Use of performance metrics for the measurement of universal coverage for maternal care in Mexico

Abstract

This study provides evidence for those working in the maternal health metrics and health system performance fields, as well as those interested in achieving universal and effective health care coverage. Based on the perspective of continuity of health care and applying quasi-experimental methods to analyse the cross-sectional 2009 National Demographic Dynamics Survey (n = 14 414 women), we estimated the middle-term effects of Mexico's new public health insurance scheme, Seguro Popular de Salud (SPS) (vs women without health insurance) on seven indicators related to maternal health care (according to official guidelines): (a) access to skilled antenatal care (ANC); (b) timely ANC; (c) frequent ANC; (d) adequate content of ANC; (e) institutional delivery; (f) postnatal consultation and (g) access to standardized comprehensive antenatal and postnatal care (or the intersection of the seven process indicators). Our results show that 94% of all pregnancies were attended by trained health personnel. However, comprehensive access to ANC declines steeply in both groups as we move along the maternal healthcare continuum. The percentage of institutional deliveries providing timely, frequent and adequate content of ANC reached 70% among SPS women (vs 64.7% in the uninsured), and only 57.4% of SPS-affiliated women received standardized comprehensive care (vs 53.7% in the uninsured group). In Mexico, access to comprehensive antenatal and postnatal care as defined by Mexican guidelines (in accordance to WHO recommendations) is far from optimal. Even though a positive influence of SPS on maternal care was documented, important challenges still remain. Our results identified key bottlenecks of the maternal healthcare continuum that should be addressed by policy makers through a combination of supply side interventions and interventions directed to social determinants of access to health care.

Health services research, maternal health, primary health care, reproductive health, women's health

Key Messages

• As suggested by WHO, developing countries need to move beyond the traditional definition of maternal care in order to guarantee effective access to those services that will eventually impact the health conditions of the population. It is imperative to further disseminate the idea of a continuous mother–newborn attention process that includes reproductive health care, antenatal care (ANC), postnatal care and child health services.

• Attainment of the SDGs demands investments in information systems that support the continuity of care. At the clinical decision-making level, such systems must register individuals and not episodes, in order to guarantee prompt and accurate follow-up through effective communication with the team of health providers. From the broader perspective of health systems performance, a crucial component of these efforts is the design of metrics that guarantee an objective follow-up and evaluation of these efforts, such as the indicators proposed in this study. Increasing the intensity of management and accountability lies at the core of proper measurement of clinical processes. These efforts require a shared agenda between academia and health services.

Introduction

One decade ago, the WHO, concerned by the slow progress made in the achievement of the Millennium Development Goals (MDGs) related to child and maternal mortality, made a call to re-orient maternal care towards a ‘continuous mother–newborn attention process’ encompassing reproductive health care, antenatal care (ANC), postnatal care and child health services (Kerber et al. 2007). Improvements in the health care process and the resulting health gains could be evaluated by measuring coverage along the stages of this continuum (WHO 2005; Kerber et al. 2007).

Recent conceptual proposals (Ng et al. 2014) and empirical evidence (Ronsmans et al. 2010) support the idea that maternal and child health could improve by increasing access to timely, high-quality health services aimed at reducing maternal and infant mortality. Almost two thirds of newborn and child deaths could be avoided if comprehensive, essential healthcare services were provided along the pregnancy, delivery and puerperium stages. Projections for expanding effective access up to 90% show a potential reduction in mortality in children under 5 of about 70% by 2035 compared with the 2010 baseline levels (Walker et al. 2013).

Regarding measurement, several studies have proposed the use of coverage indicators to monitor ANC improvements (Barros et al. 1996; VanderWeele et al. 2009), which include timeliness of ANC (initial antenatal care visit during the first trimester of pregnancy), number of visits, gestational age, and type of provider of antenatal health services (Gajate-Garrido 2013). Composed indicators that use some measures of the continuum of maternal care provide a partial view of the healthcare process (Kotelchuck 1994; Hodgins and D’Agostino 2014). However, these population-based coverage measures are lacking in patient-centered care information (as that provided by medical records), and in the integration of information about access and quality of health care along the maternal healthcare process. It is in this context that we discuss recent efforts made by Mexico to expand access to ANC and measure the expansion of coverage through the use of both standard coverage indicators and effective access indicators.

