Abstract
Background Monitoring and treating diarrheal illness often rely on individuals seeking care at hospitals or clinics. Cases that seek care through pharmacies and community health workers (CHW) are frequently excluded from disease burden estimates, which are used to allocate mitigation resources. Studies on care seeking behavior can help identify these gaps but typically focus on children under five, even though diarrheal diseases like cholera and Enterotoxigenic E. coli affect all age groups. This study aimed to estimate the proportion of individuals seeking care for themselves or their children with diarrhea, considering different age groups, case definitions, and study settings.
Methods and Findings We conducted a systematic review of population-based primary research studies published during 2000–2024 that examined care-seeking behavior for diarrhea. We included 166 studies from 62 countries. Hospitals and clinics were the most common source of care sought outside the home, with CHW and health posts rarely reported. Using a random-effects meta-analysis, we found substantial heterogeneity in care seeking between studies from low- and middle- income countries (I2 = 99.3) and estimated that the proportion of diarrhea cases seeking care at a hospital or clinic was 32.8% on average (95% Confidence Interval (CI) 28.1% to 37.9%; prediction interval 3.3% to 87.5%). Although there were trends toward higher care-seeking for children compared to adults, substantial variation existed between studies, and the differences were not significant. We estimated that the adjusted odds of seeking care at a hospital or clinic were significantly higher for severe diarrhea and cholera compared to general diarrhea (Odds Ratio 3.43; 95% CI 1.71 to 6.88).
Conclusions Our findings confirm that passive surveillance through hospitals and clinics may substantially undercount the number of people with diarrhea, particularly those with milder symptoms, although this proportion varied widely. Additionally, our findings underscore the importance of including care seeking questions across all age groups in future studies, as we cannot assume lower care seeking for adults across all settings. Our study was limited by fewer data on care- seeking from health posts, traditional healers, and CHW compared to hospitals and clinics, highlighting a need for further research on these sources of care.
Introduction
Diarrheal diseases are a leading cause of illness and death worldwide, particularly among children under five [1,2]. Diarrhea burden estimates rely partly on data from population-based surveys focused on children, which use caregiver recall of recent symptoms [1,2]. In adults and for specific pathogens, estimates are often based on passive surveillance data, such as hospital records and vital statistics [1,2]. These data are especially valuable for tracking seasonal outbreaks of pathogens like cholera that may not appear in surveys but depend on individuals seeking care for often severe symptoms [3,4]. A key challenge in accurately assessing diarrheal disease burden by age and etiology is determining how many cases go unreported due to barriers in care-seeking and identifying which populations or illnesses are most likely to be undercounted.
Estimates of diarrhea burden strongly influence investments in disease mitigation resources, including broad investments in improved water and sanitation and more pathogen-specific tools like vaccines. However, diarrheal pathogens vary in the symptoms they cause as well as the age groups they primarily affect. For example, rotavirus, Escherichia coli, and Salmonella can cause gastroenteritis with diarrhea and/or vomiting, while cholera is marked by acute watery diarrhea and/or vomiting. Rotavirus primarily affects children, whereas cholera is a leading cause of diarrhea illness and deaths across all age groups [1]. Strategies for controlling rotavirus differ from those used for cholera, as its primary mode of transmission is person-to- person via the fecal-oral route, rather than through consuming contaminated food and water. Enteric pathogens also cause a range of disease severity, from mild infection to severe dehydration requiring rehydration to prevent severe outcomes including death. Given these variations, relying solely on population-level estimates of general diarrhea in children and passive surveillance data may not optimize burden reduction efforts.
Programs such as the Demographic and Healthcare Surveys (https://dhsprogram.com/) that assess care seeking behavior for diarrhea around the world provide an opportunity to examine these biases in disease burden estimates [5–9]. Previous analyses of these data have estimated that the proportion of caregivers that seek care at health facilities for a child with diarrhea was on average 45% in sub-Saharan Africa [10] and 49-51% in West and Central Africa [11], with substantial variation by country [12] and type of care sought [13]. Importantly, these studies all focused on care seeking for children, and did not incorporate data from studies that have reported care seeking across age groups including adults [14,15]. In addition, while one of these reviews found that more people sought care for general childhood illnesses when illness was more severe [13], none of these studies examined care seeking behavior by severity of diarrheal illness or etiology, limiting the applicability of their results.
