Factors associated with non-adherence to Artemisinin-based combination therapy (ACT) to malaria in a rural population from holoendemic region of western Kenya

The current study was carried out in Southeast Alego location in Karemo division of Siaya County, Nyanza Province, Kenya. In 2009, the national housing and population census, recorded that Karemo division had a total population of 84,986 with an urban population of 19,611. The total area of the division is 235.1 km² with a population density of approximately 350 people/km²[17]. Karemo division has five locations; Siaya township, North Alego, East Alego, South Alego and Southeast Alego locations. Southeast Alego location consists of four sub-locations; Masumbi, Nyang’oma Kogelo, Mur-Ng’iya and Bar-Agulu sub-locations. The area experiences two rainy seasons; the long rainy season (between April-June) and the short rainy season (between October-December). Challenges faced by the community include low farm productivity, high rates of unemployment, and most importantly, resource-constrained health facilities due to poverty. These act as the primary drawbacks in the control of infectious disease such as malaria, with the highest rates of infections being experienced immediately after the rainy seasons [17].

In order to determine adherence level, and causes of non-adherence and accessibility of ACT in the study population at a point in time, residents living within households (n = 297) in the above described remote region of Siaya County in western Kenya, were recruited. This was a cross-sectional survey in which sampling was carried out using multi-stage and probability proportion to size sampling methods. Under these methodologies, rather than using all the elements contained in the selected clusters, the researcher randomly selected elements from each cluster. Constructing the clusters was the first stage while deciding what elements within the cluster to use was the second stage. The divisions, locations, sub-locations and the villages formed the clusters. Siaya District has three divisions namely Karemo, Boro and Uranga divisions (according to the administrative boundaries at the time of the study) (Figure 1). Karemo division has five locations namely; Siaya Township, North Alego, East Alego, South Alego and Southeast Alego was selected as study units at the division level. Southeast Alego location, which has four sub-locations namely Masumbi, Nyang’oma Kogelo, Mur Ngiya and Bar Agulu was selected at the location level. The sample size was drawn from all the four sub-locations of Southeast Alego location. The sub-locations had a total of forty two (42) villages. To ensure equal representation from each of the sub-locations and the villages, data on the number households in all the sub-locations and villages were obtained from Kenya Bureau of Statistics offices in Siaya District. The obtained data was then used in determining the proportionate sample size for each sub-location and village. This site was targeted for the study since it is one of the areas with highest malaria burden in the district. In addition, interventions such as stable supply of ACT were being implemented in this study area at the time of the study (Figure 1). These methodologies were applied since a complete list of all members of the population did not exist. This strategy was preceded by systematic random sampling in 388 households (in which a member had used an anti-malarial two weeks prior to the interview). The sample size was based on Fisher’s formula for sample size estimation [18] taking into account population size greater than 10, 000 (i.e. a total of 17,780 people were resident in the study area at the time of the study). The formula for sample size determination used the prevalence of non-adherence to anti-malarial and this was estimated at 50% since the prevalence is not known.

Multi-stage and probability proportion to size methods were used to determine the sample size per sub-location in the study area. Systematic random sampling method was then used to sample out the households to be recruited as sampling unit [19, 20]. Under this methodology, a member with a recent malaria episode from every third household on entering a village was recruited and interviewed. In the absence of a recent malaria case (within two weeks prior to the survey), the interviewer moved to the next household. From the 388 sampled households, a total of 297 who were either laboratory-diagnosed or had self-reported episodes, used ACT in treating malaria. The others (91) used other anti-malarial drugs which were not artemisinin-based. In the absence of laboratory diagnosis, malaria according to residents in the study area was defined by either the presence of/or mixed symptoms of fever, diarrhoea, vomiting and joint pains. However, if more than one individual in a household had experienced malaria and had used anti-malarial within the described period (i.e. two weeks), then focus was on the most current episode, while the rest were excluded. This strategy was adopted so as to capture the most recent pattern of anti-malarial drug use and to minimize errors associated with recall. In children <13 years, the parents/caregivers or a legal guardian were interviewed. Data was collected through interviews, using structured questions in the selected households. Information on the anti-malarial drugs taken, adherence and causes of non-adherence to ACT and accessibility to the drug were collected. Adherence level to ACT was then assessed by duration of treatment and number of tablets taken in relation to the age of the patient.

