Factors associated with the utilisation of skilled delivery services in Papua New Guinea: evidence from the 2016–2018 Demographic and Health Survey

Abstract Background We sought to determine the prevalence and factors associated with the use of skilled assistance during delivery in Papua New Guinea. Methods We analysed nationally representative data from 5210 women in Papua New Guinea using the 2016–2018 Demographic and Health survey. Both bivariate and multivariable analyses were performed. Statistical significance was set at p<0.05. Results The prevalence of skilled assistance during delivery was 57.6%. The richest women (adjusted OR [AOR]=3.503, 95% CI 2.477 to 4.954), working women (AOR=1.221, 95% CI 1.037 to 1.439), women with primary (AOR=1.342, 95% CI 1.099 to 1.639), secondary or higher education (AOR=2.030, 95% CI 1.529 to 2.695), women whose partners had a secondary or higher level of education (AOR=1.712, 95% CI 1.343 to 2.181], women who indicated distance was not a big problem in terms of healthcare (AOR=1.424, 95% CI 1.181 to 1.718), women who had ≥4 antenatal care (ANC) visits (AOR=10.63, 95% CI 8.608 to 13.140), women from the Islands region (AOR=1.305, 95% CI 1.045 to 1.628), those who read newspapers or magazines (AOR=1.310, 95% CI 1.027 to 1.669) and women who watched television (AOR=1.477, 95% CI 1.054 to 2.069) less than once a week had higher odds of utilising skilled attendants during delivery. On the contrary, women in the Momase region (AOR=0.543, 95% CI 0.438 to 0.672), women in rural areas (AOR=0.409, 95% CI 0.306 to 0.546), as well as women with a parity of 3 (AOR=0.666, 95% CI 0.505 to 0.878) or ≥4 (AOR=0.645, 95% CI 0.490 to 0.850) had lower odds of utilising skilled attendance during delivery. Conclusion There is relatively low use of skilled delivery services in Papua New Guinea. Wealth, employment status, educational level, parity and number of ANC visits, as well as access to healthcare and place of residence, influence the utilisation of skilled delivery services.


Introduction
Skilled birth attendants (SBAs), such as doctors, nurses, midwives or trained village health workers, are people with expertise in midwifery. 1 With such expertise, SBAs, through their services, have the capacity to reduce maternal mortality, as stipulated in the UN's Sustainable Development Goal 3.1 (SDG 3.1), and also ensure child survival. 2,3 However, research suggests that developing countries record a low use of services provided by SBAs. 4 This has resulted in a higher prevalence of disease and deaths related to maternity in developing countries, most of which result from otherwise preventable causes such as obstructed labour, sepsis, haemorrhage and eclampsia. 5 For instance, the United Nations 6 reports that, as of 2013, developing countries recorded about 14 times the maternal mortality recorded in highincome countries. A similar report indicated that, as of 2015, 99% of maternal deaths recorded worldwide occurred in developing countries. 7 With the current situation on the ground, it becomes imperative to explore the factors associated with SBA utilisation in lowand middle-income countries and several studies have been conducted in this regard. For example, Olakunde et al., 8 focusing on married adolescents, investigated the factors associated with SBA uptake in Nigeria. Also, focusing specifically on rural Nigeria, Solanke and Rahman 9 explored factors associated with SBA uptake during delivery. Other countries that have featured in studies of this kind include Ghana, 10,11 Ethiopia 12,13 and Kenya. 14 Generally, such studies have revealed some associations between SBA uptake and various socioeconomic variables. It has, for instance, been revealed that people with educated family members, those whose best friends use maternal care 13 and those with access to mass media 9 are highly likely to use services provided by SBAs.
As a way of contributing to the growing literature on SBA, the present study investigates factors associated with the uptake of SBAs in Papua New Guinea. This study is important for two main reasons. In the first place, the present study extends studies of this kind in terms of geography, as Papua New Guinea has not featured in any study of this kind. More importantly, research suggests that, in Papua New Guinea, huge numbers of woman deliver at home, without the services of SBAs, which results in high maternal mortality of about 733 maternal deaths per 100 000 live births. 15,16 Therefore, it is important to reveal the factors that influence the use of SBAs in Papua New Guinea, as this will inform some policy interventions aimed at reducing maternal mortality in the country of focus to help attain SDG 3.1.

