Geographic access to care is not a determinant of child mortality in a rural Kenyan setting with high health facility density

BMC Public Health, Volume 10, Year 2010

Interpretation

Background. Policy-makers evaluating country progress towards the Millennium Development Goals also examine trends in health inequities. Distance to health facilities is a known determinant of health care utilization and may drive inequalities in health outcomes; we aimed to investigate its effects on childhood mortality. Methods. The Epidemiological and Demographic Surveillance System in Kilifi District, Kenya, collects data on vital events and migrations in a population of 220,000 people. We used Geographic Information Systems to estimate pedestrian and vehicular travel times to hospitals and vaccine clinics and developed proportional-hazards models to evaluate the effects of travel time on mortality hazard in children less than 5 years of age, accounting for sex, ethnic group, maternal education, migrant status, rainfall and calendar time. Results. In 2004-6, under-5 and under-1 mortality ratios were 65 and 46 per 1,000 live-births, respectively. Median pedestrian and vehicular travel times to hospital were 193 min (inter-quartile range: 125-267) and 49 min (32-72); analogous values for vaccine clinics were 47 (25-73) and 26 min (13-40). Infant and under-5 mortality varied two-fold across geographic locations, ranging from 34.5 to 61.9 per 1000 child-years and 8.8 to 18.1 per 1000, respectively. However, distance to health facilities was not associated with mortality. Hazard Ratios (HR) were 0.99 (95% CI 0.95-1.04) per hour and 1.01 (95% CI 0.95-1.08) per half-hour of pedestrian and vehicular travel to hospital, respectively, and 1.00 (95% CI 0.99-1.04) and 0.97 (95% CI 0.92-1.05) per quarter-hour of pedestrian and vehicular travel to vaccine clinics in children <5 years of age. Conclusions. Significant spatial variations in mortality were observed across the area, but were not correlated with distance to health facilities. We conclude that given the present density of health facilities in Kenya, geographic access to curative services does not influence population-level mortality. © 2010 Mosi et al; licensee BioMed Central Ltd. In this paper, we present an analysis of the data routinely collected by the Epidemiological and Demographic Surveillance System (Epi-DSS) in Kilifi District, Kenya, a member of the INDEPTH network of demographic surveillance sites. Kilifi District is a largely rural area situated on the Indian Ocean coast of Kenya, with a tropical climate characterized by two dry seasons and two rainy seasons each year. Kilifi District Hospital in Kilifi serves as a primary care center and first-level referral facility for the district. The KEMRI-Wellcome Trust Research Programme has performed hospital and field-based epidemiological research in Kilifi for two decades. The Epi-DSS covers approximately 900 km2 around Kilifi District Hospital and was created in 2000 to track a population of over 220,000 people. After completion of the initial census in 2001, all homesteads in the Epi-DSS area were visited two to three times each year to collect information on births, deaths, in-migrations and out-migrations of household members. Beginning March 2003, pregnancy data was recorded to permit assessment of pregnancy outcomes and improved ascertainment of neonatal and early infant deaths. The census area comprises 15 administrative locations, further divided into 40 sublocations. It has been mapped using Magellan (Magellan Navigation Inc, Santa Clara, CA) and e-Trex (Garmin Ltd, Olathe, KS) Geographic Positioning Systems technology, providing detailed information on topography, footpaths and roads, and human occupation of the area, including the coordinates of all homesteads. In January 2007, we collected data on the seven matatu (local bus) routes, including speed, frequency and cost of travel. One of these routes followed the only paved road in the Epi-DSS area, the Mombasa-Malindi coastal road. All geographic data were imported via Datasend, Map Source, or DNRGarmin software into ArcGIS 9.2 (ESRI, Redlands, CA) for mapping and analysis (Figure (Figure11). Kilifi area health facilities and transport networks. A survey of health facilities in Kilifi District was conducted in September 2006. Four hospitals, 47 vaccine clinics (clinics offering childhood immunization among other preventive or curative services), and 100 other public, private or NGO-run health facilities were identified and mapped (Figure (Figure1).1). Residents of the Epi-DSS may also access inpatient care outside the district, at Malindi District Hospital, which was therefore incorporated into our analysis. Travel time to hospitals and vaccine clinics was calculated using an ArcGIS cost-distance algorithm. We divided the study area into 100-m × 100-m cells and created an impedance raster (grid) defining the speed of travel through each cell. The algorithm computes travel time from each health facility to all neighboring cells, then from these to all of their neighboring cells, proceeding iteratively until the entire area is covered. Thus, it delineates a catchment area for each health facility and obtains travel time to this facility from all cells in its catchment area. For pedestrian travel time, we assumed speeds of 5 km/hr on roads and footpaths and 2.5 km/hr off-road. In the vehicular model, matatu speeds were used on matatu routes and pedestrian speeds elsewhere. Kilifi Creek constitutes a natural barrier to travel and was attributed high impedance (1.25 km/hr speed). Changes in elevation in the Epi-DSS area are small and were not incorporated into the impedance raster. For each individual, we observed a series of dated, spatially-defined demographic events which were used to construct consecutive, non-overlapping observation periods. Each observation period was linked to residence in a homestead with known geographic coordinates. This data structure enabled us to perform survival analysis on a dynamic cohort of children entering and exiting risk sets over time. We constructed Kaplan-Meier survival curves and instantaneous hazard curves by administrative location and by stratum of travel time to hospitals and vaccine clinics, as well as by sex, ethnic group (Giriama, Chonyi, Kauma, Duruma, Luo, Jibana, and "other" which combines all groups with <40 deaths), maternal education (proportion of women 15-49 years old with any education in a given sublocation: <0.5, 0.5-<0.6, 0.6-<0.7, ≥ 0.7), migrant status (migration from outside the area between 2000 and 2006), and rainfall (total rainfall in past seven days <40 mm vs. ≥ 40 mm). We built univariate and multivariable proportional hazards models to examine the effects of travel time on mortality hazard. We included an age adjustment (indicator variables for 2-month age strata from 0 to 11 months and 6-month strata from 12 to 59 months) to control for the changing age distribution of the population over time and a calendar time adjustment (six-month time strata) to control for temporal trends. To account for spatial clustering of deaths in our models, we used a spatial bootstrap method with 50 repetitions. In each repetition, we randomly selected 40 sublocations (with replacement) and estimated the proportional hazards model on all data from the selected sublocations. Standard errors for regression coefficients were obtained as the standard deviation of coefficients across repetitions. Variables without statistically significant effects (at the 0.05 level) based on Wald tests were dropped from the multivariable models. All data were double-entered into File Maker Pro 5.5 and cleaned using Stata 9.2 (StataCorp, College Station, TX). Analyses were conducted in Stata 9.2. To ensure comparability with other demographic and epidemiological studies, the analysis was conducted for under-5 year olds, under-1 year olds and 1 to 4 year olds separately. We excluded data from the period during which death ascertainment was incomplete, restricting the analysis to 2004-6 for infants and to 2003-6 for children 1 to 4 years of age. This study was approved by the Scientific Coordinating Committee and Ethical Review Committee of the Kenya Medical Research Institute and by the Institutional Review Board of the Johns Hopkins Bloomberg School of Public Health.

