Evaluation of methods for linking household and health care provider data to estimate effective coverage of management of child illness: Results of a pilot study in Southern Province, Zambia

Journal of Global Health, Volume 8, No. 1, Year 2018

Interpretation

Background Existing population-based surveys have limited accuracy for estimating the coverage and quality of management of child illness. Linking household survey data with health care provider assessments has been proposed as a means of generating more informative population- level estimates of effective coverage, but methodological issues need to be addressed. Methods A 2016 survey estimated effective coverage of management of child illness in Southern Province, Zambia, using multiple methods for linking temporally and geographically proximate household and health care provider data. Mothers of children < 5 years were surveyed about seeking care for child illness. Information on health care providers' capacity to manage child illness, or structural quality, was assessed using case scenarios and a tool modeled on the WHO Service Availability and Readiness Assessment (SARA). Each sick child was assigned the structural quality score of their stated (exact-match) source of care. Effective coverage was calculated as the average structural quality experienced by all sick children. Children were also ecologically linked to providers using measures of geographic proximity, with and without data on non-facility providers, to assess the effects of these linking methods on effective coverage estimates. Results Data were collected on 83 providers and 385 children with fever, diarrhea, and/or symptoms of ARI in the preceding 2 weeks. Most children sought care from government facilities or community-based agents (CBAs). Effective coverage of management of child illness estimated through exact-match linking was approximately 15-points lower in each stratum than coverage of seeking skilled care due to providers' limited structural quality. Estimates generated using most measures of geographic proximity were similar to the exact-match estimate, with the exception of the kernel density estimation method in the urban area. Estimates of coverage in rural areas were greatly reduced across all methods using facility-only data if seeking care from CBAs was treated as unskilled care. Conclusions Linking household and provider data may generate more informative estimates of effective coverage of management of child illness. Ecological linking with provider data on a sample of all skilled providers may be as effective as exact-match linking in areas with low variation in structural quality within a provider category or minimal provider bypassing. The study was nested in a validation study of maternal reports of care-seeking for childhood illness [7] and was conducted in two urban and three rural health facility catchment areas in Choma District in Southern Province, Zambia, between January 18 and March 20, 2016. Choma district is primarily agrarian, although Choma town is a growing commercial hub and provincial capital [8]. Under five mortality rates in Zambia have declined dramatically over the past two decades; however, pneumonia, diarrhea, and malaria remain the leading causes of child mortality in the post-neonatal period [9]. Mothers report approximately 70 percent of children in Southern Province with fever, diarrhea, or ARI symptoms are taken for care, primarily in the public sector [9]. The Zambian government manages 90% of health facilities either directly or through service agreements with the Churches Health Association of Zambia (CHAZ). However, the private sector is growing in urban centers [10]. Health services are free for children <5 years at all government facilities, including hospitals with referral [11]. The Integrated Management of Child Illness (IMCI) approach has been implemented in all districts since the 1990s; however by the late 2000s only about 65% of health facilities had been staffed by an IMCI-trained clinician [12]. Community based health agents (CBAs) participate in task shifting at government facilities and implement a variable package of community-based interventions, including diagnosis and treatment of malaria and treatment of diarrhea with oral rehydration solution (ORS) [13]. The study area has been the site of ongoing malaria testing and treatment and mass drug administration trials [14,15]. The study included two components; 1) a household survey on care-seeking for child illness, and 2) an assessment of health care providers’ structural quality for managing child illness. Ethical approval for the study was obtained from the Institutional Review Boards of Johns Hopkins Bloomberg School of Public Health and Excellence in Research Ethics and Science (ERES) Converge in Zambia. The Zambian Ministry of Health granted permission and the Choma District Health Office provided support to survey government health facilities in the study area. In each of the urban and rural strata, 700 households were randomly sampled from the health facility catchment areas (HFCAs) of five government health facilities in and around Choma town. Households were randomly sampled from the catchment population of three rural health centers using an existing household listing created in 2014 [16]. Urban households were sampled from a census of households conducted immediately prior to the household enrollment phase. Households with a woman of reproductive age (15-49 years) with at least one biological child <59 m were eligible to participate in the study. These criteria correlate with the DHS requirements for the child questions in the Women’s Questionnaire and ensured that participating children were less than 5 years of age at the time of the household survey. A sample of 700 households per stratum was expected to yield information on 155 episodes of child illness per stratum, allowing estimation of effective coverage of management of child illness with a precision of ±6.0%, based on a type-1 error probability of 5% (two-tailed test) and an underlying standard deviation in care of 0.35. Households were enrolled in the study from January 18 to February 13, 2016, and subsequently revisited approximately four to six weeks later for the household care-seeking survey completed between March 3-20, 2016. Mothers were asked about child illness and care-seeking using a questionnaire based on the 2013-2014 Zambia DHS (ZDHS). These included questions about the presence of diarrhea, fever, or suspected ARI in each of their children <5 years in the preceding two weeks. If a child experienced an illness, mothers were asked if care was sought, the source of care, and treatment received. In addition to the series of DHS care-seeking questions, mothers answered questions to ascertain the name of the specific source of care and sequence of care-seeking events. If the name of the source of care was unknown, data collectors were instructed to probe about provider location and other identifying features. Concurrent to household enrollment, health care providers were identified and invited to participate in the provider