The quality of malaria case management in different transmission settings in Tanzania mainland, 2017–2018

Tanzania is undergoing an epidemiological transition for malaria transmission with some areas of the country having <10% (hypoendemic) and other areas 10% - 50% malaria prevalence (mesoendemic). It is not known whether there is a difference in the quality of malaria case management based on endemicity in Tanzania mainland. We examined the influence of endemicity on the quality of malaria case management at health facilities. We conducted a cross-sectional analysis of 1713 health facilities in Tanzania mainland, using data collected by the National Malaria Control Program through an assessment tool to evaluate quality of malaria case management. The data was gathered from September 2017 to December 2018. Using standard quality factors, mean scores from facilities in the different endemicity regions were compared by a Student’s t-test. Simple and multiple linear regression analyses were performed to determine the association between facility performance (score) and endemicity (mesoendemic vs. hypoendemic). Facilities in mesoendemic regions scored higher than those in hypoendemic regions on the overall quality of services [difference in mean scores (d¯) = 2.52; (95% Confidence Interval (CI) 1.12, 3.91)], site readiness [d¯ = 2.97; (95% CI 1.30, 4.61)], availability of malaria reference materials [d¯ = 4.91; (95% CI 2.05, 7.76)], availability of Health Management Information System tools [d¯ = 5.86; (95% CI 3.80, 7.92)] and patient satisfaction [d¯ = 6.61; (95% CI 3.75, 9.48)]. Predictors associated with lower facility scores included; being located in a hypoendemic region [β: -2.49; (95% CI -3.83, -1.15)] and urban area [β: -3.84; (95% CI -5.60, -2.08)]. These findings highlight the differences in quality of malaria case management based on endemicity, but there is still a need to target improvement efforts in underperforming facilities, regardless of endemicity.


Introduction
In 2021, nearly half of the world's population was at risk of malaria [1], making it the most important human public health parasitic disease. A documented new global strategy for malaria control dates back to 1993 [2], and after two decades of scaling up effective control tools to reduce morbidity and transmission, major achievements have been recorded [1,3]. In 2021, approximately 247 million cases and 619,000 deaths were reported globally. The majority of the morbidity (95%) and mortality (96%) occurred in sub-Saharan Africa [1].
A major decline in malaria prevalence has been witnessed in Tanzania. Malaria prevalence among children under the age of five has declined more than 50%, from 18.1% in 2007 to 7.9% in 2022. Some regions such as Arusha and Manyara have reached a prevalence of <1%, while in other regions prevalence has remained high at �10% [4][5][6]. The global malaria strategy emphasizes the need for stratification of areas to effectively target malaria interventions. This stratification allows for a targeted approach in developing specific case management strategies that align with the unique characteristics and needs of the different areas [7]. However, the majority (96%) of the population in Tanzania mainland is still at risk for malaria infection [4,6].
In Tanzania mainland, 80% of all patients visiting the health facilities are attended as outpatients [8,9]. Prompt diagnosis and treatment with an effective antimalarial reduces morbidity and prevents mortality [10] and is one of the core interventions in controlling malaria [7]. Appropriate management of cases might also reduce transmission by reducing the human parasite reservoir and prevent the emergence of drug resistance [10][11][12][13]. Furthermore, improving case management may also contribute to improved treatment of non-malarial febrile illnesses, which are often misdiagnosed and treated presumptively as malaria [14,15]. As countries progress toward elimination, malaria interventions should focus on rigorous case management and transmission control [16]. While there are fewer malaria fevers in these settings, patients still consult health facilities for non-malarial fevers [15]. Since fever is the entry point for malaria case management [17], the higher prevalence of and priority given to management of non-malaria fevers might compromise the quality of malaria case management [18].
The quality of malaria case management provided in health facilities in regions progressing toward elimination has not been evaluated in Tanzania mainland [15]. This study aimed to compare the quality of malaria case management provided at outpatient departments in hypoendemic versus mesoendemic settings.
The information generated from comparing the quality of malaria case management in these different settings will facilitate targeting strategies to improve the quality of services in lower performing health facilities. Thus, will ensure that all individuals receive the best possible care for malaria, regardless of where they live.

