The impact of geographic access on intuitional delivery care utilization in low and middle-income countries: Systematic Review and Meta-Analysis

Background

Access to and quality of health services are a high priority for investments designed to increase health services utilization which eventually improve health outcomes ⁽¹⁾. Healthcare access is defined in terms of geographic, financial, temporal, digital/eHealth and cultural access, and availability of services ^{(2, 3)}. The digital dimension of healthcare access addresses the digital connectivity/communication with healthcare providers, peer groups, and computerized health applications ⁽²⁾. Temporal access addresses the time required to get services (waiting time in reception, time spent in receiving treatment and time to next appointment) and its trade-offs ⁽²⁾. Availability is the opportunity to get the right type of healthcare, appropriate healthcare providers, materials and equipment ⁽³⁾. Financial accessibility focuses on the costs of services, and users’ ability and willingness to pay for, as well as to be protected from financial consequences of health care costs ^{(2, 3)}. The cultural dimension of access focuses on the acceptability of health services to individual users and communities, such as language match, trust to provider and public stigma. On the other hand, the geographic dimensions of access is the measure of the physical distance or travel time to service delivery points ^{(2, 3)}.

People living in low and middle-income countries (LMICs) tend to have less access to health care services than those in high income countries ⁽³⁾. Furthermore, within these countries, the poor and those living in rural areas have less access to these services ^{(2, 3)}. In rural areas, and for congested urban centres, the geographical dimension of access could be more important than urban centres with good transportation infrastructure; in those settings, service users might be expected to walk long distances and/or spend more time on travelling ⁽²⁾. Availability of transport services, nature of roads (seasonal impassibility), mountains and rivers may also play a role in accessing health care services ⁽⁴⁾.

Moreover, for any reason, the nearest health facility might not be the facility of choice. Not all services are provided at all levels of health facilities ⁽⁴⁾. For instance, in many countries, comprehensive emergency obstetric care (CEmOC) is not available at the lowest health facilities. In most countries, it is only available at hospitals, and for this reason pregnant women and their families may be required to travel further away for childbirth ⁽⁵⁾.

Studies from Zambia and Malawi found that the odds of facility delivery was higher among those who had physical access to higher level of care facilities ^{(6, 7)}. Analysis of the 2012 Haiti DHS and 2013 SPA data found that health facilities service readiness within 10 kilometre of radius of a given cluster was associated with an increased use of delivery services ⁽⁸⁾. Similarly, in Malawi and Zambia, living close to facility providing delivery service was significantly associated with health facility delivery ⁽⁷⁾. It was also noted that the increase in geographic distance was associated with a decrease in the use of health facility delivery ^{(6, 9)}.

However, it was found that having access to obstetric care facilities within one-hour ^{(10, 11)} and three kilometres ^{(12, 13)} of distance were not significantly associated with institutional delivery. Therefore, this study synthesized the evidence on the influence of geographic access on institutional delivery care use in low and middle-income countries.

Method

Search strategy

Our search included the following databases: MEDLINE, EMBASE, CINAHL, PsycINFO, Scopus and Maternity & Infant Care. We used multiple combinations of search terms or keywords, such as delivery care, geographic access, observational studies, low and middle-income countries, and Boolean operators (Table 1). The searching terms/keywords first used in OVID MEDLINE were adapted to the other databases. We also conducted a hand search of reference lists.

Study selection

Search results were imported into EndNote software to aggregate relevant articles and to manage duplications. Two authors independently screened the titles and abstracts to determine if the returned electronic search articles were related to the study. The respective lists of articles of both authors were combined and full-text articles were reviewed against the inclusion and exclusion criteria. Disagreements were resolved through discussion with a third reviewer, and we calculated inter-reviewer reliability; that is, the percentage of agreement between reviewers (Cohen’s kappa) ⁽¹⁴⁾.

Inclusion and exclusion criteria

We included studies only conducted in low-income and middle-income countries as defined by the World Bank ⁽¹⁵⁾ and published in English. We took studies published since 2000 that was the introduction of the Millennium Development Goals (MDGs) up to December 31, 2016. We have included quantitative cross-sectional studies, cohort and case-control studies. We excluded experimental or intervention studies and organization reports in our analysis. The studies had to report on the influence of geographic accessibility on maternal delivery service use.

Outcome measures

We selected articles that reported geographic access on delivery care use. Our outcome measure was utilization of institutional delivery.

Study quality assessment

Two authors independently assessed the methodological quality of each study using the Joanna Briggs Institute’s (JBI) critical appraisal checklist for studies reporting prevalence data ⁽¹⁶⁾ and cohort studies ⁽¹⁷⁾. Disagreements were resolved by discussion with a third reviewer.

