Medication adherence in type 2 diabetes mellitus patients during Covid-19 pandemic: a cross-sectional study from the United Arab Emirates

Background: Patients with chronic diseases often experience difficulty adhering to recommended treatments as instructed by their healthcare professionals. Recently, diabetes has been associated with the severity of the novel coronavirus disease (Covid-19), which raises the importance of improving medication adherence for diabetic patients to enhance the right use of antidiabetics amid the Covid-19 pandemic. Methods: This work assesses medication adherence among type 2 diabetes mellitus patients in the United Arab Emirates (UAE) and identifies the set of key demographic and health factors significantly associated with medication adherence. A descriptive cross-sectional study was conducted on an appropriate sample of type 2 diabetic patients in the UAE, with 180 patients of both genders and various social levels. A validated version of the eight-item Morisky Medication Adherence Scale (MMAS) was used for data collection. Results: The average MMAS score was 4.88, with 95% confidence intervals (CI) 4.6 and 5.2. 61.67% (n=111), 28.89% (n=52), and 9.44% (n=17) of patients were categorized into low, medium, and high adherent groups, respectively. These findings indicate that a high level of non-compliance to antidiabetic regimens among the population in the UAE. Conclusions : Patients demonstrated low level of compliance to antidiabetic regimens. Therefore, they must receive up-to-date knowledge about the disease and the treatment and enable easy access to their health care providers to enhance medication adherence.


Introduction
Medication adherence is the extent to which an individual takes medication as instructed by a healthcare professional. According to the World Health Organization (WHO), medication adherence is defined as "the degree to which the patient's behavior complies to the prescribed recommendations and instructions from the health care provider". 1 In this context, the concept aims to integrate the professional medical opinion and the patient's preferences and lifestyle where there is mutual cooperation between the physician and the patient to improve the outcome of the treatment and enhance the prescribed regimen's efficacy. Although adherence to medications guarantees a maximum benefit from the treatment plan, many patients do not comply with the instructions and the issue of non-adherence to medications is progressively increasing nowadays. 2 There are several types of non-adherence described in the literature. The first type is called 'primary nonadherence', in which the physician prescribes the treatment plan, but the patient does not follow the instructions from the beginning and does not initiate the plan itself. The second type of non-adherence is 'non-persistence non-adherence', where the patients start the treatment regimen but do not complete it and stop the regimen without consulting their physicians. Several factors contribute to the development of this type. For example, lack of frequent communication between the patient and the physician, the patient feeling initial improvement after taking the primary doses of the prescribed drugs, the patient cannot afford the price of the completely prescribed regimen, or difficulty in the accessibility of the drug, especially in rare diseases that requires certain drugs. The third type of non-adherence is 'non-confirming non-adherence' in which the patient does not strictly stick to the prescribed regimen. Instead, the patient changes the plan by altering doses or time, or even skipping some of them. 3 Assessing the patient's adherence is quite a challenging issue. However, there are several approaches adopted to fulfill this target. The first approach is the subjective method in which the patient, a family member, or caregiver is asked about the patient's commitment and the number of the doses taken with a specified time interval. The second method is the objective measurement, and it is achieved by counting the doses using electronic records for the medications or checking the pharmacy refill. A third method is the biochemical approach in which a non-harmful marker is added to the medication and then assessed in the organs where the drug is present or excreted, like assessing the serum or the drug's urine levels.
The patient is considered compliant if compliance exceeds 80% of the prescribed plan of treatment. 4 Recently, diabetes has been associated with the severity of the novel coronavirus disease (COVID- 19), where diabetic patients with COVID-19 are at higher risk to be admitted to intensive care unit and with higher mortality compared to nondiabetic COVID-19 patients 5-7 raises the importance of improving medication adherence for diabetic patients to enhance the right use of antidiabetics amid COVID-19 pandemic. 8 As far as the authors of this work are aware, no published studies in the literature have studied medication adherence among diabetic patients during the COVID-19 pandemic in the United Arab Emirates (UAE). This work assesses medication adherence among type 2 diabetes mellitus patients in the UAE and identifies the set of key demographic and health factors significantly associated with medication adherence during the year of 2020.

Ethical statement
Institutional ethical approvals were obtained from Ajman University by the Research Ethics committee (Approval number P-F-H-2019-Nov-28) and the Saudi German Hospital in Ajman.
Signed/verbal consent was obtained from the patients before data collection. Informed consent was given to all patients who participated in the study before answering the questionnaire and their personal information was kept in a closed closet for a certain period with full privacy was obtained from all subjects involved in the study.
Verbal consent was for patients who were not willing to complete the questionnaire in the presence of the interviewer and was completed by five senior Pharmacy graduate students as trained interviewers/data collectors. Please, note that the presence of the interviewer can cause bias, and this should be avoided.