AI Digest

Study Justification:

This study aimed to investigate the effects of distance to health facilities on childhood mortality in a rural Kenyan setting. The study was conducted to provide evidence on whether geographic access to care influences child mortality rates. This information is important for policy-makers evaluating progress towards the Millennium Development Goals and addressing health inequities.

Highlights:

– The study analyzed data collected by the Epidemiological and Demographic Surveillance System in Kilifi District, Kenya.
– The study used Geographic Information Systems to estimate travel times to hospitals and vaccine clinics.
– Under-5 and under-1 mortality ratios were 65 and 46 per 1,000 live births, respectively.
– Significant spatial variations in mortality were observed across the area.
– However, distance to health facilities was not associated with mortality.
– The study concluded that given the present density of health facilities in Kenya, geographic access to curative services does not influence population-level mortality.

Recommendations:

Based on the findings of the study, the following recommendations can be made:

1. Policy-makers should focus on other factors that contribute to child mortality, such as improving the quality of healthcare services and addressing social determinants of health.
2. Efforts should be made to ensure equitable access to healthcare services, regardless of geographic location.
3. Further research is needed to explore other potential determinants of child mortality in rural settings.

Key Role Players:

To address the recommendations, the following key role players may be needed:

1. Policy-makers and government officials responsible for healthcare planning and resource allocation.
2. Healthcare providers and professionals involved in delivering healthcare services.
3. Community leaders and organizations working on health promotion and education.
4. Researchers and academics studying child health and mortality.

Cost Items:

While the actual cost of implementing the recommendations cannot be estimated without detailed planning, the following cost items may need to be considered:

1. Infrastructure development and maintenance, including building and upgrading health facilities.
2. Human resources, including hiring and training healthcare professionals.
3. Medical equipment and supplies.
4. Health education and awareness campaigns.
5. Monitoring and evaluation systems to track progress and outcomes.
Please note that these cost items are general considerations and may vary depending on the specific context and needs of the healthcare system.

Strength of Evidence

The strength of evidence for this abstract is 8 out of 10.
The evidence in the abstract is strong, but there are some areas for improvement. The study design is robust, using Geographic Information Systems and proportional-hazards models to evaluate the effects of travel time on mortality hazard. The data collection methods are well-described, and the sample size is large. However, the abstract could benefit from providing more specific information about the results, such as the magnitude of the hazard ratios and their statistical significance. Additionally, it would be helpful to include information about potential limitations of the study, such as any confounding factors that were not accounted for. To improve the abstract, the authors could consider providing more detailed information about the results and discussing the limitations of the study.