assessment. The term health care provider will be used to refer to both individual providers such as CBAs and traditional practitioners, and health care outlets that include multiple staff such as health facilities and pharmacies. Public, private, informal, and traditional sources of care were included in the assessment. Community leaders and health workers initially provided a list of commonly utilized care providers offering medicine or alternative treatment for sick children. The list was further expanded with information from participating mothers about common sources of care for treating illness in their children <5 years collected during enrollment. All providers included in the assessment were grouped into categories of providers used in the ZDHS (Box 1) and this classification was employed in all ecological linking analyses restricted by provider category. Public • Government hospital • Government health center/post • Government CBA / fieldworker Private • Private hospital/clinic • Pharmacy Informal • Shop/market • Traditional/faith-based practitioner The provider assessment was completed among all identified health care providers (Figure 1). The provider assessment was designed to assess a provider or facility’s capacity to provide curative services for children <5 years, including presence of drugs and commodities, training, supervision, and provider case management knowledge. The assessment was designed to assess a provider’s structural quality following the Donabedian structure-process-outcome model [17] and to align with the WHO definition of provider readiness [18] as upstream measures of health care provider quality. At facilities and pharmacies with multiple staff, the questionnaire was administered to the most senior staff member and reports of the existence and functionality of physical commodities (medicines, equipment, etc) were verified by observation. Questions were modeled off the SARA general and child health questionnaire [18] and adapted for use with facility-based, community-based, public, private, and informal providers. Clinical case scenarios developed for use in the evaluation of the IMCI program were used to assess provider case management knowledge [19]. Providers were read four clinical case scenarios and asked how they would manage each hypothetical sick child. At outlets with multiple clinical staff, up to three staff members within each cadre of clinical health workers were randomly selected among those available at the time of the assessment to respond to case scenarios. Map of health care provider locations. The provider assessment was used to generate a “structural quality score” corresponding to a provider’s structure or capacity to appropriately manage a child illness. The structural quality score measured availability of services, commodities, and human resources needed to appropriately manage common child illnesses (Box 2). These indicators were considered the minimum inputs for appropriate care: the basic commodities required to diagnosis and treat common child illness, along with the human resources and clinical knowledge to apply them correctly. As such, the score reflects an upper threshold of the potential quality of care offered by a provider. A provider received one point for each indicator if requirements were met and zero if not; each domain received equal weight. The knowledge domain was calculated as an average score of provider performance on four case scenarios. An average facility knowledge score was generated when knowledge was assessed for multiple health workers at a single facility. Providers were assessed against the expected capacity for their specific provider category; for example, CBAs were not penalized for not having antibiotics in this setting where CBAs are not allowed to treat ARI. Diagnostics • Malaria Diagnostic (RDTs or microscopy) • Malnutrition Diagnostic (MUAC or Scale + Height board + Growth chart) • ARI Diagnostic (stethoscope or respiratory timer) • General microscopy (functioning microscope and slides) Basic medicines • Oral rehydration solution • Zinc • Artemisinin combination therapy (ACT) • Oral antibiotic Severe/complicated illness medicines • IV fluids • Injectable quinine or artesunate • Injectable antibiotics Human Resources • Training (at least one staff member with IMCI or relevant training) • Guidelines (IMCI guidelines or relevant guidelines or job aid available) • Supervision (received supervision visit with case management observation in past 3 months) Available services • Diagnosis and treat malaria (by pathology) • Diagnosis and treat diarrhea (by pathology) • Diagnosis and treat ARI (by pathology) • Diagnosis and treat malnutrition (by pathology) • Facilitated referral capacity Knowledge • Average performance on case scenarios The primary outcome was input-based effective coverage of management of child illness estimated through exact-match linking and ecological linking methods. Effective coverage of management of child illness was calculated as the average level of structural quality experienced by sick children based on their reported care-seeking behavior and linked source of care. Each child was assigned the structural quality score of either their specific reported source(s) of care (exact-match linking) or the closest provider(s) based on measures of geographic proximity (ecological linking). The linking was performed using provider assessment data on all health care providers, to reflect capacity among all categories of providers, and data on only health facilities, to replicate the provider data available through common provider assessments such as the SPA or SARA. We considered estimates of effective coverage generated through the exact-match linking using data on all health care providers to be the most accurate linked coverage estimate. However, we did not assess the validity of the effective coverage estimates generated through the method against a true measure of how sick children in the study area were managed.The input-based effective coverage estimate reflects an upper limit on the proportion of children that could have been correctly managed. Estimates generated using the ecological linking methods and using data on only facility-based providers were compared against the exact-match all-provider coverage estimates to assess their population-level validity, or how closely they reproduced the exact-match all-provider estimates of effective coverage. For exact-match linking, each sick child was linked to the specific source(s) of care from which care was sought, based on the name of the facility, outlet, or provider reported by the mother during the household survey. For ecological linking, each sick child was linked to the closest provider(s) based on various measures of geographic proximity. Seven methods for ecological linking were employed, depicted in Figure 2. Measures of geographic proximity employed in the ecological linking were adapted from the work of Skiles and colleagues [20]. Geographic proximity was calculated using ArcGIS 10.1 (Esri, Redlands, CA, USA). Specifications and steps for generating geographic links are presented in