Determinants of delay in care-seeking for febrile children in eastern Uganda
Corresponding Author Elizeus Rutebemberwa, Makerere University School of Public Health, P.O. Box 7072, Kampala, Uganda. E-mail: firstname.lastname@example.org; email@example.com
Objective To explore factors associated with delay in seeking treatment outside the home for febrile children under five.
Methods Using a pre-tested structured questionnaire, all 9176 children below 5 years in Iganga–Mayuge Demographic Surveillance Site were enumerated. Caretakers of children who had been ill within the previous 2 weeks were asked about presenting symptoms, type of home treatment used, timing of seeking treatment and distance to provider. Children who sought care latest after one night were compared with those who sought care later.
Results Those likely to delay came from the lowest socio-economic quintile (OR 1.45; 95% CI 1.06–1.97) or had presented with pallor (OR 1.58; 95% CI 1.10–2.25). Children less likely to delay had gone to drug shops (OR 0.70; 95% CI 0.59–0.84) or community medicine distributors (CMDs) (OR 0.33; 95% CI 0.15–0.74), had presented with fast breathing (OR 0.75; 95% CI 0.60–0.87), used tepid sponging at home (OR 0.43; 95% CI 0.27–0.68), or perceived the distance to the provider to be short (OR 0.72; 95% CI 0.60–0.87).
Conclusion Even in the presence of ‘free services’, poverty is associated with delay to seek care. Drug shops and CMDs may complement government efforts to deliver timely treatment. Health workers need to sensitize caretakers to take children for care promptly. Methods to elucidate time in population-surveys in African settings need to be evaluated.
Objectif: Explorer les facteurs associés au retard dans le recours au traitement hors du domicile pour les enfants fébriles de moins de cinq ans.
Méthodes: En utilisant un questionnaire structuré pré-testé, tous les 9176 enfants de moins de cinq ans dans le Site de Surveillance Démographique de Iganga-Ma-Yuge ont étéénumérés. Les soignants d’enfants qui ont été malades au cours des deux semaines précédentes ont été interviewés au sujet des symptômes, du type de traitement utiliséà domicile, du délai au recours à un traitement et de la distance au pourvoyeur de soins. Les enfants qui ont eu recours à des soins au plus tard après une nuit ont été comparés à ceux qui ont eu recours à des soins plus tard.
Résultats: Ceux susceptibles de recourir tardivement aux soins provenaient du quintile socioéconomique le plus bas (OR: 1,45; IC95%: 1,06 - 1,97) ou avaient présenté une pâleur (OR: 1,58; IC95%: 1,10 - 2,25). Les enfants moins susceptibles de recourir tardivement aux soins avaient eu recours à des pharmacies (OR: 0,70; IC95%: 0,59 - 0,84) ou à des pourvoyeurs de médecine de communauté (OR: 0,33; IC95%: 0,15 - 0,74), avait présenté une respiration accélérée (OR: 0,75; IC95%: 0,60 - 0,87), utilisé un épongeage tiède à domicile (OR: 0,43; IC95%: 0,27 - 0,68) ou perçu la distance au prestataire comme étant courte (OR: 0,72; IC95%: 0,60 - 0,87).
Conclusion: Même en présence de «services gratuits», la pauvreté est associée au retard à faire recours aux soins. Les pharmacies et pourvoyeurs de médecine de communauté pourraient complémenter les efforts du gouvernement à fournir des traitements à temps. Les agents de santé devraient sensibiliser les soignants à apporter rapidement les enfants pour des soins. Des méthodes permettant d’élucider le temps dans les études de population dans les cadres Africains devraient être évaluées.
Objetivo: Explorar los factores asociados con el retraso en la búsqueda de tratamiento fuera del hogar para niños menores de cinco años con fiebre.
Métodos: Utilizando un cuestionario estructurado previamente probado, se enumeraron todos los 9176 niños menores de cinco años del enclave de Iganga – Mayuge, bajo seguimiento demográfico. Los cuidadores de los niños que habían estado enfermos dentro de las dos semanas anteriores fueron interrogados acerca de los síntomas presentados, el tipo de tratamiento utilizado en el hogar, el tiempo que se tardó en buscar tratamiento y la distancia hasta el proveedor más cercano. Los niños que buscaron cuidados a más tardar después de una noche, fueron comparados con aquellos que tardaron aún más en buscar ayuda.