Different initiatives to improve reproductive and child health outcomes among the most vulnerable population have been launched in Mexico (Frenk et al. 2012; Strouse et al. 2016; Serván-Mori et al. 2016a). In 2003, a new publicly funded insurance scheme called Seguro Popular de Salud (SPS) became available for the self-employed, the un-employed and those working in the informal sector of the economy. This is a voluntary and defacto non-contributory scheme of public insurance (in-built within the Ministry of Health) for population without access to traditional social security (Frenk et al. 2006). The SPS benefit package includes a set of maternal and newborn services intended to reduce maternal deaths (Serván-Mori et al. 2016a).

Evaluations of SPS have typically focused on the impact on financial protection of its enrollees (Galárraga et al. 2010; García-Díaz and Sosa-Rubí 2011; Sosa-Rubí et al. 2011; Ávila-Burgos et al. 2013). Although there is limited evidence on the impact of this program on maternal health, it has been documented that exposure to SPS increases the probability of receiving at least four maternal consultations (adjusted OR = 1.65, P < 0.01). However, this relationship loses statistical significance once timely antenatal care is incorporated as a covariate (Serván-Mori et al. 2015). Additionally, there is evidence that SPS enrollees are more likely to receive antenatal care in an SPS-accredited unit although one-fifth of these women subsequently choose the private sector for the delivery event (Sosa-Rubí et al. 2009; González-Block et al. 2010). SPS also removes barriers to access healthcare services among women of low educational attainment, thus reducing the risk of preterm birth (Strouse et al. 2016). More recently, a study found that an increase in SPS coverage of 10 percentage points caused a decrease in the risk of miscarriage of 0.4 percentage points, which means that a hypothetical scenario of universal coverage of SPS would potentially lead to a 30% reduction in the risk of miscarriage for the target population (Pfutze 2015).

After >10 years of SPS implementation and in spite of evidence of an increasing utilization of maternal and child services, universal access to maternal care has not been reached. There are problems related to the actual delivery of services and the measurement of coverage expansion. Regarding resources for delivery, there is still an inequitable distribution of physical (infrastructure) and human resources, which in turn has limited access to basic health care and emergency services among disadvantaged populations (Grogger et al. 2014; Knaul et al. 2012; Fried et al. 2013). Persisting variability in timeliness and quality of ANC has also been documented (Fajardo-Dolci et al. 2013). This variability is expressed in differences in maternal health indicators. In 2013, the maternal mortality ratio in Nuevo Leon (the second Mexican state with the highest Human Development Index or HDI) was 17.6 per 100 000 live births while in Guerrero (one of the states with the lowest HDI) it was 63.3 (DGIS/SSa 2016).

Performance of the provision of antenatal care shows enormous variation at the state level (in terms of timeliness, frequency and content) for a given level of public health expenditure (Serván-Mori et al. 2016a). Important geographical differences in maternal health care coverage also persist, which negatively impact access to these services by women of the most disadvantaged groups (Heredia-Pi et al. 2016). Furthermore, there is evidence that suggests that access to health care in rural communities is much lower than access in urban localities (Grogger et al. 2014) as shown by the limited access to these services among indigenous, low socioeconomic status and geographically isolated populations, which calls for the urgent monitoring of the expansion and availability of infrastructure and human resources (Laurell 2013; Gakidou et al. 2016).

Health authorities and researchers in Mexico have used standard indicators to measure maternal care coverage, including onset of prenatal care, frequency of antenatal care or, in the best of cases, occurrence of several independent antenatal interventions or access to obstetric care. However, there is a shortage of metrics to measure the continuity of the maternal care services, including antenatal, obstetric and postnatal care. The lack of these metrics has possibly not allowed the identification of critical features in the maternal and child care continuum, reducing the capacity to focus on the weakest maternal care stages and allocate resources to them.

How can we better measure the continuity of maternal care? This study provides evidence for the development of maternal health metrics and health system performance fields, and for decision-makers responsible for achieving universal and effective health coverage. Following the perspective of continuity of maternal and child health care, we discuss the need to migrate to new health metrics aiming to produce innovative indicators to measure the performance of maternal health services in Mexico. We specifically analyse the middle-term effects of SPS on access to maternal health care in order to discuss how more new indicators of maternal health care effectiveness could expand available evidence on healthcare system performance.