Malaria service and data quality improvement (MSDQI) package
The MSDQI package is part of a national strategy for monitoring and improving malaria services provided in health facilities and assessing the quality of routinely collected malaria data reported from health facilities [19]. The package was developed by the National Malaria Control Program (NMCP) and development partners in line with the National Health and Social Welfare Quality Improvement Strategic Plan [20], the Tanzania Quality Improvement Framework in Health Care [21] and Situation Analysis of Quality Improvement in Health Care [22].
The MSDQI package consists of seven modules, each accompanied by a corresponding checklist. These modules assess various areas, including the outpatient department (OPD), malaria Rapid Diagnostic Test (mRDT), antenatal care (ANC) clinic, inpatient department (IPD) for severe malaria, logistics and supply chain for malaria commodities, malaria microscopy, and a Data Quality Audit (DQA) conducted in health facilities [19].
Trained Council Health Management Team (CHMT) members perform health facility supportive supervision visits, including MSDQI package supervision, at least once per quarter. The NMCP trained seven CHMT staff in every council on performing supervision visits using the MSDQI package. The trainings were conducted for a period of six days during the rollout of MSDQI [19].
MSDQI package supervision visits began in 2017. Data was collected via two means, either by electronic-based checklists on android smartphones or tablets, known as the Electronic Data System (EDS), or by paper-based checklists. Supervision visits documented by paperbased checklists were later entered into the EDS [23]. Supervisors for each module conducted supervision visits, completed the checklists, calculated quality performance scores for supervision visits conducted with paper-based checklists, and completed the quality improvement plan to address deficiencies identified through the MSDQI package during the supervision visit. Upon completion of the supervision visit, the supervisors entered the scores into the District Health Information Software 2 (DHIS2) system. The MSDQI checklists completed in the EDS were capable of automatically generating the quality scores and uploading the data to DHIS2. The MSDQI data and score results could be visualized in and retrieved from the 'Malaria Dashboard' user interface in DHIS2 [19].

Study setting and design
We conducted an analytical cross-sectional study to compare the quality of health-facility based malaria case management in malaria hypoendemic versus mesoendemic regions in Tanzania mainland.
Tanzania mainland is divided into 26 administrative Regions which are further subdivided into Councils. There are a total of 184 Councils. Councils are categorised according to population settings with 137 being rural and 47 being urban. There is one type of rural Council: a District Council. There are three types of urban Councils: Town, Municipal and City Councils [24].
This study only uses data from the MSDQI OPD checklist, which is one of the seven modules in the MSDQI package. Available MSDQI scores from health facility OPD checklists completed during supportive supervision visits conducted between September 1, 2017 and December 31, 2018 were retrieved from DHIS2 for analysis.
Malaria case management services are provided across all types of health facilities in Tanzania. A dispensary, located at village level, offers services on an outpatient basis. Health centres, at ward level, are the first referral centres for dispensaries, and also provide inpatient and diagnostic services. Hospitals are located at council and regional levels to serve as referral centres with a full complement of health services [27]. Health facilities were assigned to their respective urban or rural setting category based on the council they were located in. Implementing partners are nongovernmental organizations that provide financial and technical support to improve the provision of malaria case management services across all facility types, often with an emphasis in regions with the highest burden of malaria.
Tanzania's rainfall pattern is heterogeneous across the country due to effects of topography, winds, humidity, and tropical circulation [28]. Regional rainfall seasonality at the time (month) of health facility assessment was categorised into either dry or wet (rainy) using a cut off value of 100 mm.
Information describing the quality of malaria case management was organised by sub-factors grouped into three broad factors: site readiness, clinical management, and outcome according to the Donabedian model " Table 1" [29]. For each health facility, an overall score was calculated as the mean of the sub-factor scores for each of the broad factors.
Site readiness describes the context in which care is delivered. Information on subfactors in site readiness was obtained by interviews with health facility staff and observations.
Clinical management include all transactions made between the patients and health service providers from the time of history taking up to when the patient exits the facility. Information Table 1. Factors assessed for the quality of malaria case management in the malaria service and data quality improvement outpatient department checklist. Outcome factors refer to the effects of healthcare on the health status of patients and populations. This includes client satisfaction and changes in malaria morbidity in the region. Information on client satisfaction was obtained by exit interviews of two febrile patients. In this study, the outcome factor of malaria endemicity was used to categorize the facilities into the comparison groups i.e. those in Malaria Hypoendemic or Mesoendemic settings.

Sub-Factors Explanation
All sub-factors listed in " Table 1" are numerical variables and range from 0-100. The scores were awarded based on the performance against the standards set in the MSDQI OPD checklist. The scores were then graded into 3 categories as follows: 75-100 (Good), 50 -<75 (Average) and 0-<50 (Poor). The full MSDQI OPD checklist has been added as "S1 Checklist".

