Treatment-seeking behaviour, cost burdens and coping strategies among rural and urban households in Coastal Kenya: an equity analysis

Authors


  • 1

    A household was defined as a group of people living in the same compound, who are answerable to the same head and share a common source of food and/or income. A homestead was defined as ‘a collection of adjacent or nearby households with a single individual as an administrative head’. There may be more than one household in a homestead because people do not always share food or income.

  • 2

    Fieldworkers explained illness definitions to households at the outset and for all illnesses asked respondents if they knew the name of the problem. Where formal treatment was sought, the health card was checked but reported illnesses/symptoms were recorded because of the potential influence on health seeking behaviour.

  • 3

    The approximate exchange rate at the time the study took place was US $1=78 KES.

  • 4

    These are small clinics, usually operating in one or two rooms by a clinical officer or a nurse. A few have laboratories operated by a technician who has some ‘partnership’ with the owner of the clinic.

  • 5

    We did not locate any study that compared different categories of illnesses (acute and chronic) and their variations between rural and urban areas. Those identified in the literature tended to focus on either acute or chronic conditions only in reference to one type of disease (e.g. diabetes) or on general illnesses without classifying them into either category.

Corresponding Author Jane Chuma, Kenya Medical Research Institute (KEMRI), P.O. Box 230, Kilifi, Kenya. Tel.: +254-41-522063; Fax: +254-41-522390; E-mail: jchuma@kilifi.kemri-wellcome.org

Summary

Ill-health can inflict costs on households directly through spending on treatment and indirectly through impacting on labour productivity. The financial burden can be high and, for poor households, contributes significantly to declining welfare. We investigated socio-economic inequities in self-reported illnesses, treatment-seeking behaviour, cost burdens and coping strategies in a rural and urban setting along the Kenyan coast. We conducted a survey of 294 rural and 576 urban households, 9 FGDs and 9 in-depth interviews in each setting. Key findings were significantly higher levels of reported chronic and acute conditions in the rural setting, differences in treatment-seeking patterns by socio-economic status (SES) and by setting, and regressive cost burdens in both areas. These data suggest the need for greater governmental and non-governmental efforts towards protecting the poor from catastrophic illness cost burdens. Promising health sector options are elimination of user fees, at least in targeted hardship areas, developing more flexible charging systems, and improving quality of care in all facilities. The data also strongly support the need for a multi-sectoral approach to protecting households. Potential interventions beyond the health sector include supporting the social networks that are key to household livelihood strategies and promoting micro-finance schemes that enable small amounts of credit to be accessed with minimal interest rates.

Abstract

Objectif Investiguer les inégalités socio-économiques dans des maladies auto rapportées, le comportement de recours au traitement, les charges des dépenses et les stratégies de réaction dans un milieu rural et urbain sur la côte kenyane.

Méthodes  Plusieurs méthodes ont été employées comprenant: une surveillance de 294 ménages ruraux et 576 ménages urbains, 9 discussions focalisées de groupe et 9 entretiens détaillés dans chaque site.

Résultats  Les principales observations étaient les suivantes: taux significativement plus élevés de conditions chroniques et aiguës rapportées dans le milieu rural comparé au milieu urbain, différences dans les profils de recours au traitement par statut socio-économique et selon le milieu, et escalade des charges des dépenses dans tous les deux milieux.

Conclusions  Ces données suggèrent le besoin d'un accroissement des efforts gouvernementaux et non gouvernementaux afin de protéger les pauvres contre des charges catastrophiques des dépenses de maladie. Des options prometteuses comprennent l’élimination prudente des honoraires pour les utilisateurs, au moins dans des secteurs cibles de difficultés, en développant des systèmes de tarification plus flexibles où des honoraires sont maintenus et en améliorant la qualité des soins dans tous les services. Les données soutiennent fortement aussi la nécessité d'une approche multisectorielle pour la protection des ménages. Des interventions potentielles au delà du secteur de la santé comprennent le soutien des réseaux sociaux et sont prépondérantes aux stratégies de vie de ménage et en favorisant les schémas de microéconomie qui allouent de petits crédits accessibles à des taux d'intérêt minimaux.

Abstract

Objetivo  Investigar las inequidades socioeconómicas en enfermedades autoreportadas, en comportamiento de búsqueda de tratamiento, carga del coste y estrategias de enfrentamiento en localidades urbanas y rurales a lo largo de la costa keniata.

Métodos  Se utilizaron métodos multiples incluyendo: una encuesta en 294 hogares rurales y 576 urbanos, 9 discusiones focalizadas de grupo y 9 entrevistas en profundidad en cada localidad.

Resultados  Los principales resultados fueron: unos niveles más altos de condiciones crónicas y agudas reportadas en las zonas rurales comparadas con las urbanas, diferencias en los patrones de búsqueda de tratamiento según el estatus socioeconómico y según la localidad, y una carga de coste regresiva en ambas áreas.

Conclusiones  Estos datos sugieren la necesidad de un aumento de esfuerzos gubernamentales y no gubernamentales por proteger a los pobres de los costes catastróficos de la enfermedad. Algunas opciones prometedoras incluyen la eliminación de cuotas a los usuarios, al menos en ciertas áreas especialmente deprimidas; desarrollar sistemas de cobro más flexibles en lugares en los que se mantengan y mejorar la calidad del servicio ofrecido en los centros. Los datos también apuntan hacia la necesidad de un enfoque multi-sectorial para proteger a los hogares. Las intervenciones potenciales más allá del sector salud incluyen el apoyar las redes sociales que son una pieza clave en las estrategias de sustento de los hogares, así como promover esquemas de micro-créditos que hacen accesibles pequeñas cantidades con intereses mínimos.

