Eight-year Retrospective Analysis of Malaria Trends in Gedeo Zone, South Ethiopia (2012-2019)

Background: Informed decision making is underlined by all level of the tiers in the health system. Poor data record system coupled with under-(over)-reporting of malaria cases affects the country’s elimination activities. Thus, malaria data at health facilities and health o�ces are important particularly to monitor and evaluate malaria elimination progresses. This study was intended to assess overall reported malaria cases, spatiotemporal trends and factors associated in Gedeo zone, South Ethiopia, and compare malaria case reports by the health centers and health o�ces. Methods: Past eight years retrospective data stored in 17 health centers and 5 district health o�ces in Gedeo Zone were extracted. Malaria cases data at each health center with sociodemographic information, between 2012 and 2019, were included. Meteorological data were obtained from the national meteorology agency. The data were analyzed using Stata 13. Results: A total of 485,414 clinical suspects were examined for malaria during the previous 8 years at health centers. Of these suspects, 57,228 (11.79%) were con�rmed malaria cases. We noted, an overall under reporting of malaria, 3,758 clinical suspects and 467 con�rmed malaria cases were not captured at the health o�ces level. Based on the health centers records, Plasmodium falciparum (49.74%) was slightly higher (p = 0.795) than P. vivax (47.59%). The majority of cases were found in adults ( ≥ 15 years of age) that accounted for 11.47% of con�rmed malaria cases (p < 0.0001). There was high spatiotemporal variation: highest cases record was during autumn (12.55%) (p < 0.0001) and, the highest (18,150, 13.17%)) and lowest (5,187 (10.44%)) malaria cases were reported from Dilla town and Yirgacheffe rural district, respectively (p = 0.0002). Monthly rainfall and minimum temperature exhibited strong positive associations with the number of con�rmed malaria cases. Conclusion: A notable decline in malaria cases was observed over the eight-year period. Both P. falciparum and P. vivax co-exist; hence, control measures should continue targeting both species. The


Background
In Ethiopia, Plasmodium falciparum (P.falciparum) and P. vivax are the predominant causative species of malaria.The two coexist in almost all malarious areas at different levels of co-endemicity.Overall, large proportion of infections reported is due to P. falciparum (~ 60%) followed by P. vivax (~ 40%) [1,2] with micro-epidemiological and seasonal variation.Such co-endemicity makes malaria control and elimination more complicated in Ethiopia than in most other areas where the later species is very low [2,3].
Malaria transmission in Ethiopia is seasonal associated with precipitation and temperature changes; peaking from September to November following the large rainy season [4].However, construction of dams and irrigation-based agricultural activities sometimes modify malaria seasonal trend in Ethiopia [5,6].
The past decades witnessed a sharp decline in morbidity and mortality, putting Ethiopia among the few African countries on track to meet the global 2020 milestone of cutting incidence by 40% or more [7].These successes, encouraged the Ethiopian national malaria control program (NMCP) to stratify the country's malaria transmission into four based on annual parasite incidence (API); malaria free (API ~ 0 cases/1,000 population/year), low (API > 0 and < 5), moderate (API ≥ 5 and < 100) and high (API ≥ 100) [4], as a preparation to embark on nationwide malaria elimination.The policy and strategy shift to elimination requires data-driven decision making to tailor interventions [8,9].Thus, policy makers should be provided timely with quality and relevant data to inform national programs.
In Ethiopia, malaria data is captured through the health management information system (HMIS) at different tiers of the healthcare delivery systems.The hierarchy of data ow is from Health Posts (at kebele level) and Health Centers (HCs) to district health o ces (HOs) which in turn channels to Zonal Health Departments, then to Regional Health Bureaus and nally to the Federal Ministry of Health [10].
Therefore, health post and HC data that are organized and archived at respective district HOs are the ones that are analyzed to evaluate the spatial and temporal changes, local malaria dynamics and Plasmodia species distribution.
Although understanding malaria trend could help to recognize the progress of elimination activities, under-(over)-reporting of malaria cases could affect the country's elimination efforts.Yet, validation studies comparing data from the different tiers of the health care delivery system hardly exist in most settings and at micro-epidemiological level.Although the six malarious districts of Gedeo Zone are strati ed as elimination targeted low transmission districts by NMCP [4], little information is available to understand the overall trend of malaria and the above issues in the area.Thus, we assessed the species composition, malaria data quality, spatiotemporal trend and associated socio-demographic and climatic variables.Further, the accuracy of HO malaria records (HMIS data) was checked against the HC data (source document).