Key Informant Interviews (KIIs) were also conducted with Kenyan regulatory body-approved healthcare providers and dispensing chemist operators in the study area. This included 3 health care providers [1 at Ng’iya Health Centre (HC), 1 at Nyang’oma Kogelo Dispensary and 1 at Bar-Agulu Dispensary) and with 3 pharmacists (1 at Ng’iya shopping center (SC), 1 at Nyang’oma Kogelo SC and 1 at Bar-Agulu SC). Saturated sampling design was used to arrive at the number of the key informants. Through the KIIs, information on adherence to prescribed drugs, reasons for non-adherence and accessibility of ACT was collected. The study was approved by the ethical and scientific review committees at the Maseno University, Kenya and the Kenya National Council for Science and Technology Board.

Quantitative data was collected from 388 households [from which those using ACT (n = 297) was obtained]. The interview was carried out using structured pre-tested questionnaire, administered to the particular family member who had malaria and used the anti-malarial drugs in the household. However, in children below <13 years, parents/caretakers or a legal guardian provided the information. Using multi-stage and systematic random sampling, households with recent malaria episodes (two weeks prior to interview date) were established. In each sub-location in the study area, a sample proportion to the total household size was sampled and interviewed. Particular emphasis was on the most recent episode of self/presumptive or laboratory diagnosed malaria in the household within the last two weeks.

Prior to the actual study, field assistants’/enumerators (n = 4) who are residents of Southeast Alego location and with a minimum of fourth form level of formal education were interviewed and recruited. The enumerators were fluent in English, Swahili and Luo languages. Training of the research assistants on survey interviewing techniques was performed for one day, followed by another day of pre-testing of survey tools and methodologies in Pap Nyadiel village, Masumbi sub-location (a neighboring region exhibiting the same characteristics as the study site). This village was then excluded during the main data collection. Based on the experiences and results of the pre-test, further re-training and refining of techniques of interviewing and modification of research tool was done. The research assistants used drug charts containing samples of commonly used anti-malarial medicines to aid recall and validate reports, remaining packages or tablets were reviewed where possible. Hospital cards were also reviewed to determine the duration of treatment. Data was collected by the researcher and field assistants. Qualitative data was also collected through KIIs with healthcare providers and chemist shop owners. Written informed consent was provided by the study participants or, where participants were children, a parent or guardian.

Statistical analyses were performed using Statistical Packages for Social Sciences (SPSS, ver. 15.0). Descriptive analysis was then conducted to determine adherence level to ACT with reference to duration of treatment and number of tablets taken in relation to the age of the patient as recommended by the Ministry of Public Health and Sanitation and Ministry of Medical Services in Kenya (Table 1). In addition, a multivariate logistic regression analyses between the independent and dependant variables, was used to identify variables associated with adherence to ACT. The independent variables in the regression model included the following factors; demographic (age, gender, marital status, household head, and household size); socio-economic (source of income, household monthly income, education level and ability to read) and environmental (source of drugs, distance to source of drugs, and availability of anti-malarial drug at source). Chi-square analyses were used to determine proportionality. Statistical significance was assessed at P ≤0.050. Data collected by KIIs was organized into themes and analyzed.