Introduction

Household cost burdens and their impacts

Ill-health causes financial hardship for many households directly through spending on treatment and indirectly by limiting labour participation and undermining people's income generating activities (World Bank 1997; Barnett et al. 2001). The costs of seeking treatment and the coping strategies employed to either avoid or meet these costs are potentially catastrophic, i.e. have negative implications on the future survival households (Whitehead et al. 2001; Russell 2004). Costs of 10% or more of total health expenditure are often considered indicative of catastrophic spending (Prescott 1999; Russell 2001, 2004; Waters et al. 2004). Other studies have used cut-off values of 40–50% of household income (Xu et al. 2003; Su et al. 2006). The World Health Organization (WHO) estimates that households that spend 40% or more of their non-food expenditure on treatment are likely to be impoverished (WHO 2000). Equity in health care requires that all people benefit equally from health care services, regardless of their socio-economic status (SES) and place of residence, and that payment is based on the ability to pay (Mooney 1983; Braveman 1998). Any regressive differences between the poor and less poor in cost burdens, coping strategies and negative consequences represent inequities in health care that are both unacceptable and avoidable (Whitehead 1992). A good understanding of household costs, coping and consequences, and how and why these differ by SES and place of residence should assist the design of more equitable health systems.

Recent reviews of the literature by Russell (2004) and McIntyre et al. (2006) reveal that treatment-seeking and cost burdens vary by the type of disease. The chronic conditions of the Human Immunodeficiency Virus/Acquired Immunodeficiency Syndrome (HIV/AIDS) and tuberculosis impose significantly higher total costs on households, usually over 10% of total household expenditure, than malaria. On the basis of their reviews, these authors highlight that health service weaknesses, such as low coverage, user charges and poor quality of care contribute to high costs for patients in many countries. If household well-being is to be protected, more collective health services and resources are needed. They also argue that international decisions around disease prevention and pro-poor curative health services require further micro-economic research around costs of illness to households, household responses and the implications of costs and coping for poverty.

Several studies suggest that ill-health disproportionately affects the poor, that the poor can have particular difficulties accessing health care, and that when they seek care, they spend a greater proportion of their income on treatment than richer households (Ettling et al. 1994; Liu et al. 1996; Fabricant et al. 1999; Onwujekwe & Uzochukwu 2004). To meet the costs of illness, poor households adopt coping strategies that are potentially ‘risky’ for their future economic welfare, such as selling critical assets and sinking into inescapable debt (Russell 1996, 2001;Wilkes et al. 1998;Meessen et al. 2003). Furthermore, pro-poor policies within the public sector, for example exemptions and waivers, have been found to be relatively ineffective in protecting the poor (McIntyre et al. 1995; Asenso-Okyere et al. 1998; Nyonator and Kutzin 1999; Van der Geest et al. 2000). Such information has contributed to many governments considering a shift from charging user fees towards providing free health care services at primary levels and towards other risk pooling approaches such as community prepayment schemes and social health insurance (WHO 2001).

Kenya

The Kenyan government has long prioritized equitable access to health care services for all Kenyans (GOK 1999a, b). However, existing evidence suggests that the country's health care financing strategies have had negative implications for equity (Moses et al. 1992; Mwabu et al. 1995; Collins et al. 1996; Owino 1998; Pearson 2005). A relatively recent household expenditure and utilization survey carried out as part of the National Health Accounts (NHA) framework reported that households provide 51% of total health care expenditure through out-of-pocket payments (MOH 2003). This survey was conducted at a time when 56.8% of Kenyans were living on less than US $1 per day. Such data suggest that a majority of Kenyans face difficulties in raising money to pay for treatment. Key NHA findings regarding socio-economic and geographic differences included:

  • Urban households are more likely to report illnesses than their rural counterparts (19.5%vs. 16.9%), and are more likely to visit a provider (81.5 and 75.9%, respectively).
  • Households in the poorest quintile are less likely than those in the richest quintile to make an outpatient visit (1.7 and 2.3 visits per annum, respectively), and are more likely not to seek health care at all (33 and 16%).

These and other national level surveys provide invaluable overviews of treatment-seeking and payment patterns. However, they do not consider how treatment payment strategies — i.e. what people do to raise money to pay for health care — differ by SES and geographical location and the implications these strategies have for household assets and SES.

While there exists a wide body of literature on health care utilization and cost burdens in Kenya and elsewhere, inequities in cost burdens and coping strategies by SES, and between rural and urban settings, have received relatively little attention. In this paper we present data on socio-economic inequities in self-reported illness, treatment-seeking behaviour, cost burdens and coping strategies from a rural and an urban setting along the Kenyan coast. In doing so, we improve current understanding and contribute to national and international literature and debates on health care access and health related inequities. Such information is important for Kenya and other countries that are considering changes in health care financing and service charges.

Materials and methods

Study area

The study was conducted in two areas in Kilifi district along the Kenyan coast: Mtwapa, an urban area located along the main Mombasa–Malindi highway, 42 km south of Kilifi Town; and Ganze, a rural area 35 km inland of Kilifi Town. Mtwapa is a densely populated, low-income area typical of many cities in sub-Saharan Africa. Ganze is a remote rural area, sparsely populated with limited infrastructure and poor housing. Kilifi district is the second poorest in Kenya and Ganze is the poorest division in the country. The primary economic activities in Mtwapa are small-scale trading and skilled and unskilled work in the hotel industry, while agriculture is the main source of income in Ganze. The population in Ganze is predominantly Giriama, a sub-group of the Mijikenda ethnic group. Mtwapa houses a wide range of ethnic groups from different parts of the country.

