Are there variations in hotspots in clinical malaria in pregnancy and neonatal malaria? Evidence from Ghana

Malaria in pregnancy (MIP) is a signicant public health problem, and affects both mother and foetus. MIP has both direct and indirect effects. These direct effects attributed to malaria in pregnancy are clinical malaria in pregnancy and congenital malaria. Although there has been a steady decline of malaria in pregnancy (MIP); reducing from 9% in 2004 to 1.4% in 2016, the decline, does not seem equitable across the country. There was therefore a need to identify areas where clinical MIP and neonatal was high and target them for interventions (neonatal malaria was used because data for malaria was available only for the rst 28 days). As a result, this study sought to use routine data to identify areas of high transmission (hotspots) for clinical malaria in pregnancy and neonatal malaria (extrapolating for congenital malaria hotspots).

equitable across the country. There was therefore a need to identify areas where clinical MIP and neonatal was high and target them for interventions (neonatal malaria was used because data for malaria was available only for the rst 28 days). As a result, this study sought to use routine data to identify areas of high transmission (hotspots) for clinical malaria in pregnancy and neonatal malaria (extrapolating for congenital malaria hotspots).

Methods
Clinical MIP and neonatal malaria (suspected and con rmed) data were retrieved from the national routine database. Using expected pregnancy and all clinical malaria cases as the denominators of clinical malaria in pregnancy and clinical malaria respectively, the per 1000 cases of malaria in pregnancy and percentage neonatal malaria were calculated for both conditions. These two data sets were fed into ARC GIS software and hotspots spatially determined for both conditions from 2014 to 2016.

Results
Variations in hotspots were identi ed both chronologically and geospatially for clinical MIP and neonatal malaria respectively. Nevertheless, stable hotspots were found for clinical malaria in pregnancy for the three years and stable hotspots also identi ed for Neonatal malaria for two years. Congenital malaria was extrapolated from neonatal malaria data only for Bole District in the Northern Region of Ghana for 2015.

Conclusions:
Hotspots for malaria in pregnancy show marked variation year on year both for clinical malaria in pregnancy and neonatal malaria and vary geospatially. The differences in hotspots in clinical malaria in pregnancy and neonatal malaria also show that neonatal malaria cannot be attributed to congenital malaria and therefore is not the effect of malaria in pregnancy. However, stable hotspots for clinical malaria in pregnancy should be targeted for intervention.