Appendix S1 in Online Supplementary Document(Online Supplementary Document). Linked data sets were exported to Stata 14.2 (StataCorp LLC, College Station, TX, USA) for analysis. Illustration of ecological linking methods (household locations have been displaced in figure to protect confidentiality). The seven measures of geographic proximity used in the ecological linking can be grouped into three categories: methods linking children 1) to the single nearest provider by distance, 2) to all providers within a defined geographic unit, and 3) using kernel density estimation. Single nearest provider link: Kernal Density Estimation (KDE): The KDE method was designed to model the level of draw a provider exerts over households as they decide to seek care, based on distance decay and characteristics of the provider. KDE has been used as a means of modeling health care access [21] and service environment [20]. KDE can be used to model health care utilization in the absence of household data assuming all individuals would seek care if skilled providers are accessible, and the choice of provider is driven by provider quality and distance. KDE employs a user-specified kernel size and probability density distribution. The kernel size, or maximum radius of a provider catchment area, was selected to reflect a household preference for higher-level providers. Higher-level providers (hospitals and health facilities) had a larger catchment area than lower level providers (pharmacies and community-based providers). Within a catchment area, a provider’s draw decreased with increasing distance from a household. A provider’s structural quality score was used as the density variable, effectively modeling higher draw within their catchment area for providers with higher scores. Each child was linked exclusively or partially to a category of provider based on the level of draw exhibited by providers in the category. Information on source of care from the household survey was excluded because the method models care-seeking behavior. For the exact-match linking, each sick child was assigned the structural quality score for the specific source(s) from which care was sought. If care was reportedly sought from more than one source, the child was assigned the average score for all providers from which care was sought. If no care was sought for the illness, the child was assigned a structural quality score of zero. If the mother reported a source of care that could not be identified or included in the provider assessment, the child was assigned the average structural quality score for the source of care provider category within the stratum (urban/rural). Effective coverage was calculated as the average structural quality score (a percentage ranging from 0 to 100) across all sick children, including those who were not taken for care (who received a score of 0). For the ecological linking, each sick child was assigned the structural quality score for the source(s) of care that were closest based on various measures of geographic proximity, as described above. All non-KDE ecological linking methods maintained the reported category of source of care. In other words, a child could not be linked to a provider from a category of source of care other than the category reported by the mother (eg, a child that was reported to have been taken to a government CBA could only be linked to government CBAs, although the specific CBA(s) to which he/she was linked might vary depending on the measure of geographic proximity). Similar to the exact-match linking, children that were not taken for care were assigned a structural quality score of zero. Those that could not be linked to a provider from the reported source of care were assigned the average structural quality score for the category of source of care within the stratum. For example, when applying the 5 km radius linking approach, if a rural mother reported that her child was taken to a government health center for care but there was no government health center within 5 km of the household, the child was assigned the average of all government health centers in the rural area. If a child was linked to multiple providers, the average structural quality score for all linked providers was calculated for the child. Effective coverage was calculated as the average structural quality score (a percentage ranging from 0 to 100) across all sick children, including those who were not taken for care (who received a score of 0). To simulate the type of provider data that would typically be available when using a SPA or SARA for linking, we repeated the analysis using only facility data. Coverage was estimated using the exact-match linking method and each of the seven ecological linking methods with only facility structural quality scores. Health facilities were defined as either a government or private clinic or hospital, in line with those providers included in the SARA and SPA surveys. Presuming first-level government facilities offered the most comparable level of structural quality to government CBAs, children that reported care from a CBA were linked to one or more government health centers using the nearest provider and aggregate ecological linking methods, and assigned the average government health center structural quality score for the exact-match linking. Using the exact-match, nearest provider, and aggregate ecological linking methods, children who recieved treatment from all other sources of care (pharmacies, shops, and traditional practitioners) were treated as unskilled sources of care and assigned a structural quality score of zero, equivalent to seeking no care. All other components of the linking methodology and household data remained the same. Using the KDE methods, data on non-facility providers were excluded while modeling care-seeking behavior. A summary of all linking methods employed in the paper is presented in Table 1. Linking method summary HFCA – health facility catchment area *Children reporting seeking care from CBAs were linked 1) to government health centers (primary analysis – Table 3 or Table 2) treated as no care (Table S9b in Online Supplementary Document(Online Supplementary Document)) during facility-only analyses. All other non-facility providers treated as no care and not linked. Descriptive statistics comparing the ecological and facility-only links to the exact-match all provider links were calculated. Sensitivity analyses were also conducted to estimate effective coverage using different assumptions for children that could not be linked to a provider within the reported source of care category including: 1) assigning children that could not be linked to a provider based on geographic proximity a structural quality score of zero, and 2) assigning children that sought care from a CBA a score of zero during the facility-only analysis. These sensitivity analyses were designed to mimic the effect of service environment assessments and linking analyses that define health care access based on a capped maximum distance from a household and that ignore the contribution of CBAs in management of child illness, respectively.

