) or a ‘non-case’ (0= if healthcare service was received during the first 6 weeks of infant life). The outcome variable was examined against all potential confounding variables (figure 1). Conceptual framework adapted from Andersen’s Behavioural Model. A behavioural conceptual framework of maternal healthcare services developed by Andersen16 is frequently referenced in other studies on perinatal care services.9 17 18 As a result, our study used the Andersen16 framework as the basis for identifying key risk factors associated with non-use of PNC services in Nigeria. Figure 1 presents all potential confounding variables based on information available in the 2013 NDHS. These variables were classified into five distinct groups: community level factors (geopolitical zone and place of residence); predisposing level factors (demographic, health knowledge and social structure factors); demographic and social structure factors (household wealth index, level of mother’s education, mother’s age at delivery, level of father’s education, mother’s marital status, child’s sex and a combination of birth order and birth interval); health knowledge characteristics (frequency of reading newspaper or magazine, frequency of watching television, frequency of listening to radio and knowledge of delivery complication); enabling factors (permission to visit health services, distance to health services, presence of companion, ability to pay for health services and behaviour of health workers); need factors (delivery complications, birth size and desire for pregnancy); and previous use of health services (delivery assistance, mode of delivery and place of delivery). The prevalence of non-use of PNC services was described by conducting a frequency tabulation of all potential risk factors included in the study. Logistic regression generalized linear latent and mixed models (GLLAM) with the logit link and binomial family19 were then used for multivariable analyses that independently examined the effect of each factor, after adjusting for confounding variables. A hierarchical modelling technique20 was used in the multivariable logistic regression to allow more distal factors to be appropriately examined without interference from more proximate factors. A five-stage model was used by following a similar conceptual framework to that described by Andersen16 (figure 1). First, community level factors were entered into the baseline model to assess their relationship with the study outcome. A manually processed stepwise backwards elimination was performed and variables with p values <0.05 were retained in the model. Second, predisposing level factors were examined with the community level factors that were significantly associated with non-use of PNC, and those variables with p values <0.05 were retained. In the third stage, enabling level factors were investigated with the community and predisposing level factors that were significantly related with the study outcome. As before, those variables with p values <0.05 were retained. A similar procedure was used for need and previous use of health services level factors in the fourth and fifth stages, respectively. In our final model, we double–check for collinearity in order to reduce any statistical bias. All analyses were conducted using ‘SVY’ commands in STATA V.13.1 (STATA Corporation, College Station, Texas, USA) to adjust for the cluster sampling survey design and weights. The PAR was calculated for the significant risk factors to estimate the contribution of each risk factor to the total risk for non-use of PNC services between 2009 and 2013. We obtained PAR and 95% CIs by using the following similar method employed by Stafford et al.21 where pr is the proportion of the population exposed to risk factors, and a OR was the adjusted OR for non-use of PNC.
N/A