Study setting
Gedeo zone (Fig. 1) is 360 kilometers from Addis Ababa, it is one of the 14 zones in the Southern Nations, Nationalities and Peoples' Region.It is located 5°53'N to 6°27'N latitude, and 38°8' to 38°30' east longitude.The altitude of the zone ranges from 1,268 to 2,993 meters above sea level.The mean annual temperature is between 12.6 o C and 30 o C and the mean annual rainfall ranges from 1,001-1,800 mm.
Based on the Gedeo Zone health department report, 36.31%(423,411/1,166,163) of the population is at risk of malaria.Malaria transmission in the zone is seasonal with peak from September to November.The API of the zone in 2019 was close to 2.0.The zone is sub-divide into six districts and two town administrations [11].Districts, also known as "woredas" in Ethiopia, are the third level administrative divisions of the country, following regional states and zones and are further sub-divided into "kebeles" (the smallest administrative unit with its own jurisdiction).Six elimination-targeted settings, low transmission, by the NMCP; Dilla and Yirgacheffe towns, and Dilla zuria, Wonago, Kochore and Yirgacheffe rural districts were included [4].

Data collection
Eight-year (from 2012 to 2019) data were collected from malaria laboratory registration logbooks in the HCs and HMIS of the district HOs.Six district HOs and seventeen public HCs with at least 8 years of service which report them were covered.HC records with missing information of cases; address (kebele/district), dates of HC visit, age, sex or results of malaria diagnosis were excluded from the main part analysis.These excluded data were again analyzed separately to address the data quality issues at HCs.Data on malaria diagnosis results (negative, or positive, and infecting Plasmodium species for positives), time of diagnoses (date/month/year) and socio-demographic data (kebele/district, age and sex) were collected.In addition, meteorological data; station level monthly and annual precipitation, maximum and minimum temperatures and relative humidity was obtained from the national meteorology agency of Ethiopia.Data collectors attended adequate training to assure quality.Further, the consistency and completeness of the extracted data was checked for each HC and district.

Data analysis
Microsoft o ce excel worksheet 2019 and the Stata data software 13 (College Station, Texas 77845 USA) were used for data entry and analysis.Descriptive statistics was used to show the distribution of malaria cases with respect to months, years, sex, age, Plasmodia species and district.Pearson's chisquare (X 2 ) test was used to determine the association of the different variables with con rmed malaria cases.Logistic regression was also performed to assess the association of malaria cases with sociodemographic variables, seasons and districts.Odds ratio (OR) with the corresponding 95% con dence interval (CI) was used to assess the differences in malaria prevalence with selected predictors.Spearman correlations were used to measure the strength of association of monthly malaria cases with meteorological variables.P-value of less than 0.05 was taken as statistically signi cant.

Year-based data with missing variables at HCs
In respective order 1852 (0.38%) and 249 (0.43%) of clinical suspects and con rmed malaria cases were excluded from the downstream analysis due to incompleteness.Overall, there was a signi cant reduction in missing data between 2012 and 2019 (p = 0.001).During the rst two years there were more numbers of both suspected and con rmed cases of missing variables.Speci cally, the highest data with missing variables (413 clinically suspected and 76 con rmed cases) occurred in 2012.Whereas, the lowest records of missing variables with 107 clinically suspected and 10 con rmed cases were reported in 2019 and 2018 respectively.Generally, the number of clinical suspects with missing variable declined from 413 in 2012 to 107 in 2019 by approximately 4-fold and con rmed cases from 76 in 2012 to 11 by 7-fold in 2019 (Table 1).

Annual clinical suspects and con rmed malaria cases based on HC versus HO records
There is an evidence of statistically signi cant inconsistency (p = 0.041) in both clinical and con rmed malaria case reports between HCs and HOs, overall, the 8-years period (2012-2019).Higher clinical suspects (485,414) were recorded at HCs compared to the HOs (481,656).Similarly, the corresponding con rmed malaria cases were 57,228 (11.79%) at the HCs and 56,761 (11.78%) at HOs although establishing which one is more accurate is rather not easy.With this, the number of clinical suspects recorded by the HCs was higher by 3758.The data kept by the HOs was lower on average by about 470 each year (except 2012, 2018).This difference was pronounced in 2016.During 2012 and 2018 the numbers of clinical suspects recorded at HCs were consistent with HO data.In addition, the HMIS captured on average 58 less con rmed malaria cases each year except in 2012 and 2018.In 2012 and 2018 the numbers of con rmed cases were higher at HO records than at HC records (Table 2).
On the other hand, in winter the proportion of infection due to the two species, P. falciparum (5451 (49.51%)) and P. vivax (5312 (48.25%)) was relatively lowest (Fig. 3).Although P. falciparum was higher than P. vivax in all seasons, the difference was the smallest (1.26%) during winter, while in autumn it is 3.17%.The number of mixed infections also had same pattern throughout the four seasons.
The monthly total con rmed malaria cases showed a strong and positive association with monthly rainfall (Spearman's rho = 0.895, p < 0.0001) and monthly minimum temperature (Spearman's rho = 0.671, p = 0.017).Nevertheless, the maximum monthly temperature was negatively and weakly associated with the monthly con rmed malaria cases.Relative humidity seems not to have an effect on malaria transmission in the study area (Fig. 5).

