Mean incidence rates, seasonality, and risk of malaria
Highest burden regions and districts also hosted health facilities with the highest number of confirmed malaria cases reported. For instance, Bala health centre (HC) III in Kole district of the Lango region reported 3,317 cases during November 2015, while Bira HCII in Adjumani district of the West Nile region reported 6,697 cases during June 2016. Moreover, Barakala HCIII (highest for two consecutive years) also from West Nile in Yumbe district, reported 9,654 cases during October 2017 and 9,246 cases during July 2018. Lastly, Matany hospital in Napak district of Karamoja region reported 8,089 confirmed cases during September 2019.
This study showed spatial and temporal variation in incidence rates between regions and districts in any given region, as well as between health facility catchments within districts, both during the low (Fig. S9, Additional file 1) and high burden seasons (Fig. 2).
National incidence rates:
The model estimated 38.8 (95% CI: 37.9–40.9) million confirmed malaria cases over the study period of July, 2015 to September, 2019, highest in 2016 with 10.3 (95% CI: 9.9–10.7) million cases and lowest in 2018 with 6.5 (95% CI: 6.4–6.9) million cases among complete calendar years (Table S4, Additional file 1). Annual incidence rates reduced from 281.7 (95% CI: 274.9–296.7) in 2016 to 170.0 (95% CI: 165.9–178.8) cases per 1000 in 2018.
Monthly incidence rates showed a general declining trend in the burden of malaria from 2015 to 2019, strongest through 2018 followed by an increase in 2019 (Fig. 3). In all the years of the study, the incidence rates consistently peaked in June and July, reaching a maximum of 36.6 (95% CI: 35.7–38.5) cases per 1000 in June 2017 (Table S5, Additional file 1). Conversely, low risk periods were less consistent, although often lowest in February and March, reaching a minimum of 8.9 (95% CI: 8.7–9.4) in February 2018.
Spatial distribution of incidence rates across the country:
Overall, mean monthly regional incidence rates were highest in Acholi region (Northern Uganda) at 52.3 (95% CI: 50.3–59.6) cases per 1000 per month and lowest in Kigezi region (South Western Uganda) at 7.9 (95% CI: 7.6–8.2) cases per 1000 per month (besides Kampala).
Consistent with national trend assessments, monthly trends in regional incidence rates showed the highest peaks in June-July, highest in June, 2017 (Range: 13.4–95.6 cases per 1000) and July, 2019 (Range: 13.5–95.5 cases per 1000 in Kigezi and Acholi, respectively) and the lowest troughs in February-March of each calendar year (Fig. 3). These trends showed that Acholi, West Nile, Karamoja, East Central – Busoga, and Teso persistently recorded the highest monthly incidence rates across the entire study duration. Moreover, the greatest variability in incidence rates was also observed among these five highest burden regions of with respective estimated mean monthly incidence rates of 52.3 (SD: 17.8), 43.3 (13.9), 30.3 (10.4), 26.3 (8.6), and 23.5 (8.0) cases per 1000 per month.
Within these regions, high burden and risk districts were also identified, both during the highest and lowest burden months. During June 2017 district monthly incidence reached the maximum in Lamwo of Acholi, Moyo of West Nile, Kaabong of Karamoja, Namayingo of East Central - Busoga, and Katakwi of Teso regions, at 167.6 (95% CI: 165.6–169.8), 192.5 (95% CI: 189.9–195.1), 81.1 (95% CI: 79.6–82.5), 73.1 (95% CI: 71.9–75.0), 72.0 (95% CI: 70.9–73.1), cases per 1000 per month, respectively (Table S6, Additional file 1).
Monthly incidence rate trends among districts showed that Moyo, Lamwo, Adjumani, Pader, Nwoya, and Maracha persistently recorded the highest monthly incidence rates across the study duration (Fig. 3). Moreover, higher incidence rates were also associated with higher variability in monthly incidence rates with the mean monthly estimate in Moyo at 115.8 (SD: 36.5) and lower rates less variability with Rubanda at 1.6 (SD: 0.5) cases per 1000 (Figs. S10 and S11, Additional file 1).
