Contextual factors associated with health care service utilization for children with acute childhood illnesses in Nigeria

Objective To examine the independent contribution of individual, community and state-level factors to health care service utilization for children with acute childhood illnesses in Nigeria. Materials and methods The study was based on secondary analyses of cross-sectional population-based data from the 2013 Nigeria Demographic and Health Survey (DHS). Multilevel logistic regression models were applied to the data on 6,427 under-five children who used or did not use health care service when they were sick (level 1), nested within 896 communities (level 2) from 37 states (level 3). Results About one-quarter of the mothers were between 15 and 24 years old and almost half of them did not have formal education (47%). While only 30% of the children utilized health service when they were sick, close to 67% lived in the rural area. In the fully adjusted model, mothers with higher education attainment (Adjusted odds ratio [aOR] = 1.63; 95% credible interval [CrI] = 1.31-2.03), from rich households (aOR = 1.76; 95% CrI = 1.35-2.25), with access to media (radio, television or magazine) (aOR = 1.18; 95% CrI = 1.08-1.29), and engaging in employment (aOR = 1.18; 95% CrI = 1.02-1.37) were significantly more likely to have used healthcare services for acute childhood illnesses. On the other hand, women who experienced difficulty getting to health facilities (aOR = 0.87; 95% CrI = 0.75-0.99) were less likely to have used health service for their children. Conclusions Our findings highlight that utilization of healthcare service for acute childhood illnesses was influenced by not only maternal factors but also community-level factors, suggesting that public health strategies should recognise this complex web of individual composition and contextual composition factors to guide provision of healthcare services. Such interventions could include: increase in female school enrolment, provision of interest-free loans for small and medium scale enterprises, introduction of mobile clinics and establishment of more primary health care centres.