Each study site has one government health care dispensary, and there is a district hospital in Kilifi Town. In general, the urban area has better access to health care services: there are 21 private clinics operating in Mtwapa, a number of pharmacies and approximately 104 shops selling drugs. In Ganze, there are two private clinics and approximately 30 shops. Mtwapa is also 18 km north of Mombasa, where there is a whole range of private and public health facilities. At the time the study took place, all government health facilities in Kenya charged user fees. Patients were required to pay 10 Kenyan Shillings (KES) for each type of drug prescribed and KES 20 for an injection. Other expenses differed depending on the type and severity of illness. In July 2004, user fees in dispensaries and health centres were reduced to a total of KES 10 and KES 20, respectively. This amount serves as a consultation fee for all patients but children under the age of five are exempted from payment.

Data collection and analysis

Maps indicating the location and landmarks of every homestead (rural) and house (urban) were drawn by hand to enable the selection of survey households1. Each homestead or house was allocated a unique number and a random selection conducted on the basis of a complete list. For each randomly selected homestead/house, one household was selected depending on the location within the compound. Fourteen trained fieldworkers surveyed 294 rural and 576 urban households.

In the questionnaire, we separately considered ‘chronic illnesses’ (illnesses reported to have lasted 3 months or more are defined chronic diseases, such as diabetes or high blood pressure (WHO 1980), ‘acute illnesses’ (illnesses experienced within the last 2 weeks and not chronic), and ‘hospitalizations’ (at least one overnight stay at any source of treatment) that took place in the year preceding the survey2. For each category of illness, data were collected on symptoms, perceived severity, name of illness, treatment responses, costs of treatment and coping strategies. The questionnaire was administered to the household head or spouse or, in his/her absence, another senior member of the household. A majority of respondents were females since males were often absent. For illnesses reported among adults and children, efforts were made to interview the ill person and the primary caretakers, respectively.

The survey was complemented by in-depth interviews with nine key informants (village elders, leaders of community groups, and members of the dispensary committee) and 18 focus group discussions (FGDs) in each setting. FGDs were organized to ensure that participants were of similar gender and age. They were aimed at contributing towards designing the survey questionnaire and at gathering more qualitative data around key areas of interest including decision-making behind treatment-seeking and coping strategies and implications of different coping strategies on the households’ ability to pay for health care.

Quantitative data were entered into FoxPro (version 6.0) and later transferred to STATA version 8.0 for analysis. Direct costs were classified as all cash payments arising due to the illness. These included cash spent on consultation, drugs, tests, hospital bed costs, special foods, transport charges to and from the facility, and daily living costs for accompanying members. Indirect costs were measured in terms of the number of days that the ill person and their caretakers were unable to conduct their activities. Income days lost to both the patient and the caretaker were valued in monetary terms using an average daily income estimated from the survey as KES 33 in the rural area and KES 145 in the urban area. Cost figures were expressed in KES3 and converted into monthly estimates to enable comparison between different illness categories. The costs of acute illnesses over the last 2 weeks were converted into annual costs and divided by 12 to arrive at a monthly estimate. Costs arising from hospitalizations are converted into monthly costs to enable comparison. This approach has been used elsewhere (Russell 2001, 2004), although it has been shown that illness costs are not smooth over time with important implications for household coping and impact (Russell 2001; Chuma et al. 2006).

Households were divided into SES categories using expenditure data. Expenditure data were preferred to income because expenditure is less subject to fluctuations, and collecting income data in settings where the majority of the population work in agriculture or in the informal sector is difficult (Deaton 1997). Expenditure data were converted into per capita estimates weighted for age using adult equivalence scales. Adult equivalence scales require that total monthly expenditure per household be divided by the household size but children are given a lower weight because of the differences in resource demands for different age groups. Inequities were measured using an equity ratio to show the degree of income related inequality in treatment-seeking behaviour and cost burdens. The equity ratio is used to compare the value of the lowest and highest quintile. It identifies the gap that requires to be addressed to achieve equity (Wagstaff et al. 1991; Ettling et al. 1994; Onwujekwe and Uzochukwu 2004; Onwujekwe 2005). A ratio of one signifies perfect equity, and a ratio greater than one indicates that the variable of interest occurs more among the poor than the least poor. A recognized limitation of the equity ratio is the failure to consider values within the middle quintiles.

Data from the urban and rural area were analysed separately to assess inequities within each setting. A separate analysis for urban and rural data was necessary because contextual differences, such as income generation activity influence household SES. Differences and similarities are then explored between the two settings. Chi-squared and t-tests are used to test differences in groups and means. All individual interviews and FGDs were tape-recorded and transcribed. The data gathered were analysed manually using content analysis (Fielding 1993). All transcripts were coded and the codes were later recorded in summary sheets to identify common themes and sub-themes.

Results

Socio-demographic and economic characteristics

A total of 2162 rural and 3125 urban residents were enumerated in the survey. Mean household size was higher in the rural than in the urban area (n = 7 vs. 5; P < 0.001). The majority of rural adults have no education (53.5%). Urban adults are better educated, with only 11.2% reporting no education and 28.0% having secondary level education. The pattern of asset ownership differed in both areas, reflecting the different economic and social environments. The main types of assets owned among rural households were land (88.4%) and livestock (91.5%), while urban households were more likely to own radios (76.8%), mobile phones (35.9%) and television sets (33.8%). Both populations were relatively young, with over 50% of individuals in both areas aged below 18 years. Rural households reported significantly lower mean monthly per capita expenditure than urban households (KES 992 and KES 4338, respectively; P < 0.001). However, the equity ratio indicates similar levels of inequities within each setting: least poor households in both settings reported a mean per capita expenditure about ten times higher than that of the poorest households.

