Menu
Menu
Menu
Menu
Background: In an effort to improve population health, many low- and middle-income countries (LMICs) have expanded access to public primary care facilities and removed user fees for services in these facilities. However, a growing literature suggests that many patients bypass nearby primary care facilities to seek care at more distant or higher-level facilities. Patients in urban areas, a growing segment of the population in LMICs, generally have more options for where to seek care than patients in rural areas. However, evidence on care-seeking trajectories and bypassing patterns in urban areas remains relatively scarce. Methods: We obtained a complete list of public health facilities and interviewed randomly selected informal sector households across 31 urban areas in Lusaka District, Zambia. All households and facilities listed were geocoded, and care-seeking trajectories mapped across the entire urban area. We analyzed three types of bypassing: i) not using health centers or health posts for primary care; ii) seeking care outside of the residential neighborhood; iii) directly seeking care at teaching hospitals. Results: A total of 620 households were interviewed, linked to 88 health facilities. Among 571 adults who had recently sought non-emergency care, 65% sought care at a hospital. Among 141 children who recently sought care for diarrhea, cough, fever, or fast breathing, 34% sought care at a hospital. 71% of adults bypassed primary care facilities, 26% bypassed health centers and hospitals close to them for more distant facilities, and 8% directly sought care at a teaching hospital. Bypassing was also observed for 59% of children, who were more likely to seek care outside of the formal care sector, with 21% of children treated at drug shops or pharmacies. Conclusions: The results presented here strongly highlight the complexity of urban health systems. Most adult patients in Lusaka do not use public primary health facilities for non-emergency care, and heavily rely on pharmacies and drug shops for treatment of children. Major efforts will likely be needed if the government wants to instate health centers as the principal primary care access point in this setting.
Zambia is a lower-middle-income country in southern Africa with a life expectancy at birth of 64 years, maternal mortality rate of 213 deaths per 100,000 live births, and child mortality ratio of 62 deaths per 1,000 live births [1]. In 2019, 44% of the population lived in an urban area [1]. Lusaka district, including the capital city, has a population of approximately two million people living in an area of approximately 418 square kilometers. In Lusaka province (of which 80% is Lusaka district), average household wealth, infrastructure, education levels, and access to health care services are generally higher than in other parts of Zambia. For example, in 2018, 50% of the population of Lusaka province was in the country’s highest wealth quintile; 98% had access to an improved source of drinking water compared with 71% nationwide; the female literacy rate was 80% compared with 66% nationwide; and 91% of live births in the preceding five years were in a health facility compared with 84% nationwide [39]. The Zambian health system has a pyramid-structure with three levels. Level 1 includes health posts (with catchment areas of 500 households in rural areas and1000 households in urban areas), health centers (with catchment areas of 10,000 in rural areas and 50,000 in urban areas), mini hospitals (catchment population between 50,000 and 80,000) and district hospitals (catchment population between 80,000 and 20,000). Level 2 includes provincial level hospitals (catchment population 200,000 to 800,000) which provide secondary care and curative care in pediatrics, obstetrics and gynecology and general surgery. Level 3 includes tertiary hospitals (catchment population 800,000 and above), such as the University Teaching Hospital in Lusaka, and specialized hospitals, such as the Cancer Diseases Hospital and the National Heart Hospital. Residential neighborhoods are generally assigned to a nearby health center or health post where they are expected to go as their first point-of-contact with the health system; they may then be referred to a hospital if needed. In practice, residents may choose to go to a different health center or health post from the one they are assigned to; in these cases, they do not incur a bypassing fee because they are still accessing the system at the recommended level. However, if they seek care directly at a hospital, then they incur a bypassing fee. In addition to the public system, there are private and not-for-profit health facilities throughout Zambia. These are registered by the National Health Professions Council [40]. In Lusaka, these are mainly health centers and Level 1 hospitals. At the data of data collection, residents of Lusaka mainly used Level 1 and Level 3 care, as there were few Level 2 hospitals in the city. Since data collection, many health facilities in Lusaka have been upgraded in levels. Throughout this paper, we focus on the levels as they were at the time of data collection. This study was a cross-sectional household survey conducted in Lusaka district in Zambia from November to December 2020. The target population for the study was all adults employed in the informal sector and aged between 18–65 years who lived in Lusaka district, and their children. We define the informal sector as businesses or other economic units that are not registered with a tax or licensing authority. Those who are employed in the informal sector tend not to have contracts or entitlements. As of 2014, the informal sector accounted for about 90% of employment in Zambia [41]. To determine whether respondents were employed in the formal or informal sector, we asked whether they had a formal employment contract and contributed to the National Pension Scheme Authority (NAPSA). We used a random clustered sampling approach to select households for participation in this study. The target sample size of 700 households was chosen for the purposes of a separate analysis of health insurance participation and health system confidence. To draw the sample, we first randomly sampled 35 enumeration areas (EAs) from the 1,225 listed in the 2010 Zambia Census of Population and Housing. Within each