Ten-year trend of mortality among patients hospitalized with Malaria in Ghana,

Background Multiple interventions have been implemented over the years to decrease malaria morbidity and mortality in Ghana. After years of rolling out these interventions, assessing its effect on the trends for decision making is key. The objective of this study was to understand the trends of malaria related deaths in the country from 2005 to 2014. Methods Between September 2016 and June 2017, abstraction of retrospective data covering January 1st 2005 to December 31st, 2014 was conducted in 93 sampled health facilities providing AIDs/HIV, Tuberculosis and Malaria (ATM) services in Ghana. This paper is written out of a bigger study which looked at the mortality of ATM in Ghana. Abstraction form was used to retrieve socio-demographic and admission outcome of patients from facility registers, death certicates and inpatients registers. Data was entered using EpiData 3.1 statistical software package then exported to STATA 11 version for analysis. Bivariate analysis with Chi-square test and multiple logistic regression were done to assess factors associated with malaria related mortality at a 5% level of signicance.


Background
Malaria is endemic in Ghana with the entire population at risk with the most vulnerable being children under ve years, pregnant women and the immuno-compromised. The World Health Organization (WHO) estimated about 198 million cases of malaria and 584 000 deaths globally in 2013. Malaria mortality rates have fallen by 47% and 54% globally and in Africa respectively since 2000 [1]. Most of the malaria deaths occurred among children living in Africa where a child dies every minute from malaria. Malaria mortality rates among children in Africa by 2013 have been reduced by an estimated 58% since 2000 [1].
Data from the District Health Information Management System for Ghana (DHIMS-II) 2014, revealed that malaria was a major cause of hospital attendance, comprising about 30% of Out-Patient Department (OPD) visits, 27.9% in-patients and 7.2% deaths [2]. The total number of deaths attributable to malaria in 2014 was 2,200 out of which 1,060 (48.2%) occurred among children under 5 years [2].
There has been the roll out of several interventions in the country throughout the years to ensure reduction in malaria mortality and morbidity. These interventions such as the insecticide treated nets, seasonal malaria chemoprevention, intermittent preventive treatment in pregnant women are aimed at preventing the occurrence of the malaria disease. Strategies such as test, treat and track at the health facility and community levels also help diagnose and treat malaria cases appropriately. After several years of carrying out these interventions, it is important to assess the trends of malaria related morbidity and mortality in the context of increasing interventions. The objective of this paper is to describe the trends in the malaria-related deaths in the country from 2005 to 2014.

Study design
This study was done as part of a bigger one which looked at the mortality of AIDS/HIV, TB and malaria. A retrospective data analysis from a representative sample of health facilities providing services for the three diseases (ATM) was conducted.

Study Site
The study was conducted in all ten regions across the country. The study included teaching hospitals, regional hospitals, district hospitals and other primary level hospitals that provide inpatient services for AIDS/HIV, TB and Malaria (ATM).

Study Period
Data was collected from September 2016 to June 2017.

Sample size and sampling procedure
For a facility to be eligible for inclusion in the sampling frame, the study used the following eligibility criteria: A facility should provide diagnosis and treatment as well as inpatient services for at least one year during the study period for the three diseases: AIDS/HIV, TB and Malaria. Out of the total number of health facilities in the country, 198 health facilities provided diagnosis and treatment services for HIV/AIDS, TB, and Malaria of which 169 facilities provided additional in-patient services. These 169 facilities were used as the sampling frame. Of the total 169 health facilities which quali ed the study criteria (provided admission for ATM) 100 (59.2%) were selected for the study. Facilities were strati ed into regions and the number of health facilities in each region was allocated based on the total number of health facilities in that region.
In each region, all teaching and regional hospitals were purposively sampled and the remaining facilities were selected randomly using excel generated random numbers proportionate to the number of ART sites as shown in Table 1 (With all region having a minimum of 50% of the eligible. Extraction of data from the registers and various data forms were carried out by trained Research Assistants.

Data entry and analysis
Data was entered using a predesigned database format using EpiData 3.1 statistical software package and then exported to Stata (v11, Stata Corporation, USA) and SPSS (v20, IBM Corporation, USA) for analysis. Frequencies, proportions, and ratios were calculated to describe basic characteristics of study subjects. Trends in malaria-related mortality were determined. Bivariate analysis with Chi-square test was done to assess factors associated with ATM related mortality. A p-value of less than 5% was considered statistically signi cant. In order to explore the relative in uence of each independent variable or factor on mortality and to test signi cance of association, we conducted a binary logistic regression analysis using a multivariate model.
Meta-analysis was used to evaluate the trends in death rates among inpatients over time and to have a pooled estimate of death rate. Death rate in malaria-related admission for each year was determined at 95% CI. The point estimate of death rate, upper and lower value of the 95% CI were fed to the Metaanalysis syntax command. Weighting using random effect was applied since we expected heterogeneity across cohorts.
Limitations of the study It did not fully assess risk factors for disease speci c mortality because of limited data found in the records on variables.

