Access to a health facility and care-seeking for danger signs in children: before and after a community-based intervention in Lusaka, Zambia

Authors


Corresponding Author Satoshi Sasaki, Department of Infectious Disease Control and International Medicine, Graduate School of Medical and Dental Sciences, Niigata University, 1-757 Asahimachi-dori Chuo ward, Niigata 951-8510, Japan. Tel.: +81 25 227 2129; Fax: +81 25 227 0765; E-mail: ssasaki@med.niigata-u.ac.jp

Summary

Objective  To assess the association of accessibility to a health facility with caregivers’ care-seeking practices for children with danger signs before and after community-based intervention in Lusaka, Zambia.

Method  Health education on childhood danger signs was started in September 2003 at the monthly Growth Monitoring Program Plus (GMP+) service through various channels of health talk and one-on-one communication in a peri-urban area of Lusaka. Two repeated surveys were conducted: in 2003 to collect baseline data before the intervention and in 2006 for 3-year follow-up data. Caregivers who had perceived one or more danger signs in their children within 2 months of the surveys were eligible for the analysis. The association between appropriate and timely care-seeking practices and socio-demographic and socio-economic factors, attendance at community-based intervention and the distance to a health facility was examined with logistic regression analysis.

Results  The percentage of caregivers immediately seeking care from health professionals increased from 56.1% (106/189) at baseline to 65.8% (148/225) at follow-up 3 years later (OR = 1.51, P < 0.05). Long distance to the health facility and low-household income negatively influenced caregivers’ appropriate and timely care-seeking practices at baseline, but 3 years later, after the implementation of a community-based intervention, distance and household income were not significantly related to caregivers’ care-seeking practices.

Conclusion  Poor accessibility to health facilities was a significant barrier to care-seeking in a peri-urban area. However, when caregivers are properly educated about danger signs and appropriate responses through community-based intervention, this barrier can be overcome through behavioural change in caregivers.

Abstract

Accès à un service de santé et recours aux soins pour les signes de danger chez l’enfant: Avant et après une d’intervention basée sur la communautéà Lusaka, Zambie

Objectif:  Evaluer l’association entre l’accessibilitéà un service de santé et les pratiques des soignants dans le recours aux soins pour les enfants présentant des signes de danger, avant et après une intervention basée sur la communautéà Lusaka, en Zambie.

Méthode:  L’éducation sanitaire sur les signes de danger de l’enfance a débuté en septembre 2003, lors du service mensuel du ‘Program Plus de Surveillance de la Croissance (GMP+)’à travers différents canaux de débats sur la santé et une communication de tête-à-tête dans une zone péri-urbaine de Lusaka. Deux surveillances répétées ont été menées, en 2003 pour recueillir des données de base avant l’intervention et en 2006 pour le suivi des données après 3 ans. Les soignants qui avaient perçu un ou plusieurs signes de danger chez leurs enfants endéans deux mois de la surveillance étaient éligibles pour l’analyse. L’association entre les pratiques appropriées de recours aux soins en temps opportun et les facteurs sociodémographiques et socioéconomiques, la participation à des interventions basées sur la communauté et la distance à un service de santé, a été examinée avec l’analyse de régression logistique.

Résultats:  Le pourcentage des soignants recourant immédiatement à des soins auprès de professionnels de la santé a augmenté de 56,1% (106/189) au départ à 65,8% (148/225) au cours du suivi 3 ans plus tard (OR = 1,51; p <0,05). La distance éloignée à un service de santé et un revenu familial faible influençaient négativement les pratiques des soignants dans le recours à des soins appropriés et en temps opportun. Mais, 3 ans plus tard, après la mise en œuvre d’une intervention communautaire, la distance et le revenu du ménage n’étaient plus significativement liés aux pratiques de recours aux soins des soignants.
Conclusion: L’accessibilité limitée aux services de santéétait un obstacle important au recours aux soins dans une zone péri-urbaine. Cependant, lorsque les soignants sont bien éduqués sur les signes de danger et les réponses appropriées dans une intervention communautaire, cet obstacle peut être surmonté grâce au changement de comportement chez les soignants.

Abstract

Acceso a un centro sanitario y búsqueda de cuidados frente a signos de peligro en niños: antes y después de una intervención basada en la comunidad en Lusaka, Zambia

Objetivo:  Evaluar la asociación de acceso a un centro sanitario con la práctica de búsqueda de cuidados por parte de los cuidadores de niños con signos de peligro antes y después de una intervención basada en la comunidad en Lusaka, Zambia.

