Poor-quality antimalarials further health inequities in Uganda

Abstract Substandard and falsified medications are a major threat to public health, directly increasing the risk of treatment failure, antimicrobial resistance, morbidity, mortality and health expenditures. While antimalarial medicines are one of the most common to be of poor quality in low- and middle-income countries, their distributional impact has not been examined. This study assessed the health equity impact of substandard and falsified antimalarials among children under five in Uganda. Using a probabilistic agent-based model of paediatric malaria infection (Substandard and Falsified Antimalarial Research Impact, SAFARI model), we examine the present day distribution of the burden of poor-quality antimalarials by socio-economic status and urban/rural settings, and simulate supply chain, policy and patient education interventions. Patients incur US$26.1 million (7.8%) of the estimated total annual economic burden of substandard and falsified antimalarials, including $2.3 million (9.1%) in direct costs and $23.8 million (7.7%) in productivity losses due to early death. Poor-quality antimalarials annually cost $2.9 million to the government. The burden of the health and economic impact of malaria and poor-quality antimalarials predominantly rests on the poor (concentration index −0.28) and rural populations (98%). The number of deaths among the poorest wealth quintile due to substandard and falsified antimalarials was 12.7 times that of the wealthiest quintile, and the poor paid 12.1 times as much per person in out-of-pocket payments. Rural populations experienced 97.9% of the deaths due to poor-quality antimalarials, and paid 10.7 times as much annually in out-of-pocket expenses compared with urban populations. Our simulations demonstrated that interventions to improve medicine quality could have the greatest impact at reducing inequities, and improving adherence to antimalarials could have the largest economic impact. Substandard and falsified antimalarials have a significant health and economic impact, with greater burden of deaths, disability and costs on poor and rural populations, contributing to health inequities in Uganda.


Background
Health equity is defined as everyone having 'the opportunity to attain their full health potential' and no one being 'disadvantaged from achieving this potential because of their social position or other socially determined circumstance' (Brennan Ramirez et al., 2008). Health inequities arise when a socially determined group or groups of a population are prevented from achieving their full health potential. For example, when individuals in lower socio-economic statuses (SESs) face higher risks for poor health outcomes, it creates an inequitable division in health that is attributable to a predetermined social circumstance. Health inequities have a significant health and socio-economic impact, creating and reinforcing cycles of poverty and poor health, and slowing national development (WHO, 2018a). Preventing and addressing health inequities is important because all individuals should have the opportunity to attain their highest level of health (WHO, 2018a).
Malaria poses a significant threat to achieving health equity. The malaria burden disproportionately affects low-and middle-income countries (LMICs), and rests on children, the poor and rural populations (Filmer, 2005;Roca-Feltrer et al., 2008; Uganda Bureau of Statistics (UBOS) and ICF, 2018). The World Health Organization (WHO) estimates that globally, 219 million malaria cases and 435 000 deaths were due to malaria in 2017 (WHO, 2018c). The brunt of this burden rests on countries in Sub-Saharan Africa where 90% of these cases and 91% of these deaths occur (WHO-AFRO, 2017;WHO, 2017c). Uganda ranks third in Sub-Saharan Africa in total number of malaria cases (Roberts and Matthews, 2016). Children under five are the most vulnerable to malaria infection and death, representing two-thirds (285 000) of malaria deaths (Maitland, 2016;WHO, 2018b). As a result, malaria is one of the leading causes of under-five mortality in Uganda (Roberts and Matthews, 2016;UNICEF, 2018). The 2016 Uganda Demographics and Health Survey (DHS) found that the prevalence of malaria among children in the lowest wealth quintile were over 10 times higher than the prevalence of children in the highest quintile [Uganda Bureau of Statistics (UBOS) and ICF, 2018]. Similarly, children in rural households were three times more likely to be infected with malaria than urban children [Uganda Bureau of Statistics (UBOS) and ICF, 2018].