Resultados: Aquellos con mayor probabilidad de retraso provenían del quintil socioeconómico más bajo (OR 1.45; 95% CI 1.06 - 1.97) o se habían presentado con palidez (OR 1.58; 95% CI 1.10 - 2.25). Los niños con menor probabilidad de retraso habían acudido a farmacias (OR 0.70; 95% CI 0.59 - 0.84) o distribuidores comunitarios de medicamentos (OR 0.33; 95% CI 0.15 - 0.74), se habían presentado con respiración acelerada (OR 0.75; 95% CI 0.60 - 0.87), habían utilizado en casa baños con agua tibia para bajar la fiebre (OR 0.43; 95% CI 0.27 - 0.68), o percibían la distancia hasta el proveedor como corta (OR 0.72; 95% CI 0.60 - 0.87).
Conclusión: Aún en la presencia de “servicios gratuitos”, la pobreza está asociada con un retardo en la búsqueda de cuidados. Las farmacias y los distribuidores comunitarios de medicinas pueden complementar los esfuerzos gubernamentales a la hora de entregar a tiempo el tratamiento. Los trabajadores sanitarios deben sensibilizar a los cuidadores para que lleven a los niños a tiempo a los centros sanitarios. Es necesario evaluar nuevos métodos para calcular el tiempo en los estudios poblacionales en África.
Many children under five in low income countries die of febrile illnesses like malaria and pneumonia (Black et al. 2003; WHO/UNICEF 2004; Rowe et al. 2006). In Uganda 15% of paediatric deaths in the National Referral hospital in 1998 were due to pneumonia alone (Government of Uganda 1998) and in 2001, 44.4% of children younger than 5 years presenting in outpatient departments had malaria (Government of Uganda 2001). Malaria could have been overdiagnosed, as its symptoms overlap with those of other illnesses such as pneumonia (Kallander et al. 2004). However, this mortality and morbidity data indicates the burden of disease.
To reduce mortality from febrile illnesses, sick children not only need to get efficacious and appropriate drugs, but also need to get them in time. At the Roll Back Malaria summit in Abuja in 2000, the heads of state committed themselves to ensure that by 2005, 60% of those suffering from malaria should have correct, affordable and appropriate treatment within 24 h of symptom onset (WHO/CDS/RBM 2000). However, a study in Kenya found that only 2.3% of the children accessed first line treatment for uncomplicated malaria within 24 h (Amin et al. 2003) and in Ghana only 11% of the children suspected of having malaria received prompt treatment (Ahorlu et al. 2006). Late care seeking contributed to mortality from acute respiratory infections in Uganda (Kallander et al. 2008) and from malaria in Tanzania (de Savigny et al. 2004).
The sources of health care in Uganda include Government and Non Governmental Organization (NGO) health facilities, community medicine distributors (CMDs), traditional healers, drug shops and private clinics. With an aim to increase access to health care for people, the government removed user fees in all government health units (Burnham et al. 2004; Nabyonga et al. 2005; Xu et al. 2006) and started the Home Based Management of Fever (HBMF) in 2002 using CMDs to distribute antimalarials for free in the villages (Government of Uganda 2002). However, even when drugs are free at the community level, people do not have timely access to them (Nsungwa-Sabiiti et al. 2007).
Various studies have demonstrated that caretakers’ choice for seeking providers outside the home is associated with perceived aetiology (Mwenesi et al. 1995), perceived severity of diseases (Taffa & Chepngeno 2005), perceived quality of care (Leonard 2002), cultural and traditional beliefs (Nuwaha 2002; Nsungwa-Sabiiti et al. 2004), knowledge and symptoms of the illness (Stekelenburg et al. 2002), home treatment (Agyepong & Manderson 1994; Kallander et al. 2008) or socio-economic status (Schellenberg et al. 2003; Olaogun et al. 2006; Deressa et al. 2007). However, the relative contributions of these factors specifically to delayed care seeking are much less understood. Our study focuses on determinants of delay to get treatment from a health care provider in a rural area with relatively good access to public health facilities, private sector and CMDs. It responds to the question why caretakers delay to seek care outside the home when government health facilities and CMDs supposedly provide free care.