Table 2 presents the demographic and socio-economic characteristics of the study participants. A total of 388 households sampled were included in the analyses. Out of 388 households, 297 (76.5%) used ACT (Coartem/Artemether Lumefantrine) while 91 (23.5%) used other anti-malarial drugs. Overall, males were 186 (47.9%) while females were 202 (52.1%). A total of 162 (41.8%) were aged <13 years while 226 (58.90%) were ≥13 years. Majority of households consisted of 3–5 individuals [200 (52.5%)], others were >6 individuals [155 (40.0%)] and 1–2 individuals [33(8.5%)]. Households headed by husbands were 269 (69.3%) and wives, 119 (30.7%). Respondents who were married comprised of 105 (27.1%), single 79 (20.4%), widowed 30 (7.7%), while underage (<15 years and assumed to have not acquired marriageable age) were 169 (43.9%) (Table 1).

Amongst the parents/caretakers, 40 (10.3%) did not have any formal education, 188 (48.5%) had primary education, 116 (29.9%) had secondary education, 37 (9.5%) had post-secondary education, while 7 (1.8%) were still in school. A higher proportion of the respondents (68.3%) were able to read, while 31.7% were unable to read the written prescription (Table 2). Farming (38.9%) was the main source of income; others were petty business (20.6%), salaried employment (11.9%), casual work (11.3%), and self-employment (17.3%). The respondents with monthly income of below KShs. 4,500 (~ $56) were 204 (52.6%), between KShs. 4,500-9,000 ($56-112) were 133 (34.3%), while those with over KShs. 9,000 (>$ 112) were 55 (13.1%) (Table 2).

Results on the sources of anti-malarial drugs in Southeast Alego location demonstrated that a total of 77.8% of all anti-malarial drugs were from government institutions, 17.8% from nearby shops/pharmacies/chemists, 3.1% given by neighbor/ friend/ relative, and 1.3% were left-over drugs in the house (previously given in the hospital).

Out of those who used ACT (Coartem; n = 297), 278/297 (93.6%) obtained the drugs from government institutions, 14/297 (4.7%) were given the drugs by friend/neighbor/relative, 3/297 (1.0%) bought them from pharmacy/shop and 2/297 (0.7%) used left-over medicine in the house. Since the study mainly focused on ACT, distance to the source of ACT was measured based on the time taken to reach ACT source while walking. The respondents who used <30 minutes were 44.8% while 55.2% used >30 minutes.

In the context of the current study, adherence was defined as abiding by the recommended dose and period of usage of ACT. In order to assess level of adherence to ACT use, the number of tablets taken in relation to age and days of treatment were evaluated. Results revealed that most of the respondents who took ACT [215/297 (~72.0%)] had the recommended number of tablets according to their ages while others [82/297 (~28.0%)] either took less or more than the recommended number.

Duration of treatment was also assessed and results demonstrate that 161 (54.0%) took ACT within the recommended 3 days, 22 (8.0%) within <3 days while 114 (38.0%) within >3 days. When combined (duration of treatment and number of tablets taken), adherence level was found to be 140 (47.0%) for adhered and 157 (53.0%) for non-adherence.

Further analysis on adherence was conducted to determine the level of adherence with reference to demographic and socio-economic factors of the study population. General stratification of adherence to ACT and demographic factors revealed the following patterns: 57.9% of individuals <13 years and 42.1% of those >13 years of age adhered. Additional stratification of demographic factors [gender, marital status, household (HH) head, and HH size] and adherence was also conducted. Chi-square analyses revealed that age significantly differed between the ACT adherents and non-adherents (P = 0.011). However, gender, marital status, HH head, and HH size were comparable between the ACT adherents and non-adherents (P = 0.178, P = 0.220, P = 0.140, and P = 0.121, respectively) (Table 3).

In order to determine adherence level with reference to socio-economic factors, the study participants were stratified according to education level, ability to read, income per month and occupation. Chi-square analyses demonstrated that education level (P < 0.01), monthly income (P = 0.002) and ability to read (P < 0.01) significantly differed between adherents and non-adherents (Table 3). However, HH income source was comparable between the ACT adherents and non-adherents (P = 0.094) (Table 4).