In this study, we sought to address these knowledge gaps through a systematic review and meta-analysis of surveys conducted during 2000-2024 on diarrhea care seeking behavior across all age groups and diarrhea case definitions. Our primary research question was: what proportion of individuals of any age seek care at hospitals, clinics, or other sources when they have diarrhea? In addition, we asked: how does seeking care at hospitals or clinics vary by age group, severity of diarrheal illness, and other factors?
Methods
Systematic review
We developed the methods for this study based on the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [16] and submitted the protocol to the International Prospective Register of Systematic Reviews (PROSPERO) on February 24th, 2023 (review ID: CRD42023402435) [17].
We searched PubMed, Embase, Web of Science, and Global Index Medicus on January 27, 2023 using search terms developed by a medical librarian (AGS); searches were updated on September 3, 2024 (Supplementary Methods). We included studies that 1) included sampling on or after January 1, 2000, 2) were primary research studies, 3) were conducted in human populations, 4) included questions about diarrhea care seeking, perceptions knowledge, and/or management, 5) surveyed individuals of any age about their own care seeking for diarrheal illness, 6) reported the number of individuals surveyed about care seeking for diarrheal illness either for themselves or a child, 7) reported the number or proportion who said that they would or did seek care, and 8) were written in English, French or Spanish. We excluded studies that 1) were conducted in special populations, e.g., HIV- or malaria-infected individuals, healthcare professionals, deployed troops, travelers, or long-term care facility residents, 2) focused exclusively on non-infectious diarrhea, 3) did not separate diarrhea-related care seeking from other illnesses, 4) did not report the sampling method and study population, 5) were case studies, case-control studies where cases selected based on care seeking behavior, or qualitative studies, 6) used convenience sampling, 7) were reviews or secondary data analyses, 8) were conference abstracts, or 9) did not have full texts available.
Studies were screened by title and abstract by two blinded and independent reviewers (DJC, APS, MHM, or KEW) using Covidence (https://www.covidence.org/). If a tiebreaker was needed, a third reviewer was used, or a decision was made by consensus. This process was repeated for full text screening. For the updated search, a single reviewer (MHM) performed the screening. One of the reviewers independently extracted data from each included study using a shared Google Sheet. Extracted data included study design, sampling method, study population, diarrhea case definition, start and end dates, study catchment area matched to a GADM polygon (https://gadm.org/) when possible, how the care seeking questions were phrased, who the care was sought for (e.g., self or child), recall time in days, type of care facility sought, number of individuals surveyed about diarrheal illness, whether they “would” and/or “did” seek care, and when available, age and sex distribution of the surveyed population, diarrhea symptoms, and diarrhea severity. When care seeking data were stratified by any of these variables, we extracted the data at those stratifications. If a study reported “yes”, “no”, and “maybe” for “would/will you seek care”, we coded “maybe” as “no”.
A single reviewer (SMH) evaluated the quality of the studies that met our inclusion criteria using an adapted version of the appraisal tool for cross-sectional studies (AXIS) [18]. Specifically, we assessed whether 1) the data presented in the study were internally consistent (i.e., numbers for the proportion seeking care were identical throughout the abstract, text, and tables), 2) the manuscript included a justification for their sample size, and 3) the response rate was reported [0 = no, 1 = yes]. An overall quality score was created by summing across the three indicators for total scores ranging from 0-3, where 0 was lowest quality and 3 was highest quality. Studies were not excluded based on quality assessment criteria. We retained studies with internal inconsistencies to avoid penalizing those that provided more detailed information or in-depth discussions of results, where minor differences or ambiguities often arose. Studies with major ambiguities were excluded in full text screening for lacking care seeking data (for example, if it was unclear what the denominator was for the reported proportion seeking care).
Data cleaning
We manually reviewed studies by country, sorted by start date, to identify studies with overlapping data (i.e., conducted among the same people in the same time period with the same questions). Overlapping studies were excluded from the primary dataset, and reason for selection of the retained study was recorded in the extraction sheet. We prioritized inclusion of studies with 1) more representative sampling methods, 2) more consistent or clearly described data, 3) more specific case definition, 4) larger sample size, 5) longer sampling period, 6) more detailed age information, and 7) more detailed methods description, in that order.
Data were cleaned, checked for mistakes, and analyzed using R statistical software version 4.3.3 [19]. Extracted data were checked for correct location information (e.g., could be correctly matched to a GADM shapefile, impossible values (e.g., number seeking care greater than number surveyed, start date after end date), and inconsistent values (e.g., multiple different sample sizes entered per study or stratification). Data were matched with World Bank 2023- 2024 Income Classifications [20] and regions by country.
