http://orcid.org/0000-0001-6147-3475
Figures
Abstract
This paper examines the association between health facility quality, subjective perceptions, and utilization of obstetric care. We draw on unique survey data from Nigeria describing the quality of care at rural primary health care facilities and the utilization of obstetric care by households in the service areas of these facilities. Constructing a quality index using the detailed survey data, we show that facility quality is positively related to perceptions of quality and utilization. Disaggregating quality into structural, process and outcome dimensions, we find a consistently strong relationship only between utilization and structural measures of quality. The results suggest that efforts to improve quality may involve a trade-off between investing in dimensions that are more easily observed by households, which will influence utilization, and investing in dimensions that are more closely related to outcomes.
Citation: Peet ED, Okeke EN (2019) Utilization and quality: How the quality of care influences demand for obstetric care in Nigeria. PLoS ONE 14(2): e0211500. https://doi.org/10.1371/journal.pone.0211500
Editor: Seth Adu-Afarwuah, University of Ghana, GHANA
Received: September 26, 2018; Accepted: January 15, 2019; Published: February 7, 2019
Copyright: © 2019 Peet, Okeke. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All data files are available from openICPSR, located at: https://www.openicpsr.org/openicpsr/project/108221/version/V1/view/.
Funding: RAND Corporation is a non-profit research organization and did not play any role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript and only provided financial support in the form of the author’s (EO and EP) salaries and research materials. Funding for the survey used in this paper was provided by the International Initiative for Impact Evaluation (3ie) through Grant No. OW4/1225. EO received the award. The funder provided support in the form of salaries for authors (EO), but did not have any additional role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript. The specific roles of these authors are articulated in the ‘Author Contributions’ section.
Competing interests: The authors have declared that no competing interests exist.
Introduction
Despite global reductions in mortality and increasing life expectancies, large gaps remain, particularly in maternal and newborn health [1]. In 2015, 303,000 women died during and following pregnancy and childbirth, 4.2 million infants died within the first year of life, and 2.6 million babies were stillborn [2–4]. Nearly all maternal and child deaths occur in low- and middle-income countries: women living in poor countries are nearly 23 times more likely to die from pregnancy and childbirth-related complications than their counterparts living in developed countries [5]. Pregnancy-related morbidity and mortality has serious economic and social consequences, with estimates of the global productivity losses attributable to maternal and newborn deaths reaching approximately US $15 billion annually [6].
Proper obstetric care is considered to be a crucial determinant of birth outcomes [7–8], but remains underutilized in many low and middle-income countries. In the least developed countries that account for the majority of global maternal and newborn deaths, over 50% of births take place at home without formal assistance [5, 7, 9]. Concerns about underutilization of care have led to various policy interventions designed to stimulate demand. Many of these policies have focused on relaxing demand-side constraints such as access and cost [10–12]. However, even when demand-side factors are addressed, many women continue to choose to give birth at home [13–14]. Research suggests that traditional attendants are often viewed as being as good as, or in some cases even better than, formally trained birth attendants [13, 15], and there is evidence that formal care is not always perceived as being associated with better outcomes [16–17]. Despite being a global health priority, demand-side efforts have not yielded the expected gains in utilization in many settings. One potential reason is that utilization is also constrained on the supply side [18]. Consequently, it is becoming increasingly obvious that efforts to incentivize utilization must also address supply-side constraints.
Poor quality of care is believed to be an important supply-side constraint [19–20], and there is evidence that poor quality of care can counteract the effects of demand-side interventions [21]. Low quality may not only constrain improvement in outcomes, it may also depress demand [22–23]. The latter is of particular interest. Understanding the relationship between quality and utilization, however, is challenging in part because quality is multi-dimensional [24]. It encompasses ‘structural’ factors such as availability of resources in health facilities, ‘process’ factors such as quality of diagnostics, and ‘outcomes’—or the end result of care provided [24]. Conceptually it is not obvious that each of these dimensions will have the same effects on demand. Some aspects of quality are not easily observed (e.g., a provider’s true level of competence), or their importance may not be obvious to non-medically trained individuals. This suggests that more easily observed dimensions of quality (e.g., structural factors) may have greater effects on demand even when their correlations with outcomes might be weak [18]. Therefore, there may be an inherent policy tradeoff between the effects of an intervention on demand and its effects on outcomes. This necessitates a deeper understanding of how quality influences demand, and specifically, which dimensions are important.