AI Digest

Study Justification:

– Existing population-based surveys have limited accuracy for estimating the coverage and quality of management of child illness.
– Linking household survey data with health care provider assessments can provide more informative population-level estimates of effective coverage.
– Methodological issues need to be addressed in order to accurately link household and health care provider data.

Study Highlights:

– The study estimated effective coverage of management of child illness in Southern Province, Zambia.
– Multiple methods for linking household and health care provider data were used.
– The study assessed the effects of different linking methods on effective coverage estimates.
– Data were collected on 83 providers and 385 children with fever, diarrhea, and/or symptoms of ARI.

Study Recommendations:

– Linking household and provider data can generate more informative estimates of effective coverage of management of child illness.
– Ecological linking with provider data on a sample of all skilled providers may be as effective as exact-match linking in certain areas.
– Consideration should be given to the effects of geographic proximity and non-facility providers on effective coverage estimates.

Key Role Players:

– Researchers and data collectors
– Mothers of children under 5 years
– Health care providers (government facilities, community-based agents, private sector, etc.)
– Community leaders and health workers
– Institutional Review Boards
– Ministry of Health
– Choma District Health Office

Cost Items for Planning Recommendations:

– Research and data collection expenses
– Training and supervision of data collectors
– Ethical approval process
– Administrative support from the Ministry of Health and Choma District Health Office
– Travel and transportation costs for data collection
– Data analysis and interpretation
– Reporting and dissemination of study findings

Strength of Evidence

The strength of evidence for this abstract is 7 out of 10.
The evidence in the abstract is strong, but there are some areas for improvement. The study design and methods are clearly described, and the data collection process is well-explained. The study includes a large sample size and covers both urban and rural areas. The linking methods used to estimate effective coverage are appropriate and are compared to exact-match linking. However, the abstract could be improved by providing more information on the statistical analysis conducted and the results obtained. Additionally, it would be helpful to include information on the limitations of the study and suggestions for future research.