Data retrieval
MSDQI supervision data from the health facilities were retrieved from DHIS2 via pivot table application and downloaded as comma-separated values files. Information on health facility type, location, level, and ownership were obtained from the Health Facility registry. Regional rainfall seasonality patterns were obtained from the Tanzania Meteorological Agency Map room [30]. Data were merged in Microsoft Excel 2013 (Microsoft Office Professional Plus 2013, Seattle WA). Data used in the study has been added as "S1 Data".

Data analysis
Merged data were analysed using STATA version 15.1 (STATA Corp, Texas, USA). Descriptive statistics were summarised by frequencies, percentages, and means where appropriate. A Chisquare test was used to compare the distribution of health facilities between the MSDQI scores and endemicity. The means of scores for the facilities in malaria mesoendemic and hypoendemic settings were compared using a Student's t-test to determine whether there was a difference in the site readiness, clinical management, and outcome factors of the quality of malaria case management. A simple linear regression analysis followed by a standard multiple linear regression analysis were performed to determine the association between the quality of malaria case management (dependent) and other variables (independent) i.e., malaria endemicity, health facility location (urban or rural), facility type according to the level of care provided (hospitals, health centers, or dispensaries), facility ownership (public, private for profit, private Non-Governmental Organizations or Faith-Based Organizations), the season when the facility was surveyed (wet or dry), and whether the region has implementing partners supporting malaria case management (yes or no). Independent variables were added into the model using forward selection. To address the issue of multicollinearity, a threshold of variance inflation factor (VIF) equal to 5 was chosen. Variables with VIF values exceeding this threshold were considered to have a high level of collinearity and were subsequently removed from the regression model to ensure the stability and reliability of the estimated coefficients. The linear regression analyses were also performed for site readiness, clinical management and patient satisfaction. P-values <0.05 were considered statistically significant. Maps were drawn using Q-GIS Desktop 2.18.13 (QGIS Development Team, Switzerland). All shapefiles used are from openly available sources. (https://www.nbs.go.tz/index.php/en/census-surveys/gis/385-2012-phc-shapefiles-level-oneand-two). The shapefiles were made based on the 2012 population and housing census.

Ethical consideration
Ethical clearance was obtained from the Institutional Ethical Review Board of Muhimbili University of Health and Allied Sciences; Reference Number DA.287/298/01A. Additional permission to conduct the study and use the extracted MSDQI programmatic data was obtained from the National Malaria Control Program. None of the data included individual or patient identifying information.

Results
Between the 1 st of September 2017 and 31 st of December 2018, a total of 1713 health facilities (21.3% of the overall 8033 operating facilities located across all 26 regions) in Tanzania mainland reported data into DHIS2 which included factors of quality case management assessed through the MSDQI OPD checklists conducted by CHMTs. " "Fig 1,"" shows the distribution of health facilities reporting MSDQI data into DHIS2 during the study period by region and malaria endemicity. MSDQI data was available from 992 (57.9%) facilities reporting from regions in the Lake and Southern Zones of Tanzania. Simiyu region had the highest MSDQI OPD coverage at 74.3%, while Dar es Salaam had the lowest coverage at 0.5% "S1 Table".
The shapefile used is from an openly available source. (https://www.nbs.go.tz/index.php/ en/census-surveys/gis/385-2012-phc-shapefiles-level-one-and-two). The shapefile was made based on the 2012 population and housing census.
Out of the 1713 health facilities, more than half (57.6%) were located in malaria mesoendemic settings.
Thirteen hundred and ninety-six (81.5%) of the health facilities were dispensaries. The distribution of health facilities by managing authority and location were similar between hypoendemic and mesoendemic settings (p-values 0.466 and 0.742, respectively). Nine hundred and seventy-eight (99.1%) of the health facilities in the mesoendemic settings had implementing partners supporting malaria case management. Nine hundred and forty-two (55.0%) of the facilities were evaluated during the dry season " Table 2".
The distribution of all operating facilities in Tanzania mainland by endemicity setting are presented in "S2 Table".
Nearly all of the health facilities scored Good or Average in the overall quality of case management (95.5%) and site readiness (95.8%). Approximately one third 553 (32.3%) of the facilities scored poorly in clinical management. A higher proportion of facilities, irrespective of level or type, in the mesoendemic regions had higher good scores compared to those in the hypoendemic regions in all categories except for clinical management "  . There were no differences in the site readiness or overall quality of case management scores between the two endemicity settings (p-value = 0.994 and 0.772, respectively) " Table 6". A Significant Regression equation was found with an R 2 of 0.018 on multiple linear regression analysis. Independent predictors of the Overall Quality of Case Management score that  Table 7". Significant multiple linear regression results were found for predicting site readiness and client satisfaction scores. In facilities located in Hypoendemic regions, site readiness and client satisfaction scores were found to be 2.99 points and 6.99 points lower, respectively, compared to those in mesoendemic regions. The multiple linear regression results for predicting client satisfaction scores were not significant "Tables 8-10".