Substandard and falsified medicines also disproportionately affect LMICs due to poor pharmaceutical governance and supply chain management in these settings (Newton et al., 2010;Nayyar et al., 2012;WHO, 2017b). As defined by the WHO, substandard medicines are 'authorized medical products that fail to meet either quality standards, specifications, or both'; and falsified medicines are 'medical products that deliberately or fraudulently misrepresent their identity, composition or source' (WHO, 2017a). Antimalarials are one of the medication classes most commonly found to be of poor quality (Newton et al., 2010;WHO, 2017b). On the basis of a recent meta-analysis, 19.1% of antimalarials in LMICs were identified to be substandard or falsified (Ozawa et al., 2018), and prevalence as high as 35% have been reported in Uganda (Bate et al., 2008;WHO, 2009WHO, , 2017b. These medications contribute to the malaria burden by prolonging illness and increasing the patient's risk of treatment failure, progression to severe malaria, and death (Newton et al., 2006;Amin and Kokwaro, 2007;Newton et al., 2011;Nayyar et al., 2012;Buckley and Gostin, 2013;WHO, 2017b). By reducing clinical efficacy, substandard and falsified antimalarials also increase the risk of antimicrobial resistance and add to healthcare expenditures (Buckley and Gostin, 2013;Chaccour et al., 2015;WHO, 2017b).
Through a model that examines the heterogeneity of a paediatric population at a country level, this study assesses the inequitable divisions in health outcomes as well as direct and indirect economic costs that result from substandard and falsified antimalarials. We estimate the burden of disease, death and disability, along with the costs to patients and society of substandard and falsified antimalarials. We present results by SES and urban/rural regions to examine the burden specific to poor and rural populations. Through scenario analysis, we also examine the benefits of potential interventions to reduce the burden of health inequities for paediatric malaria in Uganda.

Materials and methods
We developed and utilized the SAFARI (Substandard and Falsified Antimalarial Research Impact) model, a dynamic agent-based model built in NetLogo (Version 6.0.2, Wilensky, 1999). The model is probabilistic, with an ability to simulate the natural heterogeneity that exists across population characteristics, risk of malaria infection, patient care-seeking rates, disease outcomes and associated costs. Key epidemiological, demographic, care seeking and cost data were derived from available literature and are presented in Table 1. The SAFARI model is explained in further detail in other publications (Ozawa et al., 2019a;Ozawa et al., 2019b), with inputs and details specific to Uganda presented here.
The model simulated 1000 children over a 1-year time horizon by 5-day increments. Each child was assigned six demographic characteristics (age, sex, location, province, SES and maternal education level) based on population distributions derived as proportions from 2014-15 Uganda Malaria Indicator Survey (MIS) [Uganda Bureau of Statistics (UBOS) and ICF, 2015]. These characteristics were chosen as significant predictors of incidence, care-seeking, disease progression and treatment outcomes, but also to facilitate analysis by demographic groups (Ozawa et al., 2019b). These demographic characteristics were used to estimate individual incidence and careseeking probabilities for each child. Every 5 days, the children faced an individual incidence of becoming infected with malaria and becoming symptomatic. Children who became symptomatic or remained symptomatic from the previous period simulated care seeking and disease progression based on their demographic characteristics. Treatment was sought from one of six locations (public facility, private facility, pharmacy, drug store, community health worker or at home/from neighbour) or not at all. Location-based    ACT, artemisinin combination-based therapy; API, active pharmaceutical ingredient; CHW, community health worker; NS, neurological sequelae; SF, substandard and falsified. care-seeking rates were derived from MIS data [Uganda Bureau of Statistics (UBOS) and ICF, 2015]. At health facilities, patients with severe malaria were treated as inpatients, while those with uncomplicated malaria were treated as outpatients. Children with uncomplicated malaria received one of three categories of antimalarials [artemisinin combination therapy (ACT), quinine or all other alternative treatments] or faced a stock out. National rates of stock outs for each location of care were extracted from Uganda ACTwatch data (ACTwatch Group and PACE, 2015).