Método:  La educación sanitaria sobre los signos de peligro en niños comenzó en Septiembre 2003 como parte del programa de monitorización mensual del crecimiento (GMP+) y a través de varios canales incluyendo charlas sanitarias y comunicaciones individuales en un área peri-urbana de Lusaka. Se realizaron dos encuestas repetitivas: en el 2003 para recolectar datos de base antes de la intervención, y en el 2006 para realizar un seguimiento 3 años después. Los cuidadores que habían percibido uno o más signos de peligro en sus niños dentro de los dos meses anteriores a las encuestas fueron considerados elegibles para el análisis. Mediante regresión logística se examinó la asociación entre una práctica apropiada y a tiempo en la búsqueda de cuidados y factores socio-demográficos, factores socio-económicos, haber participado en la intervención basada en la comunidad, y la distancia al centro sanitario.

Resultados:  El porcentaje de cuidadores que buscaron atención sanitaria profesional inmediata aumentó de 56.1% (106/189) en el estudio base a 65.8% (148/225) durante el seguimiento, 3 años más tarde (OR=1.51, p<0.05). Una distancia grande al centro sanitario y un bajo ingreso económico en el hogar tenían una influencia negativa sobre una búsqueda de cuidados sanitarios apropiada y a tiempo, pero 3 años después, tras la implementación de la intervención basada en la comunidad, la distancia y el ingreso del hogar no estaban significativamente relacionados con las prácticas de búsqueda de cuidados sanitarios de los cuidadores.

Conclusión:  El mal acceso a los centros sanitarios era una barrera significativa en la búsqueda de cuidados sanitarios en un área peri-urbana. Sin embargo, cuando mediante una intervención basada en la comunidad, los cuidadores habían recibido una educación apropiada sobre los signos de peligro y las respuestas apropiadas, esta barrera se superaba con un cambio de comportamiento de los cuidadores.

Introduction

Nearly 10 million children die worldwide by the age of five, mostly in developing countries and as a result of infectious diseases such as pneumonia, diarrhoea, measles, malaria and HIV/AIDS (Black et al. 2003; Bryce et al. 2005). A large number of child deaths can be attributed to delays in care-seeking (Reyes et al. 1997; Terra de Souza et al. 2000; Kallander et al. 2008). As health care-seeking is strongly related to particular illness symptoms and their perceived severity (Amarasiri de Silva et al. 2001; Taffa & Chepngeno 2005), the lack of recognition of children’ severe symptoms has been reported as a significant barrier to appropriate and timely care-seeking (Snow et al. 1994; Baume 2002; Hill et al. 2003).

Accessibility of health facilities is another important deterrent to health care-seeking. If it is poor, delays in appropriate care-seeking ensue (Terra de Souza et al. 2000). Accessibility includes not only geographical distances but also health service fees, the cost of transportation and the opportunity cost of lost time for earning money (Noor et al. 2003; Bazzano et al. 2008). To facilitate change in caregivers’ behaviour towards appropriate and timely care-seeking, assessing the association between accessibility and care-seeking practice and designing an intervention to mitigate the burden of accessibility is important.

In Zambia, the Ministry of Health, in collaboration with the Japan International Cooperation Agency (JICA), launched the Primary Health Care Project in 1997 to improve the health status of children in peri-urban areas of Lusaka. Community-based outreach activities called the ‘Growth Monitoring Program Plus’ (GMP+) were designed as a multifaceted activity of essential health services for children to be delivered by trained health volunteers and health professionals. During the health education sessions in the GMP+, the emphasis was put on danger signs by which caregivers could recognize children’ critical conditions and seek immediate care from health professionals. In the previous study, we indicated that the GMP+ intervention with health education about danger signs could facilitate caregivers’ appropriate and timely care-seeking for severely sick children (Fujino et al. 2009). However, the impact of the intervention which facilitated caregivers’ surmounting access barriers and financial constraints and led to appropriate care-seeking behaviour was not fully investigated.