Background
Malaria, during pregnancy, is a signi cant public health problem particularly in sub-Saharan Africa. It affects both the mother and the foetus. (1). Pregnant women constitute the main adult risk group which is serious for the foetus in malaria endemic countries. The condition may be asymptomatic or symptomatic, and manifests as follows: clinical malaria (symptomatic), maternal anaemia, spontaneous abortion, preterm birth, low birth weight, intrauterine growth retardation, preterm delivery, foetal anaemia, stillbirth and rarely congenital malaria ( may manifest as neonatal malaria) (2). These effects of malaria in pregnancy are particularly found in the primigravid and secundigravid women(1) and women living with HIV (3).
Malaria is caused by plasmodium sp and transmitted by the infective female Anopheles mosquito. The most vulnerable groups are children under ve years (accounting for 70% of all 429 000 deaths worldwide in 2015) (4), pregnant women (causing about 10,000 deaths per year) (3) and migrants to endemic countries for less than six months.
Ghana is mesoendemic for malaria and the disease is found in all areas of the country. Prevalence of malaria in Ghana averages 21% among children under ve years (indicator used to assess malaria prevalence); ranging from 25% in Northern Region to 5% in Greater Accra (5).
The National Malaria Control Programme (NMCP) has since 2003 revamped interventions to control and reduce the burden of malaria in Ghana. These efforts have contributed to reducing the burden of malaria in pregnancy over the past several years. For example, in 2016, malaria in pregnancy contributed to 1.4% of all malaria outpatient attendances as compared to 9% in 2004(6).
Nevertheless, the NMCP has postulated that the reduction may not be even in all parts of the country. Identifying areas of greatest burden and morbidity, referred to as hotspot, is necessary. It helps in determining populations with greater incidence to enable countries to design targeted control and prevention measures (7). By de nition, a hotspot is an area with a higher-than-average level of transmission intensity or burden of disease. Hotspots are not straightforward to determine but are important in ensuring targeted interventions (8). Identi cation of a hotspot could be a challenge. To address this challenge, some authors have used spatial distribution in the form of hotspots to identify areas of the highest burden [7, [8].
Some authors have used geographic information system (GIS) to identify hotspots of disease prevalence both at the national and international levels. Fobil used GIS to identify hotspots and coldspots relating to malaria and diarrhoea-related mortality in Accra (10). Kenu also used GIS to identify communities with the greatest concentration of buruli ulcer cases along the Densu River in Ghana (11).
Different authors have used various indicators to identify malaria hotspots. These include malaria incidence, asymptomatic parasite carriage, reported fever, drug use, serological responses to malariaspeci c antigens, mosquito abundance, and exposure to infected mosquitoes (8,9). Some other authors have identi ed hotspots at the community level. Bousema et al (12), used age-adjusted prevalence and density of antibodies against malaria to determine hotspots of malaria in a community. In other instances, routine data has been used to identify hotspots. Kamuliwo et al, tracked monthly reported district-level clinical cases of malaria in pregnancy using routine Zambian District Health Information System to detect malaria hotspots in the country (9). This study therefore sought to ascertain the existence of clinical malaria in pregnancy and neonatal malaria hotspots in the Ghana, and show whether the reduction in malaria in the country is inequitable. It also determined whether districts with hotspots for neonatal malaria could be attributed to hotspots for congenital malaria. This study also determined areas of high malaria morbidity (i.e. malaria in pregnancy and congenital malaria) through geospatial identi cation of hotspots. This is because clinical malaria in pregnancy and congenital malaria are conditions directly linked to malaria in pregnancy.

Study Design and Population
The study was a retrospective cross-sectional review of routine data in District Health Information Management Systems 2 (DHIMS2). The population covers pregnant women who have reported with clinical malaria in pregnancy (con rmed and uncon rmed). The second set of study population are neonates (infants less than 28 days) with con rmed and uncon rmed malaria.

Study area
Ghana is endemic for malaria with an estimated population of 28.21 million people in 2016 with population growth rate of 1.9%, projected from the 2010 Population Census. Expected pregnancy is estimated at 4% of the population (13). The rural-urban ratio is 47:53 of the total population (14). Ghana has 10 administrative regions, which are further divided into 216 districts based on the population and land size.

De nition of Variables
In this study, the term 'expected pregnancies' is de ned as the number of women in the reproductive age in the population who are expected to become pregnant within each year. Neonatal malaria refers to uncomplicated malaria suspected cases in children 28 days or less reported on the Monthly Morbidity Out-Patient-Department (OPD) returns. Clinical malaria in pregnancy is de ned as all pregnant women reporting at health facilities with clinical symptoms of uncomplicated malaria; whether con rmed or not.
The information was extracted from the population data form on the DHIMS2 platform and the 2010 Population and Housing Census.
Z-score refers to districts whose clinical malaria in pregnancy and/or neonatal malaria per 1000 respectively are zero, one (+/-1), two (+/-1.96) or three (+/-2.58) standard deviations away from the mean of each district and its neighbouring districts; number of neighbouring districts ranging from 5 to 10.

Study participants
The study participants were pregnant women with clinical malaria in pregnancy (con rmed and uncon rmed) and neonates; i.e. infants of 28 days or less with malaria symptoms which were either con rmed or uncon rmed.