Results
From the total 100 facilities selected for this study, we able to collect data from 93 (93%) of facilities  Figure 1). Malaria mortality declined by approximately 59% over the ten-year period with an average annual decline of 7 %.
The highest case fatality rate (3.0%) was recorded in Central region with Ashanti region having recorded the lowest (1.0%). Even though children under ve accounted the highest number of malaria mortalities (2,771), that age group together with those aged 5-14years recorded the lowest (0.9%) malaria case fatality rate while patients above 74 years recorded the highest CFR (7. 4%). The Regional Hospitals recorded the highest case fatality rate (3.0%). Case fatality rate was lower for insured patients (1.1%) compared with non-insured patients (2.9%). Year of admissions, age, sex, insurance status, type and ownership of facility were signi cantly associated with mortality (p < 0.001) ( Table 2). The regional hospitals have about 24% (Odds Ratio=1.24, p-<0.001) increased odds of mortality compared to teaching hospitals. Cases of malaria admitted into Private, Faith-based and Quasi-government facilities had a decreased likelihood of dying compared to government facilities (Table 3). A unit increase in age among malaria related admissions had a 0.3 increased odds of mortality (Odds Ratio=1.029, p<0.001).  (Table 3).

Discussion
Assessing trends of malaria morbidity and mortality is one of the important tools in measuring the impact of interventions over the years and data capturing gaps. Over the years, the National Malaria Control Programme has implemented multiple interventions with the help of Global Fund to decrease the malaria morbidity and mortality in Ghana. This is manifested in survey ndings which shows that households who owned at least one ITN increased from 18.7% in 2003 [3], to 73% in 2016 [4]. Healthseeking behaviour has improved over the years with advice or treatment sought for 49.4% of children with fever in the 2 weeks before the survey in 2011 [5] increasing to 72% in 2016 [4].
In this study, mortality among malaria admissions (Case Fatality Rate) showed an upward trend from 2. The trend could also be due to improved quality of malaria case management by health workers as case management trainings have increased over the years [6,7] with increase in country donor support. The cumulative effect may have started manifesting after the year 2007 as seen in other researches [8][9][10]. Teaching and Regional hospitals have increased likelihood for malaria-related mortality compared to the lower facilities. This could be resulting from the fact that regional and teaching hospitals are referral levels where complicated cases (with high risk of dying) are sent to be managed. Malaria-related admissions in government facilities had an increase likelihood of dying compared to the other types of ownership. Though the private sector and some public sector (district and sub-district hospitals) health facilities are known to be the rst point of call for cases especially malaria [11][12][13], cases tend to be referred to the higher facilities (Regional and Teaching Hospitals) which are often public. It is worthy of note that all the regional and teaching hospitals are public/government facility. The ndings may have therefore have resulted from improved quality of care in the non-governmental facilities or admission of relatively non-severe cases in the lower and private facilities compared to the regional facilities. This increases the likelihood of the government facilities having potentially fatal cases.
In the bivariate analysis, region of admission was signi cantly associated with mortality. The Central Region recording the highest CFR (3.0%) while Upper East and Northern Regions recorded the lowest (0.9%). This could be due to varying case management challenges across the regions, late reporting of cases and uneven distribution of capacity to manage such cases across the country. This is on the backdrop that Central Region is not the region with the worst Doctor: Patient ratio [14] and not the poorest in Ghana [15] which are variables that could worsen CFR rate yet the region recorded the highest CFR.
Increase in age has a signi cant 0.2% increased likelihood of malaria-related mortality. This could have resulted from the high attention given to the management of fevers in children over the years and possible delays in seeking health care by adults as compared to taking their children to the health facility when the need arose. The theory of older ages becoming more vulnerable to malaria than before due to a gradual loss of partial immunity as malaria burden decreased following intensi ed malaria control measures could also be the reason. This has been documented in some researches [16][17][18].
Among malaria-related admissions, patients with malaria as principal diagnosis have decreased odds (OR = 0.11, P < 0.001) of dying compared to having HIV/AIDS as principal diagnosis. Having any other disease as principal diagnosis using HIV/AIDS as reference also showed a decreased likelihood of dying. This may have resulted because patients admitted with HIV/AIDS as principal diagnosis may be severely ill and in an immune-compromised state which may get complicated with other diseases therefore increasing the odds of dying.
Finally, not having health insurance increases the likelihood of malaria-related mortality by 2.34 times. This could be linked to the fact that insurance holders are more likely to seek health care early and in an appropriate place and are therefore less likely to present with severe forms of the disease. They are also likely to be provided with all the resources for their management especially if there is the need for management of concomitant diseases. This health insurance nding is supported by similar trends seen in some researches [19,20].

Conclusion
Declines in malaria cases and deaths were higher in children under ve years old than in all age groups. Malaria mortality declined by approximately 59% over the ten-year period with an average annual decline of 7%. The likelihood of dying after malaria-related diagnosis in regional hospitals and government hospitals was higher than in the other type and ownership of facilities respectively. An increase in age and not having health insurance among malaria-related admissions increases the likelihood of mortality.
Targeted interventions among elderly and an intervention to increase health insurance coverage may help the malaria control program to reduce malaria-related mortality in the country. Superintendents of all the selected health facilities included in the study. All the information obtained from this study was kept con dential.

Consent for publication
Not applicable Availability of data and materials Data used in this study is available and will be provided by the lead author upon request.