Methods

Data and study population

We conducted a retrospective cohort analysis using data from the 2009 National Demographic Dynamics Survey (ENADID by its Spanish initials), a nationally representative survey developed by Mexico’s National Institute of Statistics and Geography (INEGI 2016). This survey uses a two-stage probability sample from Mexico’s 32 states. This survey includes modules covering household composition and characteristics, demographic, education and health information on 343 887 household members. In particular, ENADID includes a reproductive health module to be responded by each woman aged 12 to 54 years residing in the household (n = 112 159). However, the ANC, delivery and postpartum module was administered only to women aged 15 to 54 years.

The analytical sample included 16 267 Mexican women aged 15 to 49 (with or without SPS) that reported having had an obstetric episode from 2004 to 2009. We excluded from the sample those women aged 50 years or older, since they represent a small age group (<5%). After excluding observations with incomplete basic socio-demographic information, the final sample included 14 414 individuals (non-response rate = 11.4%), 7 118 of whom were affiliated to SPS. No differences in socioeconomic and demographic characteristics between the analytical and excluded samples were found.

Measures

Following our previous work (Heredia-Pi et al. 2016; Serván-Mori et al. 2016b), and the Mexican Official Guidelines (Secretaría de Salud 1995), we analysed seven process ANC and postnatal care indicators calculated from the fertility and maternal health survey module: (a) if the woman received any kind of ANC (by a doctor, nurse, midwife or other); (b) if the ANC was provided by a skilled birth attendant (by a doctors or nurse); (c) if the first medical visit occurred in the first trimester of pregnancy (or timeliness); (d) if the patient received at least four antenatal consultations (or frequent ANC); (e) adequate content of ANC. These indicators summarize the main procedures that women should receive from their health provider during ANC. The survey has information for eight procedures: weight measurement, blood pressure and abdominal examinations, urine, blood and HIV test, immunization against tetanus and the prescription of vitamins/mineral supplementation. We considered appropriate content as being located in the uppermost quantile of the sum of the total of procedures (Heredia-Pi et al. 2016). In line with previous analysis (Serván-Mori et al. 2016b), all interventions or procedures provided during antenatal care visits were weighted equally. We also included two indicators related to the delivery process and postnatal care: (f) if the delivery was attended by skilled personnel (either doctor or nurse) or was an institutional delivery and (g) if the newborn received a postnatal consultation within the first month of life (or timely postnatal care). We further defined seven binary outcome variables indicating the incremental access of the aforementioned interventions along the antenatal-postnatal continuum. Access to standardized comprehensive antenatal and postnatal care was then set for those women at the intersection of the seven process indicators.

Women were classified as affiliated with SPS if they answered affirmatively and exclusively to the question […] (YOU/NAME) are you affiliated to SPS? […] Those with no health insurance were classified as no-SPS. A dichotomous variable was created, with affiliation to SPS equal to one (1), and no-SPS equal to zero (0).

Covariates included the individual and household-level data shown in Table 1. Individual-level variables included woman’s age, schooling and employment at the time of the survey, marital status, parity and year of the index live birth (2004–2009). Household-level covariates included the characteristics of head of household (sex, age and schooling), indigenous condition (a household is considered indigenous if the head of the family, a spouse and/or an ascendant self-identifies as a speaker of an indigenous language) (Comisión Nacional para el Desarrollo de los Pueblos Indígenas 2009), access to social programs that provides economic support to the family (including conditional cash transfers Oportunidades program, Farmers Direct Support Program or PROCAMPO), and a continuum asset and housing index (SES) broken-down into tertiles. It implemented by analyzing the principal components through polychoric correlation matrices (Kolenikov and Angeles 2004), and designed to combine different household assets and housing infrastructure conditions (range: -4.57 to 5.05), where higher values denoted a greater number of assets and better housing conditions. The proportion of under 5 and older than 60 years members was also registered. Community or contextual characteristics include the size of the place of residence (rural, semi-urban or urban). We also calculated a state-level indicator for concentration of highly marginalized localities or localities with deprivation in order to account for differences in the households’ proximate environment.

Table 1.