Study area and study population
This cross sectional study was done in the Iganga–Mayuge Demographic Surveillance Site (DSS) in eastern Uganda between October and December 2006. The DSS was established in August 2004 and became functional in 2005. It is located at the border of the two districts of Iganga and Mayuge, about 115 km from Kampala. The DSS has a registered population of about 67 000 in about 13 000 households in 65 villages. Every 4 months, field assistants visit every household to update vital events of births, deaths and migrations and to identify pregnancies. The household and community structures have been mapped using the Global Positioning System (GPS). About 90% of the population in the DSS is rural and 10% live in peri-urban areas. Most of the inhabitants practice peasant agriculture. The major tribe are the Basoga, one of the Bantu ethnic groups. Malaria is endemic with two main transmission periods in the year; one in March and one in September. The DSS area has 13 health facilities: 10 are government facilities including the district hospital, and 3 NGO facilities. The area is also served by 122 drug shops and private clinics mainly located in small trading centres and in Iganga town. Some CMDs who provide antimalarial treatment are scattered in the villages but their activity levels are generally low.
Data was collected by the field assistants who work in the DSS during the regular update round. The field assistants have a minimum of secondary education and are trained in data collection techniques. For the purpose of this study they were trained in interview techniques, research ethics and confidentiality. The tool was translated into the local language Lusoga, pre-tested in the neighbouring area outside the DSS and adjusted when necessary. The field assistants were supervised daily by the first author and the DSS staff. During the census round, all households with children under five were exposed to a separate pre-tested structured questionnaire in addition to the routine DSS questionnaires (both available from the first author). The caretakers of children under five were asked whether the children had been sick in the previous 2 weeks. For all those who were reported to have been sick, further questions focused on the presenting symptoms, the treatment given at home immediately after the child was seen to be ill, whether the child was taken outside the home for treatment and the timing and source of care sought. At the end of the day, the forms were reviewed by the first author and the DSS staff and corrections made.
After every week of data collection the forms were entered in the computer by trained data entrants using FoxPro, data cleaned, and then linked with the DSS database. Data was then transferred to stata version 10 for analysis. For socio-economic status, we used the same group of context specific assets used by Uganda Bureau of Statistics. These items were screened for relevance and reliability testing done using Cronbach’s α (De Vellis 1991; Bernard 1995). The final list included number of sleeping rooms, type of floor material, type of roof material, wall material, fuel used for cooking and source of light. Other variables were the households having or not having the following items: a radio, a sewing machine, an electric flat iron, type of bed, charcoal flat iron, a bed net, kerosene lamp, kerosene stove, car, tea table, refrigerator, television set, sound stereo, telephone, mattress, wheel barrow, cell phone and camera. These gave a Cronbach’s α of 0.848. A principal components analysis was performed and the first principal component was scored to create an asset index that was used to group all households in the DSS into wealth quintiles (Filmer & Pritchett 2001). The socio-economic status of the households in this study was established based on these wealth quintiles.
Counting a 24-h period where the majority of the population does not use watches was a challenge. Instead we used nights as indicators of time. The children who were taken to a health care provider on the same day of fever onset or the next day after only one night were considered to have received treatment within 24 h. This definition has been used previously to assess prompt use of artemesinin-based combination therapy (Ajayi et al. 2008).
Univariate analysis was done for socio-demographic characteristics of the household heads, caretakers and children. In bivariate analysis, the children who were taken outside the home after more than one night (delay) were compared with those who had been taken on the same day or after only one night (non-delay). This was done with respect to socio-economic status of the household, the reported symptoms, the reported home treatment and the perceived distance to the provider seen. The variables that had a P-value of < 0.10 in bivariate analysis were put in the multivariate logistic model in order to identify the significant predictors for delay. Some households had more than one child under five and adjustment for clustering was performed using stata’s svy feature. The unit of analysis was the sick child episode.