Simulated health outcomes varied by severity and whether care was sought. Severe malaria cases could result in death, neurological sequelae or recovery, based on rates from the AQUAMAT clinical trial (Dondorp et al., 2010;Lubell et al., 2011;Das et al., 2013;Lubell et al., 2014). Health outcomes of children who received antimalarials were determined by the clinical efficacy of the type of antimalarial received (Checchi et al., 2004;Staedke et al., 2004;Bakyaita et al., 2005;Piola et al., 2005;Yeka et al., 2005;Bukirwa et al., 2006;Yeka et al., 2008;Achan et al., 2009;Arinaitwe et al., 2009;Bassat et al., 2009;Clark et al., 2010;Piola et al., 2010;Verret et al., 2011;Yeka et al., 2013), as well as drug quality and probability of treatment adherence (Bruxvoort et al., 2015). Prevalence of substandard and falsified medicines was estimated for each category of antimalarial based on the WWARN (Worldwide Antimalarial Resistance Network) database and a systematic literature search specific to Uganda (Bate et al., 2008;WHO, 2009;Bate et al., 2010;Bjorkman Nyqvist et al., 2012;Kitutu, 2015). Substandard and falsified antimalarials were assigned an active pharmaceutical ingredient (API) concentration where medication effectiveness decreased by API concentration (ACT Consortium Drug Quality Project Team and the IMPACT2 Study Team, 2015; Kaur et al., 2015). Patients who received a very poor-quality antimalarial or who had very poor adherence experienced complete treatment failure and faced a higher likelihood of progressing to severe malaria. All other patients who did not recover (i.e. did not have adequate clinical and parasitological response to treatment) experienced late clinical or parasitological failure. Since poor-quality antimalarials would not affect individuals who do not seek care, we focus on outcomes for those who sought care.
Costs were incurred throughout the simulation and estimated using the cost-of-illness approach (Hodgson and Meiners, 1982). Five categories of direct costs (medication, testing, consultation, transport and hospitalization) were incurred by patients out-ofpocket based on unit costs from the literature [Uganda Bureau of Statistics (UBOS) and ICF, 2009;Batwala et al., 2011;Orem et al., 2012;ACTwatch Group and PACE, 2015;Management Sciences for Health, 2016;Hansen et al., 2017]. Four forms of indirect costs were estimated: lost wages by caregivers, opportunity cost of time, productivity losses due to neurological sequelae and productivity losses due to premature death (Vos et al., 2016;The World Bank, 2018). Productivity losses were estimated using the human capital approach (Drummond et al., 2015), based on Uganda's gross domestic product, duration of lost productivity and disability-adjusted life year (DALY) weights. All costs are presented in 2017 USD. Further explanation for calculating economic outputs can be found in other publications (Ozawa et al., 2019a;Ozawa et al., 2019b).
We examined the distribution of health and economic burden of malaria in addition to the distributional impact of substandard and falsified antimalarials in Uganda. Number of hospitalizations, deaths and undiscounted DALYs were estimated as well as direct and indirect costs. The impact of substandard and falsified antimalarials was analysed by comparing the baseline estimates to a scenario where treatment efficacy was not reduced due to substandard and falsified antimalarials. The distributional impact of potential artemisinin resistance was also analysed in which the clinical efficacy of ACTs was reduced to be the same level as that of other treatments. A chi-squared test was used to examine the statistical significance of health and economic disparities assessing the burden of substandard and falsified medicines and artemisinin resistance. Additionally, a concentration index was used to quantitatively assess socio-economic inequity across wealth quintiles (O'Donnell et al., 2007). Stata 14 was used to conduct statistical analyses (StataCorp, 2015).
To account for the variation in model inputs, we recorded available ranges of modelled parameters from the literature and probabilistically ranged them altogether. Epidemiological inputs were taken from normal distributions whereas cost inputs were parameterized from gamma distributions. The overall results provide the baseline estimate as well as the 95% confidence intervals based on sensitivity analyses. Uncertainty ranges were calculated from baseline standard deviations or variance across scenarios from 10 000 model simulations.