We created three standardized categories for diarrhea case definition: 1) “diarrhea” included explicit definitions of three or more loose stools in past 24 hours and diarrhea used broadly without an explicit definition; 2) “severe diarrhea or cholera” included severe diarrhea, defined as having danger signs, dehydration, ≥1 week duration, or death, and acute watery diarrhea of any severity including deaths; 3) “gastroenteritis or other etiologies” included definitions of either diarrhea or vomiting, and specific etiologies including rotavirus, Escherichia coli, Salmonella, and giardia. A single study stratified care seeking by diarrhea cases that were “moderate to severe” (diarrhea with sunken eyes, loss of skin turgor, dysentery, intravenous rehydration, or hospitalization) and “less severe” (all other diarrhea) [21]; we categorized “moderate to severe” as “severe diarrhea or cholera” and “less severe” as “diarrhea”.
In addition, we created standardized categories of the source of care sought (Table S1). For the analysis of care seeking at hospitals or clinics, we defined “hospital or clinic” as any general hospital or clinic or explicitly defined public hospitals or clinics. We excluded explicitly defined private hospitals or clinics from this category because they are rarely part of surveillance systems.
Meta-analysis
Observations, or units for analysis, were created by aggregating extracted data by study, country, source of care sought, and potential sources of variation in care seeking (shown in Table 1). When aggregating data from studies with multiple choice questions where there were multiple different options within one category or source of care (e.g., multiple different hospitals or clinics that could have been selected), we took the maximum value across those options.
We used logistic regression models with observation-level random intercepts to pool estimates of the proportion of individuals with diarrhea that did or would seek care for diarrhea at hospitals or clinics across all included studies, implemented using the metaprop function of the meta package in R [22]. We used logistic regression models with observation-level random intercepts and fixed effects to examine the effects of diarrhea case definition, whether the study was in an urban or rural location, whether the study took place during or closely following a diarrhea outbreak, as well as other components of study methodology on care seeking in univariate and multivariate analyses, implemented using the rma function of the metafor package in R [23]. For multivariate analysis, we included variables that we identified as significantly associated with care seeking (p < 0.05) in univariate analysis but that were not significantly associated with other included variables. In addition, we performed sub-group analyses by age group and diarrhea case definition.
Data and code availability
All extracted data and source code for data cleaning and meta-analysis are available at https://github.com/wienslab/diarrhea-careseeking.
Role of the funding source
This study was supported by the Bill and Melinda Gates Foundation (INV-044865) and the National Institute of Allergy and Infectious Diseases (K22AI168389). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Results
Study characteristics
We retrieved 7069 studies from four databases, screened 5188 for eligibility following de- duplication, and included 188 studies (Figure 1). Of these, 166 studies were non-overlapping and asked participants whether they did seek care for a past diarrheal illness. For the primary analysis, we stratified these study-level data by characteristics shown in Table 1, for a total of 211 observations pertaining to care seeking outside of participants’ homes, and 177 observations pertaining to care seeking specifically at hospitals or clinics. The complete analytical datasets can be found in Supplementary Data 1.
The studies included in the primary dataset were conducted in 62 countries across 10 World Bank geographic regions, with most observations coming from East Africa (N = 60), South Asia (N = 41), Latin America and the Caribbean (N = 25), Western Africa (N = 24), and East Asia and the Pacific (N = 23). Twenty of the observations came from nationally representative studies, and 53, 78, and 60 came from studies representative of first, second, and third administrative divisions, respectively (Figure S1). Studies included sampling completion dates between 2000 and 2022, with most of the observations completed during 2005–2009 (N = 67) and 2010–2014 (N = 67) (Figure S2).
Most observations in the primary dataset were from studies that asked a caregiver about care seeking for a child with diarrhea (N = 138, 65.4%) (Table 1). Others represented a combination of children, other family members, or the respondent themselves (N = 59, 28.0%) and a small proportion represented care seeking exclusively for the survey respondents (N = 14, 6.6%) (Table 1). A small proportion of observations represented responses from caregivers of individuals at a health facility (N = 11, 5.2%) (Table 1). More of the observations were from exclusively rural areas (N = 83, 39.3%) than exclusively urban areas (N = 68, 32.2%), and two (0.9%) came from internally displaced persons or refugees (Table 1). Seventeen (8.1%) observations took place during or closely following a diarrheal disease outbreak (Table 1).