Geographical Information System (GIS) is an efficient and practical way to assess the access distance to a health facility. Geographical accessibility is represented by Euclidean distance which is defined as the straight-line distance between two points and network distance along a road (Tanser 2006). Euclidean distance has been used as a proxy measure of accessibility, especially in developing countries where a GIS infrastructure is not well established. However, these measures of distance underestimate time consumption and opportunity cost for the accessibility of health facilities (Noor et al. 2006). Measuring network distances to the nearest health facility with GIS overcomes the shortcomings of straight-line travel assumptions.

In this study, we examined the association between network distance and health facility with other socio-demographic and socio-economic factors, and caregivers’ appropriate and timely care-seeking practices when they perceived danger signs in their children.

Methods

Study area

The study area, called George Proper, is located in the northwestern outskirts of Lusaka and categorized as one of the low-income areas of the capital. The population of the study area was 40 352 in 8256 households, and 4571 households had one or more children under five, according to a household survey conducted by the Primary Health Care project in 2002. The average number of family members per household is 4.89.

The George Health Center is the nearest public health facility for residents in the study area. It provides health services for outpatients, mother and child health, maternity and laboratory services. The Ministry of Health introduced user fees as a method of contributing to financing of health services in 1997. Patients need to either pay a user fee for each consultation or subscribe to a monthly insurance scheme to be exempt from paying user charges. In Lusaka, approximately 90% of outpatients participated in the insurance scheme (Kondo & McPake 2007). However, the Ministry of Health also decided that children below the age of 6 are qualified for free health care (Hjortsberg 2003).

Intervention

We introduced health education on danger signs with the aim of facilitating caregivers’ understanding of the danger signs and behavioural change towards appropriate and timely care-seeking practices. Danger signs were defined as physiological and behavioural symptoms that indicate the immediate need to seek health care from a health professional (WHO 1997, 2005). In accordance with guidelines and modules of IMCI, we identified five symptoms of the danger signs: refusing to breastfeed or drink, vomiting everything, abnormal breathing, appearing weak or sleepy and convulsions.

Health education on the danger signs was started in September 2003 at the monthly GMP+ service through various channels, such as health talk as a topic of health education and one-on-one communication in nutrition counselling (Fujino et al. 2009). A flip chart and video were developed as visual educational aids with dramatized scenes using local terminologies. Health education was provided by community health volunteers who attended a 6-week training course on the methodologies of health education as well as teaching and counselling skills.

Data collection

The care-seeking practices and socio-demographic information of caregivers were collected through two repeated sample surveys of the study area conducted in February 2003 to establish baseline data before the intervention, and in September 2006 for 3-year follow-up data after the intervention. As the incident rates of major childhood illnesses such as diarrhoea and respiratory infection in the study area between February and September showed similar occurrence according to the LDHMT Health Management Information System, seasonal differences in disease occurrence were limited.

Five-hundred caregivers with children under five were interviewed in each survey by a systematic random sampling method in the study area. The study area was divided into 12 administrative zones. The number of caregivers interviewed in each zone was calculated in accordance with the population of the zone. A trained survey team started visiting houses at the south end and moved west along the border of the zone. The survey team collected information from households, in which children under five were taken care of after bypassing the specific number of households that were calculated by dividing the total population of the zones by the number of samples to be collected. If a household with a child or children under five was not found after bypassing the specified number, surveyors asked the adjacent household ahead until they found an appropriate household.

The socio-demographic information collected for each household included the number of children, the age and education level of caregivers, the age and sex of children and the geographical location. The number of times they attended GMP+ sessions was copied from child healthcare records for children under five. If there were more than two children under five, the information of the youngest child was recorded. The caregiver was asked to recall illness episodes of the child during the 2 months prior to the survey and their care-seeking practices in observing the episodes. If a mother reported seeking care in a hospital, a health centre, or a private clinic within 24 h of perceiving the danger signs, the case was categorized as appropriate and timely care-seeking. In cases in which caregivers observed illness episodes more than two times during the period, we collected information on the most serious episode and the caregivers’ response to it.

The sampled households’ socio-economic situations were obtained from household survey data collected in January 2003 and July 2006. Trained surveyors visited all the households in the study area and collected the demographic and socio-economic information of household members including sex, age, employment and monthly income. The household data from 2003 and 2006 that had matching information in three areas, including the name of the caregiver, the name of the child and the home-address, at baseline and in the 3-year follow-up data, respectively, were used for the analysis.