Main socio-demographic and economic characteristics of the studied population according to Seguro Popular Affiliation before and after matching process, Mexico 2009

	Mean or percentage [CI 95%]
	Before matching				After matching
	Seguro Popular de Salud		Without health insurance		Seguro Popular de Salud		Without health insurance
	(n = 7 118)		(n = 7 296)		(n = 5 128)		(n = 5 128)
Individual characteristics of women
Age (yrs)	26.0	[25.8;26.2]	25.6	[25.4;25.7]**	25.5	[25.3;25.7]	25.5	[25.3;25.7]
Schooling (yrs)	7.9	[7.79;7.95]	8.6	[8.54;8.71]**	8.2	[8.12;8.30]	8.2	[8.14;8.34]
Employed (%)	21.8	[20.8;22.8]	29.7	[28.6;30.7]**	24.8	[23.6;26.0]	24.8	[23.6;25.9]
Married (%)	50.0	[48.8;51.1]	46.0	[44.9;47.2]**	47.2	[45.8;48.5]	46.7	[45.4;48.1]
Parity (%)
No living children	1.2	[0.92;1.42]	2.5	[2.10;2.81]**	1.6	[1.22;1.90]	1.5	[1.13;1.79]
1 child	30.1	[29.0;31.2]	34.8	[33.7;35.9]**	33.3	[32.0;34.6]	34.0	[32.7;35.3]
2 children	27.2	[26.2;28.3]	29.2	[28.2;30.2]*	29.0	[27.7;30.2]	29.0	[27.7;30.2]
3 or more children	41.5	[40.4;42.7]	33.5	[32.4;34.6]**	36.1	[34.8;37.4]	35.5	[34.2;36.9]
Year of obstetric episode (%)
2004	9.9	[9.2;10.58]	14.2	[13.4;15.0]**	11.1	[10.2;11.9]	10.7	[9.80;11.5]
2005	11.9	[11.1;12.6]	16.5	[15.7;17.4]**	13.0	[12.1;13.9]	12.5	[11.6;13.4]
2006	13.6	[12.8;14.4]	19.6	[18.7;20.5]**	15.8	[14.8;16.8]	16.0	[15.0;17.0]
2007	21.8	[20.8;22.7]	19.8	[18.9;20.7]*	21.9	[20.8;23.1]	22.4	[21.3;23.5]
2008	29.4	[28.3;30.4]	22.2	[21.2;23.1]**	27.4	[26.2;28.6]	27.9	[26.7;29.1]
2009	13.5	[12.7;14.3]	7.8	[7.16;8.39]**	10.8	[9.90;11.6]	10.5	[9.60;11.3]
Household characteristics
Male household head (%)	80.2	[79.3;81.1]	78.4	[77.4;79.3]*	79.4	[78.3;80.5]	78.9	[77.8;80.0]
Household head age (yrs)	39.8	[39.5;40.2]	40.8	[40.5;41.1]**	40.1	[39.8;40.5]	40.3	[39.9;40.7]
Household head years of education	6.5	[6.38;6.55]	7.5	[7.35;7.55]**	6.9	[6.77;6.99]	6.8	[6.70;6.93]
Indigenous (%)	13.0	[12.2;13.8]	10.4	[9.7;11.1]**	11.8	[10.9;12.7]	11.6	[10.7;12.4]
Beneficiary of government program (%)	23.0	[22.0;23.9]	7.5	[6.89;8.1]**	11.3	[10.4;12.1]	10.5	[9.70;11.4]
Asset and housing indexa	−0.3	[-0.31;-0.25]	0.3	[0.23;0.3]**	−0.02	[-0.06;0.02]	−0.02	[-0.06;0.02]
Demographic composition (%)
Proportion of members <5 years	13.9	[13.8;14.1]	13.7	[13.6;13.8]+	13.9	[13.7;14.1]	13.9	[13.8;14.1]
Proportion of members >60 years	8.8	[8.68;8.99]	8.8	[8.65;8.99]	8.8	[8.58;8.95]	8.7	[8.53;8.92]
Community or contextual characteristics
Household in rural community (%)	38.5	[37.4;39.6]	23.9	[22.9;24.9]**	31.0	[29.7;32.3]	30.5	[29.3;31.8]
Household in semi-urban community (%)	35.8	[34.7;36.9]	32.0	[30.9;33.1]**	35.9	[34.6;37.2]	36.0	[34.6;37.3]
Household in urban community (%)	25.7	[24.7;26.7]	44.1	[42.9;45.2]**	33.1	[31.8;34.4]	33.5	[32.2;34.8]
Localities with high marginality levels (%)b	15.6	[15.3;15.9]	17.2	[16.9;17.5]**	16.2	[15.9;16.6]	16.1	[15.7;16.4]