Data Collection
All the data used in this study were extracted from the DHIMS2 platform for 2014-2016. The incidence per 1000 was calculated as follows: Clinical Malaria in Pregnancy: "Uncomplicated Malaria Suspected" data were extracted from Monthly Morbidity OPD Returns for the years 2014-2016. The incidence per 1000 expected pregnancies, calculated using data on the expected pregnancy for districts Population Data extrapolated from the 2010 Population and Housing Census. Both were obtained from the DHIMS2 platform and the z-scores were calculated for each districts). The study population is all cases of uncomplicated clinical malaria in pregnancy (whether tested or not) that was reported on the DHIMS2 platform and all expected pregnancies for all districts.
Data of uncomplicated neonatal malaria (con rmed and uncon rmed) incidence were retrieved for 2014 to 2016 from DHIMS 2 platform. Data on incidence of suspected malaria in children less than 28 days were extracted from the same Monthly Morbidity OPD returns form. All uncomplicated malaria suspected cases for all ages were also extracted from the same form and used as the denominator.

Statistical Analysis
Data on clinical malaria in pregnancy per 1000 and neonatal malaria percentage of all malaria cases were generated and fed separately into Aeronautical Reconnaissance Coverage Geographic Information System (ArcGIS). The mean for all districts was calculated and the z-scores in relation to neighbouring districts obtained. These were plotted spatially using the coordinates of districts.

Hotspot Analysis
The Hot Spot Analysis tool calculates the Getis-Ord Gis statistic for each district. The resultant Z-score gives an indication of which districts have either a high or low value cluster spatially. Each district was assessed in the context of neighbouring districts. A spatial weighted matrix was then created with the following spatial relationship conceptual framework: Spatial relationship was based on contiguity edges and corners (i.e. which districts share a boundary, a node, or overlap) as means of de ning neighbours. Average number of neighbours was six districts, minimum and maximum number of neighbours were 5 and 10 districts respectively.
A hotspot analysis was done using the parameters of the spatial weighted matrix with a Euclidean Distance method to determine clustering of high and low gures of malaria in pregnancy across districts geographically.
A high z-score and small p-value for a district indicated a spatial clustering of high values. A low negative z-score and small p-value indicates a spatial clustering of low values. The higher or lower the z-score, the more intense the clustering. A z-score near zero indicates no apparent spatial clustering.

Results
Demographic Characteristics Table 1 shows that total the number of cases of clinical malaria in pregnancy identi ed among girls and women

Discussion
Malaria morbidity and mortality in Ghana have been largely declining [6, [15]. This decline has also been found in malaria in pregnancy; a condition which has direct effects of clinical malaria in pregnancy and congenital malaria (using neonatal malaria as a proxy in this article). However, authors of this study doubted that the decrease in malaria in pregnancy was not equitable across the country. Consequently, this study sought to establish the existence of hotspots of malaria in pregnancy (if any) in Ghana. These hotspots were measured using incidences of clinical malaria in pregnancy and neonatal malaria.