Much of the existing evidence on quality and utilization is from qualitative studies and there are few rigorous quantitative studies [14–23, 25–28]. This paper attempts to fill this gap by quantitatively examining the relationship between health facility quality and demand for obstetric care. Importantly, we are able to disaggregate quality into its different dimensions and examine how each dimension correlates with demand. We also examine the extent to which perceptions of quality are correlated with objectively measured quality. Several studies have shown that perceptions are important determinants of utilization [29]. Understanding how this relates to objective quality is an important but understudied question [30].
Materials and methods
Ethics statement
Ethical approvals for the study were granted by the RAND Human Subjects Protection Committee and the Aminu Kano Teaching Hospital Research Ethics Committee in Nigeria. Each participant gave informed consent prior to participation in the study. The human subject data of this study were analyzed anonymously.
Data
We draw on data from the Better Obstetrics in Rural Nigeria (BORN) study. This study measured availability and quality of obstetric care and birth outcomes in rural communities in Nigeria [31]. Trained research assistants collected extensive data on service availability and quality from primary health care facilities and also surveyed women with a recent birth living in the communities served by these facilities. The health facilities in the sample consisted of rural public primary health care facilities that provided obstetric services. Primary health care facilities are the point of entry for most patients into the health care system in Nigeria. Participating facilities were spread across 12 States covering all six geopolitical regions in Nigeria. In total, 362 primary health care facilities were surveyed [31].
Data collection took place between June 2014 and January 2015. In each health facility, extensive data were collected on service delivery, staffing and availability of supplies and equipment. The research assistant also observed and recorded the physical condition of the clinic. In addition, a randomly selected midwife (or other health worker if no midwife was available) was interviewed. Health workers were asked about their qualifications, the length of time they had been employed at the clinic, and satisfaction with various aspects of their work environment. Their clinical competence was also assessed using a combination of multiple-choice questions and clinical case studies.
Additionally, research assistants surveyed a random sample of approximately 20 households in each community where the facility was located. Since a comprehensive listing of households in each community was unavailable, we randomly generated 20 random GPS coordinates within each community using a GPS-enabled tablet and special software and selected the dwelling nearest this point for interview. If there was no eligible household within the dwelling, the interviewer visited the dwelling on either side until one was found. If there were multiple eligible households within the dwelling, one was randomly chosen for interview. Eligibility for participation was based on having a female member who was pregnant within the preceding five years. This eligibility criteria was adopted in line with the commonly accepted practice of using a five-year recall period as is used in the Demographic and Health Surveys. All such women within the household were interviewed. General information was collected about household characteristics and ownership of various household assets. An individual module collected demographic information as well as information on use of antenatal care, and place of delivery for prior births.
To account for the fact that facility characteristics are likely to change over time, we restrict our attention to the deliveries that occurred in the 12 months immediately preceding the health facility survey. Restricting the sample from deliveries that occurred within the five years preceding the health facility survey to those that occurred within 1 year reduced the number of observations from 8,902 to 2,140. The 2,140 deliveries comprise our analytic sample. Summary characteristics are reported in Table 1.
Key variables.
The main dependent variable is utilization of obstetric care (antenatal and delivery care) in the primary health care facility. We define binary indicators for utilization of antenatal and delivery care at the primary health care facility. We also examine utilization of antenatal and delivery care at other public facilities, private facilities, and overall utilization of formal care. Our secondary dependent variable is perceptions of the primary health care facility quality. Women were asked to rate services in the facility as poor (1), average (2), good (3), or excellent (4). To analyze perceptions, the woman’s rating is used as the dependent variable.
