Inadequate access to health care is a complex, multi-dimensional problem (Mamdani and Bangser, 2004). On the supply-side, availability of appropriate interventions such as drugs or vaccines, quality of services, and affordability all affect the uptake of health interventions. Demand-side factors such as acceptability of interventions, health education, and treatment seeking behavior can also affect access (Ensor and Cooper, 2004, Krause and Sauerborn, 2000). Physical accessibility plays an important role in utilization of health services (Bell et al., 2005, Rosero-Bixby, 2004, Gething et al., 2004, Noor et al., 2003, Mamdani and Bangser, 2004). The investment and effort required to access health services including distance or time required to travel, loss of productivity due to time away from work, and availability and cost of transportation are crucial factors in the decision about whether and when to seek treatment.

The effectiveness of primary care in reducing malaria disease reflects the impact of both early treatment as well as prevention. It is likely that many families living further from primary care facilities waited longer to seek care in the public sector for their febrile child than those living nearby, leading to an increase in the progression to disease requiring inpatient care. Most of the facilities in the analysis also distribute insecticide treated nets (ITNs) at highly subsidized rates. If those children who are close to a facility are more likely to attend, then they are also more likely to have an ITN. Thus, the trend in malaria as a function of travel time to the clinic may be influenced by ITN coverage as well as treatment of clinical cases.

The protective effect of primary care facilities in reducing hospitalized clinical disease was 33%. This estimate reflects the actual reduction in disease burden in a typical rural district facing significant challenges to delivering health care. Evaluating the effectiveness of the primary health care system based on disease burden incorporates many different components of access to health care including physical distance, availability of appropriate medicines, quality of care, treatment seeking patterns, and utilization rates. It provides an integrated measure of overall effectiveness of the available health infrastructure. Importantly, the protection afforded by access to treatment and prevention delivered through imperfect health systems is equivalent to that offered by alternative interventions such as anti-malaria vaccine candidates (Alonso et al., 2004) and intermittent preventive treatment (Meremikwu et al., 2008) in carefully controlled efficacy studies. This suggests that improvements in access to primary care have a tremendous potential to increase the effectiveness of existing interventions.

Our results are consistent with other studies which have related the prevalence of infection and odds of progressing to severe disease to access to a health facility. For example, school children in Cote d’Ivoire were less likely to be infected with P. falciparum in a cross-sectional survey if there was a health care facility in their village (Raso et al., 2005). In Papua New Guinea, the prevalence of infection was lower across all age groups in those living within a one hours walk to a health facility compared to those living further away (Mueller et al., 2005). A case-controlled study in Yemen showed that close proximity (less than 2 kilometers) to a health facility reduced the odds of progression from mild to severe malaria (Al-Taiar et al., 2008). Our findings extend these observations by demonstrating that the overall incidence of clinical malaria is a continuous function of access to care.

Non-malaria admissions also declined near the primary care facilities, as would be expected, particularly if the Integrated Management of Childhood Illnesses program was being implemented as it is in Kilifi. Antibiotics, oral rehydration therapy, and routine immunizations, among other things, can all be delivered effectively at the primary care level and prevent hospitalization with acute disease. However, some illnesses cannot be treated at the primary level, and therefore access to primary care may not reduce hospitalization from tuberculosis, neonatal complications, or severe malnutrition, for example. As a result, the incidence of non-malaria admissions was influenced less by access to primary care than malaria, for which effective antimalarials and preventive interventions can be delivered at the primary care level.

Socioeconomic status has been shown in some studies to influence the incidence of clinical malaria, although not the prevalence of infection (Raso et al., 2005). If risk of infection varies little with wealth, but disease varies considerably (Barat et al., 2004), variations in disease incidence may reflect events downstream of infection, such as delaying treatment due to affordability or distance from the health facility. If socioeconomic status declines away from the health facility, this may contribute to the direct correlation between the burden of malaria disease and distance from the facility. Although this is a potential confounding factor in interpreting these results, it may also mean that changes in the trend of malaria admissions might reflect a more equitable distribution of treatment access amongst those at risk. In other words, it could be an indicator of the distribution of health services in the community, both geographically and amongst economic groups. It should be noted, however, that Kilifi district is amongst the most impoverished communities in Kenya and gradients of socioeconomic status are not likely to be very pronounced, particularly if the town center is excluded as has been done for this analysis.

This analysis assumes that malaria transmission is relatively homogenous over the scale of two hours walk from the clinic. Although it is known that transmission in Kilifi district is heterogeneous (Mwangi et al., 2005, Mbogo et al., 1995), and that incidence of disease can vary from one homestead to the next (Mackinnon et al., 2005), averaging incidence over all the health centers is likely to smooth out these small-scale heterogeneities and reveal robust trends. We have not included private facilities or access to treatment in the retail sector in our analysis. Previous work in Kilifi shows that most mothers purchase biomedical treatment in the retail sector first, but more attend public rather than private facilities for their child’s fever (Molyneux et al., 1999). Evaluating the impact of the retail sector may be an important extension of this work, particularly if artemisinin combination therapies become available over-the-counter.