Data and sampling design
This study analysed data from the 2016-2018 Papua New Guinea Demographic and Health Survey (PNGDHS), which were collected from October 2016 to December 2018. Among the aims of the PNGDHS is to give current information on basic demographic and health pointers. The survey specifically gathered information on fertility, awareness and the use of family planning methods, breastfeeding practices, the nutritional status of children, maternal and child health, childhood immunisation, adult and childhood mortality, women's empowerment, domestic violence, malaria, awareness and behaviour regarding HIV/AIDS and other sexually transmitted infections, as well as other health-related issues. Technical assistance for the survey was offered by inner city fund (ICF) through the Demographic and Health Survey Programme. Financial assistance was given by the Government of Papua New Guinea, the Australian Government Department of Foreign Affairs and Trade, the United Nations Population Fund (UNFPA) and UNICEF. 17 The survey used the list of census units (CUs) from the 2011 Papua New Guinea National Population and Housing Census as the sampling frame.
The survey adopted a two-stage stratified sampling technique. The provinces in the country of focus were further divided into 43 strata, paying attention to urban-rural differentials; however, the National Capital District did not have any rural strata. Each stratum provided samples of CUs, and this was done independently in two stages. The first stage involved the use of probability proportional-to-size sampling. The second stage of sampling involved the selection of 24 households from each of the clusters, using an equal probability systematic selection, with the resulting sample consisting of about 19 200 households. During the survey, the enumerators were able to cover 16 745 out of the 17 505; 16 021 of the occupied households were interviewed, with a response rate of 96%; 18 175 women of reproductive age were identified in the interviewed households for individual interviews, with 15 198 women completing the interviews at a response rate of 84%. The sample for the present study comprised 5210 women who had given birth to live babies within the 3 y prior to the survey. We realised that some women had given birth to more than one live birth during the selected period; in such cases, we only focused on the most recent birth. Details of the methodology, pretesting, training of field workers, the sampling design and selection are available in the PNGDHS final report, which is available at https://dhsprogram.com/publications/ publication-fr364-dhs-final-reports.cfm. The dataset can be accessed at https://dhsprogram.com/data/dataset/Papua-New-Guinea_Standard-DHS_2017.cfm?flag=0.

Dependent variable
The binary response-whether or not a woman had given birth with the assistance of an SBA-was considered to be the outcome variable. 17 From the PNGDHS, a skilled attendant delivery is a birth delivered with the assistance of doctors, midwives, nurses (including trained community health workers) or trained village health volunteers (p. 138). 17 In this study, skilled delivery, supervised delivery and skilled provider at birth are used interchangeably.

Independent variables
Seventeen explanatory variables were considered in this study, based on their availability in the dataset 17 and conclusions drawn from them associated with skilled delivery in previous studies. 9,10,13,18-21 The variables comprised maternal age, wealth, working status, education, partner's education, marital status, place of residence, region of residence, parity (birth order), getting money for treatment, distance to health facility, antenatal care (ANC) attendance, exposure to mass media (radio, television, newspapers) and gender of the head of the household. Some of these variables were recoded. ANC attendance was recoded into 0, 1, 2, 3 and ≥4 visits. Parity was categorised as 1, 2, 3 or ≥4 births. Education and partner's education were classified into three categories: no education, primary education and secondary education/higher education. Occupation was captured as working or not working and the decision-maker on healthcare was captured as either alone or not alone (Table 1).

Statistical analyses
Three key steps were followed to analyse the data. First, descriptive statistics, such as frequency with %, were executed to represent the background characteristics of study participants and the prevalence of skilled delivery services utilisation. Second, a bivariate analysis using χ 2 was employed to select candidate variables for the regression analysis. Variables with p<0.05 were moved to the regression analysis stage. At the regression stage, crude and adjusted models were employed. The crude model was estimated to examine the effect of each independent variable on the outcome variable, while multivariable logistic regression was used to examine the effect of all the significant independent variables at the crude level on utilisation of skilled services during delivery. The output was reported as crude ORs (CORs) and adjusted ORs (AORs) with their corresponding 95% CIs. Using the variance inflation factor (VIF), the multicollinearity test showed that there was no evidence of collinearity among the independent variables (mean VIF=1.5, max. VIF=1.99, min. VIF=1.02). We applied sample weight to correct for oversampling and undersampling to ensure generalisation of the findings. The survey (svy) command was applied to take care of the complex sampling procedure involved in the demographic and health surveys. In other words, the svy command was used to declare the data survey data. We carried out the analyses with stata version 14.2 for MacOS (Stata Corporation, College Station, TX, USA).