For both the primary and secondary dependent variables, our main independent variable is care quality. Drawing on the Donabedian framework, we measure care quality along three dimensions: structure, process, and outcomes [24]. Using our rich survey data, we identify key measures in each of these domains. We conceptualize each of these measures as mapping loosely to a sub-domain of care quality. We use principal component analysis to extract an index for each sub-domain, each domain—structure, process, and outcomes—and to create an index of overall quality. Principal component analysis (PCA) is a procedure that generates orthogonal eigenvectors of a covariance matrix, and retains maximum variation [32]. While we acknowledge that there are limitations to assessing aggregated measures of quality generated by PCA, the objective of the study is to understand how quality, generally, influences utilization decisions rather than how one specific measure of quality influences utilization.
Analysis.
Our objective is to relate variation in facility quality to utilization of obstetric care. We estimate the following using a linear probability model. We also verify that the results are robust to alternative specifications, including a logit model.
Ui denotes the utilization outcome of interest for birth i and Qi is the measure of quality. Xi is a vector of covariates including individual and household characteristics such as maternal education, maternal literacy, mother’s age at child birth, mother’s ethnicity, mother’s belief in the necessity of a facility birth, household assets, cost of delivery and drugs, distance to facility in minutes, cost to reach facility, minutes to a referral hospital (each described in Table 1). Finally, εi describes the error term which is clustered at the facility-level. The coefficient βQ describes the influence of overall facility quality on the outcome of interest. If βQ = 0 then the care quality provided at the facility is unrelated to the outcome; in contrast, if then utilization of the facility is associated with the facility’s quality.
The next step in our analysis is to assess the effect of each of the domains of care quality–structure, process and outcomes–on utilization. We denote the structural index as , the process index as
, and the outcomes index as
. We can then write the following model:
(2)
, and
therefore describe how structural quality, process quality, and outcome quality influence facility utilization. We also explore the contributions of each sub-domain (results are in the Appendix, S1 File).
Given that we exploit cross-sectional variation in facility quality, there is a concern that because higher-quality facilities are not randomly distributed, there may be a correlation between facility quality and household and individual characteristics that confound the relationship estimated by Eqs (1) and (2). To try to address this, we make use of propensity score-based methods. The idea is to compare similar women exposed to different levels of quality. Methods based on the propensity score have long been used for causal inference in observational studies; they are easy to use and can reduce the bias caused by non-random treatment assignment. There is evidence that in certain situations the propensity score method produces more reliable estimates of causal effects than other estimation methods [33–34]. Among the various propensity score methods developed for continuously valued treatments, we estimate the relationships described in Eqs (1) and (2) using the inverse second moment weighting method (ISMW) [35], and the weighted regression (WTRG) [36]. The ISMW approach is an extension of the inverse probability treatment weight approach using a marginal structural model that applies weights derived from propensity scores and generates a pseudo-population that mimics the properties of the overall population to correct distortions that arise from selection [37]. The WTRG method is a doubly robust estimator that uses inverse probability weights designed to give consistent estimates in a model for outcome prediction [36]. Further discussion is provided in the Appendix, S1 File. These propensity score-based estimators serve as a robustness checks for our main estimates generated using ordinary least squares. If the estimates are similar, the results are suggestive of a causal relationship. The results of these analyses are in the Appendix, S1 File.
Results
Summary characteristics
Table 1 reports means and standard deviations for the birth sample. Slightly over half the mothers received at least some secondary education and are literate. The mean age of mothers at the birth of the child is 26.7 years. 15% of the mothers believe that giving birth at a formal facility is unnecessary. For 65% of births, women utilized antenatal care at the primary health care facility. 51% of deliveries also occurred in the primary health facility (28% occurred at home). Table 2 describes the components of the quality indices and the mean and standard deviation of the variables comprising each index. Each component of the three indices of quality is measured at the facility level except for measures of antenatal and postnatal care quality, which come from the women’s survey. Neonatal, obstetric, and maternal outcomes are given as rates relative to the number of deliveries at the facility during the 12 months preceding the survey. The correlations between the overall index, the domain indices, the sub-domain indices and the indicators are shown in the Appendix, S1 File. For ease of interpretation, we standardize each index.