Notes: **P < 0.01, *P < 0.05 and ⁺P < 0.10.

a

Continuous variable computed using principal component analysis with polychoric matrices.

b

In order to account for differences in the households’ proximate environment, a state-level indicator for concentration of highly marginalized localities was used. Data obtained using the marginality index by state and municipality. Mean bias before matching: 15.3% (p50 = 13.2). Mean bias after matching: 0.9% (p50 = 1.0%). The single-nearest neighbor algorithm (caliper size = 0.001) was used to match the treated and comparison groups. Source: National Demographic Dynamics Survey 2009.

Statistical approach

We describe sociodemographic and ANC characteristics by type of insurance (SPS and no-SPS). To assess independence between groups, unadjusted linear regression models were constructed for continuous variables, and chi-square tests were performed for binary and categorical variables.

Next, taking into account self-selection aspect of enrollment to SPS (Sosa-Rubí et al. 2011; Wirtz et al. 2012), we used a quasi-experimental matching method through propensity score matching (Rosenbaum and Rubin 1983; Abadie and Imbens 2006). Matching can improve causal inference in observational studies, assess the robustness of results, reduce potential bias due to the non-random assignment to treatment, and allows to build a comparison group to control for the potential effect of self-selection bias and confusion due to observable variables. We estimated a propensity score with the Stata module psmatch2 (Leuven and Sianesi 2003). Covariates selected to estimate the SPS propensity score were those previously used to determine SPS enrollment (Sosa-Rubí et al. 2009; García-Díaz and Sosa-Rubí 2011; Wirtz et al. 2012; Ávila-Burgos et al. 2013) (see Appendix S1 in Supplementary material). A nearest neighbour matching algorithm 1-1 with common support and caliper equal to 0.001 was performed. The matching analytic sample consisted of 10 256 women aged 15-49 (71% of the prior analytical sample). Bivariate tests suggest well-balanced groups (Table 1 and Appendix S2 in Supplementary material); the average absolute percent bias after matching was 0.9% (compared with 15.3% prior to matching), the interquartile range of the propensity score on the region of common support was (0.39–0.60). We reported the average treatment effect on the treated or the average gain from treatment for those who actually were affiliated to SPS (expressed in percentage points).

Empirical results

Descriptive analysis

Table 1 shows the descriptive statistics of the covariates mentioned above before matching. We found systematic differences between SPS and non-SPS individuals: women enrolled in SPS, as expected, were more prone to social and health vulnerability. Untreated women (n = 7118) had, on average, 0.7 years less of educational attainment compared with controls (n = 7296), and were less likely to be formally employed at the time of the survey (22% vs 30%). 42% of the exposed population had ≥3 living children, in contrast to 34% of the unexposed. Women in the exposed group were also more likely to have had their last obstetric event in the previous 2 years. Regarding household characteristics, there were important differences in educational attainment of the household head (6.5 vs 7.5 years), exposure to government programs (23% vs 7.5%), and likelihood of living in a rural community (39% vs 24%). Although no important differences were found in terms of marginality at the community-level (16% of households living in localities with high and very-high marginality index, vs 17% in the control group), an important difference in the household assets index (-0.30 vs 0.30) was found.

The independent and conditional analyses of the antenatal and postnatal care coverage among unmatched women are shown in Table 2. In both groups, all women reported having received some kind of prenatal consultation, and 94% had access to skilled personnel in the prenatal stage and 93% in the delivery stage (panel 1). The exposed group outranked its uninsured counterparts in the following indicators: ANC frequency (91% vs 88%) and ANC adequate content (92% vs 89%). However, they were less likely to having had received timely postnatal care (77% vs 79%).

Table 2.