The majority of the observations came from surveys where respondents were asked about care seeking for general diarrhea (N = 156, 73.8%), followed by either gastroenteritis or non-cholera etiologies (N = 33, 15.6%), and severe diarrhea or cholera including deaths (N = 22, 10.4%) (Table 1). Most observations focused on care seeking at any time during illness (N = 175, 82.9%) and most used recall periods of 14-30 days (N = 143, 67.8%) (Table 1).
Trends in care seeking at hospitals or clinics in the unadjusted data
Our primary interest was care-seeking at hospitals or clinics, where people can receive treatment and are most often counted in disease surveillance efforts. Since care seeking behavior among individuals at a clinic likely differs from care seeking behavior in the broader community, we restricted the data in the analyses below to non-clinic-based studies unless otherwise noted.
We observed variations in care-seeking by geographic region (Figure S3a) and country income group (Figure S4a). These trends were inconsistent between case definitions, which also varied by income group (Figure S3b, Figure S4b). Specifically, in high-income countries (HICs), the most common case definitions were gastroenteritis or non-cholera etiologies, while severe diarrhea and cholera were more often studied in low- and middle-income countries (LMICs). Given a relatively low number of observations from high-income countries, we focus on care- seeking behavior in LMICs in the main analyses.
We found that the reported proportion seeking care at a hospital or clinic in 145 observations from LMICs was higher for observations from studies that surveyed participants about care seeking at any time during illness (median of 34.7%; N = 125; IQR, 18.9% to 55.0%) compared to their first source of care (median of 21.2%; N = 20; IQR, 10.0% to 33.5%) (Figure 2a) as well as studies conducted during or following a diarrheal disease outbreak (median of 62.7%; N = 11; Interquartile Range (IQR), 44.2% to 65.5%) compared to those that were not (median of 31.0%; N = 134; IQR, 15.3% to 49.9%) (Figure 2c). Care seeking at hospitals or clinics was also higher for severe diarrhea or cholera (median of 64.1%; N = 15; IQR, 45.4% to 70.0%) compared to general diarrhea (median of 31.0%; N = 122; IQR, 15.3% to 48.2%) and gastroenteritis or other etiologies (median of 33.3%; N = 8; IQR, 25.5% to 44.1%) (Figure 2b).
Although we found no significant correlation between overall study quality and reported care seeking at hospitals or clinics (Spearman rho = 0.07; 95% Confidence Interval -0.09 to 0.23; Figure S5a), studies with internal inconsistencies identified during study quality assessment more often reported lower care seeking than those without (Figure S5b).
Studies that surveyed participants about their own care-seeking behavior reported lower levels of care-seeking on average compared to those focused on children (Figure 2d). Yet, across 134 observations in LMICs that reported the age distribution of participants, we found no significant correlation between the proportion of the study population under five years of age and the proportion that sought care at hospitals or clinics (Spearman rho = 0.10; 95% Confidence Interval -0.08 to 0.26; Figure 3a). Studies conducted only among children reported higher care seeking at hospitals or clinics (median of 34.2%; N = 98; IQR, 17.5% to 55.0%) compared to studies conducted only among adults or individuals over five (median of 22.4%; N = 22; IQR, 14.5% to 49.3%) (Figure 3c), but there was wide variation between studies and this trend held for diarrhea and gastroenteritis but not for severe diarrhea or cholera (Figure 3b,d).
Trends in care seeking at other sources outside the home in LMICs
Hospitals and clinics were the most common source of care sought across studies in LMICs (median of 32.3%; N = 145; IQR, 17.0% to 52.6%) (Figure 4a), though they were less commonly the first source of care (median of 21.2%; N = 20; IQR, 10.0% to 33.5%) (Figure 4b). Seeking care from community health workers (CHW) was reported more often as first sources of care (median of 16.9%; N = 9; IQR, 4.5% to 18.9%) than the main or primary of care (median of 6.2%; N = 32; IQR, 1.7% to 14.7%) (Figure 4b). Seeking care from CHW and health posts were also more frequently reported in observations from low-income compared to middle-income countries (Figure S6) and for general diarrhea compared to severe diarrhea or cholera (Figure S7). Traditional healers were rarely reported as a source of care (median of 4.0%; N = 48; IQR, 1.0% to 8.7%) (Figure 4). These less common sources were also less frequently included or reported as options in survey questions than hospitals or clinics (Figure 4).