Discussion
In this study, we found that health facilities located in malaria hypoendemic regions scored lower than those in mesoendemic regions in terms of the overall quality of services provided for malaria, overall site readiness to provide care, availability of malaria reference materials, availability of HMIS tools, and patient satisfaction. However, there was no difference in the clinical management factors between facilities in hypoendemic compared to mesoendemic regions. Facilities located in urban areas were also found to score lower than those in rural areas, and those without implementing partner support scored lower during the MSDQI assessments. Nevertheless, this comparison may help decision makers to identify areas that need to be targeted for improvement in future analyses, especially considering that supervision is provided to only a limited number of facilities. Possible areas to focus improvements based on our analysis are site readiness, especially staff training, and clinical management.

Site readiness
Facilities located in malaria mesoendemic regions were found to have higher readiness scores compared to those in hypoendemic regions; this is consistent with findings from a study among 826 health facilities in Kenya, Namibia, and Senegal [31]. This might be explained by having a higher proportion of facilities located in rural settings included in the analytical Table 5. Difference in mean scores between facilities with (n = 1540) and without (n = 173) case management implementing partners. samples. Malaria has traditionally been a rural disease; the higher malaria burden may lead to improved facility readiness in mesoendemic settings [31,32]. The type of health facility did not influence the readiness to provide care for malaria patients. This contradicts previous findings from studies conducted in low and middle-income countries, including Tanzania [33], which showed there were differences in readiness by the type of the facility. This might be due to having a different composition in the types of facilities; more than three quarters (81.5%) of the facilities involved in this study were dispensaries (i.e., lowest facility level). Another possible reason might be that malaria is among the most common diagnoses attended in health facilities in the country and the standards for operating facilities require malaria services to be delivered in facilities at all levels; this might explain the high level of facility readiness in our analysis regardless of facility type.

Factor With Partner Support Mean (SD) Without Partner Support Mean (SD) Difference in Means (95% CI) P-value
Facilities in mesoendemic regions were found to have a higher availability of reference materials. A possible explanation for the observed difference could be due to the presence of implementing partners providing case management support, which includes printing and distribution of reference materials to health facilities. A higher proportion of facilities in the mesoendemic regions had implementing partner support compared to those in hypoendemic regions. Another reason could be due to the number of reference materials used in the calculation of the indicator. This indicator was comprised of five different reference materials, some of which might not be frequently used in facilities in hypoendemic regions, resulting in lower scores during the evaluation.
In both endemicity settings, health facilities scored best in the availability of HMIS tools compared to other factors. This could be explained by the recent health systems strengthening efforts for improving the HMIS in Tanzania [34]. The tools evaluated by this indicator were not specific for malaria, they included the HMIS registers, tally sheets and monthly summary forms, which are standard throughout the country. Staff training on appropriate malaria case management scored lowest among the site readiness factors in both endemicity settings. We expected the scores to be high as the institutions (universities and colleges) training clinicians in Tanzania have standardised curricula that include the management of malaria. The low scores might be caused by how the assessment is conducted for this indicator. The MSDQI OPD checklist documents the number of available clinical staff, the number of staff who had received formal, or on-the-job training on malaria case management, including artesunate injectable prescription, and the mean of these scores is calculated. Injectable artesunate was introduced in Tanzania in 2013 and incorporated in the National Guidelines for Diagnosis and Treatment of Malaria in 2014 [35]. Facilities with a higher proportion of clinical staff who had graduated more than 8 years ago, before the introduction of artesunate, might score low in the first part of the checklist. Although on-the-job trainings on injection artesunate preparation and administration were provided when it was being introduced, a recent study shows that health care worker knowledge is still low [36].