Geographical information

We used Lusaka’s digital base map which was developed by the JICA Primary Health Care Project. The map was digitized based on satellite imagery (SPOT 5 with 2 m resolution) using ArcView software (ESRI, USA), and it included streets, major official buildings, public health facilities and households with addresses (Sasaki et al. 2008). The locations of the interviewed households were mapped by means of an address matching method.

Distances from the sampled households to George Health Center, which was the nearest health facility, were measured with a network analysis method. Footpaths were digitized from aerial photographs (Ministry of Land, Zambia) and satellite images (QuickBird with 60 cm resolution). A network dataset was built, and the network distances were measured using Network Analyst (ESRI).

Data analysis

The association between appropriate and timely care-seeking practices and the access distance to a health facility and, socio-demographic and socio-economic factors, was examined with logistic regression analysis. Within the five danger signs for health education, four more critical and apparent symptoms, including refusal to breastfeed or drink, vomiting everything, convulsions and abnormal breathing, were identified for the analysis. To determine the predictor variables for timely and appropriate care-seeking practices, households which had reported one or more danger signs were selected. Explanatory variables, including child age, child sex, caregiver’s age, caregiver’s education level, number of children, whether a household is female-headed, the daytime presence of a caregiver at a house, monthly income and distances to the health facility were all used as factors in the analysis. The distances to the health facilities were divided by tertiles of households (0–679, 680–983, 984–1581 m). Monthly income was divided into two groups of <200 000 and ≥2 000 000 Kwacha. Logistic regression was applied for crude and adjusted analysis to the selected samples of the baseline and 3-year follow-up data. The odds ratio (OR), 95% confidence interval (CI) and P value were calculated using spss ver.17.0 (SPSS, USA).

To examine the effects of the intervention on appropriate and timely care-seeking practices with due consideration of distances to the facility and monthly income, we calculated chi-square statistics for differences in appropriate and timely care-seeking practices between baseline and 3-year follow-up data in the three categories of distance stratified by two groups of monthly income levels.

Ethical approval

The study protocol was approved by the management committee of Lusaka District Health Management Team, Ministry of Health, Zambia. In the field, data were collected with participants’ verbal consent.

Results

Of the 500 interviewed caregivers at baseline and 3-year follow-up, 189 (37.8%) and 225 (45%) had perceived one or more danger signs. The children at baseline were older than those in the 3-year follow-up (P < 0.05) (Table 1). Education levels of caregivers at baseline were higher than those of the 3-year follow-up as well (P < 0.01). The economic circumstances of households improved in the 3-year follow-up (P < 0.01). With regard to the mean distance from the households to George Health Center, there was no significant difference between baseline and 3-year follow-up (P = 0.80).

Table 1.   Socio-demographic characteristics and other variables of caregivers in the baseline and 3-year follow-up data
Characteristics of variablesBaseline (n = 189) N (%)3-year follow-up (n = 225) n (%)P
  1. Note: SD, standard deviation; GMP+, Growth Monitoring Program Plus.

  2. *Student’s t-tests were used for comparison between two surveys of variables.

  3. †Chi-square test was used for comparison between two surveys of variables.

  4. ‡Primary level is grade 1–7, secondary level is grade 8 and 9, >secondary is grade 10 and higher education.

  5. §The number of GMP+ sessions attended in the last 6 months prior to the survey.

Child age (month)
 Mean ± SD26.59 ± 11.5324.00 ± 12.36<0.05*
Child sex
 Male89 (47.1)118 (52.4)0.28†
 Female100 (52.9)107 (47.6)
Age of caregiver (years)
 Mean ± SD28.50 ± 8.2429.11 ± 7.600.44*
 <2016 (8.5)19 (8.4)
 20-29103 (54.5)101 (44.9)
 30-3956 (29.6)80 (35.6)
 ≥4014 (7.4)25 (11.1)
Educational level of caregiver‡
 Mean ± SD6.17 ± 2.687.08 ± 2.41<0.01*
 No education19 (10.1)13 (5.8)
 Primary131 (69.3)136 (60.4)
 Secondary or above39 (20.6)76 (33.8)
Number of children
 Mean ± SD2.60 ± 1.722.94 ± 2.12 0.77*
Head of household
 Women22 (11.6)33 (14.7)0.37†
 Other167 (88.4)192 (85.3)
Caregivers working out in a day time
 Working out76 (40.2)73 (32.4)0.10†
 None113 (59.8)152 (67.6)
Monthly income (Kwacha)
 Mean ± SD212 937 ± 121 518274 684 ± 148 162<0.01*
 <200 00090 (47.6)82 (36.4)
 ≥200 00099 (52.4)143 (63.6)
GMP+ attendance§ (times)
 Mean ± SD2.67 ± 2.173.09 ± 2.320.06*
 None57 (30.2)58 (25.8)
 1-354 (28.6)53 (23.6)
 4-678 (41.3)114 (50.7)
Distance to the health facility (metres)
 Mean ± SD812.88 ± 313.97821.20 ± 339.960.80*
 0-67962 (32.8)76 (33.8)
 680-98368 (36.0)70 (31.1)
 984-158159 (31.2)79 (35.1)