The study received ethical approval from the institutional review board of Makerere University School of Public Health and the Uganda National Council of Science and Technology. Permission was granted by the management of the Iganga–Mayuge Demographic Surveillance Site and the village local council chairpersons. Verbal consent was obtained from each caretaker.
A total of 9176 children under five were enumerated. Of those 4991 (54.4%) had presented with fever in the previous 2 weeks of whom 2389 (47.9%) had sought care outside the home. Of these, a total of 1319 (55.2%) were taken for care outside the home after two or more nights from the time the caretakers noticed that the child was febrile while 1070 (44.8%) had been taken for outside care after only one night of sickness or before nightfall. A comparison of the two groups with respect to sex, age and highest education attained by the household head and the caretaker, as well as the sex and age of the child was done. There was no significant difference between the two groups (Table 1).
Table 1. Socio-demographic characteristics of the household heads, caretakers and febrile children seeking care within 24 h of illness onset or later
|Sex of the household head||n = 1319 (%)||n = 1070 (%)|
|Male||1200 (91.0)||946 (88.4)|
|Female||119 (9.0)||124 (11.6)|
|Age of the household head||n = 1319 (%)||n = 1070 (%)|
|≤29||251 (19.0)||170 (15.9)|
|30–39||531 (40.3)||462 (43.2)|
|40–49||316 (24.0)||261 (24.4)|
|50–59||122 (9.2)||96 (9.0)|
|≥60||99 (7.5)||81 (7.6)|
|Education of household head||n = 1256 (%)||n = 1013 (%)|
|No education||150 (12.0)||111 (11.0)|
|Primary 1–4||266 (21.2)||199 (19.6)|
|Primary 5–7||511 (40.7)||406 (40.1)|
|O-level secondary||292 (23.3)||263 (26.0)|
|A-level secondary||36 (2.9)||34 (3.4)|
|Sex of the caretaker||n = 1319 (%)||n = 1070 (%)|
|Male||263 (19.9)||202 (18.9)|
|Female||1056 (80.1)||868 (81.1)|
|Age of the caretaker (years)||n = 1319 (%)||n = 1070 (%)|
|<20||66 (5.0)||51 (4.8)|
|20–29||527 (40.0)||404 (37.8)|
|30–39||449 (34.0)||389 (36.4)|
|40–49||181 (13.7)||146 (13.6)|
|50–59||55 (4.2)||47 (4.4)|
|≥60||41 (3.1)||33 (3.1)|
|Education of caretaker||n = 1225 (%)||n = 980 (%)|
|No education||199 (16.2)||151 (15.4)|
|Primary 1–4||282 (23.0)||218 (22.2)|
|Primary 5–7||516 (42.1)||402 (41.0)|
|O-level secondary||219 (17.9)||196 (20.0)|
|A level secondary||9 (0.7)||13 (1.3)|
|Sex of the child||n = 1319 (%)||n = 1070 (%)|
|Male||649 (49.2)||532 (49.7)|
|Female||670 (50.8)||538 (50.3)|
|Child age (months)||n = 1319 (%)||n = 1070 (%)|
|<2||15 (1.1)||23 (2.1)|
|2–11||180 (13.6)||149 (13.9)|
|12–23||273 (20.7)||182 (17.0)|
|24–35||227 (17.2)||199 (18.6)|
|36–47||234 (17.7)||181 (16.9)|
|48–59||390 (29.6)||336 (31.4)|
Of those caretakers who had sought care outside the home, 1282 (54.2%) had gone to health facilities; 1027 (43.3%) to drug shops; 30 (1.3%) to CMDs; 9 (0.4%) to neighbours and 18 (0.8%) to other providers such as traditional healers or spiritualists. Those who had gone to CMDs (OR 0.29; 95% CI 0.13–0.63) and drug shops (OR 0.68; 95% CI 0.57–0.80) delayed less often than those who went to health facilities (Table 2). The caretakers who perceived the distance between their homesteads and the provider they contacted to be <1 km more often did not delay compared to those who perceived it to be further away (OR 0.72; 95% CI 0.60–0.87).