Does facility quality influence utilization of obstetric care?
Table 3 relates overall quality of care to utilization of antenatal and delivery care. In the first column, we examine how overall quality in a primary health facility relates to utilization of antenatal and delivery care in that facility. If higher quality is associated with greater utilization, a secondary question of interest is whether this might be because women are substituting from other health facilities or whether some of this increase is driven by new users. A priori, it is unclear whether higher quality at the primary health facility will lead women who would otherwise have traveled further away to use services to switch to the primary health care facility (particularly because care at private facilities and hospitals is more expensive), or whether higher quality in the primary health facility may draw in women who would not have used care at all (i.e., users of informal care). In the remaining columns in Table 3 we examine how quality at the primary health facility relates to utilization of obstetric care at other health facilities. Specifically, we examine how overall quality at the primary health facility relates to obstetric care utilization at other public facilities including general hospitals, at private facilities, and at any formal health facility. Obtaining a significant coefficient in the latter regression would indicate that higher quality in the primary health facility is associated with an overall increase in utilization.
We observe that higher overall quality is associated with significantly greater utilization of antenatal and delivery care at the primary health facility. A 1 standard deviation increase on the overall quality index is associated with a 4.1% and 2.9% increase in utilization of antenatal and delivery care, respectively at the primary health facility. There is some evidence of potential substitution behavior: higher quality at the primary health facility is associated with decreased utilization of care in other facilities (where care is more expensive). There is, however, no evidence of an increase in overall utilization at any formal facility. The propensity score estimates described in the Appendix (S1 File and Fig 1A and 1B) coincide to support the overall conclusion that higher facility quality leads to increased utilization of care, and provide evidence that the estimates are robust to model specification.
The figure shows the OLS, ISMW, and WTRG estimates and 95% confidence intervals of the relationship between facility quality and antenatal and delivery care utilization based off n = 2,140 deliveries that occurred within 1 year of the facility survey.
In Table 4, we examine whether the effect of quality on utilization of delivery care is stronger for women who attended antenatal care in the study facility. Women who attended antenatal care in the facility might gain first-hand information about facility quality and use that information when deciding where to obtain delivery care. We may therefore see a larger effect for such women. We indeed find evidence consistent with this. For women who received antenatal care at the primary health facility, a 1 standard deviation increase in overall quality of antenatal care is associated with 4.8% greater utilization of delivery care. This is driven by the process of care index. For women who received antenatal care at the primary health facility, a 1 standard deviation increase on the process index is associated with 7.5% greater utilization of delivery care.
What dimensions of quality matter for utilization?
In Table 5 we examine how each domain of quality relates to care utilization. For both antenatal and delivery care, structure is the dimension of quality that exhibits the largest and most statistically significant relationship with utilization. The estimates indicate that a 1 standard deviation increase on the structural quality index is associated with a 3.4% and 2.4% increase, respectively in utilization of antenatal and delivery care. In contrast, there is no observed relationship between the process and outcomes indices and care utilization. The coefficients in both cases are small and not statistically significant. The propensity score estimates described in the Appendix (and Fig 2A and 2B) confirm that the relationship between quality and utilization is driven by structural quality, and provide evidence that the estimates are robust to model specification.
The figure shows the OLS, ISMW, and WTRG estimates and 95% confidence intervals of the relationship between structure, process, and outcomes measures of facility quality and antenatal and delivery care utilization based off n = 2,140 deliveries that occurred within 1 year of the facility survey.
Is objective quality related to subjective perceptions of quality?