The percentage of caregivers immediately seeking care from health professionals for children with danger signs significantly increased from 56.1% (106/189) at baseline to 65.8% (148/225) in the 3-year follow-up (OR = 1.51, P < 0.05). The location of households which had reported symptoms of danger signs at baseline and 3-year follow-up data was mapped using an address matching method (Figure 1). Of the caregivers immediately seeking care from health professionals, 99% (105/106) and 98.6% (146/148) sought care from the nearest health facility, George Health Center, at baseline and 3-year follow-up, respectively. The most reported symptom was vomiting everything (52.9% at baseline and 46.2% in the 3-year follow-up), followed by abnormal breathing (24.9% at baseline and 20.4% at 3-year follow-up) (Table 2). The perceived cases of convulsions (14.8% at baseline and 31.6% at 3-year follow-up) and refusal to breastfeed or drink (16.4% at baseline and 20.4% at 3-year follow-up) increased.

Figure 1.

 Maps of study areas with location of households. Circle (•) indicates a caregiver with appropriate and timely care-seeking response and triangle (bsl00066) indicates a caregiver with non appropriate and timely response. (a) Locations of households in the baseline data. (b) Locations of households in the 3-year follow-up data.

Table 2.   Comparison of perceived symptoms of danger signs and caregivers’ appropriate and timely care-seeking responses
Symptom of danger signsPerceiving danger signsAppropriate and timely care-seeking responses†
Baseline, n (%)3-year follow-up, n (%)P*Baseline, n (%)3-year follow-up, n (%)P*
  1. *Chi-square test was used for comparison between the two surveys of variables.

  2. †Caregivers who took their children with danger signs to health facilities within 24 h of observing the signs.

Vomiting everything100/189 (52.9)104/225 (46.2)0.1855/100 (55.0)61/104 (58.7)0.60
Abnormal breathing47/189 (24.9)46/225 (20.4)0.2830/47 (63.8)34/46 (73.9)0.29
Convulsions28/189 (14.8)71/225 (31.6)<0.0115/28 (53.6)55/71 (77.5)<0.05
Caregivers who took their chil31/189 (16.4)46/225 (20.4)0.2917/31 (54.8)23/46 (50.0)0.68

A logistic regression analysis of the association between caregivers’ appropriate and timely care-seeking and independent variables for the baseline data indicated that lower income (<2 000 000 Kwacha) negatively influenced the caregivers’ responses (OR = 0.48, 95% CI = 0.27–0.86, P < 0.05) (Table 3). Compared to the nearest distance group (0–679 m), distance to the health facility was a significant impediment to caregivers’ appropriate and timely care-seeking for the households in the farthest distance group (984–1581 m) (OR = 0.35, 95%CI = 0.17–074, P < 0.01). Similarly, multivariate analysis after controlling confounding factors indicated that the lower-income and the farthest distance groups were negatively associated with caregivers’ appropriate and timely care-seeking (OR = 0.47, 95% CI = 0.25–087, P < 0.05, and OR = 0.30, 95% CI = 0.13–0.66, P < 0.01, respectively). Other variables were not significantly associated with caregivers’ appropriate and timely care-seeking practices in the crude and multivariate analysis.

Table 3.   Association between potential factors and caregivers’ appropriate and timely care-seeking in baseline and 3-year follow-up data
 Immediate response* (%)CrudeAdjusted
OR95% CIPOR95% CIP
  1. Note: OR, odds ratio; CI, confidence interval; GMP+, Growth Monitoring Program Plus.

  2. *Caregivers sought health care from a health professional within 24 h of perceiving danger signs in their children.