Table 2. Determinants of seeking care for febrile children outside the home within 24 h
|Type of provider||n = 1313 (%)||n = 1053 (%) REF|| || || || |
|Health facility||768 (58.5)||514 (48.8)||1.000|| || || |
|CMD||9 (0.7)||21 (2.0)||0.29 (0.13–0.63)||0.001*||0.33 (0.15-0.74)||0.007*|
|Neighbour||6 (0.5)||3 (0.3)||1.34 (0.33–5.38)||0.680||2.72 (0.52-14.1)||0.234|
|Drug Shops||517 (39.4)||510 (48.4)||0.68 (0.57–0.80)||0.000*||0.70 (0.59-0.84)||0.000*|
|Other||13 (1.0)||5 (0.5)||1.74 (0.62–4.91)||0.290||1.72 (0.61-4.91)||0.307|
|Distance to the provider||n = 1309 (%)||n = 1050 (%)|| || || || |
|≤1 km||493 (37.7)||478 (45.5)||0.72 (0.60–0.87)||0.001*||0.77 (0.64-0.92)||0.003*|
|Socio-economic status||n = 1256 (%)||n = 1011 (%)|| || || || |
|Fifth quintile||192 (15.3)||161 (15.9)||1.00|| || || |
|Fourth quintile||266 (21.2)||242 (23.9)||0.92 (0.70–1.21)||0.558||0.93 (0.70-1.23)||0.617|
|Thirrd quintile||290 (23.1)||230 (22.7)||1.06 (0.81–1.39)||0.688||1.10 (0.83-1.46)||0.506|
|Second quintile||287 (22.9)||247 (24.4)||0.97 (0.74–1.28)||0.850||1.06 (0.80-1.40)||0.675|
|First quintile||221 (17.6)||131 (13.0)||1.41 (1.05–1.91)||0.024*||1.45 (1.06-1.97)||0.019*|
|Type of symptom presenting†||n = 1319 (%)||n = 1070 (%)|| || || || |
|Cough||1073 (81.3)||856 (80.0)||1.09 (0.88–1.35)||0.428||N/A|| |
|Running Nose||1128 (85.5)||906 (84.7)||1.07 (0.84–1.35)||0.580||N/A|| |
|Vomiting||353 (26.8)||285 (26.6)||1.01 (0.83–1.22)||0.947||N/A|| |
|Fast breathing||277 (21.0)||269 (25.1)||0.79 (0.65–0.97)||0.025*||0.75 (0.61-0.93)||0.007*|
|Pallor||103 (7.8)||59 (5.5)||1.45 (1.03–2.04)||0.032*||1.58 (1.10-2.25)||0.012*|
|Fainted||20 (1.5)||15 (1.4)||1.08 (0.55–2.13)||0.815||N/A|| |
|Convulsion||84 (6.4)||78 (7.3)||0.87 (0.62–1.21)||0.393||N/A|| |
|Diarrhoea||498 (37.8)||379 (35.4)||1.11 (0.93–1.32)||0.268||N/A|| |
|Treatment given at home†||n = 1313 (%)||n = 1068 (%)|| || || || |
|Gave herbal medicine||188 (14.3)||133 (12.5)||1.17 (0.90–1.53)||0.230||N/A|| |
|Chloroquine||165 (12.6)||108 (10.1)||1.28 (0.97–1.69)||0.086||1.26 (0.96-1.66)||0.091|
|Fansidar||22 (1.7)||12 (1.1)||1.50 (0.72–3.16)||0.287||N/A|| |
|Pain killer||425 (32.4)||327 (30.6)||1.09 (0.90–1.31)||0.394||N/A|| |
|Tepid sponging||38 (2.9)||60 (5.6)||0.50 (0.33–0.76)||0.001*||0.43 (0.27–0.68)||0.000*|
|Nothing||611 (46.5)||529 (49.6)||0.89 (0.74–1.05)||0.170||N/A|| |
The lowest socio-economic quintile was more likely to delay (OR 1.41; 95% CI 1.05–1.91) than the highest quintile. Generally with decrease in wealth, there was an increase in the likelihood of delayed care seeking with the Chi square for linear trend at 4.52; P = 0.033.
With respect to reported symptoms, those who presented with fast breathing were less likely to delay (OR 0.79; 95% CI 0.65–0.97) while those who presented with pallor were more likely (OR 1.45; 95% CI 1.03–2.04). Those who had done tepid sponging were less likely to delay (OR 0.50; 95% CI 0.33–0.76) (Table 2).