Abstract

In this paper, we present an analysis of the data routinely collected by the Epidemiological and Demographic Surveillance System (Epi-DSS) in Kilifi District, Kenya, a member of the INDEPTH network of demographic surveillance sites. Kilifi District is a largely rural area situated on the Indian Ocean coast of Kenya, with a tropical climate characterized by two dry seasons and two rainy seasons each year. Kilifi District Hospital in Kilifi serves as a primary care center and first-level referral facility for the district. The KEMRI-Wellcome Trust Research Programme has performed hospital and field-based epidemiological research in Kilifi for two decades. The Epi-DSS covers approximately 900 km2 around Kilifi District Hospital and was created in 2000 to track a population of over 220,000 people. After completion of the initial census in 2001, all homesteads in the Epi-DSS area were visited two to three times each year to collect information on births, deaths, in-migrations and out-migrations of household members. Beginning March 2003, pregnancy data was recorded to permit assessment of pregnancy outcomes and improved ascertainment of neonatal and early infant deaths. The census area comprises 15 administrative locations, further divided into 40 sublocations. It has been mapped using Magellan (Magellan Navigation Inc, Santa Clara, CA) and e-Trex (Garmin Ltd, Olathe, KS) Geographic Positioning Systems technology, providing detailed information on topography, footpaths and roads, and human occupation of the area, including the coordinates of all homesteads. In January 2007, we collected data on the seven matatu (local bus) routes, including speed, frequency and cost of travel. One of these routes followed the only paved road in the Epi-DSS area, the Mombasa-Malindi coastal road. All geographic data were imported via Datasend, Map Source, or DNRGarmin software into ArcGIS 9.2 (ESRI, Redlands, CA) for mapping and analysis (Figure (Figure11). Kilifi area health facilities and transport networks. A survey of health facilities in Kilifi District was conducted in September 2006. Four hospitals, 47 vaccine clinics (clinics offering childhood immunization among other preventive or curative services), and 100 other public, private or NGO-run health facilities were identified and mapped (Figure (Figure1).1). Residents of the Epi-DSS may also access inpatient care outside the district, at Malindi District Hospital, which was therefore incorporated into our analysis. Travel time to hospitals and vaccine clinics was calculated using an ArcGIS cost-distance algorithm. We divided the study area into 100-m × 100-m cells and created an impedance raster (grid) defining the speed of travel through each cell. The algorithm computes travel time from each health facility to all neighboring cells, then from these to all of their neighboring cells, proceeding iteratively until the entire area is covered. Thus, it delineates a catchment area for each health facility and obtains travel time to this facility from all cells in its catchment area. For pedestrian travel time, we assumed speeds of 5 km/hr on roads and footpaths and 2.5 km/hr off-road. In the vehicular model, matatu speeds were used on matatu routes and pedestrian speeds elsewhere. Kilifi Creek constitutes a natural barrier to travel and was attributed high impedance (1.25 km/hr speed). Changes in elevation in the Epi-DSS area are small and were not incorporated into the impedance raster. For each individual, we observed a series of dated, spatially-defined demographic events which were used to construct consecutive, non-overlapping observation periods. Each observation period was linked to residence in a homestead with known geographic coordinates. This data structure enabled us to perform survival analysis on a dynamic cohort of children entering and exiting risk sets over time. We constructed Kaplan-Meier survival curves and instantaneous hazard curves by administrative location and by stratum of travel time to hospitals and vaccine clinics, as well as by sex, ethnic group (Giriama, Chonyi, Kauma, Duruma, Luo, Jibana, and "other" which combines all groups with <40 deaths), maternal education (proportion of women 15-49 years old with any education in a given sublocation: <0.5, 0.5-<0.6, 0.6-<0.7, ≥ 0.7), migrant status (migration from outside the area between 2000 and 2006), and rainfall (total rainfall in past seven days <40 mm vs. ≥ 40 mm). We built univariate and multivariable proportional hazards models to examine the effects of travel time on mortality hazard. We included an age adjustment (indicator variables for 2-month age strata from 0 to 11 months and 6-month strata from 12 to 59 months) to control for the changing age distribution of the population over time and a calendar time adjustment (six-month time strata) to control for temporal trends. To account for spatial clustering of deaths in our models, we used a spatial bootstrap method with 50 repetitions. In each repetition, we randomly selected 40 sublocations (with replacement) and estimated the proportional hazards model on all data from the selected sublocations. Standard errors for regression coefficients were obtained as the standard deviation of coefficients across repetitions. Variables without statistically significant effects (at the 0.05 level) based on Wald tests were dropped from the multivariable models. All data were double-entered into File Maker Pro 5.5 and cleaned using Stata 9.2 (StataCorp, College Station, TX). Analyses were conducted in Stata 9.2. To ensure comparability with other demographic and epidemiological studies, the analysis was conducted for under-5 year olds, under-1 year olds and 1 to 4 year olds separately. We excluded data from the period during which death ascertainment was incomplete, restricting the analysis to 2004-6 for infants and to 2003-6 for children 1 to 4 years of age. This study was approved by the Scientific Coordinating Committee and Ethical Review Committee of the Kenya Medical Research Institute and by the Institutional Review Board of the Johns Hopkins Bloomberg School of Public Health.

Materials

N/A

Innovations

Based on the information provided, it appears that the study did not find a correlation between distance to health facilities and child mortality in a rural Kenyan setting. Therefore, the study does not provide specific innovations or recommendations for improving access to maternal health. However, it does highlight the importance of considering other factors, such as socioeconomic status, education, and healthcare infrastructure, when addressing maternal health issues.

AI Innovations Description

The recommendation that can be developed into an innovation to improve access to maternal health based on the information provided is to focus on addressing other determinants of health care utilization and health outcomes, rather than solely relying on geographic access to health facilities.

The study mentioned in the description found that distance to health facilities was not associated with mortality in children under 5 years of age in a rural Kenyan setting with high health facility density. This suggests that there may be other factors influencing health outcomes, such as socioeconomic status, education, cultural beliefs, and availability of healthcare services.