Finally, six interventions were simulated that impact the supply chain, antimalarial treatment policies or caregiver education to examine their impact in reducing health inequities. These scenarios included: (1) Having no antimalarial stock outs at public facilities; (2) Enacting policies so that only ACTs are offered to treat malaria; (3) Enacting policies to ensure that all ACTs in Uganda are of good quality; (4) Enacting policies to ensure that all antimalarials in Uganda are of good quality; (5) Educating patients to take ACTs over all other antimalarials; and (6) Educating patients to improve treatment adherence.

Results
We estimated that there are annually $3.3 million cases of malaria in children under five in Uganda who seek treatment, resulting in 12 893 (95% CI 12 668-13 117) deaths. The resulting economic impact amounts to $333 million on patients and $32 million on the government (2017 USD). Table 2 presents the distribution of the health and economic burden of malaria in Ugandan children under five from a patient perspective, broken down by the MIS distribution of SES quintile and rural/urban location. The health burden of malaria in DALYs was significantly unevenly distributed across SES quintiles (P < 0.01) and concentrated in the lower SES quintiles (concentration index À0.30). The lowest SES quintile carried the largest proportion of the epidemiological impact, containing approximately a third of malaria cases (35.3%), deaths (28.6%) and DALYs (33.6%). Conversely, the highest SES quintile experienced less than 2% of the malaria cases and DALYs (1.6% and 1.8%, respectively), and only 2.2% of deaths. The health burden was also significantly concentrated in rural (P < 0.01) populations, with 82% of the population living in rural areas but comprising over 98% of the cases, deaths, and DALYs. Similar trends are observed in relation to the economic burden, with lower SES and rural populations experiencing a majority of the direct (concentration index À0.29) and indirect costs (concentration index À0.31). The lowest SES quintile accounted for 32.5% of out-of-pocket expenses while the highest SES quintile only accounted for 2.6%. Populations in rural areas accounted for 97.9% of the total out-of-pocket expenditures.
Substandard and falsified antimalarials were annually responsible for 1121 (8.7%) of these deaths, and 26.1 million USD (7.8%) of the economic impact of malaria through treatment costs, transportation costs, productive time lost of caregivers and years of life lost due to early death. Substandard and falsified antimalarials  contributed to 2.9 million USD (8.8%) in direct costs to the Ugandan government. The distribution of the health and economic impact of substandard and falsified antimalarials is presented in Table 3. The health impact of poor-quality antimalarials in terms of DALYs lost was unequally distributed (P < 0.01), and had the largest impact on lower SES quintiles (concentration index À0.26) and rural populations (98%). In comparison to the highest SES quintile, the lowest SES quintile is predicted to experience 23 times as many hospitalizations and 13 times as many deaths due to poor-quality antimalarials each year. Rural populations in Uganda are also expected to experience >98% of the hospitalizations and deaths due to substandard and falsified medicines, while making up 82% of the population. We simulated 60 times as many hospitalizations and 47 times as many deaths in rural compared with urban areas. The distribution was similar for economic impact of substandard and falsified antimalarials with a substantial burden on the poor (concentration index À0.28) and rural populations (98.7%). The per person out-of-pocket expenses due to poor-quality antimalarials for the lowest SES quintile was 12.2 times as high as the highest SES quintile. Rural populations paid 10.8 times that of urban populations in per person in out-of-pocket expenses due to poor-quality antimalarials.