Abstract

The study was nested in a validation study of maternal reports of care-seeking for childhood illness [7] and was conducted in two urban and three rural health facility catchment areas in Choma District in Southern Province, Zambia, between January 18 and March 20, 2016. Choma district is primarily agrarian, although Choma town is a growing commercial hub and provincial capital [8]. Under five mortality rates in Zambia have declined dramatically over the past two decades; however, pneumonia, diarrhea, and malaria remain the leading causes of child mortality in the post-neonatal period [9]. Mothers report approximately 70 percent of children in Southern Province with fever, diarrhea, or ARI symptoms are taken for care, primarily in the public sector [9]. The Zambian government manages 90% of health facilities either directly or through service agreements with the Churches Health Association of Zambia (CHAZ). However, the private sector is growing in urban centers [10]. Health services are free for children <5 years at all government facilities, including hospitals with referral [11]. The Integrated Management of Child Illness (IMCI) approach has been implemented in all districts since the 1990s; however by the late 2000s only about 65% of health facilities had been staffed by an IMCI-trained clinician [12]. Community based health agents (CBAs) participate in task shifting at government facilities and implement a variable package of community-based interventions, including diagnosis and treatment of malaria and treatment of diarrhea with oral rehydration solution (ORS) [13]. The study area has been the site of ongoing malaria testing and treatment and mass drug administration trials [14,15]. The study included two components; 1) a household survey on care-seeking for child illness, and 2) an assessment of health care providers’ structural quality for managing child illness. Ethical approval for the study was obtained from the Institutional Review Boards of Johns Hopkins Bloomberg School of Public Health and Excellence in Research Ethics and Science (ERES) Converge in Zambia. The Zambian Ministry of Health granted permission and the Choma District Health Office provided support to survey government health facilities in the study area. In each of the urban and rural strata, 700 households were randomly sampled from the health facility catchment areas (HFCAs) of five government health facilities in and around Choma town. Households were randomly sampled from the catchment population of three rural health centers using an existing household listing created in 2014 [16]. Urban households were sampled from a census of households conducted immediately prior to the household enrollment phase. Households with a woman of reproductive age (15-49 years) with at least one biological child <59 m were eligible to participate in the study. These criteria correlate with the DHS requirements for the child questions in the Women’s Questionnaire and ensured that participating children were less than 5 years of age at the time of the household survey. A sample of 700 households per stratum was expected to yield information on 155 episodes of child illness per stratum, allowing estimation of effective coverage of management of child illness with a precision of ±6.0%, based on a type-1 error probability of 5% (two-tailed test) and an underlying standard deviation in care of 0.35. Households were enrolled in the study from January 18 to February 13, 2016, and subsequently revisited approximately four to six weeks later for the household care-seeking survey completed between March 3-20, 2016. Mothers were asked about child illness and care-seeking using a questionnaire based on the 2013-2014 Zambia DHS (ZDHS). These included questions about the presence of diarrhea, fever, or suspected ARI in each of their children <5 years in the preceding two weeks. If a child experienced an illness, mothers were asked if care was sought, the source of care, and treatment received. In addition to the series of DHS care-seeking questions, mothers answered questions to ascertain the name of the specific source of care and sequence of care-seeking events. If the name of the source of care was unknown, data collectors were instructed to probe about provider location and other identifying features. Concurrent to household enrollment, health care providers were identified and invited to participate in the provider assessment. The term health care provider will be used to refer to both individual providers such as CBAs and traditional practitioners, and health care outlets that include multiple staff such as health facilities and pharmacies. Public, private, informal, and traditional sources of care were included in the assessment. Community leaders and health workers initially provided a list of commonly utilized care providers offering medicine or alternative treatment for sick children. The list was further expanded with information from participating mothers about common sources of care for treating illness in their children <5 years collected during enrollment. All providers included in the assessment were grouped into categories of providers used in the ZDHS (Box 1) and this classification was employed in all ecological linking analyses restricted by provider category. Public • Government hospital • Government health center/post • Government CBA / fieldworker Private • Private hospital/clinic • Pharmacy Informal • Shop/market • Traditional/faith-based practitioner The provider assessment was completed among all identified health care providers (Figure 1). The provider assessment was designed to assess a provider or facility’s capacity to provide curative services for children <5 years, including presence of drugs and commodities, training, supervision, and provider case management knowledge. The assessment was designed to assess a provider’s structural quality following the Donabedian structure-process-outcome model [17] and to align with the WHO definition of provider readiness [18] as upstream measures of health care provider quality. At facilities and pharmacies with multiple staff, the questionnaire was administered to the most senior staff member and reports of the existence and functionality of physical commodities (medicines, equipment, etc) were verified by observation. Questions were modeled off the SARA general and child health questionnaire [18] and adapted for use with facility-based, community-based, public, private, and informal providers. Clinical case scenarios developed for use in the evaluation of the IMCI program were used to assess provider case management knowledge [19]. Providers were read four clinical case scenarios and asked how they would manage each hypothetical sick child. At outlets with multiple clinical staff, up to three staff members within each cadre of clinical health workers were randomly selected among those available at the time of the assessment to respond to case scenarios. Map of health care provider locations. The provider assessment was used to generate a “structural quality score” corresponding to a provider’s structure or capacity to appropriately manage a child illness. The structural quality score measured availability of services, commodities, and human resources needed to appropriately manage common child illnesses (Box 2). These indicators were considered the minimum inputs for appropriate care: the basic commodities required to diagnosis and treat common child illness, along with the human resources and clinical knowledge to apply them correctly. As such, the score reflects an upper threshold of the potential quality of care offered by a provider. A provider received one point for each indicator if requirements were met and zero if not; each domain received equal weight. The knowledge domain was calculated as an average score of provider performance on four case scenarios. An average facility knowledge score was generated when knowledge was assessed for multiple health workers at a single facility. Providers were assessed against the expected capacity for their specific provider category; for example, CBAs were not penalized for not having antibiotics in this setting where CBAs are not allowed to treat ARI. Diagnostics • Malaria Diagnostic (RDTs or microscopy) • Malnutrition Diagnostic (MUAC or Scale + Height board + Growth chart) • ARI Diagnostic (stethoscope or respiratory timer) • General microscopy (functioning microscope and slides) Basic medicines • Oral rehydration solution • Zinc • Artemisinin combination therapy (ACT) • Oral antibiotic Severe/complicated illness medicines • IV fluids • Injectable quinine or artesunate • Injectable antibiotics Human Resources • Training (at least one staff member with IMCI or relevant training) • Guidelines (IMCI guidelines or relevant guidelines or job aid available) • Supervision (received supervision visit with case management observation in past 3 months) Available services • Diagnosis and treat malaria (by pathology) • Diagnosis and treat diarrhea (by pathology) • Diagnosis and treat ARI (by pathology) • Diagnosis and treat malnutrition (by pathology) • Facilitated referral capacity Knowledge • Average performance on case scenarios The primary outcome was input-based effective coverage of management of child illness estimated through exact-match linking and ecological linking