Independent and conditional analyses of the coverage of the dimensions of antenatal and postnatal care, according to Seguro Popular Affiliation, Mexico 2009

	Seguro Popular de Salud		Without health insurance
	(n = 7 118)		(n = 7 296)
Panel 1: Independent coverage
Received antenatal care (A)	100	[100;100]	100	[100;100]
ANC by skilled personnel (B)	94.4	[93.9;94.9]	94.0	[93.4;94.5]
Timely antenatal care (C)**	83.5	[82.7;84.4]	85.3	[84.5;86.2]
Frequent ANC (D)**	91.0	[90.3;91.7]	88.0	[87.2;88.7]
Adequate content of ANC (E)**	91.8	[91.2;92.5]	88.6	[87.8;89.3]
Institutional delivery (F)	93.3	[92.7;93.8]	93.2	[92.7;93.8]
Timely postnatal care (G)**	76.6	[75.6;77.6]	78.7	[77.8;79.6]
Panel 2: Conditional coverage
ANC by skilled personnel (A to B)	94.4	[93.9;94.9]	94.0	[93.4;94.5]
Timely antenatal care (A to C)*	79.2	[78.3;80.2]	80.8	[79.9;81.7]
Frequent ANC (A to D)	74.0	[73.0;75.1]	73.8	[72.8;74.8]
Adequate content of ANC (A to E)*	70.9	[69.8;71.9]	69.2	[68.1;70.3]
Institutional delivery (A to F)+	68.4	[67.3;69.5]	67.0	[65.9;68.1]
Standardized comprehensive antenatal and postnatal care (A to G)	55.5	[54.4;56.7]	56.1	[54.9;57.2]

Notes: **P < 0.01, *P < 0.05, and +P < 0.10.

(A) received ANC from either doctor, nurse, midwife or others, (B) received ANC from doctor or nurse, (C) first ANC visit during the first trimester of pregnancy, (D) at least four ANC visits, (E) Receiving at least 6 of the following interventions: weight measurement, blood pressure and abdominal examinations, urine, blood and HIV test, immunization against tetanus, and prescription of vitamins/mineral supplementation, (F) delivery attended by a doctor or nurse, (G) if the newborn received a postnatal consultation within the first month of life. Source: National Demographic Dynamics Survey 2009.

The conditional coverage shows an entirely different panorama (panel 2). Once we condition the indicators on the occurrence of the previous step of the maternal health continuum, steep declines were observed in the composite measures. For instance, 91% of the treated and 88% of the untreated reported frequent ANC, but these percentages drop to 74% for both groups when skilled and timely ANC are incorporated. Furthermore, only 56% of women had access to standardized comprehensive antenatal and postnatal care, both in the treated and untreated groups.

Continuum of maternal care analysis

Figure 1 shows association between SPS and the percentage of women covered among the matched sample. The horizontal axis displays the maternal health indicators and the vertical axis, the conditional coverage along with its CI95%. SPS enrollees consistently show higher coverage across the spectrum of indicators. SPS group showed two percentage points higher probability of receiving ANC by skilled personnel, 2.7 percentage points higher probability on likelihood of frequent ANC, and 4.6 percentage points higher probability on adequate content of care. In contrast, timely ANC was similar, regardless of insurance status. Yet, the largest loss of coverage across the continuum of health care is due to the delay of ANC in the first trimester, and timely postnatal care. We estimate that SPS had a positive effect of 5.3 percentage points on institutional delivery and 3.7 percentage points on the standardized comprehensive coverage of antenatal and postnatal care. Despite this positive effect, the access indicator was heterogeneous among Mexican states. In Figure 2, we show the geographical distribution in quartiles of the comprehensive coverage indicator in the matched treatment group. The difference of the average state located between the uppermost and the lowest quartiles was 17 percentage points (65.2% vs 48.4%). Women in four states (Chiapas, Durango, Guerrero and Jalisco) reached <50% of standardized comprehensive coverage of antenatal and postnatal care.

Figure 1.

Influence of Seguro Popular de Salud on antenatal and postnatal coverage indicators, Mexico 2009. Note: Estimated on a matched sample of 10,256 individuals. Mean bias before matching: 15.3% (p50 = 13.2). Mean bias after matching: 0.9% (p50 = 1.0%). The single-nearest neighbor algorithm (caliper size = 0.001) was used to match the treated and comparison groups (in parenthesis matched sample = 5,128). Access to the standardized comprehensive antenatal and postnatal care was then set for those women at the intersection of the seven process indicators

Figure 2.