Data extraction

We extracted data on influence of geographic accessibility on maternal delivery care use. We used data extraction form that include general information (publication details and country), and specific information (study setting, study design, study population, sample size, main findings) (Table 2). We created a summary matrix with the data extracted from all individual studies. Two authors independently extracted the data from the included studies into the constructed matrices. Discrepancies were resolved by discussion and the original study was reviewed to resolve further discrepancies.

Analysis

The results of studies were extracted, reviewed and reported in a systematic format. A Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA ) checklist ⁽¹⁸⁾ was used to synthesize and report the findings. A meta-analytic procedure was used to compute and aggregate effect sizes. The pooled effect size (Odds Ratio – OR) was calculated using a random effects model. The Q statistics, I² and Tau squared (τ²) were used to examine heterogeneity of studies. The analysis was done using ProMeta software, version 3.0.

Results

Search Results

We retrieved 394 articles, of which 33 duplicates were removed. Three hundred ten articles were excluded based on title and abstract. The remaining 51 articles were reviewed using a full-text article. Twenty articles were excluded; for instance, four were focused only on skilled birth attendant ^(19-22), and were not for the general population ^{(23, 24)} (Fig 1). Thirty-one studies were identified based on the eligibility criteria; 16 included in qualitative synthesis (systematic review) and 15 in quantitative synthesis (meta-analysis) (Fig 1). Data extracted from the 31 studies are shown in Table 2. Four of the 31 studies were a linked analysis of population and health facility surveys ^{(6, 7, 25, 26)} (Table 2). With the exception of three studies ^(10-12), all the other studies showed a significant association between physical access and delivery care use.

Study quality

We evaluated the quality of the included studies using the JBI’s critical appraisal checklist for cross-sectional and cohort studies ^{(16, 17)} and it turned out with an average score of 75.16%. Of the included studies, only three of them were graded as poor quality ^(27-29).

Heterogeneity test

Studies included in estimating the pooled effect of access to obstetric care facility beyond 60 minutes of travel showed that they were statistically heterogeneous. The Q-value was 23.23 with 8 degrees of freedom and P-value = 0.003. The I² statistic (a measure of the proportion of the variance in the observed effects that is due to the variance in the true effects), was 65.56, which demonstrates that about 66% of the variance in the observed effects was due to the variance in the true effects. Tau squared (τ²) is the variance of the true effect sizes, whereas Tau (τ) is the standard deviation of the true effects (both in log units). The estimated τ² and τwere 0.24 and 0.49, respectively. The prediction interval was from 0.06 to 0.77. Therefore, in most populations, we would expect that the odds ratio for delivery care use would fall from 0.06 (94% for those who had no access) to 0.77 (23% who relatively had access to obstetric care facilities).

In case of access to obstetric health facility within five kilometre of their usual place of residence, the included studies were not statistically heterogeneous. The estimated Q-value was 10.71 with 6 degrees of freedom and P-value = 0.098. The estimated I² was 43.99, which tells us that about 44% of the variance in the observed effects was due to the variance in the true effects. The estimated τ² and τwere 0.03 and 0.19, respectively. The prediction interval was from 1.34 to 3.85. Therefore, in most population, we would expect that the odds ratio for delivery care use would fall from 1.34 (34% for those who had no access) to 3.85 (285% who relatively had access to obstetric care facilities).

Fig 1: Systematic Review and Meta-Analysis flow diagram adapted from the 2009 PRISMA Statement ⁽¹⁸⁾

Impact of geographic access on delivery care use

Having no access to obstetric care facility within 60 minutes of walk

The pooled estimates (Odds Ratio) showed that impact of geographic access on institutional delivery care use was 0.21 (95% Confidence interval = 0.14, 0.31). This indicates that pregnant women who had no access to obstetric care facilities within 60 minutes of walk were 79% less likely to give birth at health institutions (Fig 2). The Trim and Fill analysis found that there is no need of additional studies to balance the symmetry of the funnel plot. However, the Funnel plot showed a slightly asymmetrical effect (Fig 3) but the Egger’s test found that there is no publication bias (P-value = 0.21).

Fig 2: Forest plot of effect sizes and 95% confidence intervals showing the effect of one hour and more travel time on delivery care use

Fig 3: Funnel plot on having no access to obstetric care facility within one hour travel time

Having access to obstetric care facility within 5 kilometre

The pooled estimate found that pregnant women who had access to obstetric care facilities within 5km were 2.27 times more likely to give birth at health institutions (95% Confidence interval = 2.27; 1.82, 2.82) (Fig 4). The Trim and Fill analysis found out that no more studies are required to balance the symmetry. Both the Funnel plot (Fig 5) and the Egger’s test showed that there is no publication bias in the included studies (P-value = 0.74).