In the event that artemisinin resistance emerges in Uganda, we estimated that the number of deaths and total economic impact on patients could increase by 6.9% and 11.6%, respectively. The potential health and economic impact of artemisinin resistance was unequally distributed (P < 0.01) with the majority of the potential health impact in terms of DALYs still falling on low SES (concentration index À0.32) and rural populations (98.4%). The overall economic impact was similarly placed on the poor (concentration index À0.31) and rural populations (98.3%). Compared with the burden of malaria and substandard and falsified antimalarials, the burden of artemisinin resistance was shown to result in an even greater concentration on poorer groups (Figure 1). Among our simulated interventions, improving the quality of all antimalarials had the greatest impact, reducing the number of deaths by 8.7% (P < 0.01) and total economic impact by 7.8% (P < 0.01). Improving the quality of ACTs alone would address 71.3% of the economic impact of all poor-quality antimalarials. Ensuring that public health facilities never face a stock out had a large health and economic impact, reducing the number of deaths by 4.9% and total economic impact by $9.9 million USD. Improving treatment adherence also significantly reduced the health impact by 1425 (11%) deaths and had the largest economic impact of $30.7 million USD (9.5%). Educating caregivers to select ACTs over other treatments could potentially reduce the economic impact by $16.1 million USD (4.8%), while having health facilities only offer ACTs for malaria treatment reduced the economic impact by $17 million USD (5.1%). Figure 2 compares the potential economic impact of the six supply chain, policy and caregiver education interventions across SES quintiles. Each modelled intervention is expected to have an impact on equity by benefiting lower SES quintiles more (concentration indices: À0.18 to À0.28), but vary in the amount of overall economic effect in terms of direct costs to caregivers and lifetime productivity costs. The greatest overall economic impact ($30.7 million USD, 9.5%) would come from ensuring patients adhered perfectly to all anti-malarial regimens, which would require increasing the proportion of under-five perfect adherence by 25%. Improving antimalarial quality (concentration index ¼ À0.28) and educating caregivers to reject non-ACT treatments (concentration index ¼ À0.28) resulted in the greatest health equity effect benefitting low SES patients.

Discussion
This study demonstrated for the first time that poor-quality antimalarials disproportionately affect low SES and rural populations. Populations in LMICs are at the greatest risk from substandard and falsified medicines due to poor supply chain management, regulations and surveillance (Newton et al., 2010;Nayyar et al., 2012;WHO, 2017b). Further, populations in lower SESs are more likely to seek care from the informal sector, and as a result are more likely to obtain poor-quality medicines (Ogunlesi and Olanrewaju, 2010;Almuzaini et al., 2013). This along with the disproportionate burden of malaria in poor populations resulted in these populations consuming a majority of poor-quality antimalarials in our simulation. This finding demonstrated that poor and rural children are at greater risk from the effects of poor-quality antimalarials. This highlights the need for pharmacovigilance and efforts to improve medicine quality by strengthening the medicine distribution chain, using quality-assured drugs, improving pharmaceutical governance and strengthening the technical capacity of regulatory laboratories, especially in poor and rural communities (Buckley and Gostin, 2013;WHO, 2017b). Moreover, our results indicate that improving the quality of antimalarials in Uganda could also work to address health inequities. Among the scenarios we examined, removing substandard and falsified medicines and educating caregivers to reject non-ACT treatments had the largest effect on health equity, benefitting the poor and rural populations the most. Treatment adherence was shown to be important to reducing the economic burden of malaria. Even without improving medicine quality, perfect adherence of under-five patients to antimalarials could reduce the economic burden of malaria on average 9.2% across all SES levels. Improving the availability of antimalarials at public facilities resulted in a reduction in deaths (4.9%) and economic impact (3%) of malaria. To reduce stock outs and achieve improved distribution of medicines, the WHO recommends decentralizing the supply chain and implementing rigorous monitoring and evaluation systems throughout (Mabirirzi et al., 2014). These efforts increase access to antimalarials and reduce stock outs by identifying stock problems early, reducing not just the impact of malaria but also addressing health disparities (Buckley and Gostin, 2013;Mabirirzi et al., 2014). Further, efforts to improve supply chain management, and monitoring and evaluation systems could help address substandard and falsified antimalarials (Buckley and Gostin, 2013). Poor stock availability increases medication costs and creates a demand for medications in informal sectors, which in turn opens the door for substandard and falsified medicines (Buckley and Gostin, 2013). Strong monitoring and evaluation systems are necessary to prevent poor-quality medicines from penetrating the market, targeting those most at need, driving up healthcare costs, and exacerbating health disparities (Buckley and Gostin, 2013).