Study design
This work is a descriptive cross-sectional study conducted among a convenience sample of type 2 diabetic patients in the UAE. A validated version of the eight-item Morisky Medication Adherence Scale (MMAS-8) was used for data collection. 9

REVISED Amendments from Version 1
New recent references that link COVID-19 severity with diabetes have been added. Minor change in the conclusion to more clear and concise was made.
Any further responses from the reviewers can be found at the end of the article

Participants
The criteria used to collect data for the research were composed of inclusion/exclusion criteria. For inclusion, the data collected was from randomly selected diabetic patients, who must be above the age of 18 years old and must understand Arabic or English. Questionnaires were available for participants at the reception of the hospital and clinics in envelopes for patients waiting to see their doctors and they were invited to take part in the study while waiting. Exclusion criteria included pregnant women, people with mental illness, and/or those who refused to participate.
In a systematic review published in 2015, an average of 65.8 percent (ranging from 38.5 to 93.1%) were found to be adherent to their diabetic medication. As such, we need to find if in our study there is a significant difference from 65.8%.
The sample size of 384 was determined by the use of Survey Monkey sample size calculator. The proportion of the targeted characteristics in the Northern UAE could not be estimated. Therefore, we used a calculation formula to yield the maximum number of required sample size. The following formula was used based on the sample required to estimate a proportion with an approximate 95% confidence level n ¼ z 2 pq=d 2 where n: the desired sample size z: the standard normal deviation, set at 1.96 (corresponding to the 95% confidence level) p: the proportion of the targeted characteristics in the region of Northern United Arab Emirates. Since there was no estimate available, it was set at 50% (or 0.50) q: 1-p d: absolute precision or accuracy, set at 0.05 Therefore, the research sample size is n ¼ ð1:96Þ 2 ð0:5Þ=0:05 2 ¼ 384:16 Due to the current COVID-19 pandemic situation and lockdown, it was challenging to reach the targeted sample of 384. A total number of 180 participants was achieved.

Data collection
The data was collected manually in the Saudi German Hospital in Ajman, six different clinics in Ajman and Sharjah, and via an online survey for other Northern Emirates. Physical questionnaires were available at the receptions of the hospital and clinics for participants who agreed to take part. Data collection took place from March 2020 to August 2020.
The self-reported eight-item MMAS was used to evaluate medication adherence among type 2 diabetes mellitus patients. 26 Questions 1 through 7 have categorical responses (yes/no). Item 8 has five-points Likert scale. All the questions except question 5 are reverse coded to avoid participants responding in the same manner to a series of questions irrespective of their content; each "No" is coded "1" and each "yes" is coded "0". For question 8, if the participant selects "0", the given score is "1", while if they select response "4", the given score is "0". Categorical responses "1, 2, 3" are respectively coded "0.25, 0.75, 0.75". The total MMAS-8 scores range from 0-8. The adherence level is considered low if the MMAS-8 score is less than 6 (score < 6), medium if in the range of 6-7 and high if equal to 8.
Missing total scores on multi-item instruments are equivalent to missing scores on a single-item instrument. Commonly used methods to deal with such missingness are complete-case analysis, mean imputation, or single-regression imputation. More advanced techniques that account for missing data uncertainty are multiple imputation or maximum likelihood estimation. Specific methods have also been developed for missing item scores in multi-item instruments, for example, person mean imputation, two-way imputation, response-function imputation, and multivariate normal imputation.

Statistical analysis
The data were analyzed by International Business Machines Corporation-Statistical Package for the Social Science (IBM-SPSS) version 25 (Chicago, USA). Before statistical analysis, all the dependent and independent variables were checked for any data entry errors or missing data. The number of responses and percentages were used to describe categorical variables and means AE relative standard deviation (RSD) was used to describe continuous variables. For instance, to identify outliers the distance between data point and the center of all data points to ensure data points do not fall within three SD of the mean. The normality of variables was tested using the Kolmogorov-Smirnov and Shapiro-Wilk test. Non-normally distributed data were stated using medians, interquartile ranges (IQR) and mean ranks. Qualitative variables were summarized using frequencies and percentages. The Chi-square and Fischer Exact tests were used to compare differences in proportions of qualitative variables. Univariate and multivariate logistic regression analysis was used to determine the significant factors associated with medication adherence. The stepwise method was used for variable selection and model building. A p-value < 0.05 was chosen to make statistical significance decisions.