To improve access to maternal health, it is important to consider a holistic approach that takes into account these various determinants. This could involve implementing interventions such as:

1. Improving health education and awareness: Providing information and education to women and communities about the importance of maternal health, including prenatal care, safe delivery practices, and postnatal care. This can help address cultural beliefs and misconceptions that may hinder access to maternal health services.

2. Strengthening healthcare systems: Investing in healthcare infrastructure, equipment, and supplies to ensure that health facilities are adequately equipped to provide quality maternal health services. This includes training healthcare providers and ensuring the availability of essential medicines and equipment.

3. Addressing socioeconomic barriers: Implementing policies and programs that address socioeconomic barriers to accessing maternal health services, such as poverty, lack of transportation, and cost of healthcare. This could involve providing financial support, transportation services, and reducing or eliminating user fees for maternal health services.

4. Promoting community engagement: Engaging with local communities and community leaders to raise awareness about maternal health and involve them in the planning and implementation of maternal health programs. This can help ensure that interventions are culturally appropriate and accepted by the community.

5. Utilizing technology and innovation: Exploring the use of technology, such as telemedicine and mobile health applications, to improve access to maternal health services in remote or underserved areas. This can help overcome geographical barriers and provide access to healthcare information and services.

By addressing these various determinants of health care utilization and health outcomes, it is possible to develop innovative solutions that can improve access to maternal health and ultimately reduce maternal and child mortality rates.

AI Innovations Methodology

Based on the information provided, here are some potential recommendations for improving access to maternal health:

1. Mobile clinics: Implementing mobile clinics that can travel to remote areas and provide essential maternal health services, including prenatal care, vaccinations, and postnatal care.

2. Telemedicine: Utilizing telemedicine technology to connect pregnant women in remote areas with healthcare professionals who can provide virtual consultations, guidance, and support.

3. Community health workers: Expanding the role of community health workers to provide maternal health education, screenings, and referrals in underserved areas.

4. Transportation support: Establishing transportation programs or subsidies to help pregnant women in remote areas access healthcare facilities for prenatal visits, delivery, and emergency care.

5. Maternity waiting homes: Building maternity waiting homes near healthcare facilities to accommodate pregnant women who live far away, allowing them to stay closer to the facility as their due date approaches.

To simulate the impact of these recommendations on improving access to maternal health, a methodology could include the following steps:

1. Define the target population: Identify the specific population that would benefit from the recommendations, such as pregnant women in rural areas of a particular region or country.

2. Collect baseline data: Gather data on the current access to maternal health services, including the distance to healthcare facilities, utilization rates, and health outcomes.

3. Model the impact: Use mathematical modeling techniques to simulate the potential impact of each recommendation on improving access to maternal health. This could involve estimating the number of additional women who would have access to care, the reduction in travel time, and the potential improvement in health outcomes.

4. Consider contextual factors: Take into account contextual factors that may influence the effectiveness of the recommendations, such as cultural beliefs, infrastructure limitations, and availability of healthcare resources.

5. Sensitivity analysis: Conduct sensitivity analysis to assess the robustness of the results and explore different scenarios or assumptions.

6. Evaluate cost-effectiveness: Assess the cost-effectiveness of implementing the recommendations by comparing the estimated benefits with the associated costs, including infrastructure development, training, and ongoing operational expenses.

7. Communicate findings: Present the findings of the simulation study to policymakers, healthcare providers, and other stakeholders to inform decision-making and prioritize interventions that have the greatest potential impact on improving access to maternal health.

Authors & Co-Authors

Statistics:

Citations: 37

Authors: 8

Identifiers:

DOI: 10.1186/1471-2458-10-142

Research Areas:

Disparities, Environmental, Health System and Policy, Maternal Access, Maternal and Child Health, Quality of Care, Sexual and Reproductive Health, Social Determinants, Technology and Innovations

Study Countries:

Kenya

Study Design:

Cohort Study, Cross Sectional Study

Study Approach:

Quantitative

Participants Gender:

Female