Morbidity patterns

A total of 1093 acute illnesses and 831 chronic illnesses were reported. A majority of households in both areas reported at least one acute illness in any household member (78.6% rural, 61.8% urban; P < 0.001). Fewer households reported at least one chronic illness, particularly in the urban area (62.5% rural and 39.9% urban; P < 0.001). The distribution of self-reported illnesses by SES quintiles reveal that more illness episodes were reported among the poorest groups (equity ratio >1). Differences are only significant, however, for acute illnesses and hospitalizations in the urban area (P = 0.01).

Fever (typically called homa, mwili moto or malaria) and cold (mafua) were the main types of acute illnesses reported in both areas, while respiratory or breathing problems, often referred to as asthma or pumu, were the main chronic illnesses reported. Other common chronic conditions reported included ‘chest problems’ in the rural area, and ‘diabetes’ and ‘high blood pressure’ in the urban area. Acute illnesses were concentrated among children under 10 years in the rural area (43.5% of all acute illnesses) and among those aged 18–35 in the urban area (34.5%).

Treatment-seeking patterns

Most illnesses were treated using a single action or type of health care provider. Only 17.9% of acute illnesses and 7.9% of chronic illnesses reported among rural households were treated using more than one action/provider. A similar pattern was observed among urban households (18.5 and 7.5%, respectively). Self-treatment using drugs bought from shops or private pharmacies was the main source of treatment for acute illnesses (54.7% of all rural actions and 50.1% of all urban actions), and often the only one. Cases of self-treatment for chronic illnesses were much lower (36.4% rural and 39.0% urban actions) because most were not treated at all. Results from FGDs revealed that self-treatment was taken as a first action while people ‘waited to see’ if the symptoms would go away, preferring to seek treatment from health facilities if illnesses persisted:

‘You will see the symptoms of fever and buy him/her [ill child] some drugs…we use such drugs knowing that that may well be enough for him/her. That is what we do. If he/she uses the drugs for 2 days and has not yet recovered… we take the child to the hospital.’ (rural FGD)

People used formal health care facilities less often. In general, people used private clinics4 (10.3% rural and 24.4% urban actions) more than government facilities (8.6% rural and 10.6% urban actions) in both settings. Government facilities were the main sources of in-patient care in both settings (61.4% of all hospitalizations in rural and 55.4% in urban). Residents report concerns about private clinics including cost, training and profit motivation of private providers. However, private clinics were preferred over the government dispensary in both settings for reasons such as more speedy services in private clinics, low levels of trust in the services and staff at government facilities, and poor relationships and interpersonal handling of patients in government facilities:

‘In fact people are doubting the drugs at the government dispensary. They seem to be less effective than those in the private clinics. In private clinics you take the drugs and the doctor tells you to go back the following day. When people take drugs, are checked, and are told they are improving, they have faith in that provider.’ (FGD rural)

‘We run to private facilities not because we have too much money but because they are quick. Even if it means going for credit. At the government facilities you will end up losing your fare and you will not receive the much needed assistance. You will be asked to buy a card, injections and all types of things… What you need is life, but you will go to the government hospitals and leave with nothing.’ (FGD urban)

‘At public facilities, the nurse might not even touch your child. They just ask you how the child is without even looking at him/her…how can I know what type of fever my child is suffering from?’ (FGD urban)

Table 1 shows the proportion of illnesses that were not treated, the various reasons given for not seeking treatment and how this was distributed across SES. Among acute illnesses reported 2 weeks prior to the survey, more illnesses reported in the rural area were not treated compared with those reported in the urban area (20.1% and 9.3% were not treated, respectively; P < 0.001). A similar pattern was observed in chronic conditions, with rural households being more likely not to seek treatment than their urban counterparts (56.9 and 44.9%, respectively; P < 0.001). Lack of cash was the main factor that prevented people from seeking treatment for both acute and chronic illnesses in both settings. Other reasons given included: ‘illness not serious enough’; ‘no drugs available’; and ‘drugs are ineffective’. The proportion of people who reported cash as the main factor hindering treatment differed significantly across quintiles for chronic conditions in both settings (P < 0.001 rural and P = 0.01 urban). For acute illnesses, SES differences were only significant in the urban area. The equity ratio was higher for both categories of illness in the urban area (acute = 20.0, chronic = 4.2), implying wider inequities in treatment-seeking behaviour in the urban than in the rural population.

Table 1.   Proportion of illnesses not treated and reason for failing to take action
VariableRuralUrban
Acute (n = 513)Chronic (n = 455)Acute (n = 580)Chronic (n = 376)
  1. †Does not include illnesses where data on expenditure (thus SES) was missing.