In multivariable analysis children more likely to delay were coming from the lowest socio-economic status or presented with pallor. Children who were less likely to delay had been taken to drug shops (OR 0.70; 95% CI 0.59–0.84) or CMDs (OR 0.33; 95% CI 0.15–0.74) and their caretakers perceived the distance to the provider to be <1 km. Having done tepid sponging immediately after they noticed that the child was sick (OR 0.43; 95% CI 0.27–0.68) and the child presenting with fast breathing predicted prompt care-seeking (OR 0.75; 95% CI 0.61–0.93) (Table 2).
In an area with relatively good access to public health services, no official user-fees, and free drugs available from CMDs, 43% of febrile children taken outside their homes for care were still taken to the private drug shops or clinics, and 44% sought care within 24 h. We found that the predictors for prompt care-seeking were distance to, as well as type of provider, socioeconomic status, presenting symptoms and home treatment given.
While studies report on the proportion of children getting care within 24 h (Amin et al. 2003; Ahorlu et al. 2006), they are generally silent on the methodological problems of retrospectively eliciting time information in a community where most of the members do not use watches to count time. Similar to Ajayi et al. (2008), we found that counting nights of illness was the closest approximation of a 24-h interval and the most practically possible way to estimate time. Caretakers reporting giving care the same day the child fell ill, or after one night of illness, were approximated to have sought care within 24 h. Those who reported seeking care after two or more nights of illness were considered to have waited more than 24 h. We call for further research to test reliability and validity of different methods to approximate time in retrospective surveys, and suggest that indicators and survey methods to, e.g. measure progress towards the Abuja target (WHO/CDS/RBM 2000) on prompt care-seeking be standardized.
The delay in care-seeking increased as one moved from the least poor to the most poor. This concurs with several other studies which indicate that health care seeking among the poor is worse than that of the least poor. In the Uganda Demographic Health Survey 2006, the percentage of children who had had fever in the previous 2 weeks and had received an antimalarial the same day or the next was 28.9. This increased as one moved from the lowest to the highest quintile (Uganda Bureau of Statistics and ORC Macro 2007). Studies from Tanzania (Schellenberg et al. 2003; Njau et al. 2006), Ethiopia (Deressa et al. 2007) and Nigeria (Onwujekwe et al. 2008) indicate that the most poor are less likely to receive antimalarials than others. Studies from Kenya (Taffa et al. 2005) and Nigeria (Olaogun et al. 2006) link accessibility to antimalarials with household finances and the mother’s occupation. Our study demonstrates that the poor are disadvantaged also in the timing of the care they receive. Therefore even in the presence of ‘free’ and ‘pro-poor’ interventions the poorest still have a comparative disadvantage in accessing timely care outside the home. This highlights the importance of looking closely at which care is geographically and financially accessible to them and how to improve its quality.
Caretakers who went to drug shops and CMDs were more likely to have sought care promptly compared to those who went to government health facilities. The geographical proximity to the sick children and their caretakers could have contributed to this promptness, as drug shops and CMDs are generally closer to the community than health facilities (Konde-Lule 2006). This conclusion is backed up by the finding that those who perceived the distance to the provider to be <1 km were also more likely to be prompt in their seeking care outside the home. However, distance may not entirely explain the association. Even after controlling for distance in multivariate analysis, caretakers still went more promptly to the drug shops and CMDs. Dissatisfaction with the public facilities, especially the lack of drugs and poor staff attitudes, could have contributed to preference for private providers (Williams & Jones 2004). To enable the caretakers to receive not just timely but also adequate treatment the informal private sector may need to be included in quality improvement and social marketing (Marsh et al. 2004). Interventions involving the private sector should include a comprehensive analysis of the legal and market environment including strategies for training and supervision (Goodman et al. 2007). Monitoring performance is also essential as interventions may sometimes not reduce costs to the poor (Rowe et al. 2005). Working with the formal private sector for the start may be more feasible as involving the informal sector is resource intensive, especially for countries where the public sector itself is under-funded (Mills et al. 2002). Increasing use of CMDs through training of Community Health Workers would provide cheaper treatment very close to the community (Sirima et al. 2003).