The findings of this study are consistent with previously reported associations between rates of malaria infection and SES or location (Roca-Feltrer et al., 2008;Nahum et al., 2010;Gahutu et al., 2011;Roberts and Matthews, 2016). Our distribution of the malaria burden is reflective of the distribution found by the 2016 DHS and 2014-15 MIS (Uganda Bureau of Statistics (UBOS) and ICF, 2018). The WHO estimates that 3.8-8.9% of all malaria deaths result from poor-quality antimalarials in Sub-Saharan Africa, and children under five account for a high proportion (45%) of all malaria deaths (WHO, 2017b). Our estimate for under-five malaria deaths in Uganda due to substandard and falsified antimalarials (8.7%) falls in the upper end of this range.
The burden of paediatric malaria in Uganda disproportionately rests on low SES and rural populations. The reasons for the association between malaria and SES are broadly described as attributable to differences in housing structure, nutrition and education, and malaria prevention methods (Nahum et al., 2010;Gahutu et al., 2011;Roberts and Matthews, 2016). Other studies have found that people of low SES not only have substantially higher malaria prevalence but also have greater risk of infection and catastrophic healthcare expenditures (Castillo-Riquelme et al., 2008). As a result, malaria disproportionately affects the poor globally, with the poorest 20% of the world's population accounting for a majority of malaria cases (Breman et al., 2004). Similar trends are observed between rural and urban populations, with children in urban areas having significantly lower risk of malaria infection [Onwujekwe et al., 2009; Uganda Bureau of Statistics (UBOS) and ICF, 2018]. This distribution may be explained by urban populations having higher income, education and access to healthcare and preventative measures (Onwujekwe et al., 2009).
Our model findings on the impact of interventions are specific to Uganda. Applications of the SAFARI model found other interventions to be more efficacious at reducing the burden of substandard and falsified antimalarials in other countries (Ozawa et al., 2019a). This highlights the importance of tailoring interventions to the specific needs of a population. As such, caution should be used in extrapolating the recommended interventions to other countries. Uganda adopted the WHO recommendation of using ACTs as the first line treatment for malaria in 2004 (WHO, 2016). As of 2016, ACTs were used to treat almost 90% of under-five malaria cases [Uganda Bureau of Statistics (UBOS) and ICF, 2018]. This success in ensuring children receive ACTs also means that poor-quality ACTs, as opposed to other antimalarials, are accounting for most of the total health and economic impact of substandard and falsified antimalarials in Uganda. Our model suggests that 71.3% of the economic impact from substandard and falsified medicines is due to low quality ACTs. Ensuring that all ACTs offered in Uganda are of high quality would significantly reduce the total burden of malaria and bring about a health equity benefit.
All statistical models are limited by the quality of their inputs and their ability to account for naturally occurring heterogeneity (Xie, 2011;Gomez-Ramirez and Sanz, 2013). We developed and utilized a probabilistic model where model parameters were derived from the most recent and highest quality published data. Further, since substandard and falsified antimalarials only affect individuals seeking treatment, this analysis primarily reports the malaria burden among cases that sought medical care. The model is limited by the lack of available data on the prevalence of substandard or falsified antimalarials. As a result the prevalence of poor-quality antimalarials for each category was not varied by care sector. We did vary incidence and care-seeking by SES and other variables. Additional data on the prevalence of substandard or falsified antimalarials by different population groups would improve our understanding of the health equity impact. Our economic impact focused on costs directly incurred by patients and caregivers to examine the distributional burden, and did not include costs incurred by government facilities. Our analysis focused only on the potential impact on the burden of malaria and did not account for the cost of implementing the interventions.

Conclusion
We demonstrate that the burden of malaria and effects of poor-quality antimalarials disproportionately rest on poor and rural populations in Uganda. This unequal distribution exacerbates health inequities and places these populations further behind in attaining their health potential. These findings are important to inform the government, policy makers, donors and the malaria community of the issue and the impact of potential interventions to address them.
Our findings indicate that efforts should focus on improving supply chain management and monitoring and evaluation systems. These efforts could improve access to medicines by increasing stock availability and reducing the prevalence of substandard and falsified antimalarials. Improving access to medicines and addressing substandard and falsified antimalarials are essential to address health inequities and reduce the health and economic burden of malaria.