Discussion
In the present study there was a signi cant reduction in data incompleteness; with missing variables between 2012 and 2019.Speci cally, the number of clinical suspects with missing data variables declined between 2012 and 2019 by approximately 4-fold and con rmed cases with about 7-fold.Decreasing trend in data incompleteness overtime, is plausible indication of huge improvement in data quality which thereby an implication for improving programmatic performance over the years.
There was an overall notable difference in both total number of clinical suspects (3758) and con rmed malaria cases (467) recorded at HCs (source data) and HOs (HMIS data).For the 8-years period, discrepancies in clinically suspected and con rmed cases were observed between the HOs compared to the HCs.During 2012 and 2018 the numbers of clinical suspects recorded at HCs were consistent with HO data.In these years the numbers of con rmed cases were higher at HO records than HC records.
Yirgacheffe rural district had the highest deviation in terms of con rmed malaria records between HC and HO data followed by Dilla town and Dilla zuria district, whereas Wonago district was the least.Similar to our nding, a three-month facility-based study comprising of various settings conducted in southern Ethiopia showed that majority of facilities under-reported total malaria (both con rmed and clinical malaria) cases [10].This deviation in malaria data between the two systems, HC and HO, could be due to errors during entering the data from the sources (HCs) into recording formats of HMIS, lack of cross-checking and proo ng habits, training gaps on HMIS data use and unintentional/intentional false reports.
In addition, limited computer access and skill, inadequate technical support [12], poor data management skills and limited functionality of electronic data management systems [13] might be the likely reasons for HMIS implementation challenges.
Concerning the diagnostic performance, the proportions of clinically suspected cases against con rmed cases were increased from 6.91 in 2012 to 16.32 in 2019.The highest proportion was recorded in 2019, whilst the lowest was during 2013.This revealed that the 'non-malarial' febrile cases were increasing from 2012 to 2019, which can be an indication of declining lab capacity of detecting malaria parasites.This declining lab capacity of detecting malaria might be due to the fact that the increased false negative reports associated with reduced sensitivity of microscopy with decreasing parasite densities [14], unable to detect sequestered P. falciparum parasites [15] and low competency of microscopists [16].In the other way, such increased number of non-malarial febrile illnesses might be related to other febrile cases including yellow fever virus [17,18] and typhoid fever [19] infections, as per the studies conducted in southern Ethiopia.In addition, this high number of non-malarial febrile illness might be due to fevers among positive individuals with malaria where the fever is coexisted with but not caused by the Plasmodia infection [20].If laboratory performance percent con rmed declines it means; laboratory performance was decreasing over the years or something causing febrile illness in the area is increasing.
Misdiagnosis and incorrect treatment of such non-malarial febrile illnesses with antimalarial drugs is possibly to contribute to rapid emergence of antimalarial drug resistance [21,22] in the study area.
There was an overall reduction of malaria case from 2012 to 2019.According to data from HCs (source data), a maximum of 16,037 and a minimum of 2,546 of cases were observed during 2013 and 2018, respectively with 8.34 percent reduction.2013 and 2016 were the exceptions to the declining trend as there were small epidemics in these periods in some parts of the Zone.Over the eight years period, overall, there was malaria positivity rate of 11.79%, data from the HCs.This positivity rate was comparable with certain studies done in Ethiopia including from Batu town (12.43%) [23], Arsi Negelle (11.40%) [24] and Halaba special district (9.47%) [25].In contrast, higher overall malaria positivity rates were reported from related studies conducted in south-central Ethiopia [26], southern Ethiopia [27] and abroad in Dakar, Senegal [28] with 33.83%, 21.79% and 19.68% respectively.On the other hand, the present gure was higher than records in other local studies [29,30].These differences might be due to the variation in quality of laboratory diagnoses, difference in intervention measures, microclimatic/altitudinal differences, and presence of constructions responsible for occurrence of temporary and permanent dams and drug and insecticide resistances.The possible contributing factors for the peaks/epidemics of malaria cases in 2013 and 2016 could be associated with feeble intervention activities in certain areas of the Zone.
P. falciparum and P. vivax were detected where equivalent; congruent results were reported in some other parts of Ethiopia [26,27].While other local studies [23,25,31] documented that the dominant species was P. vivax.The proportion of mixed infection in this study was congruent with other studies [26, 30], whereas inconsistent with other reports [27,31].The likely reason for the slightly higher proportion of P.
falciparum over P. vivax could be related to temperature; that is temperatures more than 18 °C for P. falciparum and more than 15 °C for P. vivax is suitable for the growth of these two species in human host and mosquito vectors [32,33].Apparently, in the current study area the average mean temperature during the eight years period found to be above 18 °C.The higher proportion of P. vivax against the national gures could also be an implication for the ability of repeated relapse cases and early emergence of gametocytes during blood-stage infection.In addition, there could be heterogeneity of the Duffy phenotype and the high number of vulnerable Duffy-positive individuals that associated with population movement [34] in the study area.Environmental uctuations that change target mosquito species abundance might have an impact on Plasmodia species occurrence [35].The issue demands additional study.The possible reason for scarcity of mixed infections in this co-endemic area might be a competitive or an antagonistic effect of one Plasmodium species over the other within the human host during co-infection [26,35].