Table 6 relates overall quality to women’s perceptions of the quality of the primary health facility. For the overall quality index, the estimate is positive and significant, indicating that better objective facility quality is associated with higher subjective perceptions of facility quality. The coefficient indicates that a 1 standard deviation increase in facility quality is associated with a 0.19 standard deviation increase in women’s reported perception of the quality of care provided at the primary health facility. Table 6 also reports how the three dimensions of quality–structure, process, and outcomes–relate to women’s perception of the quality of care provided at the facility. While all three dimensions are positively related to perceived quality, only structure and process are statistically significant. The coefficients indicate that a 1 standard deviation increase in the structural dimension of quality is associated with a 0.097 standard deviation increase in perceived quality, and a 1 standard deviation increase in the process dimension of quality is associated with a 0.145 standard deviation increase in the perceived quality. The propensity score estimates described in the Appendix (and Figs 3 and 4.) confirm that perceived quality is significantly related to structural and process quality, and provide evidence that the estimates are robust to model specification.
The figure shows the OLS, ISMW, and WTRG estimates and 95% confidence intervals of the relationship between facility quality and women’s perception of quality. The sample includes n = 2,140 deliveries that occurred within 1 year of the facility survey.
The figure shows the OLS, ISMW, and WTRG estimates and 95% confidence intervals of the relationship between structure, process, and outcomes measures of facility quality and women’s perception of quality. The sample includes n = 2,140 deliveries that occurred within 1 year of the facility survey.
Discussion
This paper addresses three questions: first, does quality of care influence utilization of obstetric care; second, if so, what dimensions of quality matter; and third, is objective quality related to subjective perceptions? Our results indicate that quality is positively correlated with utilization. Disaggregating by domains of quality suggests that the structural dimension of quality is the key driver of utilization. We also find that subjective perceptions are strongly related to objectively measured quality. We observe statistically significant relationships for structure and process but not for outcomes. The results are consistent across multiple empirical methods. Compared to the estimates generated using ordinary least squares described above, the propensity score estimates described in the Appendix only differ slightly in magnitude. The lack of model dependence lends credibility to the estimates, if not a causal interpretation. A causal interpretation of these estimates relies on the weak unconfoundedness assumption; specifically that assignment to treatment is as good as random within subpopulations that are homogenous in observed pre-treatment variables [38]. If this assumption holds, the estimates can be interpreted as causal.
The majority of studies that examine factors related to the utilization of obstetric care focus on sociocultural factors, perceived benefit/need of skilled attendance, economic accessibility, and physical accessibility [23, 39]. Few quantitative studies assess the relationship between quality of care and utilization. Both [25–26] found no significant association between structural quality indicators like the number of health workers on formal delivery care, while [27] found no relationship with availability of obstetric equipment. Our results align more closely with that of [22] and [28] who found a correlation between number of doctors and utilization. This study improves on earlier work by collecting considerably more detailed data on facility quality allowing us to not only better measure quality, but also to separate it into its various components to provide policy makers with evidence on which particular dimensions drive utilization.
A potential narrative for the results in this paper is that decisions regarding utilization of antenatal and delivery care are made based on what women can observe with greater certainty. Because the expected outcomes from utilizing antenatal and delivery care are uncertain prior to delivery, structural elements of the facility that can be easily observed (e.g. number of providers, general infrastructure) are more closely related to utilization. This narrative is supported by the fact that for women who used antenatal care at the facility, the process index exerts greater influence on their subsequent decisions to utilize delivery care. We may not observe a significant relationship between facility outcomes and utilization because these are much more difficult to observe and may also be more difficult to interpret. Outcomes are also a function of underlying risk characteristics.