Bias
Four sources of error have been described that can threaten the precision and accuracy (i.e., reliability) of survey results and must be evaluated by readers. The first type of error, coverage error (sampling bias), occurs when there is a discrepancy between the target population and the population from which the sample was derived. This type of error can compromise the ability to generalize study results.
Sampling error (or random error) occurs when the researcher surveys only a subset (sample) of all possible subjects within the population of interest. Sample error is reported usually as the mean AE1 standard error from the mean (SEM).
Measurement error (response bias) occurs when the collection of data is influenced by the interviewer or when the survey item itself is unclear from the respondent's point of view. Parallel forms (usually consisting of alternatively worded items placed throughout the survey) of either specific survey items or the entire survey instrument have been used to increase reliability of mail survey research.
Accurate assessment of measurement error relies on the provision of the questionnaire or tool used to collect data so that readers may analyze wording. Unfortunately, however, many articles relating to the results of survey research do not include the actual questionnaire used in the survey due to space and ownership issues.
Finally, nonresponse error (nonresponse bias) occurs when a significant number of subjects in a sample do not respond to the survey when responders differ from non-responders in a way that influences, or could influence, the results. Since the lower response rate may increase the potential for higher nonresponse bias, and in order to minimize the possibility of nonresponse error we could manage to increase the number of participants by including more patients from different clinics in more geographical areas in UAE. To minimize bias from the interviewers, five pharmacy graduate students were recruited and trained to conduct the interviews.
The results of statistical modeling showed that patient's age, marital status, type of antidiabetic medications, and being confident about taking diabetes medication were strong determinants of medication adherence among patients with type 2 diabetes mellitus. Significantly decreased medication adherence was observed in those with single marital status (OR 0.366; 95%, CI 0.139-0.962), patients aged 50-59 years (OR 0.238; 95%, CI 0.058-0.983), patients aged 40-49 years (OR 0.195; 95%, CI 0.044-0.857) and patients aged 30-39 years (OR 0.195; 95%, CI 0.035-0.648). On the other hand, significantly increased medication adherence was observed in the patients who had non-insulin antidiabetic medication therapy (OR 3.085; 95%, CI 1.125-8.457) and those who were confident about taking diabetes medication (OR 8.200; 95%, CI 1.036-64.908). Table 2 shows the lifestyle and health information of the participants. Most of the participants exercise two times/week or less (n = 145, 80.6%) and 115 (63.9%) were nonsmokers.

Lifestyle and health characteristics
Of the total participants, 112 (62.2%) had one chronic disease, 35 (19.4%) had two chronic diseases, and 33 (18.3%) had ≥ three chronic diseases. The years since diagnosis as diabetic were as follows: 44 (24.4%) diagnosed as diabetic under one year ago, 55 (30.6%) within two to five years, and 81 (45%) diagnosed more than five years ago. Among the participants, 122 (67.8%) had a diabetic family history. Half of the study participants (n = 90, 50.0%) had one antidiabetic medication. The frequency of taking antidiabetic medications was as follows: 68 (37.8%) took the medication once, 83 (46.1%) twice, and 29 (16.1%) ≥ 3 times. The type of antidiabetic medication was detailed as follows: 38 (21.1%) had insulin therapy, 114 (63.3%) had non-insulin therapy, and 28 (15.6%) had both insulin and non-insulin therapy. The majority of the participants (n = 164, 91.1%) were confident about taking diabetes medications, and 120 (66.7%) believed that it is possible to develop complications if they do not take diabetes medication as instructed.

Assessment of medication adherence
The average MMAS score was 4.88, with 95% confidence intervals of 4.6 and 5.2. Using the grading system mentioned in the Methods section, it can be claimed that the overall level of medication adherence was poor. Of the total 180 diabetic patients, (n = 111, 61.7%), (n = 52, 28.9%), and (n = 17, 9.4%) were low, medium, and high adherent groups, respectively. Table 3 presents the results of each item related to the MMAS-8.  Table 4 shows the distribution of medication adherent groups according to demographic information. There was a statistically significant association between participants' age and medication adherence, with low medication adherence scores among patients aged below 40 years compared to older patients (P = 0.020). Married patients were more likely to have better medication adherence scores than single participants (P = 0.018). Emirati patients were more likely to have better medication adherence scores than non-Emirati patients (P = 0.025). Table 5 shows the distribution of medication adherent groups according to lifestyle and health information. Patients who were confident about taking diabetes medication were more likely to have better medication adherence scores than those who weren't confident about taking diabetes medication patients (P = 0.021). To select the set of factors that jointly influence the medication adherence, we used the stepwise procedure applied to the multivariate logistic regression model. The results of this procedure showed that patient's age, marital status, type of On the other hand, significantly increased medication adherence was observed in the patients who had non-insulin antidiabetic medication therapy (OR 3.085; 95%, CI 1.125-8.457), and those who were confident about taking diabetes medication (OR 8.200; 95%, CI 1.036-64.908).