In studies that surveyed participants at a hospital or clinic, most participants reported hospitals/clinics as their first source of care (median of 89.3%; N = 4; IQR, 74.8% to 97.7%) (Figure S8a). In studies that surveyed participants about where they sought care before their hospital or clinic visit, hospitals/clinics were the most common previous sources of care, (median of 35.1%; N = 2; IQR, 31.2% to 39.0%), followed by pharmacies (median of 20.7%; N = 2; IQR, 18.7% to 20.4%), private hospitals/clinics (median of 10.6%; N = 2; IQR, 9.4% to 11.9%), and traditional healers (median of 5.6%; N = 2; IQR, 5.4% to 5.8%) (Figure S8b).
Adjusted estimates of care seeking at hospitals or clinics
Using a random-effects meta-analysis, we estimated that on average 32.8% (95% Confidence Interval (CI) 28.1% to 37.9%) of individuals sought care at hospitals or clinics across 145 observations in LMICs (Figure S9). In addition, we predicted that care seeking for a future study would be in the range of 3.3% to 87.5% (Figure S9). This wide prediction interval reflects the substantial variation we found between studies in care seeking (τ2 = 1.8; I2 = 99.3; 95% CI, 99.2 to 99.3), with adjusted study-level estimates of care seeking ranging from 0.7% (95% CI 0.5% to 0.9%) to 94.4% (95% CI 90.6% to 97.0%) (Figure S9).
We estimated that the proportion of people that would seek care during a hypothetical diarrheal episode, as opposed to an actual previous episode, in LMICs was substantially higher, with 71.2% (95% CI 57.1% to 83.2%; prediction interval 5.8% to 99.1%) intending to seek care at a hospital or clinic if they or their child had diarrhea across 29 observations (Figure S10). Care seeking was on average lower in HICs, with 23.9% (95% CI 16.2% to 33.8%; prediction interval 2.5% to 79.6%) seeking care at a hospital or clinic when they or their child had a recent diarrheal episode (Figure S11).
Factors associated with variation in care seeking at hospitals or clinics in LMICs
We next examined factors associated with variation in care seeking in LMICs. In univariate analyses, we found that the odds of seeking health care at a hospital or clinic were higher for severe diarrhea or cholera than general diarrhea, as well as during or after an outbreak and when a recall period exceeded 30 days (Table S2). If the question was restricted to the first source of care sought, the odds of seeking care at a hospital or clinic were lower (Table S2). We found no significant differences in care seeking between observations conducted among only children under five compared to those conducted only among individuals over five (OR 1.06 (95% CI 0.55 to 2.01)) (Table 2). In a subset of eight studies with results stratified by age, differences in care seeking for individuals under five years of age compared to over five years remained non-significant, though there was a trend towards higher care seeking for children under five (OR 1.24 (95% CI 0.29 to 5.36)) (Table 2).
To examine potential confounding between the sources of variation in our dataset, we additionally tested for associations between the variables related to study design and setting specific attributes included in the univariate analyses above. We found that studies with severe diarrhea or cholera case definitions were associated with longer recall periods and were more likely to have been conducted during or after an outbreak compared with studies on general diarrhea (Table S3). In sub-analysis, we found that there was no longer an effect of being in an outbreak setting when subsetting to studies with severe diarrhea or cholera case definitions (Table S4). In addition, there was no longer an effect of recall period when subsetting the data to observations with general diarrhea case definitions (Table S4).
Since we could not separate the effects of case definition from recall period or outbreak context, we examined the effects of case definition on care seeking at hospitals or clinics adjusting only for the timing of care seeking (i.e., whether they were surveyed about their first source of care or care at any time during illness). The adjusted odds of seeking care for severe diarrhea or cholera remained significantly higher than for general diarrhea (OR 3.43 (95% CI 1.71 to 6.88)) (Table 3), and residual heterogeneity remained high (τ2 = 1.57; I2 = 99.5; R2 = 10.2%).
Correspondingly, when we stratified the meta-analysis by case definition, we estimated that 58.6% (95% CI 39.9% to 75.2%; Prediction Interval 5.2% to 97.3%) of individuals sought care at a hospital or clinic when they or their child had severe diarrhea or cholera (Figure S13) compared to 29.9% (95% CI 25.3% to 35.1%; Prediction Interval 3.1% to 84.9%) for general diarrhea (Figure S12) and 35.8% (95% CI 23.7% to 50.0%; Prediction Interval 6.5% to 81.8%) for gastroenteritis and non-cholera etiologies (Figure S14).