Our findings indicate that presenting with pallor was associated with delay in seeking care. After ascertaining whether the child had been sick in the previous 2 weeks, caretakers were asked symptoms which they had observed in the child. These symptoms had been translated in the local language and included pallor. The presence of pallor being associated with delay differs from a study done in Ghana where it was associated with prompt treatment (Ahorlu et al. 2006). The study in Ghana qualitatively explored the perception that pallor results from shortness of blood, which prompted caretakers to take their children for care promptly. Causes of pallor were not explored in this study but a possibility for the delay could be due to non-association of pallor as a complication of fever. Various studies have documented caretakers not linking complications of fever to the febrile illness (Mwenesi et al. 1995; Nsungwa-Sabiiti et al. 2004). In a study done in Tanzania, the caretakers did not relate pallor to malaria (Winch et al. 1996). In settings like this with endemic malaria, pallor could be chronic and it seems less likely that this sign if detected would be associated with urgent need for treatment.
Some symptoms were associated with prompt care seeking. Children who presented with fast breathing were more likely to go early. The practice of taking children with fast breathing promptly for care is similar to what was reported in other studies (Nsungwa-Sabiiti et al. 2004; Hildenwall et al. 2007). Fast breathing has also been found to be easier for caretakers to recognize than other chest symptoms like chest in-drawing (Kauchali et al. 2004). This could possibly explain why caretakers could react promptly to it.
Caretakers who had reportedly used tepid sponging immediately when they noticed that the child was ill were less likely to delay to seek treatment outside the home. Tepid sponging is when the caretakers use a piece of cloth soaked in water to reduce the temperature of the febrile child. A caretaker who uses tepid sponging may have been exposed to health care messages. It has been shown that tepid sponging is considered first aid remedy for febrile illness (Nsungwa-Sabiiti et al. 2004) and it is possible that the use of a wet cloth is just but a step in the journey to get more treatment.
Having a DSS data base enabled us to use wealth quintiles constructed for all the households in the DSS rather than from a sample. This makes the wealth ranking of households in the sample closer to those in the DSS population. However, some of the households were new so they did not have their assets indicated in the database. Since we used an illness recall of 2 weeks, some information could have been misreported because the caretakers did not remember the details. This was mitigated by interviewing only the caretakers who had nursed the child during the illness episode. In order to reduce over reporting of practices more acceptable to the biomedical researcher, we used local field assistants from the same villages as the interviewers who know the local situation and who are trusted by the respondents. We did not study promptness of home-care among the high proportion of children not seeking care outside the home. During data collection, fevers were not categorized into ‘resolved’ or ‘unresolved’, ‘acute’ or ‘chronic’. This could have distorted the types of illness captured in the ‘prompt’ or ‘not prompt’ groups as unresolved fevers could still be in the process of care seeking. However, since the focus was on first treatment outside the home and most children have acute fever, the distortion could have been minimal. It was not possible to measure the exact distances between the households and the health care providers and we used a proxy of less than a km to indicate ‘being near’. A better way would have been to use GPS coordinates as used by Kazembe et al. (2007) and Noor et al. (2003, 2006). In the absence of exact distances, perception of less than 1 km was taken to represent proximity as figures given by caretakers could not be ascertained as valid.
Even in an area of relatively good physical access to care less than half of the febrile children were taken outside their homes for care, and less than half of these were taken promptly within 24 h. The socio-economic status of the household, the type of provider, the perception of distance to the provider, the recognition of symptoms at home and the type of home treatment given affect the promptness with which the child will be taken for treatment outside the home. Drug shops and CMDs may complement government efforts to deliver timely treatment to the children under five. Behaviour change communication aiming at encouraging caretakers to take the children to providers even after giving home treatment is needed. Methods to estimate time in retrospective surveys of people not living by the clock need to be validated and standardized.
Many thanks go to Max Petzold and Fred Makumbi for their assistance on statistical issues. This study received financial support from Sida/SAREC and from UNICEF/UNDP/World Bank/WHO Special Program for Research and Training in Tropical Diseases.