Under-utilization of antenatal care services in Timor-Leste: results from Demographic and Health Survey 2009–2010

This study used the dataset from the Timor-Leste Demographic and Health Survey (TDHS) 2009–2010, which was the second survey conducted since independence. The TDHS was a two stage cluster survey covering the entire nation of 1,163 enumeration areas. During the first stage, 455 enumeration areas were selected by probability sampling proportionate to size. A fixed number of 27 households were then randomly selected from each enumeration area [14]. The national survey has ethics approval from ICF Macro International and Ministry of Health, Timor-Leste [14]. Participants gave consent for themselves whereas mothers or care takers provided consent for minors including infants. The data collection process adhered to World Health Organization’s ethical and safety recommendations ([14], p. 227). The first author applied to the measureDHS and obtained permission to use the publicly available data for research and teaching learning purpose. Further details can be obtained from the website of measureDHS (https://​dhsprogram.​com/​data/​Using-DataSets-for-Analysis.​cfm).

The TDHS 2009–2010 collected data from 11,463 households representative of the entire nation. The overall response rate was 93.5 %. In total, 13,137 women and 4076 men aged 15–49 years were interviewed. The survey used three sets of validated questionnaire (household, men, women) to collect information and the data were subsequently merged in separate datasets for further analyses. The Childrenrecode dataset [20] recorded all 9828 childbirths that occurred within the past five years (URL: http://​dhsprogram.​com/​data/​Dataset-Types.​cfm). The present study included 5,895 mother-child pairs concerning the last born child only. Two missing and 38 ‘do not know’ cases were subsequently removed from the analyses.

Figure 1 presents the underlying conceptual framework. The outcome variable was ‘under-utilization of ANC’, i.e. coded 1 if < 4 ANC visits and coded 0 if ≥ 4 ANC visits were made. Independent variables were selected according to the literature and their availability from the database [21], and were grouped into: external factors, pre-disposing factors (socio-demographic and health knowledge factors), enabling factors and need factors [6]. The sole external factor considered was residential location (rural versus urban). Socio-demographic factors included: maternal age (≤19 years; 20–34 years; and 35–49 years), wealth status (rich/upper 20 %; middle 40 %; and lower 40 %) [22]; maternal and paternal education (no education; primary school; secondary or higher education), maternal occupation (not working; agriculture; paid job: professional, clerical, business, service; manual job); religion (Roman Catholic; others: Muslim, Protestant, Hindu); maternal final say (decision) on her own health (herself; herself and others; others only); and birth order (rank) (1; 2 to 3; ≥ 4). The health knowledge factor was taken to be the frequency of watching television (not at all; less than once a week; at least once a week) following previous studies. Enabling factors were whether there was a ‘big problem’ (major problem) in (a) permission to visit health facility; (b) money to pay for health services; (c) distance to health services; (d) transportation to seek service; (e) presence of companion; and (f) availability of care by female health workers [21]. Need factors referred to pregnancy complications (severe abdominal pain, severe headache, convulsion, blurred vision and swelling of hands and feet, vaginal bleeding, and other problems that required immediate care) which were classified as either no complication or with complications. Desire for children was recoded into: wanted then; wanted later; and unwanted.

Statistical analyses were performed using the Statistical Package for Social Sciences (SPSS version 20). Data were first summarized by cross tabulation and chi-square tests. Number of ANC visits was presented as frequency distributions and corresponding 95 % confidence intervals (CI). Logistic regression analyses were then undertaken to ascertain the associations between under-utilization of ANC and the aforementioned factors, with adjusted odds ratios (OR) to assess the magnitude of such associations.

A hierarchical modelling strategy was adopted [21‐23], with external factors initially entered into the logistic regression model as shown in Fig. 1. Socio-demographic factors were then entered into the model, followed by health knowledge factors, along with significant external and socio-demographic factors. In the final step, enabling factors were added to the model, as the need factors were deemed non-significant according to univariate chi-square tests. Complex sample analyses technique was adopted throughout to account for the study design and sample weight [24].