Clinical management
In both endemicity settings, health facilities scored the lowest in clinical management factors compared to the others assessed.
Facilities scored poorly in history taking and clinical examinations conducted for febrile patients in both endemicity settings. This might be explained by the human resource for health crisis in the country leading to a high workload [37]. The evaluation tool required the observer to assess fine details in history taking and physical examination that might have been skipped in a clinical setting with a high burden or turnover of patients. Patients might be referred quickly for laboratory confirmation without a thorough history and physical examination. Health facilities scored higher in the testing rate for patients suspected to have malaria. However, there was no statistical difference in scores between the endemicity settings. This was not the case in Angola during a study which found there was a lower testing rate in low transmission settings [38].
In both endemicity settings, approximately 25% of the malaria test results (mRDT and microscopy) were not properly interpreted or the patient was diagnosed clinically to have malaria (i.e., without a test result). This proportion is lower compared to previous findings from the Service Provision Assessment for Malaria conducted in 2016 [39], but still reflects an unacceptably high number of facilities prescribing anti-malarial drugs for patients either not tested or having negative test results. A review of the HMIS data for the same period that this study was conducted showed that across the country approximately 3% of all malaria cases were diagnosed clinically without testing [40]. This practice is contrary to the national guidelines on malaria diagnosis and treatment, and the recommendations of WHO, which require treatment with anti-malarial drugs be limited to patients with parasitological confirmation of malaria infection [14,17,35].
Clinicians should improve on history taking skills and the performance of thorough physical examinations, refer for or conduct diagnostic tests for suspected malaria cases, and properly interpret and adhere to the test results before prescribing an appropriate antimalarial.

Outcome factor
The level of patient satisfaction was high, comparable to that found in the Service Provision Assessment for Malaria, 2016, which reported satisfaction levels ranging from 60% to 85% [39]. It is important to note, however, that the Service Provision Assessment for Malaria did not calculate an overall satisfaction score. Patient satisfaction levels could have been influenced by differences in questionnaire length, content, and the potential for courtesy biases in the exit interviews conducted in both assessments. Patients were more satisfied with the services received in the mesoendemic regions compared to the hypoendemic regions. The checklist also gave scores when the patient could correctly explain the use of dispensed drugs at home, a situation that might favour the facilities in mesoendemic regions given they were more likely to prescribe medication as they attend more malaria cases.

Strengths and limitations
Results from this study illustrated the quality of malaria case management services provided in outpatient departments across Tanzania mainland by utilising routine health facility assessment data obtained through the MSDQI package. This data was collected using a standardized approach that is sustainable. The approach aids in the collection, monitoring, and evaluation of malaria performance indicators at the facility level that provides timely, accurate information and data for decision making. Findings from this study may serve as a baseline that could facilitate ongoing monitoring of case management performance at health facilities as MSDQI assessment data collection continues. Countrywide analysis using the MSDQI OPD data before this study could not be performed. We merged MSDQI data from several sources including those that were collected using both paper-based and electronic tools that were initially stored in separate servers.
We assigned facilities to urban or rural locations; however, misclassification might have occurred in differentiating between facilities located in peri-urban settings.
Misclassification might have also occurred in differentiating endemicity settings by malaria prevalence estimates at the regional level. There is heterogeneity within regions. A better approach might be to perform the analysis at the district level and further differentiate the endemicity settings into more strata as previously defined using Tanzania HMIS and population survey data [41].
Cross-sectional health facility performance data that was collected between September 2017 and December 2018 was linked to estimates of malaria survey data conducted between October and December 2017. A better approach could have been to use data from real-time surveillance that would have enabled the detection of hot spots, especially in the hypoendemic regions.
The study analysed data from less than a quarter of all operating facilities in Tanzania mainland. The distribution of facilities across the country that were included was not uniform as some regions started the implementation of MSDQI supervisions earlier or had a higher rate of conducting supervisions than others. This might be influenced by the presence of implementing partners in the regions and the accessibility of the facilities by road for supervision during the dry and wet seasons. These might affect the generalisability of the results. As the coverage of the MSDQI assessments increases, more frequent analyses using data from all modules are warranted.

Conclusion
Over 95% of all health facilities in our analysis sample scored Good or Average in the overall quality of case management. While there were differences observed in the quality of case management by endemicity, there is a need to target quality improvement, particularly in staff training and clinical management in the lower performing facilities, regardless of endemicity setting. Our methods and these findings could serve as a baseline for future analyses of Tanzania's MSDQI data.