Background: Artemisinin-based combination therapies (ACTs) are most effective for the treatment of malaria if patients adhere to their prescribed treatment regimen. Most adherence studies have focused on artemether-lumefantrine (AL), with limited research on adherence to the fixed-dose co-formulation of amodiaquine-artesunate (AQAS). To address this gap in evidence, this thesis aimed to: 1) measure and compare the level of adherence to ACTs at the population and to AL and AQAS at the health facility level in Sierra Leone; and 2) identify factors associated with adherence to these ACTs. Methods: A mixed-methods approach was taken to address four specific objectives. First, data from a nationwide cross-sectional household survey were used to estimate adherence and factors associated with adherence to ACTs at the population level. Second, a randomised controlled trial (RCT) was conducted to compare the level of adherence to AL and AQAS in children aged < 5 years at two public health facilities in Freetown, Sierra Leone. Third, factors associated with non-adherence to ACTs were identified using data from the RCT. Finally, in-depth interviews with caregivers enrolled in the RCT were conducted to explore the barriers, facilitators, and contextual factors that may influence adherence. Results: In the nationwide survey, 1,641 children under-five with fever were identified. Of these, only 467 received treatment with an ACT; 220 (47.2%) received the recommended 3-day treatment. In contrast, adherence to ACTs was much higher in the RCT. Of the 784 children enrolled and randomised into the trial, 660 (85.6%) were included in the per protocol analysis (340 AL, 340 AQAS). Definite adherence (self-reported adherence plus empty package) was higher for AL than AQAS at both sites (Site 1: 79.4% AL vs 63.4% AQAS, OR 2.16, 95% CI 1.34–3.49, p=0.001; Site 2: 52.1% AL vs 37.5% AQAS, OR 1.53, 95% CI 1.00–2.33, p=0.049). Self-reported adherence (ignoring drug package inspection) was higher for both regimens at both sites and there was no strong evidence of variation by treatment (Site 1: 96.6% AL vs 95.9% AQAS, OR 1.19, 95% CI 0.39–3.63, p=0.753; Site 2: 91.5% AL vs 96.4% AQAS, OR 0.40, 95% CI 0.15–1.07, p=0.067). AL was less likely to be taken correctly at one site, but was better tolerated than AQAS at both sites. In an adjusted multinomial regression analysis of the RCT data, the relative risk ratio (RRR) of non-adherence compared to definite adherence was significantly higher if the caregiver reported their child disliked the drug (RRR=8.04; 95% CI 2.69–23.98; p<0.001) or experienced adverse events/side effects (RRR=4.48; 95%CI 1.78–11.24; p=0.001) versus caregivers who did not. Logistic regression models assessing factors associated with self-reported non-adherence and incorrect treatment revealed strong associations with receiving AL, disliking the medication, and experiencing an adverse event. Interviews with 49 caregivers highlighted three key factors that influenced access to medications and adherence to treatment: (1) characteristics of the medications; (2) health system-related factors; and (3) caregivers’ previous experience with malaria treatment. Conclusions: Although adherence to ACTs in the national survey was low, adherence to both AL and AQAS in the RCT was much higher, but was influenced by the criteria used to define adherence. Higher number of tablets or daily doses (such as those required for AL), dislike of the medication (including bitter taste), and perceived side effects may contribute to poor adherence. Child-friendly formulations and patient-centred services may positively impact adherence to ACTs as may prior caregiver experience with ACTs. The responsibility to maximize adherence to ACTs lies not only with the patient and caregiver, but also more broadly with the health workers and the health system. This thesis contributes to the knowledge base on adherence by providing a population estimate for ACT adherence in Sierra Leone, comparative estimates of two co-formulated ACTs, explores factors associated with non-adherence, as well as expands on the methodological challenges highlighting the need to standardize the methodology for defining and measuring adherence.