There are five caveats with the data used, for which the impact on analysis of inequality is unclear. First, the administrative area of location used to calculate the SES of the facility is a proxy measure for catchment area, and may not accurately measure the SES of the catchment area for facility users or the area in which most facility users reside. However, the median area of 16km2 for the location appears quite consistent with typical distances users travel to government facilities (5-6km) [31, 32]. Second, SES data are from relatively old surveys (1997/99) , and the relative SES of facility locations may have changed overtime. More recent SES data were available from a 2005–6 household consumption survey  but the data were only disaggregated to the district level using the old boundaries for 69 districts, and therefore were of limited value in providing information on the specific SES characteristics of individual health facilities.
Third, the creation of input availability indicators involved some decisions on cut offs (at least 4 qualified staff, at least 2 support staff, and at least 15 out of 18 equipment items), which are somewhat arbitrary. We tested the robustness of our findings to variations in these cut-offs, but found no changes in the significance of the univariate resultse. Fourth, the equipment, drug, family planning and vaccine indicators imply that all individual commodities in these indicators are of equal importance, whereas in reality, for example, the lack of certain drugs on the tracer list may be much more significant for health outcomes than the lack of others.
Finally we were unable to conduct the survey at 24 of the facilities originally selected (4.6% of the sampling frame of facilities in selected districts) because they were not operational, not officially recognized by the government, or did not have any qualified staff. These exclusion criteria were used because the survey was conducted as the baseline for the evaluation of a new health financing mechanism for which only fully functioning facilities were eligible. Therefore, we are likely to have excluded some of the poorest performing facilities; if these were concentrated in poorer areas, this could have led to an under-estimation of inequalities. However, of the 24 dispensaries excluded, poverty data were available for 12 facilitiesf, and these were relatively evenly distributed across the SES quintiles of surveyed facilities (2 each in poorest and least poor quintiles, 3 in 2nd quintile and 5 in 3rd quintile).
In addition there are 3 limitations to the study design, which may mean that inequalities are underestimated in this analysis. First, our approach does not consider areas that had no facility at all within reasonable reach. Again it is likely that these areas are relatively poor and their populations therefore have no access to services. Secondly, we only included health centers and dispensaries, excluding public hospitals, as the expectation of inputs and services in the latter is markedly different. However, many hospitals do offer similar outpatient services to the smaller facilities, and as they are generally located in larger centers, which are likely to be relatively well off, this may have led us to underestimate inequalities in primary care provision.
Thirdly the analysis focused exclusively on government facilities, which can be considered appropriate given their use of public resources. Faith based facilities are important care providers in some areas of the country, and if the SES distribution and pattern of service and input availability differed between government and faith based facilities, their inclusion could affect overall inequity. However, the Kenya Service Provision Assessment indicated that relative input and service availability between faith based and public facilities is mixed across indicators, with no clear pattern overall. [36, 37]. Moreover, based on poverty data of the locations of faith based facilities, we did not find evidence that they were concentrated in poorer areas than government facilities. The proportion of the population above the poverty line was 27.9% in the poorest 20% of government facilities compared with 29.7% in the poorest 20% of faith based facilities, and 70.6% in the richest 20% of government facilities compared with 75.5% in the richest 20% of faith based facilities.
Addressing inequity in access to health care is a key concern in Kenya and is documented as a policy priority. This paper assessed equity of facility input and service availability in public primary health facilities across the country. For most input and availability indicators in primary health care facilities, we found no indication of variation by SES of the facility’s location. This could be considered an encouraging finding, indicating that availability of care is relatively equitable across this diverse country.
In addition, availability of some inputs and services was generally high. Close to 90% of facilities had a functioning water supply and almost all offered family planning services. More than 80% offered VCT/PITC, PMTCT and ITN services. However, the lack of inequality does not imply that availability of inputs and services were invariably high. For example, only around three quarters of facilities had all family planning or all vaccine commodities in stock, or offered antenatal care services every weekday. Only around two thirds had at least 15 of 18 key equipment items available, or had at least 2 support staff. Under half offered delivery services, all childhood immunizations every weekday, outreach services, or had at least 4 qualified staff. Only a little more than one-third had laboratory services or electricity available, and less than 20% had all drugs on the tracer list in stock. These results were similar to those for government facilities from the Kenya Service Provision Assessment Survey (KSPA) conducted in the same year for electricity and family planning, though we found higher availability for water supply, perhaps reflecting a broader definition that was used in our analysis .
For specific indicators, we found clear evidence of pro-rich inequalities for electricity and laboratory services, and some indications of pro-rich inequalities for availability of drugs and qualified staff. Drug availability and qualified staff are central inputs into both technical quality of care and demand for facility services, and laboratories are important for accurate disease diagnostics. Electricity is important for enhancing quality of care to patients at the facility in several ways. Although fridges can be run on gas cylinders where there is no electricity (79% of facilities without functioning electricity had functioning fridges), electric lighting is useful for providing care at night and for security, and in some cases is required for equipment such as microscopes, autoclaves and computers. It seems plausible that electricity and laboratory services would be more available in better off areas because they often rely on local funding from user fees, which better-off clients are more willing and able to pay. Electricity bills for health centers and dispensaries generally have to be paid from facility revenues, and laboratories are often established partly as an income-generating activity for the facility, with fees per service provided . However, we did not find pro-rich inequalities in other locally funded inputs such as water and support staff. Kenya is currently introducing a new system of financing health centers and dispensaries called the Health Sector Services Fund (HSSF), which transfers money directly into facility bank accounts to be managed by the health facility committee, and used for a range of operational expenses . It is possible that this additional funding controlled at the facility level will allow more facilities in relatively poor areas to improve availability of electricity and laboratory services in future, and enable all facilities to expand their outreach activities, support staff and equipment availability.