Multivariable analysis of the determinants of skilled delivery in Papua New Guinea
The results from the study of the factors associated with the use of skilled delivery services in Papua New Guinea are presented. Compared with the poorest women, middle, richer and richest women were, respectively, 1.5 times (AOR=1.452, 95% CI 1.149 to 1.835), 1.9 times (AOR=1.887, 95% CI 1.464 to 2.433) and 3.5 times (AOR=3.503, 95% CI 2.477 to 4.954] more likely to have utilised skilled delivery during their most recent birth. Women who were working (AOR=1.221, 95% CI 1.037 to 1.439] had 1.2 higher odds of utilising skilled providers during delivery compared with those not working. Concerning education, respondents' and their partners' levels of education were significantly associated with skilled providers during delivery. For example, compared with mothers with no formal education, those who had completed primary (AOR=1.342, 95% CI 1.099 to 1.639) and secondary or higher (AOR=2.030, 95% CI 1.529 to 2.695) had higher odds (AOR=1.712, 95% CI 1.343 to 2.181) of using skilled providers during delivery. Women whose partners had a secondary or higher level of education had higher odds of utilising skilled providers during delivery compared with those women whose partners had no formal education. In terms of regional variations, compared with those in the southern region, women in the Momase region (AOR=0.543, 95% CI 0.438 to 0.672) had lower odds of skilled delivery while those in the Islands region had higher odds (AOR=1.305, 95% CI 1.045 to 1.628). Relatedly, women in rural areas had lower odds of skilled delivery compared with those in urban centres (AOR=0.409, 95% CI 0.306 to 0.546). In terms of parity, women with a parity of 3 (AOR=0.666, 95% CI 0.505 to 0.878) and ≥4 (AOR=0.645, 95% CI 0.490 to 0.850) had lower odds of skilled delivery compared with those women with a parity of 1. Women who indicated that distance was not a big problem in terms of healthcare had higher odds (AOR=1.424, 95% CI 1.181 to 1.718) compared with those women who indicated that distance was a big problem. The odds of skilled delivery use increased with the number of ANC visits. Specifically, women who had ≥4 ANC visits had about 10.6 times higher odds of utilising skilled providers during delivery compared with those who never attended ANC. Those who read newspapers or magazines (AOR=1.310, 95% CI 1.027 to 1.669) and watched television (AOR=1.477, 95% CI 1.054 to 2.069) less than once a week had higher odds of utilising skilled attendance during delivery compared with those who did not watch television or read no newspapers or magazines at all (Table 2).