In order to determine factors associated with adherence to ACT in Southeast Alego location, multivariate logistic regression analyses were performed. The independent variables included demographic (age, gender, marital status, household head, and household size), socio-economic (household source of income, household monthly income, education level and ability to read) and environmental factors (source of drugs, distance to source of drugs, and availability at source). The above factors were regressed against the outcome variable; adherence. The full regression model of the demographic factors revealed that age was a significant factor in determining ACT adherence since many respondents (57.9%) who were <13 years (OR, 0.571; 95%CI 0.306-0.905; P = 0.017) adhered as opposed to respondents who were ≥13 years (42.1%) (Table 3). Consistently, logistic regression model demonstrated that low age (OR, 0.571, 95% CI, 0.360-0.905; P = 0.017), higher education level (OR, 0.074; 95% CI 0.017-0.322; P < 0.01), ability to read (OR, 0.285, 95% CI, 0.167-0.486; P < 0.01) and higher income (Ksh. > 9000; OR, 0.340; 95% CI, 0.167-0.694; P = 0.003) were associated with ACT adherence (Tables 3–5). However, adherence was not significantly associated with gender, HH size, HH head, marital status, source of ACT and distance to the source (Tables 3–5).

Accessibility was assessed by distance to drug source, drug availability at the source, affordability and acceptability of ACT. Assessment of distance from the drug source revealed that majority of the respondents cover longer distances (55.2%) to reach the source of ACT. Moreover, 52.9% of the respondents reported that ACT was not always available at the source. Further analyses revealed that drug availability (P = 0.020) and distance to drug source (P < 0.01) significantly varied in the sub-locations (Table 6).

In order to determine affordability, Key Informants were asked about the cost of malaria treatment at the health facilities. The respondents were also asked to comment on ACT treatment. The results revealed that ACT was more affordable at government health institutions as compared to other sources. In these facilities, with an average amount of KShs. 40, the respondents could get ACT and analgesics. “What complicates this is the distance to these health facilities”, noted one of the key informants. At the dispensing chemist, ACT was sold at KShs. 250 and KShs. 450 (for child and adult dose, respectively). In addition, the dispensing chemist operators reported that as a result of the high costs, they do not stock a lot of ACT, since they end up expiring on the shelves. However, there were ACT (generics) which retail at KShs. 40 in one chemist, and these were affordable to a majority of the study population.

In addition, acceptability level to ACT in treatment of malaria was also examined with reference to the respondents’ perception on its usage. The results revealed that acceptability was high (95.4%) since only 4.4% of those who used ACT were not content with it. Others complained of the number of tablets taken (28.7%), duration of treatment (9.5%) and the smell, taste and color (29.1%) to be discouraging, however, they said that when taken, it cures malaria (Table 7).

The study was based on both self-reported and medically-diagnosed episodes of malaria experienced in the previous two weeks prior to the interview. Hence, we could not verify the authenticity of the information obtained from the respondents. Secondly, the case-specific definition of malaria could not be ascertained in this study, since some sought to over-the-counter drugs when they had symptoms similar to those of malaria. Moreover, there was recall bias in this report. Participants were asked questions based on a two-week recall period. Some of the participants would therefore, not vividly remember whether they actually had malaria/fever or how they took the drugs in the last two weeks preceding the interview. To increase accuracy on duration of treatment, hospital cards were reviewed and the drug sachets observed and number of tablets remaining or taken were counted. This certainly affected the quality of this study. However, the two week recall period is the reference period used by WHO surveys and has been adopted by many researchers for similar surveys. Therefore, findings in this report are of comparable quality. Similarly, structured interview-based studies are limited by the pre-determined responses, which are susceptible to response bias. However, attempts were made to minimize this potential source of bias by testing the tool, training field assistants in its administration, and amending it to make the wording familiar and culturally-appropriate. Finally, not asking study participants how they took Co-Artem or what advice they were given about how to take their Co-Artem was the other major limitation of the study since we could not asses the level of failure rates as a result of these components. This follows previous observations that absorption of hydrophobic lipophilic lumefantrine is greatly increased by a small amount of fat (1.2 g) co-administration [35] and failure to implement this component during prescribing result in increased failure rates [36].