  2. P-value compares differences across SES groups within each setting.

No (%) no treatment (TMT)103 (20.1)259 (56.9)54 (9.3)168 (44.9)
No (%) no TMT
 Cash61 (59.2)176 (68.0)33 (59.3)83 (49.4)
 Severity33 (31.1)45 (17.4)10 (18.5)43 (25.6)
 Drugs not effective2 (1.9)5 (1.9) 0 (0.0)  5 (3.0)
No cash by SES†
 1 (poorest)12 (21.1)51 (32.3)20 (64.5)25 (35.2)
 211 (19.3)28 (17.7) 2 (6.5)25 (35.2)
 310 (17.5)33 (20.9) 6 (19.4)13 (18.3)
 417 (29.8)29 (18.4) 2 (6.5)  2 (2.8)
 5 (least poor)7 (12.3)17 (10.8) 1 (3.2)  6 (8.5)
Equity ratio1.73.020.0  4.2
P-value‡0.06<0.001 0.03  0.01

The use of health care providers as the first, and usually only, action taken by households in response to illnesses is shown in Table 2. Equity ratios indicate that the poorer the households, the more likely they were to use shops, government dispensaries and herbs. Least poor households were more likely to use private clinics. However, significant differences in the type of action taken by SES were observed among individuals who used private clinics in both settings (P = 0.03 rural, 0.02 urban) and among those who used herbs in the rural area (P = 0.02). Among individuals who sought treatment for chronic illnesses in the month preceding the survey, actions did not differ significantly by SES in either setting.

Table 2.   Distribution of main types of first treatment-seeking action by household SES
Type of actionAcute (%)Chronic (%)
RuralUrbanRuralUrban
  1. P-value compares differences in use of each type of action across all SES groups.

  2. ‡Numbers too small to enable any comparison across SES groups.

Shops/pharmacy(n = 260)(n = 283)(n = 46)(n = 63)
 148 (19)66 (23)13 (28)12 (19)
 247 (18)57 (20) 6 (13)14 (22)
 367 (26)68 (24)13 (28)15 (24)
 457 (22)47 (17) 8 (17)12 (19)
 541 (16)45 (16) 6 (13)10 (16)
Equity ratio 1.2 1.5 2.2 1.2
P-value† 0.129 0.698 0.31 0.42
Dispensary(n = 26)(n = 47)(n = 9)(n = 14)
 1 7 (27)15 (32) 0 (0) 7 (50)
 2 4 (15) 9 (19) 2 (22) 3 (21)
 3 6 (23) 8 (17) 3 (33) 1 (7)
 4 4 (15)10 (21) 3 (33) 1 (7)
 5 5 (19) 5 (11) 1 (11) 2 (14)
Equity ratio 1.4 3.0 0 3.5
P-value 0.85 0.38 0.46 0.11
Private clinics(n = 28)(n = 71)(n = 33)(n = 45)
 1 1 (4)11 (16) 4 (12) 6 (13)
 2 9 (35) 8 (11) 8 (24)11 (24)
 3 8 (29)17 (24) 8 (24) 6 (13)
 4 3 (11)21 (30) 6 (18)12 (27)
 5 7 (25)14 (20) 7 (21)10 (22)
Equity ratio 0.1 0.8 0.6 0.6
P-value 0.03 0.02 0.51 0.45
Herbs(n = 30)(n = 18)(n = 0)(n = 1)‡
 111 (37) 5 (28)  
 2 7 (23) 7 (39)  
 3 7 (23) 2 (11)  
 4 4 (13) 3 (17)  
 5 1 (3) 1 (6)  
Equity ratio 11 5  
P-value 0.02 0.18  
Government hospital(n = 2)‡(n = 8)‡(n = 8)‡(n = 12)‡
Private hospital(n = 1)‡(n = 3)‡(n = 2)‡(n = 7)‡

Health care expenditure and cost burdens as proportion of household income

Table 3 shows mean direct cost burdens for each category of illness and how costs differ by SES. Cost levels differed by type of action taken and by setting. Shops were the cheapest source of treatment in both settings for both acute and chronic conditions, while a visit to a private clinic was the most expensive. The mean costs of a visit to the shops and private clinics were significantly lower in the rural than in the urban area (P < 0.001). Costs of visiting a government dispensary were similar in the two settings (P = 0.25). The mean costs incurred for treating chronic conditions were significantly higher than those of acute illnesses in the urban area (P < 0.001), but in the rural area, mean costs for acute and chronic illnesses were similar.

Table 3.   Mean direct costs for each category of illness
 AcuteChronicHospitalization
RuralUrbanRuralUrbanRuralUrban
  1. †Only one action taken at the pharmacy in the rural area.

  2. ‡Does not apply

Mean cost per visit to a provider in KES
 Shops2937295416633487
 PharmacyNA†1798036718118402
 Dispensary167124128133NANA
 Private clinic3936163901245  
 Government hospital3938565971090  
 Private hospital500336918002840  
 Healers5340225175  
Mean cost per household in absolute terms (KES)330.21125194.2609.1157714
Mean monthly costs per household as % of monthly expenditure9.98.65.05.75.85.2
Mean monthly costs per household as % of monthly expenditure by SES
 115.011.89.611.813.87.9
 214.29.55.03.66.87.2
 37.66.83.85.53.23.5
 47.67.52.73.94.74.5
 54.16.03.02.81.43.8
Equity ratio3.62.03.14.29.92.1

When mean cost burdens were considered at a household level, irrespective of the types of action taken, urban households spent significantly higher amounts of money than their rural counterparts for all categories of illnesses (P < 0.001). However, when costs are expressed as proportion of monthly expenditure, there are no significant differences. The poorest households in both settings incurred the highest cost burdens for all categories of illnesses. For example, households in the lowest SES group spent over 10% of their total expenditure on acute illnesses while least poor households spent as little as 4.1%.