Overall, the results generate some useful policy implications. A key implication is that if policy makers wish to improve utilization, quality investments should be made in areas that can be observed by households. This, of course, will need to be carefully weighed against investments that are more closely tied to outcomes. With regards to increasing utilization, it is noteworthy that in this study we do not find that higher quality in the primary health facility leads to an increase in overall levels of formal care utilization, but instead appears to lead to substitution from other health facilities (though given the cross-sectional nature of the data this is only suggestive not conclusive). This suggests that interventions targeting the demand-side will still be needed if the goal is to increase overall utilization. We emphasize, however, that while policy makers might be more exercised by overall levels, substitution may also be beneficial to the extent that households spend less on care allowing these savings to be reallocated to other goods. Substitution is also beneficial if households are substituting higher quality facilities for lower quality ones.
This study has some limitations. The first limitation is that the data on quality are cross-sectional. One concern is reverse causation. This could be the case if facilities with greater utilization receive more resources that can then be invested in structural quality. However, we do observe that, for plausibly informed women, process quality also influences utilization suggesting that causality runs from quality to utilization. The second limitation is that our measures of facility quality may not perfectly capture true quality. We use indices derived from multiple observed characteristics of the facility to measure quality. If the variables used to construct these indices are not accurate, the estimates may be biased. For instance, the estimated association between outcome quality and utilization could be biased toward zero if the clinical records of neonatal deaths, obstetric complications, and maternal deaths are inaccurate. The third limitation is that our data are observational and we do not have exogenous variation in facility quality. As a result, the estimates could be biased by unobserved heterogeneity. While we use multiple empirical methods that demonstrate similar relationships between quality and care utilization, the causal interpretation of each estimate relies on the same assumption of weak unconfoundedness. If the quality of the facilities varies systematically with unobserved characteristics of the sample, the point estimates will reflect these correlations.
Conclusion
Proper obstetric care is critical for birth outcomes, yet remains underutilized in many low and middle-income countries. Using matching facility and household survey data from Nigeria, this paper has examined the association between health facility quality and utilization of obstetric care. We also explored whether subjective perceptions of quality carry information about objective quality. We find that quality of care is an important determinant of obstetric care utilization, though this relationship appears to be explained by more observable, structural dimensions of quality. We also find that perceptions of quality are strongly related to actual quality suggesting that perceptions carry important information. The results in this paper provide useful evidence to policy makers on the relationship between facility quality and demand for obstetric care.
Supporting information
S2 File. Better Obstetrics in Rural Nigeria (BORN) health facility survey.
https://doi.org/10.1371/journal.pone.0211500.s002
(PDF)
S3 File. Better Obstetrics in Rural Nigeria (BORN) household survey.
https://doi.org/10.1371/journal.pone.0211500.s003
(PDF)
Acknowledgments
We thank all our colleagues who provided helpful comments on previous versions of this document. EO gratefully acknowledges grant support from the International Initiative for Impact Evaluation (3ie).
References
- 1. Lozano R, Naghavi M, Foreman K, Lim S, Shibuya K, Aboyans V, et al. Global and regional mortality from 235 causes of death for 20 age groups in 1990 and 2010: a systematic analysis for the global burden of disease study. 2010. Lancet. 2012; 380(9859): 2095–2128. pmid:23245604
- 2.
World Health Organization (WHO). Maternal Mortality. Accessed: 27 Aug 2018a. Available from: http://www.who.int/news-room/fact-sheets/detail/maternal-mortality
- 3.
World Health Organization (WHO). Infant Mortality: Situations and Trends. Accessed: 27 Aug 2018b. Available from: http://www.who.int/gho/child_health/mortality/neonatal_infant_text/en/
- 4.
World Health Organization (WHO). Stillbirths. Accessed: 27 Aug 2018c. Available from: http://www.who.int/maternal_child_adolescent/epidemiology/stillbirth/en/
- 5.
World Health Organization (WHO). Strategies towards ending preventable maternal mortality (EPMM) 2015 Available from: http://who.int/reproductivehealth/topics/maternal_perinatal/epmm/en/.
- 6.
USAID, 2001. USAID Congressional Budget Justification FY2002: program, performance and prospects–the global health pillar. United States Agency for International Development: Washington, DC.
