Table 1.

Characteristics of the defined hospital catchment areas communities for 17 hospitals

Hospital location	Total Population 1999 in catchment [Surface area Km2 ]	Projected population <15 years in 2004 (Mid-point of surveillance) [inter-censal growth rate] % p.a.	% of Population residing in Areas ≥ 1500 m above sea level2	% of Population residing in Areas with EVI <0.33	Average annual rainfall 1999-2008 (mm)4	% population living in areas with modelled Pf PR2-10 <5% in 2009	% Population living in areas with modelled Pf PR2-10 5-39% in 2009	% Population living in areas with modelled Pf PR2-10 ≥ 40% in 2009
Western/Lakeside
Busia	369,552 [9.3]	204,650 [3.4]	0.0	0.0	1,584	0.0	0.0	100.0
Bungoma	569,953 [12.5]	339,862 [4.3]	24.4	0.0	1,431	0.0	53.5	46.5
Bondo	153,864 [6.0]	75,862 [1.8]	0.0	1.27	1,4125	0.0	44.7	55.2
Homa Bay	253,496 [8.5]	130,559 [2.7]	0.9	11.0	1,4125	0.0	77.5	22.5
Kisumu	541,002 [8.4]	251,398 [2.0]	13.0	15.8	1,4125	0.0	89.6	10.4
Siaya	243,149 [7.0]	112,039 [0.9]	0.0	0.7	1,4125	0.0	1.3	98.7
Highlands
Kericho	406,491 [15.6]	198,951 [2.4]	55.2	0.00	2,121	84.9	15.3	0.0
Kisii	405,331 [4.7]	197,580 [2.0]	88.5	0.00	2,119	10.3	89.7	0.0
Kitale	765,943 [26.9]	435,428 [3.8]	84.1	2.8	1,264	69.8	30.2	0.0
Coastal
Kilifi	338,112 [17.6]	183,651 [3.1]	0.0	5.3	1,3496	28.8	71.2	0.0
Malindi	201,049 [12.7]	114,865 [3.4]	0.0	18.8	1,077	61.7	38.3	0.0
Msambweni	241,647 [14.5]	101,802 [2.6]	0.0	1.15	887 (5)	0.0	93.3	6.8
Arid & semi-arid
Narok	167,519 [35.0]	96,886 [3.3]	44.7	33.9	7227	100.0	0.0	0.0
Hola	52,622 [31.3]	29,264 [3.4]	0.0	73.2	492	6.0	94.0	0.0
Voi	159,230 [14.3]	69,613 [1.7]	21.9	34.4	4917	94.8	5.2	0.0
Makueni	293,997 [20.4]	159,381 [2.8]	15.2	48.3	540	100.0	0.0	0.0
Wajir	287,617 [446.8]1	224,625 [9.6]	0.0	100.0	312	100.0	0.0	0.0

1. Wajir is a district that is populated by pastoralists with very little access to health services and scattered across a wide spatial area. The predicted catchment area therefore was considerably larger than all other more densely populated settled communities in other districts.

2. A digital elevation map (DEM) for Kenya was used that has a resolution of 30 meters http://www.vterrain.org/Elevation/SRTM/

3. EVI - Enhanced vegetation index - threshold values were computed to define malaria-relevant EVI categories that corresponded to accepted definitions of aridity based on annual rainfall [53]. In Kenya this approximated to an EVI threshold of 0.3 below which average annual rainfall was less than 1000 mm corresponding to historical descriptions of malaria in Kenya referred to as 'only malarious near water' [27].

4. Trend analysis of rainfall showed a relatively stable pattern across most of the sites without significant rises or declines in monthly precipitation over the 10 years of observation. Exceptions were at Msambweni where a significant, but small downward trend was observed (trend 0.44; P value for t statistic = 0.018,) and at Hola and Busia significant increased precipitation was observed over the 10 years; trend = 0.27 (P = 0.042) and trend = 0.79 (P = 0.002) respectively.

5. A single meteorological station was used for all the four hospitals immediately surrounding Lake Victoria, based at Kisumu, this met station had complete records for the period of observation and there were no other met stations located close to the catchments of the other three hospital sites. This station is however located close, between 39 and 54 km, to the other three catchment areas.

6. Rainfall data from the Kilifi Met station were missing for the period June 2008 to December 2008 and were supplemented for this period with data from Malindi met station located 24 km from the Kilifi catchment area.

7. A total of 13 (0.6%) of 2,040 monthly rainfall observations were not recorded. An average of neighbouring non-missing values was used to estimate single months of missing rainfall data in two sites (Voi and Narok).