Discussion
The services of SBAs are needed not only during pregnancy but also during childbirth and the postpartum period, as such services help to improve maternal and neonatal survival. The present study assessed the prevalence and factors associated with the utilisation of SBA services in Papua New Guinea. The study reveals that 57.6% of childbearing women in Papua New Guinea utilised skilled assistance during delivery. A similar prevalence of skilled delivery was reported in the PNGDHS 2016-2018, where 56.5% of  childbearing women had a 'skilled provider' at the birth. 17 Previous studies in the country by O'keefe et al. 22 and Mola and Kirby 23 reported SBA usage of 28% and 39.1%, respectively. One of the possible reasons for the increased prevalence of skilled delivery in the country could be an improved skilled workforce in health facilities. For example, since 2009, the quality of midwifery training in Papua New Guinea has experienced a dramatic improvement and the number of practising midwives has almost tripled, helping to curb the staff shortage the country had been experiencing. 24 Kep et al. 25 found that midwives in the country offer some training to birth attendants who work in villages, so as to help increase skilled assistance during delivery. Another possible reason for this increase in SBA uptake could be that pregnant women are sensitised on the need for SBA services during delivery. Notwithstanding, there is still a need to encourage more pregnant women in the country to utilise SBAs during delivery. We also found that high socioeconomic status (wealth, education and employment) increased the chances of SBA usage during delivery. Specifically, women with a higher wealth status, those with higher levels of education, those whose partners had higher levels of education and those who were employed had a higher likelihood of SBA uptake. Similar studies using nationally representative data in other countries also found that high so-cioeconomic status increases the likelihood of skilled attendance during delivery. 14,20,21 Women with higher wealth status and employment are highly likely to use SBA services during delivery, because such women often have the financial empowerment to access skilled attendance during delivery. 26,27 Such women could be in a better position to pay for transportation costs and the cost of skilled delivery services, if required, due to their financial position. The link between a higher level of education and use of skilled attendance during delivery is not surprising, given that with higher education, women are more likely to be conscious of their health. 20 Such women are more likely to understand the need for SBA use. This also explains the association between access to mass media and SBA uptake, as revealed by the present study.
Place of residence was found to influence the likelihood of SBA uptake, with rural dwellers, especially residents of the Momase region, having lower odds of skilled delivery compared with urban dwellers. Similar revelations were reported in previous studies. 9,26,28 This finding could be attributed to the lack of advanced health facilities with highly qualified personnel in the rural areas, which hinders access to skilled health providers for rural women. 29,30 The health workforce in the country is also characterised by an ageing workforce, with few midwives and International Health community health workers, who are not even given any attractive remuneration to motivate them. 30 This finding can also be explained in terms of the socioeconomic situation in rural areas, often characterised by low employment, low income levels and low education levels, 9 which may hinder rural dwellers from accessing the services provided by SBAs. Another reason for the disparity between urban and rural areas in the use of SBAs could be a result of the long travelling distances associated with health facilities in rural areas, especially in Papua New Guinea. 31,32 In fact, as the study reveals, women who did not consider the distance to the health facility as being a big problem reported a higher use of SBAs compared with those women who did consider the distance to the health facility as being a big problem.
Utilisation of SBAs decreased with higher parity. This finding corroborates findings from studies in other countries. 9,33 In this regard, women with high parity, with confidence, may feel that even in the absence of skilled providers they can still deliver their babies. 34 This is more likely to be predominant among those women whose previous deliveries were supervised by unskilled providers. It is, therefore, necessary to educate such women that their status does not reduce the risk associated with unskilled assistance during delivery. Skilled delivery increased with more ANC visits, as reported elsewhere. [35][36][37] This finding highlights the need for a renewal of efforts towards strengthening ANC programmes, so as to enhance skilled delivery in the country. Such efforts should, for example, include expanding coverage to enhance visits.
At the crude level, exposure to mass media was associated with higher odds of utilisation of skilled delivery services. Nonetheless, the results were attenuated after controlling for other covariates. In the adjusted model, women who were exposed to newspapers and watched television less than once a week had higher odds of SBA utilisation compared with those who were not exposed to newspapers or television at all. This is in line with a previous study by Fatema and Lariscy 38 conducted in South Asia. Viswanath et al. 39 postulated that mass media can influence attitudes towards healthy behaviour changes through community functions of information, tools, social control and healthy communication. Specifically, mass media can help ensure the update of health behaviours by consistently broadcasting programmes, public service announcements and advertisements that describe new drugs, novel treatments or risk factors. In Bangladesh, for example, the TV drama series Ujan Ganger Naiya showed maternal healthcare-related issues, including the advantages of delivery by SBAs, seeking ANC, adequate and prompt birth preparedness, postpartum check-ups, appropriate nutrition and essential newborn care, by entertaining audiences with stories set in a rural village. 38,40 The media also offers health information, such as announcements of dates, times and places relating to when and where to seek healthcare. In south Asian countries, an educational media campaign about safe motherhood by the UNFPA included information about the symptoms of labour pain, as well as where and who would be best for delivering a baby. 41

Strengths and limitations
At this point, we highlight some strengths and limitations of the study. In terms of strengths, given that the present study used nationally representative data, with its methodology validated, the findings of the study can be generalised to all women in the country. On the other hand, the data for the study are crosssectional in nature, which makes it nearly impossible to ascertain causal inferences or temporal relationships between the studied variables. Again, the possibility of recall bias and social desirability among the study participants also presents another limitation to the study.

Conclusion and policy implications
There is relatively low utilisation of skilled delivery services among women in Papua New Guinea. This study also revealed that individual-level factors such as wealth, employment status, educational level, parity and the number of ANC visits, as well as community-level factors such as access to healthcare facilities and place of residence, influence the utilisation of skilled delivery services in Papua New Guinea. In light of this, it is important for policies and programmes aimed at enhancing skilled delivery and reducing maternal mortality in the country to take these factors into consideration, so as to enhance the attainment of SGD 3.1. Such programmes should pay special attention to women who are resident in rural areas, as such women are more likely to use unskilled delivery services, due to their low socioeconomic status.