- 7. Bhutta ZA, Das JK, Bahl R, Lawn JE, Salam RA, Paul VK, et al. Can available interventions end preventable deaths in mothers, newborn babies, and stillbirths, and at what cost? Lancet (London, England). 2014;384(9940):347–70. Epub 2014/05/24.
- 8. Lee AC, Cousens S, Darmstadt GL, Blencowe H, Pattinson R, Moran NF, Hofmeyr GJ, Haws RA, Bhutta SZ, Lawn JE. Care during labor and birth for the prevention of intrapartum-related neonatal deaths: a systematic review and Delphi estimation of mortality effect. BMC public health. 2011 Dec;11(3):S10.
- 9. Montagu D, Yamey G, Visconti A, Harding A, Yoong JK. Where Do Poor Women in Developing Countries Give Birth?. Plos one. 2011 Feb;6(2):1–8.
- 10. Fink G, Robyn PJ, Sié A, Sauerborn R. Does health insurance improve health?: Evidence from a randomized community-based insurance rollout in rural Burkina Faso. Journal of health economics. 2013 Dec 1;32(6):1043–56. pmid:24103498
- 11. Lagarde M, Haines A, Palmer N. Conditional cash transfers for improving uptake of health interventions in low-and middle-income countries: a systematic review. Jama. 2007 Oct 24;298(16):1900–10. pmid:17954541
- 12. Murray SF, Hunter BM, Bisht R, Ensor T, Bick D. Effects of demand-side financing on utilisation, experiences and outcomes of maternity care in low-and middle-income countries: a systematic review. BMC pregnancy and childbirth. 2014 Dec;14(1):30.
- 13. Shiferaw S, Spigt M, Godefrooij M, Melkamu Y, Tekie M. Why do women prefer home births in Ethiopia?. BMC pregnancy and childbirth. 2013 Dec;13(1):5.
- 14. Titaley CR, Hunter CL, Dibley MJ, Heywood P. Why do some women still prefer traditional birth attendants and home delivery?: a qualitative study on delivery care services in West Java Province, Indonesia. BMC pregnancy and childbirth. 2010 Dec;10(1):43.
- 15. Sarker BK, Rahman M, Rahman T, Hossain J, Reichenbach L, Mitra DK. Reasons for preference of home delivery with traditional birth attendants (TBAs) in rural Bangladesh: a qualitative exploration. PloS one. 2016 Jan 5;11(1):e0146161. pmid:26731276
- 16.
Okeke EN, Chari A. Home and dry? Bad weather, place of birth, and birth outcomes. Working Paper, RAND Corporation. 2016.
- 17. Roro MA, Hassen EM, Lemma AM, Gebreyesus SH, Afework MF. Why do women not deliver in health facilities: a qualitative study of the community perspectives in south central Ethiopia?. BMC research notes. 2014 Dec;7(1):556.
- 18. Leslie HH, Fink G, Nsona H, Kruk ME. Obstetric facility quality and newborn mortality in Malawi: a cross-sectional study. PLoS medicine. 2016 Oct 18;13(10):e1002151. pmid:27755547
- 19. Kruk ME, Larson E, Twum-Danso NA. Time for a quality revolution in global health. The Lancet Global health. 2016 Sep 1;4(9):e594–6. pmid:27539798
- 20. Godlonton S, Okeke EN. Does a ban on informal health providers save lives? Evidence from Malawi. Journal of development economics. 2016 Jan 1;118:112–32. pmid:26681821
- 21. Robyn PJ, Fink G, Sié A, Sauerborn R. Health insurance and health-seeking behavior: Evidence from a randomized community-based insurance rollout in rural Burkina Faso. Social Science & Medicine. 2012 Aug 1;75(4):595–603.