Males were slightly more infected (51.51%) by malaria than females (48.49%) over the eight-year period.This was paralleled with other studies conducted in different parts of Ethiopia [23,26,27,31].However, this nding was not consistent with other reports in southern Ethiopia [25] and elsewhere in Mozambique [36] where higher malaria cases in females were documented.Individuals in the age group of 15 and above were also more signi cantly affected.This was in line with other local studies [23,31].Inconsistent result was observed in southern Ethiopia [24] arguing that malaria cases cluster among the under-5.And a nding in Metema, northwest Ethiopia by Ferede et al. [37] showed that 5-14 years old were more infected.Possible justi cations for the higher occurrence of malaria among males and older age group could be their engagement in various outdoor activities and staying outdoors during the nights [38].Apart from outdoor exposures, differences in treatment-seeking behavior, access to health facilities and travel history [39] might be the possible contributors for the sex-and age-based variations of malaria cases.In addition, a review report revealed that adult females are better protected from parasitic diseases than males due to genetic and biological (hormonal) factors [40].
The peak number of con rmed malaria cases was recorded during autumn followed by spring, summer and winter with a statistically signi cant variation.This seasonal peak in malaria cases in autumn deviates from various studies in Ethiopia which is during spring after the main rain season [24,26,27,31].In addition, nationally the main malaria transmission season is from September to December following the peak rain season [1,4].This trade-off in seasonal peak of malaria cases might be as a result of varying climatological conditions (rainfall pattern and temperature changes) in the area against other settings.Despite the high prevalence of malaria cases during the three seasons, there were a substantial number of con rmed malaria cases in the dry (winter) season in this study.These, absence of signi cant variation in the proportion of the P. vivax and P. falciparum burden and almost, year-round presence of malaria might suggest the presence of suitable local environments for mosquitoes.The comparable proportion of P. vivax against P. falciparum in winter might be explained by the fact that P. vivax has ability to relapse rather than new infections.Since such traits could affect the temporal patterns of P. vivax infections.
Overall, high number of clinical suspects and con rmed malaria cases were documented in Dilla town (urban) and Dilla zuria district (sub-urban) as compared to other districts.Except in 2013 and 2019, Dilla town annual malaria cases remained the highest all over the 8-year period.The highest con rmed cases during these two years were overtaken by Kochore in 2013 and Dilla zuria districts in 2019.While the lowest was from Yirgacheffe rural district.In general, though an overall declining trend of con rmed malaria cases from 2012 to 2019, peaks were recorded in Kochore during 2013 and Dilla town in 2016.Although there is expectation of a better documentation, treatment-seeking behavior, access to health facilities, community knowledge and coverage of intervention activities in urban settings, the current data pointed to the contrary.Thus, in this study, high burden of urban and suburban malaria was noted.This could be because of massive construction activities (like road, house and small dams) and presence of coffee processing sites in Dilla town and its vicinity that could create suitable habitat for mosquito breeding.Travel history [39], differences in the competence and skills of the laboratory personnel and relatively good reporting system might also be the main responsible factors in uencing the prevalence of malaria in Dilla town compared to rural districts.There have been healthy ongoing malaria control activities incorporating environmental management, indoor residual spraying (IRS), long-lasting insecticide-treated nets (LLINs) and artemisinin-based combination therapy in the area.These intervention activities could be attributed for the decreasing trends of malaria in other sites of the Zone.In addition, micro-environmental variations, micro-climatic situations [35] and changes in intervention (like IRS and LLINs) periods might have effect for these spatial differences of malaria cases.
Monthly rainfall and minimum temperature demonstrated statistically signi cant positive correlation with malaria cases.Previous studies in Ethiopia [41,42] and elsewhere [43,44] documented similar ndings.
However, the result of the current study on the association of rainfall and malaria cases was deviating from previous ndings, stating higher rainfall does not necessarily in uence the malaria changes [45,46].
In contrast to our nding, minimum temperature was weakly correlated with malaria cases in southeast Ethiopia [42].Ideally, rainfall and minimum temperature play a vital role in breeding and survival of malaria vectors and the respective parasites.Moreover, average monthly maximum temperature and relative humidity were weakly correlated with malaria cases.In disagreement to our nding, studies conducted in Jimma, Ethiopia by Alemu and others [41] and Sena and colleagues [42] in Gilgel-Gibe, southwest Ethiopia reported that inter-monthly relative humidity was signi cantly associated with monthly malaria cases.
The limitation of this study was incompleteness of patient data in the register with missed variables and only 8-year data were available during the data collection time at the HCs.Missing of asymptomatic cases and poor competence of the laboratory personnel at HCs could be the other limitations.Furthermore, clinically treated patients' (without laboratory con rmation) data and malaria mortality data were not recorded in the laboratory registration logbooks.Hence, interpretation of the nding should be with caution.