  3. †The number of GMP+ sessions attended in the last 6 months prior to the survey.

  4. ‡Distance to George Health Center.

Baseline
Monthly income (Kwacha)
 <200 00042/90 (46.7)0.480.27–0.86<0.050.470.25–0.87<0.50
 ≥200 00064/99 (60.4)1.00  1.00  
GMP+ attendance† (times)
 None32/57 (56.1)1.00  1.00  
 1–333/54 (61.1)1.230.58–2.620.601.220.52–2.850.65
 4–641/78 (52.6)0.870.44–1.720.680.810.37–1.760.59
Distance to the health facility‡ (metres)
 0–67941/62 (66.1)1.00  1.00  
 680–98341/68 (60.3)0.780.38–1.590.490.710.33–1.540.39
 984–158124/59 (40.7)0.350.17–0.74<0.010.300.13–0.66<0.01
3-year follow-up
Monthly income (Kwacha)
 <200 00052/82 (63.4)0.850.48–1.500.570.750.40–1.400.36
 ≥200 00096/143 (67.1)1.00  1.00  
GMP+ attendance† (times)
 None30/58 (51.7)0.330.17–0.65<0.010.310.15–0.65<0.01
 1–331/53 (58.5)0.430.22–0.88<0.050.410.20–0.88<0.05
 4–687/114 (76.3)1.00  1.00  
Distance to the health facility‡ (metres)
 0–67951/76 (67.1)1.00  1.00  
 680–98346/70 (65.7)0.940.47–1.870.861.180.56–2.490.66
 984–158151/79 (64.6)0.890.46–1.740.740.880.42–1.830.73

However, analysis of the 3-year follow-up data revealed that both income level and distance to the health facility were no longer significantly related to caregivers’ appropriate and timely care-seeking practices for children with danger signs in crude and multivariate analysis. Moreover, caregivers who had attended GMP+ sessions more frequently (4–6 times) in the 6 months prior to the 3-year follow-up survey were more likely to immediately resort to the health facility than those who had attended fewer sessions (OR = 0.33, 95%CI = 0.17–0.65, P < 0.01 for no attendance, and OR = 0.43, 95%CI = 0.22–0.88, P < 0.05 for 1–3 times attended). Likewise, multivariate models showed that frequent attendance at the GMP+ sessions (4–6 times) was positively related to appropriate and timely care-seeking practice compared to fewer attendance (OR = 0.31, 95%CI = 0.15–0.65, P < 0.01 for no attendance, and OR = 0.43, 95%CI = 0.22–0.88, P < 0.05 for 1–3 times attended).

The prevalence of appropriate and timely care-seeking practices in the households with lower incomes at baseline was more strongly influenced by distance to the health facility than that of higher income households (Figure 2). Of the households in the lower income level, compared to a 65.5% (19/29) prevalence of appropriate and timely care-seeking at the nearest distances, only 26.7% (8/30) of caregivers in the farthest distances resorted to the health facility immediately. A chi-square test for the difference in prevalence of appropriate and timely response between the baseline and 3-year follow-up data indicated that the prevalence for households with lower incomes at the farthest distance significantly increased at the 3-year follow-up.

Figure 2.

 Comparison of the prevalence of appropriate and timely care-seeking between the baseline and 3-year follow-up surveys. Note: n.s = not significant. *P < 0.05.

Discussion

In our analysis of the baseline data, caregivers who had low incomes and lived further away from the health centre tended to be less likely to seek appropriate and timely care for children with danger signs. Specifically, caregivers who were located at distances of more than 984 m from the health facility, which might be considered the threshold for the level of access by walking, were significantly delayed in response to children’ danger signs. Distance to a health facility is related to time consumption and the cost of transportation which have been pointed out as major deterrents to care-seeking practices (Terra de Souza et al. 2000; Hjortsberg & Mwikisa 2002; Buor 2003; Rutebemberwa et al. 2009). As caregivers living at farther distances from the health facility were more likely to pay to use a taxi or other means of transportation to take sick children to the facility, monetary cost for reaching the facility was a critical burden that hindered caregivers from seeking care immediately. This was especially evident in households with lower incomes because of extremely limited affordability. Thus, in the underprivileged peri-urban area, accessibility to the health facility in terms of distance and monetary cost was an important factor affecting the immediate care-seeking decisions of caregivers.