- 22. Mathole T, Lindmark G, Majoko F, Ahlberg BM. A qualitative study of women's perspectives of antenatal care in a rural area of Zimbabwe. Midwifery. 2004 Jun 1;20(2):122–32. pmid:15177855
- 23. Simkhada B, Teijlingen ER, Porter M, Simkhada P. Factors affecting the utilization of antenatal care in developing countries: systematic review of the literature. Journal of advanced nursing. 2008 Feb;61(3):244–60. pmid:18197860
- 24.
Donabedian A. The definition of quality and approaches to its assessment. Ann Arbor: Michigan Health Administration Press; 1980.
- 25. Hotchkiss DR, Krasovec K, Driss Zine‐Eddine El‐Idrissi M, Eckert E, Karim AM. The role of user charges and structural attributes of quality on the use of maternal health services in Morocco. The International journal of health planning and management. 2005 Apr;20(2):113–35. pmid:15991458
- 26. Hounton S, Chapman G, Menten J, De Brouwere V, Ensor T, Sombié I, et al. Accessibility and utilisation of delivery care within a Skilled Care Initiative in rural Burkina Faso. Tropical medicine & international health. 2008 Jul;13:44–52.
- 27. Mayhew M, Hansen PM, Peters DH, Edward A, Singh LP, Dwivedi V, et al. Determinants of skilled birth attendant utilization in Afghanistan: a cross-sectional study. American Journal of Public Health. 2008 Oct;98(10):1849–1856. pmid:18703445
- 28.
Tsui AO, Ukwuani F, Guilkey D, Angeles G. Health program effects on individual use of services. MEASURE Evaluation, Carolina Population Center, University of North Carolina at Chapel Hill; 2002 Oct.
- 29. Moyer CA, Adongo PB, Aborigo RA, Hodgson A, Engmann CM, DeVries R. “It’s up to the woman’s people”: how social factors influence facility-based delivery in Rural Northern Ghana. Maternal and child health journal. 2014 Jan 1;18(1):109–19. pmid:23423857
- 30. Glick P. How reliable are surveys of client satisfaction with healthcare services? Evidence from matched facility and household data in Madagascar. Social Science & Medicine. 2009 Jan 1;68(2):368–79.
- 31.
Okeke EN, Glick P, Abubakar IS, Chari A, Pitchforth E, Exley J, et al. The Better Obstetrics in Rural Nigeria (BORN) Study: An impact evaluation of the Nigerian Midwives Service Scheme. RAND Corporation Research Report RR-1215. 2015.
- 32. Abdi H, Williams LJ. Principal component analysis. Wiley interdisciplinary reviews: computational statistics. 2002;2(4):433–459.
- 33. Dehejia RH, Wahba S. Causal effects in nonexperimental studies: Reevaluating the evaluation of training programs. Journal of the American statistical Association. 1999 Dec 1;94(448):1053–62.
- 34. Imai K, Van Dyk DA. Causal inference with general treatment regimes: Generalizing the propensity score. Journal of the American Statistical Association. 2004 Sep 1;99(467):854–66.
- 35. Hirano K, Imbens GW. The propensity score with continuous treatments. Applied Bayesian modeling and causal inference from incomplete-data perspectives. 2004 Feb 7;226164:73–84.
- 36.
Wang J. Exposure–Response Modeling: Methods and Practical Implementation. CRC press. 2015.
- 37.
Galagate D. Causal Inference With a Continuous Treatment And Outcome: Alternative Estimators for Parametric Dose Response Functions With Applications. Doctoral Dissertation, University of Maryland. 2016. Available at: https://drum.lib.umd.edu/bitstream/handle/1903/18170/Galagate_umd_0117E_16898.pdf?sequence=1
- 38. Yang S, Imbens GW, Cui Z, Faries DE, Kadziola Z. Propensity score matching and subclassification in observational studies with multi‐level treatments. Biometrics. 2016 Dec;72(4):1055–65. pmid:26991040
- 39. Gabrysch S, Campbell OM. Still too far to walk: literature review of the determinants of delivery service use. BMC pregnancy and childbirth. 2009 Dec;9(1):34.