Access to the standardized comprehensive antenatal and postnatal care by state (among affiliates of Seguro Popular de Salud), Mexico 2009

Discussion and conclusion

The agreement of the Global Strategy for Women’s, Children’s and Adolescents’ Health (2016–2030) (UN 2015a) and the endorsement of the third goal of the Sustainable Development Goals (SDGs) has highlighted the importance of having universal health systems, as well as implementing efforts to improve the responsiveness and financial fairness of these health systems (UN 2015b). However, international evidence and official reports suggest that achievements in the fulfillment of the MDGs are insufficient, particularly in low- and middle-income countries (LMIC) (UN 2015b).

The ultimate goal of universal health care coverage (UHC) initiatives is to guarantee regular access to comprehensive and effective health care services with financial protection to all the population. However, most initiatives intended to increase the levels of coverage in LMIC have focused on the expansion of health care with little concern for timeliness, comprehensiveness and quality of care (QoC). This will eventually be reflected on health outcomes, which could show limited progress. In these countries, including Mexico, increases in coverage are expected to be translated into health gains. With increasing access to health services, a higher proportion of avoidable maternal have deaths moved to health facilities. In this context, poor QoC becomes a key issue to end preventable maternal mortality and morbidity (Graham et al. 2013; Tunçalp et al. 2015).

Countries seeking to achieve UHC should review and expand the definition of antenatal and maternal care, to include features such as timeliness and adequate content of care as a priority element to achieve higher performance and better and more equitable health outcomes. The design and implementation of comprehensive metrics of healthcare coverage are critical to consolidate an effective and universal health system, and to reduce social gaps.

Based on the perspective of continuity of health care and applying quasi-experimental methods, this paper has provided an example that shows how innovative and context-adjusted indicators allow us to identify gaps in maternal health care. Although access to information about the performance of health systems is relevant to respond to the health needs of the populations they cover, the generation of routinely accurate information about the performance of health systems still represents a challenge for LMIC.

The study showed declines in effective coverage of maternal health interventions after controlling for systematic differences between treatment (SPS beneficiaries) and control groups of women without health insurance. All women (94%) surveyed had received care by medical staff, but only 56% of them received the complete package of maternal health care services. The largest drops in coverage occurred in timely antenatal care (14 percentage points drop between groups, on average) and postnatal care within the first month after delivery (12 percentage points).

Results also showed that even though major efforts have been made to expand access to maternal and child health care services in Mexico through SPS and vertical health programs that intend to improve access to specific health services, there are still bottlenecks that limit access to comprehensive antenatal and postnatal care. Thus, additional efforts are needed to guarantee the continuity of care, specifically in the dimensions of opportunity of antenatal care and follow-up after delivery, as recommended by WHO and national policies.

This study offers several lessons for LMIC. First, as suggested by WHO, developing countries need to move beyond the traditional definition of maternal care in order to guarantee effective access to those services that will eventually impact the health conditions of the population. It is imperative to further disseminate the idea of a continuous mother-newborn attention process that includes reproductive health care, ANC, postnatal care and child health services. Previous studies showed that there is a significant correlation between frequency of ANC visits and timeliness of care (Serván-Mori et al. 2015), and frequency of ANC visits and content of care, and that comprehensive ANC may render large marginal gains on the weight of the low-weight births (Serván-Mori et al. 2016b).

Second, implementation of a comprehensive, individual-centered approach requires the strengthening of basic health services and the establishment of organizational and monitoring mechanisms to guarantee the continuity of health care among units and levels of care, focused particularly in the postpartum stage (in which 52% of maternal deaths in LAC countries occur) (Say et al. 2014). The design of policies aimed at improving maternal health care also requires more programmatic research on maternal healthcare performance. Interventions aimed at increasing QoC should be accompanied by a set of process and result metrics, which ought to be tailored according to specific needs of different population groups and service delivery models. These fine point metrics should be consistent with indicators associated to system-level objectives, such as those outlined in this paper.

Third, attainment of the SDGs demands investments in information systems that support the continuity of care. At the clinical decision-making, such systems must register individuals and not episodes, in order to assist prompt and accurate follow-up through effective communication with the team of healthcare providers. From the perspective of health systems performance, a crucial component of these efforts is the design of metrics that guarantee an in-depth follow-up and evaluation of these efforts, such as the metrics proposed in this study (Tunçalp et al. 2015). Increasing the intensity of management and accountability lies at the core of proper measurement of clinical processes (WHO 2013). These efforts require a shared agenda between academia and healthcare services planning and provision.