Nigeria is located in the western part of Africa and shares borders with Niger, Chad, Cameroon and Benin. It comprises 36 states, 6 geo-political zones and its population at the last Census exercise was over 140 million [17]. At independence in 1960, the main income for the country was generated from agriculture. However, attention has shifted from agriculture since the discovery of oil. The earnings from the sale of oil contributed immensely to developments in the country. Over the years, the health sector witnessed major developments in terms of establishment of hospitals, provision of equipment and drugs and introduction of programmes to promote the well-being of the citizens. Presently, the country operates a National Health Insurance Scheme which aims at making people have access to health services, relieving family of high medical expenses and ensuring equitable distribution of health care services [18]. The scheme covers both the formal and informal sectors. The formal sector comprises the public sector, organized private sector and the armed forces including the police and uniformed services. In the public sector scheme, 3.25% and 1.75% of the employee’s salary are paid by the employer and employee respectively [18]. In the private sector category, the employee pays 5% while the employer pays 10%. These contributions are meant to cover the health care expenses incurred by the employee, his/her spouse and four children below 18 years of age. The scheme has three categories in the informal sector: the tertiary social health insurance programme which makes provision to cater for the health care expenses of students in higher institutions; community-based social health insurance programme which is a voluntary scheme that enables communities to enjoy health services through the payment of flat rate per household or individual household member and; public primary school social health insurance which is targeted at primary school pupils from middle and lower socioeconomic status. This study was based on analyses of secondary data set from the Nigeria Demographic and Health Survey (DHS) 2013 which is cross-sectional and covers all the geo-political zones in the country. Details of the methods used in the DHS have been published elsewhere [17]. Briefly, the survey involved a three-stage cluster sampling technique. Nigeria was divided into 36 States and the Federal Capital Territory (FCT), Abuja making 37 districts in total. The primary sample unit (PSU) was based on 2006 Nigeria population census enumeration areas (EAs). The first stage involved selecting 896 localities (clusters). In the second stage, one EA was randomly selected from most localities. A total of 904 EAs were selected, with 372 in urban areas and 532 in rural areas. The third stage involved the selection of a fixed number of 45 households in every urban and rural geographical area. The total number of selected households was 40,680, with urban areas accounting for 16,740 and 23,940 from rural areas. The methods for data collection have been published elsewhere [17]. In brief, data were collected through household visitation and interviews with individual participants in the selected localities. Information on socio-demographic characteristics, wealth, reproduction, child health, knowledge of HIV/AIDS, domestic violence, household and environmental characteristics was obtained from the participants. This study was based on secondary analysis of existing survey datasets from the archive of the DHS Program who granted us permission for its usage after all the identifying information have been removed. The instruments and conduct of the 2013 Nigeria DHS was approved by the Institutional Review Board (IRB) of ICF Macro International in the United States and Nigeria Health Research Ethics Committee (NHREC) of the Federal Ministry of Health (FMOH). This research is limited to the use of previously collected anonymised data. Users/non users of health services; children under-five who had episode of diarrhea and/or fever or cough in the preceding 2 weeks before the survey and who sought consultation from a health care provider (either public or private) were defined as ‘users’; not seeking care were categorized as ‘non-users’. We adopted behavioural model developed by Andersen to understand the dynamic inter-relations among people and environmental factors associated with health services use for a sick child (Fig 1) [19]. We subsequently grouped such factors into individual-, community- and state-level factors. Figure adapted from Andersen RM. Revisiting the Behavioral Model and Access to Medical Care: Does it Matter? Journal of Health and Social Behavior. 1995; 36: 1–10 In this study, we considered the following variables: age of mother, educational attainment and marital status, mother’s occupation, sex of the child, wealth status and media access. Age of mother (the respondent) was categorised as 15–24, 25–34 and 35–49. The level of education attained by mother was defined as no education, primary, and secondary or higher education. Respondents’ current occupation was categorised into unemployed and employed (professional, technical and managerial, services, agricultural, skilled and unskilled manual and others). Marital status was dichotomized as ever married (i.e. currently married, living with partner, widowed, divorced, separated) and never married. Sex of the child was categorised into Male/Female. Wealth index is measured in the DHS surveys in terms of assets, rather than income. Ownership of consumer items such as a radio or car as well as dwelling characteristics such as floor or roof type, place of cooking, cooking fuel, electricity, toilet facilities and water source were the items that constituted the concept of poverty. This concept has been used by the World Bank to categorise households and their members into different wealth quintiles, through the use of principal components analysis (PCA) [20, 21]. For easy analysis, we re-categorized the weighted scores of five quintiles to three tertiles to allow for nonlinear effects and provide results that would be more readily interpretable in the policy domain. The resultant three tertiles expressed as categorical variables include poor, middle and rich. Access to media was measured as a set of additive scale (from 0 to 3) that counted the number of domains in which each of the respondents was regarding having access to various types of media (radio, television and magazine). This resulted into the following categories: no access, have access to 1 outlet, have access to 2 outlets and have access to all outlets. At community level, we included place of residence, distance to health facility, community socioeconomic status and ethnicity diversity index. Place of residence was categorized into urban and rural. Distance to health facility was grouped into two: those who experienced difficulty in reaching health facility were categorised as ‘a problem’ and those who did not experience difficulty were categorised as ‘not a problem’. Community socioeconomic disadvantage was an index created from the compositional education, wealth and occupation of people within the same PSU. Community socioeconomic disadvantage was obtained through a principal component that consisted of the proportion of respondents with: no education (illiterate), unemployed, and living below the poverty level (asset index below 20% poorest quintile) in the same PSU. This resulted in the generation of a standardized score with mean 0 and standard deviation 1; with higher scores indicative of lower socioeconomic position. We divided the resultant scores into three equal tertiles. The ethnicity of the children was computed by using ethnicity diversity index. We obtained the index by using a formula derived by Vyas and Kumaranayake [22] that captures both the number of different ethnic groups in an area and the relative representation of each group as follows Where: xi = population of ethnic group i of the area, y = total population of the area, n = number of ethnic groups in the area. Scores can range from 0 to approximately 1. For easy interpretation, we multiplied each diversity index by 100; the higher the index score, the greater the diversity in the area. An area with zero diversity indicates that all the people in the area belong to one ethnic group. As the index moves close to 100, the population becomes more evenly distributed into ethnic groups. State socioeconomic disadvantage was an index created from the compositional education, wealth and occupation. State socioeconomic disadvantage was operationalized with a principal component comprised of the proportion of respondents with: no education (illiterate), unemployed, and living below the poverty level (asset index below 20% poorest quintile) within the state. A standardized score with mean 0 and standard deviation 1 was generated from this index; with higher scores indicative of lower socioeconomic position. We divided the resultant scores into three tertiles; tertile 1(least economically disadvantaged), tertile 2 and tertile 3(most economically disadvantaged) In the descriptive statistics, the respondents’ characteristics at different levels were expressed as numbers and percentages. We specified a three-level model with individuals clustered within the communities and the communities clustered within the states in order to estimate the effects of variables at all three levels on utilization of health services. We constructed four models. The first model is a univariable model of the individual-level factors. The second model contained community-level variables while the third model contained the state-level variable at the univariable level. The fourth model adjusted for the individual-level, community-level and state-level variables respectively. P value of < 0.05 was used to define statistical significance. We presented the results of fixed effects as odds ratios (OR) with their corresponding 95% credible intervals (CrI). Random effects were measured through intra-cluster correlation (ICC), variance partition coefficient (VPC) and median odds ratio (MOR). Median odds ratio, which reflects the unexplained cluster heterogeneity, measures the area variance as odds ratios. Details of the procedure used for calculating MOR have been published elsewhere [23, 24]. While we applied Bayesian Deviance Information Criterion (DIC) to assess the goodness-of-fit of the model, Variance Inflation Factor (VIF) was used to check for multicollinearity among the independent variables. All multilevel modelling operations were performed using MLwiN 2.36 [25] calling Stata statistical software for windows version 14 [26]. The Bayesian approach with Markov Chain Monte Carlo (MCMC) estimation was used [27] for the multilevel logistic regression models.

N/A

• adedokun, sulaimon t.
• adekanmbi, victor
• uthman, olalekan a.
• lilford, richard j.