Clinical Malaria in Pregnancy
The average proportion of clinical malaria in pregnancy was 347.60 per 1000 women. The proportion was highest in 2016 compared to 2014 and 2015. The high incidence for the 2016 year may be attributed to better reporting and not an increase in incidence. Indeed, the introduction of routine data reporting on the platform DHIMS2 has improved to a large extent the quality of data collated from health facilities. Therefore, an increase in proportion of clinical malaria in pregnancy could be linked more to improved reporting. Correspondingly, the assertion of better reporting has also been made by Rouamba  zero z-score for these three years is consistent with ndings from Aregawi et al (15). That is, Aregawi et al (15) identi ed that among all ages, outpatient malaria cases dropped by 57% (95% CI, 47-66%) by rst half of 2015.
Nevertheless, hotspots were identi ed in the country. Some hotspots were stable while others showed variation both geospatially and chronologically year on year. This nding contrasts the results of Bousema at al (12) who identi ed stable hotspots in his study.
Stable hotspots were identi ed in the southern part of Western Region for the three years. They include Prestea Huni Valley, Tarkwa Nsuem, Wassa East and Mporhor. Findings of stable hotspot are consistent with those documented by Bousema and others (12) and Kamuliwo et al (9) for clinical malaria in pregnancy. Kamuliwo et al (9) also found stable hotspots (z-score > 2.58) in the north-eastern and southeastern parts of Zambia.
Another point of interest is that, the districts in which these hotspots were found are largely rural. This nding is consistent with literature, in which greater levels of malaria transmission are found in rural areas compared to urban areas. For example, a desk review, triangulated data from entomological studies and the 2011 Malaria Indicator Cluster Survey (MICS) in the 2012 Ghana Malaria Urban Study (17). The study con rmed higher malaria transmission in rural areas . Stable hotspots in these rural areas may be due to poor health seeking behaviour because of barriers with respect to geographical access. Such an assertion has also been made by Wiru, Oppong, Gyaase, Agyei, Abubakari, Amenga-Etego, Zandoh & Asante(18). Wiru et al(18) attributed malaria mortality hotspots in their study to districts being mostly rural.
A second reason for these stable hotspots was that they were found mostly in the Forest Ecological Zone. These districts have had their environment largely degraded by illegal mining, resulting in large pools of water suitable for the breeding of the anopheles mosquito (19). The proliferation of such collections of water is likely to lead to an increased number of mosquitoes. Therefore, this increase in mosquitoes, would result in elevated entomological inoculation rates and its related upsurge in malaria transmission rates (20).
In spite of the stability in some hotspots, there were also variability in other hotspots for clinical malaria in pregnancy. In Western Region for example, hotspots reduced from 15 in 2014 to 7 in 2016. This nding is consistent with the reduction of parasite prevalence of malaria (among children less than 5 years) in country-wide surveys of 2014 MICS and 2016 MIS. These two surveys also showed a reduction of parasite prevalence from 39% (2014 MICS) to 22% (2016 MIS) (21). It would imply that malaria transmission has reduced signi cantly within the two years.
In fact, variability in hotspots in those areas may also be attributed to the level e ciency of the health system; using its six building blocks (i.e. leadership/governance, healthcare nancing, health work force, service delivery, health information system and medical products and technologies (22)). Of particular relevance to this study is the improved availability of malaria logistics (in this case, the availability and use of ITNs and sulphadoxine-pyrimethamine (SP) for intermittent preventive treatment (IPTp)).
Unfortunately, factors such as supply chain blockages have been found to be a challenge. For example, availability of SP varies year-on-year depending on whether there is accessibility, shortages or maldistribution at the facility level (23). For instance, in 2016 there were stock outs of SP at the facility level but there were stocks at the regional level (6). Indeed, Bhushan and Bhardwaj (24) identi ed that some health system effects such as quality of care and critical shortages in commodities had impact on transmission, further impeding reduction in maternal and newborn mortality.
Lastly, variability in hotspots may also be linked to pathogen's natural history (20). For instance, Paull et al (20) identi ed geospatial heterogeneity as being responsible for transmission of diseases such as SARS and Typhoid. They however concluded that such variability was caused by host factors in 20% of the cases and transmission events in 80% of cases.