methods. Effective coverage of management of child illness was calculated as the average level of structural quality experienced by sick children based on their reported care-seeking behavior and linked source of care. Each child was assigned the structural quality score of either their specific reported source(s) of care (exact-match linking) or the closest provider(s) based on measures of geographic proximity (ecological linking). The linking was performed using provider assessment data on all health care providers, to reflect capacity among all categories of providers, and data on only health facilities, to replicate the provider data available through common provider assessments such as the SPA or SARA. We considered estimates of effective coverage generated through the exact-match linking using data on all health care providers to be the most accurate linked coverage estimate. However, we did not assess the validity of the effective coverage estimates generated through the method against a true measure of how sick children in the study area were managed.The input-based effective coverage estimate reflects an upper limit on the proportion of children that could have been correctly managed. Estimates generated using the ecological linking methods and using data on only facility-based providers were compared against the exact-match all-provider coverage estimates to assess their population-level validity, or how closely they reproduced the exact-match all-provider estimates of effective coverage. For exact-match linking, each sick child was linked to the specific source(s) of care from which care was sought, based on the name of the facility, outlet, or provider reported by the mother during the household survey. For ecological linking, each sick child was linked to the closest provider(s) based on various measures of geographic proximity. Seven methods for ecological linking were employed, depicted in Figure 2. Measures of geographic proximity employed in the ecological linking were adapted from the work of Skiles and colleagues [20]. Geographic proximity was calculated using ArcGIS 10.1 (Esri, Redlands, CA, USA). Specifications and steps for generating geographic links are presented in Appendix S1 in Online Supplementary Document(Online Supplementary Document). Linked data sets were exported to Stata 14.2 (StataCorp LLC, College Station, TX, USA) for analysis. Illustration of ecological linking methods (household locations have been displaced in figure to protect confidentiality). The seven measures of geographic proximity used in the ecological linking can be grouped into three categories: methods linking children 1) to the single nearest provider by distance, 2) to all providers within a defined geographic unit, and 3) using kernel density estimation. Single nearest provider link: Kernal Density Estimation (KDE): The KDE method was designed to model the level of draw a provider exerts over households as they decide to seek care, based on distance decay and characteristics of the provider. KDE has been used as a means of modeling health care access [21] and service environment [20]. KDE can be used to model health care utilization in the absence of household data assuming all individuals would seek care if skilled providers are accessible, and the choice of provider is driven by provider quality and distance. KDE employs a user-specified kernel size and probability density distribution. The kernel size, or maximum radius of a provider catchment area, was selected to reflect a household preference for higher-level providers. Higher-level providers (hospitals and health facilities) had a larger catchment area than lower level providers (pharmacies and community-based providers). Within a catchment area, a provider’s draw decreased with increasing distance from a household. A provider’s structural quality score was used as the density variable, effectively modeling higher draw within their catchment area for providers with higher scores. Each child was linked exclusively or partially to a category of provider based on the level of draw exhibited by providers in the category. Information on source of care from the household survey was excluded because the method models care-seeking behavior. For the exact-match linking, each sick child was assigned the structural quality score for the specific source(s) from which care was sought. If care was reportedly sought from more than one source, the child was assigned the average score for all providers from which care was sought. If no care was sought for the illness, the child was assigned a structural quality score of zero. If the mother reported a source of care that could not be identified or included in the provider assessment, the child was assigned the average structural quality score for the source of care provider category within the stratum (urban/rural). Effective coverage was calculated as the average structural quality score (a percentage ranging from 0 to 100) across all sick children, including those who were not taken for care (who received a score of 0). For the ecological linking, each sick child was assigned the structural quality score for the source(s) of care that were closest based on various measures of geographic proximity, as described above. All non-KDE ecological linking methods maintained the reported category of source of care. In other words, a child could not be linked to a provider from a category of source of care other than the category reported by the mother (eg, a child that was reported to have been taken to a government CBA could only be linked to government CBAs, although the specific CBA(s) to which he/she was linked might vary depending on the measure of geographic proximity). Similar to the exact-match linking, children that were not taken for care were assigned a structural quality score of zero. Those that could not be linked to a provider from the reported source of care were assigned the average structural quality score for the category of source of care within the stratum. For example, when applying the 5 km radius linking approach, if a rural mother reported that her child was taken to a government health center for care but there was no government health center within 5 km of the household, the child was assigned the average of all government health centers in the rural area. If a child was linked to multiple providers, the average structural quality score for all linked providers was calculated for the child. Effective coverage was calculated as the average structural quality score (a percentage ranging from 0 to 100) across all sick children, including those who were not taken for care (who received a score of 0). To simulate the type of provider data that would typically be available when using a SPA or SARA for linking, we repeated the analysis using only facility data. Coverage was estimated using the exact-match linking method and each of the seven ecological linking methods with only facility structural quality scores. Health facilities were defined as either a government or private clinic or hospital, in line with those providers included in the SARA and SPA surveys. Presuming first-level government facilities offered the most comparable level of structural quality to government CBAs, children that reported care from a CBA were linked to one or more government health centers using the nearest provider and aggregate ecological linking methods, and assigned the average government health center structural quality score for the exact-match linking. Using the exact-match, nearest provider, and aggregate ecological linking methods, children who recieved treatment from all other sources of care (pharmacies, shops, and traditional practitioners) were treated as unskilled sources of care and assigned a structural quality score of zero, equivalent to seeking no care. All other components of the linking methodology and household data remained the same. Using the KDE methods, data on non-facility providers were excluded while modeling care-seeking behavior. A summary of all linking methods employed in the paper is presented in Table 1. Linking method summary HFCA – health facility catchment area *Children reporting seeking care from CBAs were linked 1) to government health centers (primary analysis – Table 3 or Table 2) treated as no care (Table S9b in Online Supplementary Document(Online Supplementary Document)) during facility-only analyses. All other non-facility providers treated as no care and not linked. Descriptive statistics comparing the ecological and facility-only links to the exact-match all provider links were calculated. Sensitivity analyses were also conducted to estimate effective coverage using different assumptions for children that could not be linked to a provider within the reported source of care category including: 1) assigning children that could not be linked to a provider based on geographic proximity a structural quality score of zero, and 2) assigning children that sought care from a CBA a score of zero during the facility-only analysis. These sensitivity analyses were designed to mimic the effect of service environment assessments and linking analyses that define health care access based on a capped maximum distance from a household and that ignore the contribution of CBAs in management of child illness, respectively.