Analysis of the 3-year follow-up indicated to us that the distance and monetary barrier were reduced after the community intervention. Exposure to health education on symptoms of danger signs and appropriate and timely care-seeking practices through attending GMP+ sessions or other means of intervention played a major role in facilitating caregivers’ behavioural change in care-seeking (Awasthi et al. 2009; Fujino et al. 2009). A comparison of prevalence between the baseline and 3-year follow-up data indicated that there was significant improvement among caregivers who were located at further distances from the health facility with lower household incomes. These caregivers may be more likely to benefit from learning to discern which symptoms really need to be taken to the health facility through health education. This was because their opportunity cost to reach the health facility is relatively higher than that of caregivers at nearer distances. As lower-income households incurred significantly higher costs in proportion to their monthly expenditure (Chuma et al. 2007), the monetary factor still influenced decisions regarding immediate care-seeking mostly for the underprivileged. This led us to assume that the effectiveness of health education on facilitating caregivers’ behavioural change could be strengthened with a multispectral approach to protect severely underprivileged households.

This study also showed that frequent attendance at health education sessions was a major contributing factor to caregivers’ appropriate and timely care-seeking practices. A study in India has indicated that appropriate health education on danger signs enhances caregivers’ ability to recognize these signs and that behavioural change in care-seeking requires prolonged exposure to the educational message (Mohan et al. 2004). The advantage of the GMP+ was the fact that it was easy for the caregivers to access and receive necessary child health services, because the GMP+ was packaged with required basic health services, and the sessions were delivered close to the families every month within the smallest administrative areas. We assume that this convenience led caregivers to attend the GMP+ sessions more frequently and regularly, and hence the caregivers were exposed to the health messages on danger signs repeatedly.

Caregivers perceiving convulsions as danger signs and seeking care immediately from health facilities significantly increased after the intervention. Health education delivered in local terminology with visual aids such as a video and flip charts helped caregivers understand the symptoms that they were unfamiliar with. Caregivers with both primary and secondary education showed improvement in care-seeking behaviour, whereas caregivers without formal education were less likely to benefit from health education. For the reason that the impact on appropriate and timely care-seeking response varied according to the symptoms of danger signs and education level of caregivers, further improvement in educational materials and methods is required.

The impact of income growth of households on increasing the prevalence of appropriate and timely care-seeking practice cannot be disregarded. The average monthly income increased by nearly 30% from 2003 to 2006. As the proportion of households with a monthly income of more than 200 000 Kwacha increased, more caregivers sought immediate health care for their sick children accordingly. However, because the monthly cost of food for a family of six increased in 2006 to 1.4 times higher than that in 2003 (The Jesuit Centre for Theological Reflection 2005), we assume that the income gain was offset by inflation in the Zambian economy which limited the impact of increased income on the increased prevalence of appropriate and timely care-seeking practices. Further analysis is required to determine the impact of income growth on care-seeking practices with due consideration of indirect monetary costs.

Our study had limitations. The study area was a relatively small urban area, and the longest distance from a household to the health facility was no more than 1514 m. As the average radius distance from a health facility to the nearest neighbouring health facility was 1368 m, our analysis contained a large group of caregivers within a health centre catchment area in an urban city. Furthermore, our accessibility findings cannot be applied to rural villages without modification of the analysis method. Therefore, further analysis in a larger study area and a rural settlement is required to reveal the impact of longer distances to health facilities on caregivers’ care-seeking practices. Our study did not apply case control study to verify the effect of community-based intervention. However, our previous study proved the improvement of caregivers’ care-seeking practices through intervention with cross-sectional repeated study. Therefore, we focused on the association between accessibility and caregivers’ behaviour change. Moreover, the survey data relied on the memory of the caregivers, and thus, the study was subject to recall bias. However, because the period of retrospection was only 2 months and the severest episode seemed to be engraved on caregiver’s memory, the recall bias could be minimized.

In conclusion, we found that poor accessibility of the health facility was a significant care-seeking barrier in a peri-urban area at the first stage. However, when health education messages on danger signs are properly delivered to caregivers through appropriate community-based intervention, the accessibility deterrent can be mitigated through behavioural change in the caregivers in a peri-urban setting.

Acknowledgements

We thank the staff of the Lusaka District PHC Project assigned by LDHMT and JICA. We are also grateful to all the respondents and participants who willingly gave their time to respond to two repeated surveys.

Ancillary