Finally, in LMIC, the implementation of more stringent criteria to evaluate health programs faces several challenges. First, it has recognized the need for administrative registries that support decision making (Kendall & Langer 2015). It is critical to move towards information systems that are reliable, portable, secure and manageable, and design metrics that monitor broader dimensions of care. The routine registries should be oriented to capture patients and not events or visits. The individual records should be the input of the administrative registries. There is evidence that shows that this type of information contribute to more intensive doctor-patient interaction, treatment adherence, and compliance with treatment guidelines (Chaudhry et al. 2006). Data should be gathered at the facility level to better deal with technical aspects of QoC that cannot be assessed through health survey data (Graham and Varghese 2012). Second, the focus has to shift away from measuring simple coverage of ANC services or ANC single interventions (Dettrick et al. 2013; Kendall and Langer 2015) to metrics that include both access and QoC at different stages of the maternal healthcare process. This raises an important question about the measurement of outcomes and the way to link them to health results, taking into account demand and supply perspectives. For example, frequent interaction between patients and doctors (and other health personnel cadres) contribute more to perceived quality than the number of available doctors at the facility (Dettrick et al. 2013). Institutional and organizational arrangements are critical for the continuity and accessibility of care (Graham and Varghese 2012). Third, countries should assess the feasibility of implementing supply-side incentives and shared responsibility mechanisms and incorporate the evidence of pay-for-performance experiences (Das et al. 2016), non-pecuniary incentives, such as public recognition of health providers (Giuffrida et al. 2009), and conditional cash transfers to pregnant women (Lim et al. 2010).

This study has limitations. First, although ENADID is a high-quality, population-based survey, it may have sampled a non-representative group of SPS population, which would limit the generalizability of our results. Second, we used self-reported measures of outcome variables, SPS exposure, and co-variables. In particular, outcome variables may be subject to recall bias. Women in our sample were asked about current exposure, so our measure may not have captured those who had been exposed but left SPS. Third, despite the use of a rigorous methods that reduce potential biases in the selection of participation of SPS individuals due to observed factors (Diaz and Handa 2006), it is possible that the consistency of our results was affected by the existence of unobservable elements in the examined relationship which could influence the decision to seek health care (e.g. moral hazard due to known health problems). Ideally, this problem could be solved by using instrumental variables, such as the percentage of SPS members at the municipal level (Sosa-Rubí et al. 2009). However, ENADID did not publish information on the place of residence of the surveyed population. Therefore, authors do not claim that the conclusions of this article have a strong causal inference and results should be considered a conservative estimation of the influence of SPS. Fourth, ENADID does not provide detailed information on the content or timeliness of postnatal care. This limited our ability to identify more accurately the quality of postpartum care according to international guidelines (WHO 2014). Further studies, surveys and information systems aimed at keeping track of the continuum of maternal health care should make an effort to better capture this important stage in order to have more reliable metrics.

In order to achieve the SDGs, the Mexican government has made major efforts to expand access to a basic set of healthcare services and achieve UHC. However, the presupposition that UHC can be readily translated into effective access is at least naïve (Cataife and Courtemanche 2014). This study suggests that users of healthcare services offered by the Ministry of Health have reasonable access to antenatal care but access to additional interventions included in the continuum of antenatal and postnatal care described above is still limited. QoC should move up strongly in the agenda of the Mexican health system, with a focus in content of care, continuity and timeliness.

Data availability statement

Data do not contain identifying or sensitive subject information. There are no ethical restrictions on requesting access to the data. All data files are available from the web page http://www.inegi.org.mx/est/contenidos/proyectos/.

Authors’ contributions

ESM and DCL designed the study, outlined the idea for the paper and wrote the first draft of the manuscript. DCL analysed the data. OGD, GN, SGSR and RL substantially revised the drafts. All authors were involved in the revision of the final version of the article. All authors read and approved the final article.

Funding

No funding was secured for this study.

Supplementary Data

Supplementary data are available at HEAPOL online

Conflict of interest statement. None declared.