To show the proportion of households incurring potentially catastrophic burdens, costs incurred by each household for acute, chronic and hospitalizations were summed to arrive at a total monthly direct cost estimate. The total mean monthly direct cost burdens for households reporting at least one illness in the rural area was 12.8% of monthly expenditure, compared with 10.7% for urban households (P = 0.25). The distribution of total monthly direct cost burdens across households reporting illness within each area is indicated in Table 4. A good proportion of these households reported direct cost burdens below 5% of their monthly expenditure in both areas (51.5% rural and 57.2% urban). However, 31.1% of rural households and 28.1% of urban households facing illnesses incurred potentially catastrophic cost burdens. Among the households recording costs ≥10%, almost half fell in the two lowest SES groups in both settings (47.2% rural and 49.5% urban). As a result of different methodological approaches used to estimate costs and catastrophic spending, it is difficult to compare these findings with other studies, but the findings of studies that used similar methodology fell within the range of around 20% (Russell 2001, 2004).

Table 4.   Distribution of direct costs across households reporting illness
SES groupMean monthly direct cost burdens as % of monthly expenditure
<55–<1010–<15≥15Total
  1. †Does not include households whose data on expenditure and/or household size was missing. If this data were to be included, the total number of households = 258 rural and 437 urban.

Rural
 1191221750
 223931247
 32849849
 42575744
 52485441
Total119 (51.5)40 (17.3)24 (10.4)48 (20.7)231 (100)†
Urban
 139962478
 2441051776
 3441391076
 4401081270
 5471331174
Total214 (57.2)55 (14.7)31 (8.3)74 (19.8)374 (100)

Ability to conduct activities during the period of illness

Individuals reporting an illness were asked if they were able to conduct their income and non-income activities during the period of illness, or whether children were able to go to school. Similar questions were asked for caretakers. Significant differences were observed between settings in the number of ill adults reporting loss of income days. Rural adults reporting an acute illness were more likely to lose at least one income day (50.8%) as compared with 27.9% of ill adults in the urban area (P < 0.001). A similar pattern was observed among ill adults reporting a loss of non-income days (46.4% of ill adult individuals in rural and 36.7% in urban; P = 0.03). Mean number of income days lost due to an acute illness was 5.5 in the rural area and 5.7 in the urban area, while the mean number of non-income days lost were 4.9 and 5.6, respectively. The mean number of school-going days lost per illness episode amounted to 3.4 in the rural area and 4.6 in the urban area. When lost days are valued in monetary terms for sick individuals and their caretakers, urban households incurred significantly higher mean monthly indirect costs than rural households (KES 496 in rural and KES 2156 urban; P < 0.001). It was not possible to estimate the number of days lost due to chronic illnesses because some chronic conditions affect people's lives for many days, with some leading to permanent incapacitation.

Coping with the costs of illness

Individuals who sought treatment were asked where the money to pay for treatment came from. The results (Table 5) show that cash savings were usually available in the house to treat acute and chronic conditions, especially for drugs from shops and pharmacies. Cash savings were usually small amounts put aside to meet daily requirements, and the use of this money was often reported as a ‘dissaving’ (i.e. using money saved for other basic items such as food). In cases of cash shortages, strategies adopted were similar in both settings and across the types of treatment sought. The most frequently reported strategy was borrowing from relatives and friends. Other less common strategies included gifts, sale of labour on farms, selling assets and receiving treatment on credit. Similar patterns were observed for hospitalizations.

Table 5.   Sources of money to meet costs of main types of treatment
Type of treatmentAcuteChronic
Rural (%)Urban (%)Rural (%)Urban (%)
  1. †Total adds up to >100 because some actions were financed using more than one coping strategy.

Shops/pharmacy(n = 251)(n = 311)(n = 50)(n = 76)
 Cash savings188 (75)258 (83)35 (70)64 (84)
 Borrowing29 (12)27 (9)7 (14)6 (8)
 Gifts20 (8)11 (4)5 (10)7 (10)
 Casual labour16 (6)0 (0)4 (8)0 (0)
 Credit4 (2)4 (1)2 (4)0 (0)
 Sell assets2 (1)4 (1)0 (0)1 (1)
 Other10 (4)15 (5)1 (2)1 (1)
Government dispensary(n = 37)(n = 67)(n = 11)(n = 15)
 Cash savings18 (49)51 (76)6 (55)11 (73)
 Borrowing9 (24)15 (22)4 (37)1 (7)
 Gifts7 (19)5 (8)1 (9)0 (0)
 Casual labour3 (8)1 (2)0 (0)0 (0)
 Credit2 (5)1 (2)0 (0)0 (0)
 Sell assets1 (3)0 (0)0 (0)1 (7)
 Other0 (0)2 (3)1 (9)2 (13)
Private clinics(n = 49)(n = 151)(n = 37)(n = 46)
 Cash savings22 (45)102 (68)22 (60)35 (76.1)
 Borrowing12 (25)17 (11)5 (14)9 (19.6)
 Gifts2 (4)9 (6)7 (19)4 (8.7)
 Casual labour4 (8)0 (0)2 (5.4)0 (0.0)
 Credit6 (12)18 (12)1 (2.7)7 (15.2)
 Assets2 (4)3 (2)6 (16)0 (0.0)
 Other3 (6)15 (10)3 (8)4 (8.7)

Quantitative data revealed that borrowing was preferred to other strategies because it was the easiest and quickest way to raise money. Other strategies such as sale of assets required time to get a buyer, possibly leading to poor prices being fetched or delays in seeking treatment. Borrowing and lending was primarily on the basis of reciprocity and trust that repayment would be made:

‘One can sell chickens…one can even sell a goat. But this is difficult because you are sick, so who will hawk the goat for you? And there will be nobody to buy it…that is why you will be forced to borrow. It is the easiest.’ (FGD rural)