Within individual districts, a wide distribution of incidence rates was estimated among health facility catchments both during the lowest and highest burden months. From the 3446 catchment areas identified across the country, mean monthly incidence rate reached a maximum of 569.8 (95% CI: 555.2–584.3) cases per 1000 per month in Namayingo district of East Central – Busoga region and minimum of 0.13 (95% CI: 0.10–0.17) cases per 1000 per month in Rukungiri district of Kigezi region, excluding Kampala. Also, higher incidence rates within catchments were associated with higher viability in monthly incidence rates and lower incidence rates with less variability (Fig. S10, Additional file 1). Among health facility catchments, variability in incidence rates reached a maximum standard deviation (SD) = 142.4 cases per 1000 in highest incidence rate catchment located in Namayingo and a minimum SD = 0.1 among the lowest burden catchments in Arua and Kasese districts.
Spatial distribution of relative risk across the country:
Consistent with incidence rates, relative risk of malaria was highest among the highest burden regions of Acholi, West Nile, Karamoja, East Central – Busoga, and Teso, both during the lowest and highest burden months, maintaining their rank of risk at both times (Table S7, Additional file 1). During the highest burden month of June 2017, the relative risk of malaria among these regions ranged from 1.18 (95% CI: 1.17–1.19) to 2.6 (95% CI: 2.6–2.8)-times higher than national average in Teso and Acholi, respectively. Moreover, while mean relative risk among districts within these regions was higher during the highest burden month at 1.8 (95% Confidence Interval:1.5–2.1) than the lowest at 1.7 (95% Conf. I:1.4–2.0), the difference was not significant (p = 0.676) by a two-sample t-test.
Spatial and temporal variation in relative risk observed between regions, and districts within regions (largely informative at programmatic or NMCP levels), was also present between catchments within districts (informative for district health managers). Relative risk remained consistent among the 15 regions, between low and high burden seasons, but showed additional variability among districts and health facility catchments across the two seasons (Fig. 4).
Results showed that catchment risk ranged from 0 to 24.9 (95% CI: 24.4–24.9) times higher than national average during the highest burden month and from 0 to 50.5 (95% CI: 49.0–50.8) during the lowest burden month. Moreover, a non-linear association of catchment risk was observed between the lowest and highest burden months further confirming this rising risk during lower burden months (Fig. S16, Additional file 1). However, the highest risk catchments at the two time points were neither identical nor located in the same district or region.
Spatial clustering of risk:
Assessment for spatial autocorrelation of incidence and/or risk showed consistent levels of moderate global autocorrelation between both districts (Moran’s I range by month: 0.4 to 0.6, p < 0.001) and health facility catchments (0.3 to 0.5, p < 0.001). Both during the highest (June-2017) and lowest (February-2018) burden months, global autocorrelation between districts was very similar (Moran’s I = 0.5, p < 0.001) (Figs. 18 and 19, Additional file 1) but slight difference between health facility catchments (Moran’s I = 0.4 and 0.3, p < 0.001, respectively) (Figs. S20 and S21, Additional file 1).
Analysis of local spatial autocorrelation at two levels of significance (p < = 0.05 and p < = 0.01) identified substantial significant high-high clustering in Acholi and West Nile regions in the North, as well as East Central – Busoga region in the South East of the country, both during the highest and lowest burden seasons (Fig. 5). Similarly, large low-low clustering was identified in the Southern regions of the country. Moreover, outlier catchments typically had significantly lower risk than their neighbours in the north, and higher risk than their neighbours in the rest of the country. Significant monthly high-high clusters were comprised of between 191 health facility catchments during February 2018 and 236 during June 2017 and 2019 (Fig. S22, Additional file 1).