Materials

https://efashare.b-cdn.net/materials/207722be2151bd3805a9def82ae26144e29cfdef067827da49b1df950fdd9472/jogh-08-010607-s001.pdf

Innovations

Based on the provided information, here are some potential innovations that could be used to improve access to maternal health:

1. Mobile Health (mHealth) Solutions: Develop and implement mobile health applications that provide information and resources related to maternal health, such as prenatal care, nutrition, and postnatal care. These apps can be easily accessible to pregnant women and new mothers, providing them with important information and reminders.

2. Telemedicine: Establish telemedicine services that allow pregnant women in remote or underserved areas to consult with healthcare providers remotely. This can help overcome geographical barriers and provide access to quality prenatal care and consultations.

3. Community Health Workers: Train and deploy community health workers (CHWs) who can provide basic maternal health services and education in rural or underserved areas. CHWs can conduct prenatal visits, provide health education, and refer women to higher-level healthcare facilities when necessary.

4. Transportation Support: Develop innovative transportation solutions to address the challenge of accessing healthcare facilities in remote areas. This could include initiatives such as mobile clinics, community transportation services, or partnerships with ride-sharing companies to provide affordable transportation for pregnant women.

5. Financial Incentives: Implement financial incentives or subsidies to encourage pregnant women to seek timely and appropriate maternal healthcare. This could involve providing financial support for transportation costs, reducing or eliminating fees for maternal health services, or offering incentives for attending prenatal visits.