Discussion
According to the MMAS, our study showed that the prevalence of low adherence to antidiabetic medication was 61.67%, medium adherence was 28.89%, and high adherence was 9.44%. These findings reveal a high level of non-compliance to antidiabetic regimens among the UAE people.  12 Other studies reported that the level of non-adherence in Malaysia ranges from 40% to 50%. 16,17 In India, there were contradicting results that ranged from 30% to 60% of non-compliance. 18,19 It can be seen from these results that the level of non-adherence is higher amongst the Arab population compared to Asian or American populations. One possible reason for the increased level of medication non-compliance among Arabs is that the level of diabetes, hypertension, and other chronic diseases is higher among Arab citizens, which increases the number of individuals diagnosed with these diseases and increases the probability of non-adherent individuals. Furthermore, the Arab people share other factors that directly contribute to higher levels of non-compliance among them. This may require more investigations to determine the contributing factors such as socioeconomic and lifestyle factors. Studies have shown that several factors are associated with the increased level of non-compliance to antidiabetic medications. Our study showed that age is a significant determinant of the level of adherence. We report that the level of adherence significantly decreased with increased age up to the age of 60 years old. There was no statistically significant decrease in the level of adherence among patients older than 60 years old. These results were slightly different to Al Haj and their coworkers' findings, who concluded that the level of adherence is increased with age. 20 Aloudah et al., also reported a lower level of adherence among the younger population compared to the older population. 13 However, Mukherjee et al., reported similar results to our findings and concluded that the level of compliance significantly decreases with increased age. 18 In addition, Ahmad et al., reported similar results about the impact of age on the level of adherence. 17 A possible explanation for the decreased level of adherence with increased age of patients is that patients can forget about their medications and several studies have reported that forgetfulness is one of the leading causes behind non-adherence. 18,21 Another factor that influences the level of non-compliance is marital status. Our study showed low adherence to antidiabetic medication among unmarried patients compared to married ones. These findings were consistent with Ahmed et al., 11 who reported similar findings of the significant effect of marital status on adherence level. Gelaw et al., 22 and Thakrar et al., concluded that married persons are more adherent to medications than single ones. 19 The possible reason behind this relationship is that married individuals are more supported by their families to adhere to their regimens. In addition, forgetfulness is less likely when the family reminds the patient of his/her schedule and medication time. However, the effect of marital status was not proven in all the literature that studied the topic. Several studies reported no significant relationship between marital status and drug adherence. 12,14,18,23  These contradicting results signify that further investigations need to be conducted in order to find out the real relationship between the variables.
The regimen of antidiabetic medications was thoroughly analyzed in several publications. Studies discussed the types and number of prescribed antidiabetic agents and their effect on the level of adherence. In our study, we found out that the level of adherence is increased with non-insulin antidiabetic medications compared to the insulin group. Our results were consistent with most of the previous literature. Mukhurjee et al., reported that a low adherence level is found among the patient group who were prescribed insulin only or insulin combined with oral hypoglycemic agents (OHAs) compared to people who use OHAs only. 18 The same findings were reported by Khan et al., 14 Chua et al., 16 Ahmad et al., 17 Al Haj et al., 20 Aminde et al., 24 and Bali et al. 25 Each of the previously mentioned studies confirmed that insulin only, or in combination with OHAs, is a significant factor in decreasing adherence and patients are more adherent when only OHAs are prescribed. A possible explanation for this finding is that people prefer to take their drugs via oral routes rather than injections. Hence, they are more adherent when taking oral pills. In addition, a combination of insulin and pills means that the patient must take two different drugs at different times; this increases non-adherence, as they have to stick to two (or more) drugs rather than only one. The previous hypothesis is supported by most of the previously mentioned papers,   11 Kalyango et al., also reported no significant effect of either the number of administrated drugs or the administration route on the level of patient's adherence to the medications. 11 The mainstay of literature is consistent with our findings, which supports our hypothesis that the number of drugs and the administration route do affect patients' compliance with antidiabetic medications.
We also found that the patient's knowledge and persuasion to take the medication were significantly associated with higher adherence to the medication. These results were consistent with other literature which supported our findings and reported a higher level of adherence associated with increased patient knowledge of the drugs and the complications of the disease. 17,18 The reason behind this finding is quite obvious. It is expected that patients' adherence increases when they realize the importance of the drug on the prognosis of their condition, complications of the diseases, and the burden of their pathology on the family, community, and themselves in the first place.
Interestingly, we found no significant relationship between the level of adherence to antidiabetic medications and factors such as sex, educational level, monthly income, exercise, smoking, and the duration of diabetes. Each of these factors was discussed in one of the previously mentioned literature and one paper discussed a significant effect between each of these factors and the level of adherence. 13 In this study, low level of adherence is associated with young and male people, and that people with high education and monthly are more adherent to their medication. It was also found that the duration of diabetes when exceeding 10 years is also associated with low levels of adherence.