EA, we then approached every fourth household until we reached a sample of 20 informal sector households. Eligible heads of households or their spouse were provided information about the study and those who consented were interviewed using the questionnaire. For the purposes of this analysis, we defined the adult analytic sample to include all adults whose most recent health visit was for care for a chronic condition, a check-up, or a new (acute) health issue. We excluded adults whose most recent health visit was an emergency. We defined the child sample to include all children aged five and under who had received care in the past two weeks for fever, diarrhea, cough, or fast breathing. Interviewers were trained and supervised directly by a member of the study team (DOA). Household interviews were conducted from November 6 to December 19, 2020. During interviews, adults in the sample were asked about their own care-seeking during their most recent health visit, as well as care-seeking for fever, diarrhea, cough, or fast breathing in the past two weeks for children aged five and under in their household (up to a total of five children per household). All data were collected using the Open Data Kit (ODK) software package on hand-held tablets. Survey tools were developed in English and then translated to local languages by the survey team. Interviews were conducted in the respondent’s preferred language (English, Nyanja, or Bemba). Residential coordinates for all households were collected directly through the tablets using a geolocation function integrated into ODK. In addition, we collected information on the locations of health facilities in Lusaka. An initial list of facilities as well as their geolocations was obtained from the Zambian Ministry of Health. This list included public facilities as well as private and not-for-profit (e.g., religious) health facilities. It did not include pharmacies or drug shops. Geocodes of all facilities in the sample were verified by one of the authors (DOA) in January 2021 through a combination of online mapping resources (January 10–15) [42] and personal visits to facilities (January 17–22). We obtained ethical clearance from the University of Zambia Social Sciences and Humanities Ethical Clearance Committee (HSSREC-2020-SEP-012) and authority to conduct research from the National Health Research Authority (NHRA00018/15/10/2020). We also obtained ethical clearance from the Ethikkommission Nordwest- und Zentralschweiz (EKNZ) in Switzerland (AO_2020-00,029). The primary outcome was bypassing. We used three definitions of bypassing (Table (Table1).1). These definitions are not mutually exclusive, but each measure different bypassing constructs with different interpretations. First, we defined “primary care bypassing” as using a health facility other than a health center or health post for any non-emergency care. This strict definition of bypassing aligns with guidelines from Zambia’s Ministry of Health. Second, we defined “horizontal bypassing” as using a distant health facility or a pharmacy rather than a nearby facility for non-emergency care – this type of bypassing implies additional transport time and cost, and is likely a reflection of households anticipating to find higher quality of care outside of their residential areas. To identify nearby facilities, we asked all subjects in each neighborhood about the facility their neighborhood belonged to. In most cases, the large majority of respondents agreed on one specific facility. In some cases, two primary facilities were mentioned. We defined nearby facilities as the one (if only one was mentioned) or two (if two were mentioned) facilities that respondents mentioned, as well as the facility that was spatially closest to the respondent (if this was different from the one or two facilities mentioned). Of note, Ministry of Health guidelines do not specify which specific health facility people should go to for care, so horizontal bypassing can in principle be in line with Ministry of Health guidelines as long as people seek care for non-emergency conditions at a health centre or health post rather than a hospital. In practice, many patients seeking care outside of their residential area seek care at higher level facilities, in which case horizontal bypassing also implies primary care bypassing. Last, we defined “two-level” bypassing as using a teaching hospital (Level 3) for non-emergency care. Patients who do this are bypassing not only the available primary health care facilities but also the regular (Level 1, non-teaching) hospitals. Definitions of bypassing We began our analysis by describing the characteristics of the adult and child analytic samples. We described respondents’ demographic characteristics (e.g., gender and age) as well as the landscape of health facilities in the area the where respondents lived. To describe the landscape of health facilities, we calculated the number of health facilities within 1 km and within 5 km of where each respondent lived using Euclidean distance and then took the average across respondents. Next, we mapped and described the spatial distribution of the health facilities in Lusaka and the types of facilities that adults and children in the study sample visited. Mapping included any facilities on the Ministry of Health’s list of health facilities, but it did not include pharmacies or drug shops, even though some respondents sought care in these locations. We then calculated the rate of bypassing (using all three definitions above) for adults and children in the sample, disaggregated by the reason for their health visit. We mapped care-seeking patterns for each study participant meeting each of the three definitions of bypassing using QGIS Version 3 [43]. In addition, we examined how bypassing patterns varied across constituencies. Constituencies are administrative areas that contain multiple EAs; Lusaka has 7 constituencies covering 1,125 EAs. Finally, we used logistic regression to analyze associations between study participant characteristics (including sex, age, marital status, education level, wealth measured using an asset score, and reason for seeking care) and each of the three types of bypassing. We fit models in the adult and child samples separately. We clustered standard errors at the EA level. All analyses were conducted using Stata 16 [44].