Fig 4: Forest plot of effect sizes and 95% confidence intervals showing the effect of geographic access within 5km on delivery care use

Fig 5: Funnel plot on geographic access to obstetric care facility within 5km

Furthermore, the qualitative synthesis of most of the studies showed that geographic access had an effect on obstetric care use. Physical access to obstetric care facilities was assessed in two ways; that is, in terms of geographic distance and travel time to health care facilities. Looking at travel time to reach nearby obstetric care facilities, when the travelling time goes down to half an hour and less, pregnant women were more likely to have health facility delivery ⁽³⁰⁾. However, in another two studies, having access to delivery care facility within one-hour ⁽¹⁰⁾ and half an hour ⁽¹¹⁾ were not associated with institutional delivery. One the other hand, every one-hour increase in travel time to the nearby obstetric care facilities was associated with a 20% decrease in the odds of facility delivery ⁽²⁶⁾.

In addition to travel time, a one-kilometre increase in walking distance to obstetric care facilities was significantly associated with a significant decrease in health facility delivery ^{(9, 25, 31-33)}. In rural Zambia, every doubling of distance to the nearest obstetric care facility was significantly associated with a 29% decrease in health facility delivery ⁽⁶⁾. Moreover, the odds of health facility delivery was decreased by 65% in Malawi and 27% in Zambia for every 10 kilometre increase in distance to the closest obstetric care facility ⁽⁷⁾.

The odds of health facility delivery was higher among pregnant women who had physical access to obstetric care facilities within 10 ^{(25, 34)} and five ⁽²⁸⁾ kilometres.  However, having access to obstetric care facilities within three ⁽¹²⁾ and two kilometres ⁽¹³⁾ were not significantly associated with institutional delivery care use. In Burkina Faso, pregnant women who were residing seven or more kilometres away from obstetric care facilities were more likely to have home births ⁽³⁵⁾.

Discussion

The main findings of this meta-analysis and systematic review were that geographic access to obstetric care facilities, measured in either physical distance and/or travel time, had an impact on intuitional delivery use. Pregnant women who were living within a five-kilometre of an obstetric care facility had the higher odds of intuitional delivery as compared to those living beyond a physical distance of five-kilometre. In terms of walking time, the odds of using institutional delivery was low among pregnant women who had no access to obstetric care facilities within one-hour walking time. This implies that long distance has a dual influence on institutional delivery service utilization. It can be a barrier for both reaching obstetric care facilities and discouraging to seek care. The problem further worsens for rural pregnant women who often have no access to reliable transportation systems.

Furthermore, it was observed that both a single kilometre and a one-hour increase in accessing obstetric care facilities were associated with lower odds of intuitional delivery. This was consistent with the concept of distance decay ⁽³⁶⁾, where interactions varied inversely with distance. The greater the distance, the greater the barrier and the less the interaction would be. Therefore, the further away a pregnant woman lives from an obstetric care facility, the less likely she will use institutional delivery service.

Even though the study was the first of its kind, it had several limitations. It took only one of the three delays model – delay in reaching health care facilities ⁽³⁷⁾ and measures of healthcare access ^{(2, 3)}; even it did not address the different means of transportation, travel costs and terrains. There were also variations in the operationalization and measurement of geographic access to obstetric care facilities. It was due to the unavailability of a universally agreed clear cut-off point, in either a geographic distance and/or travel time, for a health facility to be accessible or not. The World Health Organization has put distance and travel time as a measure of physical accessibility ⁽³⁸⁾; however, there was no clear cut-off point for its measurement. Different countries are using different cut-off points; for instance, Ethiopia uses a 10 km ⁽³⁹⁾ distance whereas the United States of America and the United Kingdom use 30 minutes of travel time ⁽⁴⁰⁾ for measuring geographic access to health care services.

Even though the analysis was done using both distance and travel time, still it was subjected to measurement errors. The physical distance used in the studies was not uniformly measured; whilst some studies used a direct physical distance, the rest used walking tracks. Making comparisons and judgement based on measured physical distance is subjected to errors where the geography and transportation infrastructure largely vary within and among countries. Moreover, though the World Health Organization recommends using travel time, instead of physical distance, for assessing geographic accessibility ⁽³⁸⁾ unless stratifications are used, it could pose problems in making a comparison. The value of and actual travelling time varies depending on people; for instance, age and condition of the person, and the transportation mode used, and thus its strength as access barrier varies too.

In conclusion, it was found that the more closely the obstetric care facilities, the more likely that pregnant women to use institutional deliveries. Researchers should account the different measures of geographic accessibility, taking into consideration of means of transportation, travel cost and terrains, for measuring how obstetric care accessibility and utilization of intuitional deliveries interact with each other. Furthermore, future research is also needed to compare each of the measures of health care access how they could influence utilization of obstetric care services.

Table 1: Search strategy for impact of geographic access on delivery care use