‘When in need of help, you have to go to someone you know, a son-in-law, a brother, a neighbour or any other relative. It is difficult to get help from someone you do not know. There is no trust.’ (FGD rural)

‘When you get credit and you repay it, people will trust you…but if not, you get no help.’ (FGD urban)

In some cases, the type of treatment sought was in itself a coping strategy:

‘One can have KES 100 to buy food and also to cover the hospital costs. That will not be enough to cover hospital costs so instead I will take two shillings and buy aspirins so that my child gets some relief and can have some porridge. There is no money to go to hospital.’ (FGD urban)

Regarding coping with time costs, ill individuals who were unable to conduct their income generating activities were asked whether anything was done to avoid or minimize income loss. In the rural area, 27 (29.7%) individuals adopted a coping strategy as compared with 14 (16.1%) of their urban counterparts (P = 0.03). The main types of time management strategies adopted were receiving help from friends and relatives, substituting labour within the household, and hiring people to conduct their activities.

Discussion

In this paper, we set out to examine inequities in reported illnesses, treatment-seeking behaviour, cost burdens and how these variables differ by SES and between rural and urban settings. Key findings were significantly higher levels of reported chronic and acute illness in the rural as compared with the urban setting, differences in treatment-seeking patterns between and within the rural and urban setting, and regressive cost burdens in both areas. Here we discuss these findings in more detail, comparing them to the wider literature and commenting on their implications.

Levels of reported illness and treatment-seeking patterns

Chronic and acute illnesses were reported significantly more often among rural households as compared with urban households, while hospital admissions were reported more often by urban households. Various patterns have been reported in the literature5: for chronic conditions, levels of self-reported illness have generally been higher in urban as compared with rural settings (McLarty et al. 1989; Ceesay et al. 1997); for acute illnesses, findings have been mixed, with some studies documenting higher rates of reported illness in urban areas (Balarajan et al. 1987; Thompson et al. 2003) and others the opposite (Kloos et al. 1987; Gupta 2003). The underlying reasons for rural households reporting more illnesses might be due to socio-economic and population differences between the two settings. Rural households had significantly lower expenditure (income) than their urban counterparts, suggesting that the urban poor are better off than the rural poor. If the data were pooled, the majority of rural households would have fallen into the lowest SES quintile. In addition, urban households were better educated and were relatively wealthy in terms of the kind of assets owned. Other reasons might be differences in population densities and disease transmission patterns. It was beyond the scope of our study to explore this. Regarding reported chronic illnesses, urban households might have sought prompt effective treatment more often, reducing the chances of prolonged illnesses. However, caution should be taken when interpreting chronic illness data since definitions were not based on clinical diagnosis.

There were no significant differences in levels of reported illnesses by SES for all categories of illness in the rural area. However, significant differences by SES were observed among households reporting acute illnesses and in the number of hospitalizations in the urban area. Other studies have observed that reported number of illnesses does not differ significantly by SES and that the differences in cost by SES are likely to relate more to the type and timing of the actions taken in response to an illness (Onwujekwe et al. 2003; Filmer 2005; Onwujekwe & Uzochukwu 2005; Worrall et al. 2005). Reported acute illnesses differed significantly by SES within the urban area. This might reflect greater inequities in terms of wealth and well-being in this area. High levels of inequity are increasingly recognized within urban settings, with some writers concerned that the ‘urban poor’ can be worse off in many respects as compared with the rural poor (Baharoglu and Kessides 2000; Harpham & Molyneux 2001).

Shops were the main source of treatment for acute and chronic illnesses in both areas, and for all SES groups. Self-treatment as a main response to illnesses has been observed repeatedly all over Sub-Saharan Africa and in the district of study (Foster 1995; Akenso-Okyere & Dzator 1999; Molyneux et al. 1999; Nyamongo 2002; Onwujekwe & Uzochukwu 2005; Chuma et al. 2006). For government dispensaries and private clinics, there was significantly higher use among the least poor in both settings, suggesting better ability to pay. Urban residents used private facilities more often than rural residents, a factor that can be attributed to higher income levels, better resource endowments and greater proximity to private clinics in the urban area. Similar patterns of treatment-seeking have been observed in other parts of Africa and Asia (Russell 2001; Sudha et al. 2003; Zere & McIntyre 2003; Onwujekwe & Uzochukwu 2004; Onwujekwe 2005Raso et al. 2005).

An important influence over treatment-seeking behaviour is household ability to pay for health care. The number of people failing to seek treatment due to cash shortages was particularly high in the rural area. About half of chronic illnesses reported in the rural and urban area did not receive any regular treatment. This suggests that households may choose not to seek health care rather than cope with impoverishment. The findings were in contrast with those from a low-income urban Sri Lanka where 72.9% of reported chronic conditions received regular treatment (Russell 2001, 2004). Russell associates the high levels of regular treatment with a well-functioning free public health system that was trusted by most people, especially regarding chronic conditions, paediatric care and other illnesses requiring potentially large amounts of money. In the Kenyan study, government health services are relatively cheap, but people often prefer private providers because of perceived higher quality of services and a lack of trust in government services. It has been shown elsewhere that poor households can bypass even free government services where these are perceived to be of poor quality and seek services from private clinics (Bitran 1989, 1995). In doing so, they incur costs that could have been avoided. Our findings indicate the potential the public health system has to protect poor households from high cost burdens. Current charging systems at government facilities, perceived weaknesses in quality of care, and the relatively ‘high’ costs of private providers apparently deter many people from seeking any formal care, particularly for chronic conditions.