6. Public-Private Partnerships: Foster collaborations between the public and private sectors to improve access to maternal health services. This could involve leveraging private sector resources and expertise to expand healthcare infrastructure, improve service delivery, and increase access to quality maternal healthcare.

7. Health Information Systems: Strengthen health information systems to improve data collection, monitoring, and evaluation of maternal health services. This can help identify gaps in access and quality of care, inform decision-making, and track progress towards improving maternal health outcomes.

These are just a few examples of potential innovations that could be explored to improve access to maternal health. It is important to consider the local context, resources, and needs when implementing these innovations to ensure their effectiveness and sustainability.

AI Innovations Description

The study mentioned in the description focuses on evaluating methods for linking household and health care provider data to estimate effective coverage of management of child illness in Southern Province, Zambia. The goal is to generate more informative population-level estimates of effective coverage by combining data from household surveys and health care provider assessments.

The study collected data on 83 health care providers and 385 children with fever, diarrhea, and/or symptoms of acute respiratory infection (ARI) in the preceding 2 weeks. The majority of children sought care from government facilities or community-based agents (CBAs). The study estimated effective coverage of management of child illness by calculating the average structural quality experienced by all sick children. Structural quality was assessed using case scenarios and a tool based on the WHO Service Availability and Readiness Assessment (SARA).

The study compared different methods of linking children to providers, including exact-match linking (assigning children the structural quality score of their stated source of care) and ecological linking (assigning children the structural quality score of the closest provider based on geographic proximity). The study also assessed the effects of including data on non-facility providers in the linking methods.

The findings of the study showed that linking household and provider data can generate more informative estimates of effective coverage of management of child illness. Ecological linking with provider data on a sample of all skilled providers was found to be as effective as exact-match linking in areas with low variation in structural quality within a provider category or minimal provider bypassing.

Overall, the study provides valuable insights into methods for improving access to maternal health by estimating effective coverage of management of child illness. By combining household and provider data, policymakers and health practitioners can gain a better understanding of the quality of care received by children and identify areas for improvement in maternal health services.

AI Innovations Methodology

Based on the provided description, here are some potential recommendations for innovations to improve access to maternal health:

1. Mobile Health (mHealth) Applications: Develop mobile applications that provide pregnant women and new mothers with information and resources related to maternal health, such as prenatal care guidelines, nutrition advice, and reminders for appointments and vaccinations.

2. Telemedicine Services: Implement telemedicine services that allow pregnant women in remote or underserved areas to consult with healthcare providers through video calls or phone calls, reducing the need for travel and improving access to medical advice and support.

3. Community Health Workers (CHWs): Train and deploy CHWs in rural and underserved areas to provide basic maternal health services, including prenatal care, health education, and referrals to healthcare facilities for more specialized care.

4. Maternal Health Vouchers: Introduce voucher programs that provide pregnant women with financial assistance to cover the costs of maternal health services, including prenatal care, delivery, and postnatal care, making these services more affordable and accessible.

To simulate the impact of these recommendations on improving access to maternal health, a methodology could be developed as follows:

1. Define the indicators: Identify key indicators that measure access to maternal health, such as the percentage of pregnant women receiving prenatal care, the percentage of deliveries attended by skilled birth attendants, and the percentage of postnatal check-ups.

2. Collect baseline data: Gather data on the current status of these indicators in the target population or area.

3. Implement the innovations: Introduce the recommended innovations, such as mHealth applications, telemedicine services, CHW programs, or maternal health voucher programs.

4. Monitor and evaluate: Continuously monitor the implementation of the innovations and collect data on the indicators to assess their impact on improving access to maternal health.

5. Analyze the data: Analyze the collected data to determine the changes in the indicators after the implementation of the innovations. Compare the post-intervention data with the baseline data to measure the impact.

6. Adjust and refine: Based on the findings of the analysis, make any necessary adjustments or refinements to the innovations to further improve their effectiveness in enhancing access to maternal health.

7. Scale-up and replicate: If the innovations prove to be successful in improving access to maternal health, consider scaling up the interventions to reach a larger population or replicating them in other similar settings.

By following this methodology, it would be possible to simulate the impact of the recommended innovations on improving access to maternal health and make evidence-based decisions on their implementation and scalability.

Author

Dima Health

Statistics:

Citations: 9

Identifiers:

DOI: 10.7189/jogh.08.010607

ISSN: 20472978

Research Areas:

Community Interventions, Food Security, Health System and Policy, Infectious Diseases, Maternal Access, Maternal and Child Health, Quality of Care, Sexual and Reproductive Health

Study Countries:

Zambia

Study Design:

Cross Sectional Study

Study Approach:

Quantitative

Participants Gender:

Female