Neonatal Malaria
Studies on neonatal malaria are scarce because neonatal malaria is not thought to be common.
Therefore, studies to identify hotspots of neonatal malaria are not common either. Albeit, Makhtar (25) in his review of literature reported that neonatal malaria is not as rare as was originally anticipated. Indeed, a meta-analysis by Park Hence, it was not surprising that the review of neonatal malaria incidence for the years 2014-2016 from routine data showed that neonatal malaria in Ghana existed. The review further showed that there were less male neonatal malaria cases reported than females (1:1.15) ( Table 2). Lower proportion of males than females with neonatal malaria contradicts ndings by Runsewe-Abiodun et al in Nigeria; who found more males than females (1.6:1) had neonatal malaria (28). Runsewe-Abiodun et al's study however reported that the infection was not sex-linked and hence the difference in sex distribution in the two studies cannot be linked to the disease (28).
Our study also revealed that Ghana largely had a zero z-score for neonatal malaria; although hotspots were identi ed in all the three years under review. Nevertheless, these hotspots for neonatal malaria showed variability year-on-year as with those found for clinical malaria in pregnancy. Hotspots for neonatal malaria were found in the forest ecological zone and rural areas. Hotspots in these identi ed areas are in tandem with literature because forest zones and rural areas have been documented to have high transmission rates (17).
These hotspots even increased in number in 2016. The increase in hotspots may also be due to better reporting as has been stated above. The increase in hotspots may also be attributed to increased resistance of malaria parasites to the prevailing antimalarial drugs and increased virulence (29).
In spite of hotspot variability, some districts were consistent with hotspots for neonatal malaria for 2015 and 2016 in Ashanti and Brong Ahafo Regions. In addition, it is noted that there was a persistent hotspot in Bole in the western border of Northern Region for two consecutive years (2014-2015). Since Bole is in the western border with la Côte d'Ivoire, it could be that there was consistent transmission across borders.
In 2016, some parts of Northern Ghana (Kumbungu, Tolon and Wa West) had hotspots for neonatal malaria. Some of these districts had the intervention of Indoor Residual Spraying (IRS) discontinued. The emergence of these hotspots may thus be due to suspension of IRS; leading to a rebound effect of malaria cases; particularly for the neonates. This is of particular importance because malaria transmission in Northern Ghana tends to be more seasonal and therefore children in these part are more vulnerable.
Notwithstanding, Kumbungu continues to receive IRS and therefore hotspot can not be completely attributed to a rebound malaria incidence. Additionally, the attribution may not necessarily be associated with cessation of IRS because ndings from Kamuliwo et al's work (9), showed the protective effect of IRS.
One must note, albeit, that the neonatal hotspots are few and therefore largely IRS has had protective effect in other districts particularly because there is signi cant reduction in prevalence of malaria in the three regions in the north (5). It will be informative to undertake further studies to ascertain the persistent case of hotspots in Northern Ghana.

Interaction of Clinical Malaria in Pregnancy and Neonatal Malaria Hotspots
Generally, districts with hotspots for clinical malaria in pregnancy and neonatal malaria respectively were not the same. This buttresses the ndings of Menendez and Mayor (30) that neonatal malaria may have been and identi ed within the rst few days of life.
Yet, of signi cant interest is that of Bole District, which was found to have elevated levels of both clinical malaria in pregnancy and neonatal malaria for the year 2015. It is therefore supposed that a signi cant proportion of the clinical malaria may have been vertically transmitted from mothers to the neonates; giving incidences of congenital malaria. This would be recorded as neonatal malaria because the DHIMS 2 data collection platform does not segregate data into rst 7 days and above. This assertion is supported by the meta-analysis by Park et al (26). Indeed, Nhama, Varo and Bassat (31) in their commentary have called for further investigation into incidences of suspected congenital and neonatal malaria cases as their burden may be more than may have been originally thought of. Further investigation into effects of stillbirths, congenital anaemia from the neonate in relation to the women with clinical malaria in pregnancy is needed to con rm if reported neonatal malaria in Bole can be attributed to congenital malaria.

Limitations
At the time of data mining, the country had some challenges with data of con rmed cases and therefore it was decided that presumptive diagnosis of malaria be used. This gave a uniform description of clinical malaria in pregnancy and neonatal malaria in pregnancy. Albeit, the ndings may not re ect data of con rmed cases.

Conclusion
The country averagely had zero z-score regarding levels of clinical malaria in pregnancy and neonatal malaria. Hotspots were identi ed for all three years under review but showed year-on-year variations. In the case of clinical malaria in pregnancy, there were stable hotspots in parts of the Western Region and few districts in Central Region. Also, stable hotspots of neonatal malaria were found in parts of the Ashanti and Brong Ahafo Regions and in some parts of the northern regions. It must be noted however that these hotspots were in different parts of the country for clinical malaria in pregnancy and neonatal malaria respectively for the different years.
An exception was found in Bole in 2011, where a hotspot was found for both clinical malaria in pregnancy and neonatal malaria. Therefore, it is suspected that for the year 2015, some neonatal malaria cases reported may have been congenital malaria. It is recommended that Bole be monitored and if a