Conclusion
Over the entire period, 2012 to 2019, the program data quality improved over the years; underreporting and proportion of data incompleteness declined signi cantly, and the burden dropped continuously except in 2016.Yet, equivalent P. falciparum and P. vivax malaria existed, thus intervention actions should target both species.The high malaria prevalence in urban setting (Dilla town) and its vicinity and in autumn season necessitates spatiotemporal consideration by the control campaign.The intervention strategies need also consider older age groups.In general, malaria still remains a public health problem in the area, which demands strengthening of interventions and short-term forecasting based on local meteorological factors to achieve elimination goals in the area.In addition, the data recorded at the HCs and district HOs should be monitored for consistency.
EM designed the study, involved in data collection, analyzed and interpreted the data, and drafted the manuscript.SWB contributed in data analysis works.FGT initiated the idea, involved in the data entry template development and gave feedback on the data collection protocol.SD was responsible for revising the draft manuscript.EG initiated the idea, guided the process of data collection and substantively revised the manuscript.HM participated in guidance of the data collection progresses and critically commented the whole section of the paper.All authors read and approved the nal manuscript.Map of the study districts; Gedeo Zone Southern Nationals Nationalities and Peoples Regional state, Ethiopia

Figures Figure 1
Figures

Figure 2 Annual
Figure 2

Table 2
Numbers of clinical suspects (N) and con rmed malaria cases (n) by year; HCs and HOs data, Gedeo zone, South Ethiopia, 2012-2019 from 6.91 in 2012 to 16.32 in 2019.The highest proportion (16.32) was recorded in 2019, whilst the lowest (5.59) was during 2013.This shows the proportion of 'non-malarial' febrile cases were increasing from 2012 to 2019 except in 2016 (Table3).
(30,: nu11er of cases; %: percentage (n/N*100), Pf: Plasmodium falciparum; Pv: Plasmodium vivax Overall, a notable decline in malaria, HCs (source data), was observed during the eight-year period, Diagnostic performance; data from HCs Regarding the diagnostic performance, the yearly trend of clinical suspects was directly proportional to the con rmed cases in each year.But the proportions of examined clinical suspects against con rmed cases were increased Malaria cases number by sex and age; data from HCs Slightly more males (29,480 (11.34%)) were malaria positive (p = 0.236) than females(27,748 (12.30%)) (Fig.2).The above 15 years age group was the most affected(30,406,11.47%)thantheother age groups, followed by under 5 children which accounted for 15116 (13.83%) of the cases.The above 15 years age group was twice more likely (OR = 2.00, 95% CI: 1.90, 2.11, p < 0.0001) to have malaria compared to the under 5 (Table4).