Cost burdens and coping strategies

Cost burdens were regressive among the urban and rural households. The poorest households in both settings incurred significantly higher costs as a proportion of their monthly expenditure for all categories of illnesses. Other studies have reported significantly higher costs in rural areas as compared with urban areas (Thompson et al. 2003; Onwujekwe & Uzochukwu 2005), but we found no significant differences in mean cost burdens. This is explained by greater use of more expensive facilities in the urban setting and the higher costs charged in town for similar services compared to the rural setting.

About half of the households in both settings reported mean total direct costs below 5% of their monthly expenditure. Although this suggests low cost burdens for many households, we found that treatment-seeking behaviour itself was often a coping strategy: households often opted for cheaper alternatives in order to manage potential cost burdens. Overall around one-third of households in both settings incurred potentially catastrophic costs, defined here as above 10% of their expenditure. Almost half of these households were in the two lowest SES groups in both settings, implying that the poorest households are at a relatively high risk of being impoverished by illness. Of interest would be to design a study examining to what extent these expenses are necessary or, in fact, represent wasted resources. Furthermore, whether or not costs incurred lead to impoverishment will depend on other factors such as household socio-economic and demographic characteristics, asset base, and ability to cope with costs through using less risky strategies. The timing of health care costs, how costs are managed within household budgets, and the functioning of the health system are also important (Russell 2004; Van Damme et al. 2004; McIntyre et al. 2006). In poor communities, even relatively small amounts of money have been found to cause financial hardship and to contribute to accumulating debts, which can become catastrophic (Van Damme et al. 2004). Low cost burdens in our study setting should not be taken to indicate a community that is managing or less needy but rather to indicate the cost prevention strategies among people struggling despite illnesses. Whether or not these cost burdens actually lead to impoverishment is explored elsewhere (see for example Chuma et al. 2006).

Regarding indirect costs, rural households were more likely to report loss of income or non-income days as compared with urban residents. However, urban residents reported a higher mean number of lost days. This pattern might indicate that rural residents’ illnesses were relatively minor, that they were more able to draw on others to substitute their labour (for a brief period), or that they ‘could not afford’ to take more time off from work. A combination of these factors is likely and will be explored in subsequent analyses.

A good proportion of households did not have cash readily available to pay for treatment in both settings. This proportion differed depending on where the treatment was sought. Individuals who self-treated often had enough cash to meet arising costs, but those who consulted a formal health care provider were more likely to mobilize additional resources. Borrowing and gifts were the main types of coping strategies adopted in both settings. The important role of borrowing from informal sources (friends, neighbours, relatives) in meeting the costs of illness has been identified elsewhere (Sauerborn et al. 1996a; Russell 2001, 2004). Borrowing is much more readily available to households who are more likely to have relatively well-off friends and who are less likely to defy or delay repayment (Lucas and Nuwagaba 1999; Chuma et al. 2006). Although borrowing is considered a low risk strategy, accumulated debts can have negative implications for household economic and social status, especially if debts are not repaid on time. In other settings, the sale of assets, particularly livestock or jewellery, has been reported as the main strategy to raise money for health care (Sauerborn et al. 1996b; Mock et al. 2001). In our setting, the sale of assets was rarely reported because the market was limited and the liquidity of livestock was low at the time of our work. These contrasting findings suggest that the nature and range of coping strategies vary by context and over time.

Conclusions

The Kenyan government is considering ways of protecting poor households from catastrophic health care costs. For example, user fees in dispensaries and health centers were reduced in 2004, and debates on a social health insurance programme are ongoing. Our study indicates that efforts to protect the poor are essential, and that a multi-sectoral approach to improve household SES is required if this is to have any hope of being achieved. Eliminating all user fees in government facilities (including hospitals) would be a positive move towards protecting households from high costs of illness. Such an approach, however, requires additional resources to meet the expected rise in demand and to ensure that quality of care is maintained (Chuma et al. 2006). Changes have to be carefully planned and implemented to avoid negative implications (Gilson & McIntyre 2005). Where resources to support abolishing user fees are inadequate, geographical targeting is recommended to remove the consultation fee in remote rural areas such as Ganze, where communities are predominantly poor.

Other alternatives might be to promote community prepayment schemes and to strengthen the public health system in order to attract people and prevent them from incurring higher costs at private providers, or failing to seek treatment altogether. Russell (2004) argues that even if health services are improved, they cannot protect households from all illness costs. He recommends that health policy research and debates should be broadened to include interventions beyond the health sector; interventions that protect the poor and increase their incomes. We support this argument and recommend interventions such as supporting micro-credit schemes that encourage people to save a little amount of money each week or month. Supporting asset accumulations through, for example, strengthening markets for livestock products or through food for work or other safety net programmes might also be considered. Such interventions are essential if out-of-pocket payment remains the main source of health care financing and if the poor are to participate in any risk sharing mechanism such as social health insurance.

The challenge with interventions within and beyond the health sector remains as to how to ensure that these pro-poor policies actually benefit the poor. Evidence shows that the rich tend to benefit from policies at the expense of the poor (Abdulla et al. 2001; Onwujekwe 2003; Schellenberg 2003). There is need for further research on why this is the case. Such studies should demonstrate the various factors that act as barriers for the poor, including non-financial factors, and identify measures that could be put in place when designing policies and interventions that target the poor.

Acknowledgements

This study was made possible through a fellowship grant awarded to CM by The Wellcome Trust, UK, and that was supported by the Kenya Medical Research Institute. We are grateful to the study communities and to the personnel team that worked tirelessly to make fieldwork successful. This paper is published with the permission of the Director of KEMRI.

Ancillary