Discussion
Here we estimated that most people with diarrhea (on average 67%) do not seek care at hospitals or clinics where they can receive treatment when needed and be counted in typical surveillance systems. We found substantial variation between studies, even after accounting for sources of variation such as case definition and timing of care seeking. Hospitals and clinics were the most common source of care investigated and reported, with pharmacies, CHW, and health posts less commonly reported, and traditional healers very rarely cited as a source of care. Studies that phrased questions in terms of hypothetical care seeking reported substantially higher care seeking than those that asked about actual diarrheal episodes. We found no significant differences in care seeking between children under five and older individuals, and the odds of seeking care for severe diarrhea and cholera were significantly higher compared to general diarrhea.
Our findings shed light on the extent to which we may underestimate diarrhea burden when we rely solely on passive surveillance data, with the greatest underestimates occurring, unsurprisingly, among individuals with mild or moderate symptoms. Additionally, our findings suggest that the relationship between age and care-seeking behavior varies by location, context, and etiology. While we found trends towards adults seeking care less often for themselves than for children with diarrhea, we found that, on average, adults sought care for severe diarrhea and cholera at similar rates as for children. Given these results and the considerable heterogeneity across studies, we recommend that local data on care-seeking behavior, stratified by age and either symptom severity or etiology, be used to adjust burden estimates whenever possible. In the absence of local data or for global analyses, the case- definition-stratified estimates from this review may provide useful approximations of missed symptomatic cases.
Although we estimated lower average care-seeking rates than previous analyses of DHS data (33% here vs. 45-51% in DHS data)—possibly due to our inclusion of a broader range of study settings and populations—our findings of very low reported care-seeking at CHW and traditional healers align with those studies [5–9]. This may reflect true patterns of care-seeking behavior, but it could also indicate a research bias, as hospitals and clinics were more frequently included in survey questionnaires than other sources of care. Additionally, only a minority of studies reported results where multiple sources of care were selected, and individuals may be less likely to report CHW or traditional healers as their sole or primary source of care. This is supported by our finding that CHW were reported more frequently, and hospitals and clinics less frequently, in studies that examined the first source of care rather than overall source of care. These limitations could have led us and others to under-estimate the role of CHW and traditional healers in diarrhea case management.
This study has several additional limitations. Only a small proportion of the included studies were conducted among adults or individuals over five years old and care seeking in those studies varied widely, which may have prevented us from identifying significant trends.
Methodology and questionnaires also varied between studies, which we may not have fully captured in our analysis, and there were several variables we could not examine due to data scarcity or potential confounding (e.g., income group, outbreak setting, and recall period).
We also estimated that more than twice as many people would seek care as actually did, but we were not able to examine factors that shape care seeking behavior in detail. Decisions about whether and where to seek care are complex, with tradeoffs and determining factors that vary by location, time, and individuals. Existing theories for health service utilization describe a wide range of factors influencing care seeking, including environmental and health system factors (e.g., provider and treatment availability, transportation and access), population and socio- demographic factors (e.g., age, gender, income, culture, community support, disease prevalence), and individual behavior (e.g., perceptions about disease risk and severity, health care costs, and benefits) [24]. The huge heterogeneity we found in care seeking, even after accounting for different aspects of study methodology, may result from the propagation of individual heterogeneity in decision making and contextual factors. An important area for future research will be to examine these factors to identify ways in which care seeking could be improved, especially for severe cases.
Given our findings, and their limitations, we propose several areas for future work. First, it will be important for future studies to include questions about care seeking across ages and diarrhea case definitions, include a broad range of sources of care including health posts and CHW, and focus on actual rather than hypothetical care seeking where possible. In addition, studies that explicitly examine the order that individuals seek care at different sources (for example, [25–27]) can help inform when and why there may be delays in seeking care at hospitals or clinics and how this varies by age and disease severity. Finally, synthesizing qualitative studies that examine reasons why individuals do or do not seek care could provide insight into non-hospital- based care seeking as well as what needs to be done to reduce barriers for populations with the least interaction with the health system.
In summary, our findings suggest that care-seeking estimates for broadly defined diarrhea may not reflect true patterns, particularly for severe symptoms. Conclusions about care-seeking behaviors across age groups were less straightforward, as we found trends but no significant differences between children under five and older individuals. DHS data may reasonably approximate general care-seeking patterns for non-severe, non-etiology-specific diarrhea for the locations in which they were collected and in the absence of additional information. When possible, assumptions about the relationship between age and care seeking should be setting- specific, as adult care-seeking was not universally equivalent or lower.
Funding
This study was supported by the Bill and Melinda Gates Foundation (INV-044865 to ASA) and the National Institute of Allergy and Infectious Diseases (K22AI168389 to KEW).
Conflicts of interest
The authors declare no conflicts of interest.