Limitations
A possible limitation to our findings is the low number of participants enrolled in the study. Due to the current pandemic situation, we only managed to enroll 180 patients while most of the previous studies used the data of more than our number. Further investigations need to be conducted to reassess the effect of these factors on the level of patient's adherence to antidiabetic medications.

Conclusions
Patient adherence to antidiabetic medications is crucial in maintaining low blood sugar levels. With the recent findings that link diabetes with the severity of COVID-19, it is essential to test the willingness of patients to improve adherence to their antidiabetic regimens during the pandemic. Our study showed that the prevalence of low adherence to antidiabetic medication is 61.67%, medium adherence is 28.89%, and high adherence is 9.44%. These results revealed a high degree of non-compliance to antidiabetic regimens among UAE individuals. Non-adherence can occur as a result of the patient intentionally disregarding their treatment schedule, or as a result of carelessness and/or forgetfulness, whereby patients often omit their medication from their daily routine or take the medication later than necessary. We should ensure that patients receive up-to-date knowledge about the disease and drugs, apply prescription treatment activities, provide written and oral information to the patient, and enable patients to contact their health care providers regularly to enhance medication adherence.

Data availability
Underlying data The questionnaire responses and interviews for those not willing to complete the questionnaire themselves that were collected/conducted from/with the participants are not openly available to protect the confidentiality and privacy of the participants. All document files were eradicated after data analysis. De-identified questionnaire responses are available in English and Arabic. The data can be obtained by applying to the Ajman University Ethical Committee through rec@ajman.ac.ae. Alternatively, please contact the corresponding author Dr. Akram Ashames at a.ashames@ajman.ac.ae who can facilitate this process. The ethical committee will study data requests on a case-by-case basis to ensure integrity, objectivity, confidentiality, and professional behavior.

Mukhtar Ansari
Department of Clinical Pharmacy, College of Pharmacy, University of Hail, Hail, Saudi Arabia Overall, the topic of research is interesting and relevant, however, several studies have already been performed in gulf as well as other part of the world. The manuscript requires major revisions both in context as well as in grammar. It is advisable to copy edit the the manuscript after addressing the reviewers' comments. Abstract: More details are required on how data was collected, who collected the data and how they were analyzed? Result section is missing with pertinent data from table 6. Conclusion should be based on the findings rather than generalization.
○ Introduction: Can further be improved in terms of stating the problem statement and justification.

Method:
It is not clear whether Signed or verbal consent was taken.
○ Convenient sampling is a non-probability sampling, so the findings can not be generalized. However, more details are required about how convenient sampling was used.
○ It is confusing, to mention 'convenient sampling' and 'randomly selected diabetic patients'. \it is not clear whether English or Arabic version of the questionnaire was used? ○ This sentence "In a systematic review published in 2015, an average of 65.8 percent (ranging from 38.5 to 93.1%) were found to be adherent to their diabetic medication. As such, we need to find if in our study there is a significant difference from 65.8%" should go in introduction section. Authors have reported non-compliance with Diabetic medicines. The study was carried out during COVID pandemic so the sample size was insufficient. this has been reported by authors as a limitation of the study. The study has been well designed for hypothesis testing for noncompliance.
which platform did you use for the data collection.
○ which tools did you use to distribute the online survey among the participants? ○

Results:
Results section is clearly and well-presented.

Discussion:
Please draw a concise conclusion from this study and present the limitations and future research and provide further information about study implications and future research.