COPING WITH HUMAN IMMUNODEFICIENCY DISEASE AS PREDICTOR OF ADHERENCE WITH ANTIRETROVIRAL MEDICATION

Infection by Human Immunodeficiency Virus (HIV) is a multifactor disease process in which the patient must confront an array of physiological, sociocultural, economic, and psychological stressors that have the collective potential for triggering major stress responses and psychological dysfunction. People's reactions to and the outcomes of traumatic events are mediated by their subjective style of coping. Coping is defined as a person's "constantly changing cognitive and behavioral efforts to manage specific external and internal demands that are appraised as taxing or exceeding the resources of a person". Coping, an important psychological construct has been shown to affect outcomes such as disease progression and quality of life in HIV infected patients. But the subjective styles of coping that patients use to cope with HIV have rarely been assessed as predictors of medication adherence. This study aims at determining the role of coping styles as determined by the "Ways of coping questionnaire" modified by Dunkel-Schetter et al. to suit their study of cancer patients as predictors of medication adherence in patients infected with HIV. The five dimensions of coping identified by Dunkel-Schetter et al. were the first to be identified with a large sample of cancer patients and may be representative of the universal dimensions of coping. This study is the first to utilize the dimensions of coping as described by Dunkel-Schetter et al. to predict adherence in HIV patients. The few studies on coping styles and adherence as an outcome use varied coping scales to assess coping like the Billings and Moos coping inventory and original Ways of coping ( questionnaire. Their results have shown that poor copmg strategies like avoidance coping were associated with non-adherence. Methods: The sample for the study consisted of 145 patients who were currently prescribed medication for HIV. The medication adherence shown by patients with anti-retroviral drugs and protease-inhibitor drugs was assessed separately. A total of 137 patients were on anti-retroviral drugs, while 77 patients were on protease-inhibitor drugs. Medication adherence was determined by using the "percentage of doses missed in the past three months and "Medication Adherence Scale. Coping strategies used by the patients were assessed by the "Ways of coping questionnaire" developed by Lazarus and colleagues and later modified by Dunkel-Schetter et al. to suit their study of cancer patients. The scale comprised of five coping sub-scales: seeking social support, distancing, focusing on positive, behavioral escape avoidance and cognitive escape avoidance and assessed the frequency of use of each coping style. Several demographic variables as well as clinical variables, which are known to affect medication adherence, were examined. Logistic Regression analyses were used to determine whether the coping strategies were predictive of medication adherence controlling for the confounding factors . Results: In agreement with previous research that shows that poor coping was associated with non-adherence, for the patients prescribed antiretroviral medications, behavioral escape avoidance was found to be significantly and inversely associated with adherence. Seeking social support, distancing, focusing on positive and cognitive escape avoidance were not found to be significantly associated with medication adherence. ACKNOWLEDGEMENTS This thesis would not have materialized but for the invaluable inputs of many people whom I am indebted to. I would like to express my thanks to Dr. Cynthia Willey, my major advisor who helped me with this thesis right from the conception of the idea to the very end. Her expertise and authority in the field of Epidemiology and her advice always led me on the right path. It is a privilege to be associated with her as her student. I would like to congratulate Dr. Stephen Kogut for successfully defending his Doctoral Thesis and being inducted as faculty member in the Applied Pharmaceutical Sciences. I cannot thank him enough for his tremendous inputs and help. He was always approachable and more like a friend than an instructor. I express my thanks to Dr. King for being so enthusiastic about my research and for being on the defense committee. I express my gratitude to Dr. Keykavous Parang for chairing my defense. I would like to express my thanks to all my friends, including Sridhar, Anil, Amogh ,Rakesh and Sanjeev who were great support for me in a new country. A special thanks to Aniruddha for the numerous brainstorming sessions. Above all, I am deeply indebted to my family who has always been there for me.


List of Tables
Until quite recently, the disease was considered to carry an almost certain debilitating, downward course leading to early death from opportunistic infections. A variety of medications were used to treat HIV related diseases, and some such as Zidovudine could temporarily suppress levels of HIV responsible for immune compromise. However the treatment only produced transient benefits because the circulating HIV remained in enormous quantity and the virus has a rapid error prone replication cycle that allows it to quickly evolve resistance to any single drug. The nature of medical care changed dramatically in 1996 with the development and wide use of treatment regimen that added a new class of antiretroviral medication called protease inhibitors in combination with other antiretroviral medications (2). Highly Active Anti-retroviral Therapy (HAART), usually a protease inhibitor combined with at least two other drugs, controls the viral replication by targeting specific viral enzymes. There are currently two distinct groups of anti HIV drugs that are targeted at different viral enzymes.
These are reverse -transcriptase inhibitors and protease inhibitors (3). HAART has enormous potential to delay disease progression and death (4). HAART is designed to suppress HIV viral replication, which results in increases in CD4 cell count, improved immune function, delayed clinical progression, and prolonged survival (5,6). Successful treatment of HIV with HAART requires that patient maintain nearly perfect adherence to the prescribed regimen. Adherence, often used interchangeably with compliance, is "the act, or quality of being consistent with administration of prescribed medication". Non-adherence may mean not taking medication at all, taking reduced amounts, not taking doses at prescribed frequencies or intervals or not matching medication to the food requirements (7).

A] Important of Adherence
Adherence to HAAR T is the single most important factor for achieving maximum and durable HIV plasma viral load suppression. Several studies have demonstrated that lapses in anti-retroviral adherence lower the likelihood of suppressing viral loads below detectable limit (8). Non-adherence leads to increased mortality and morbidity. A study by Hoggs et al. (9) reported a 16% rise in mortality for every 10% drop in adherence.
Strict adherence to HAART is imperative because the therapy is "unforgiving" in two respects. First, m non-adherent patients, resistant viral strains develop because of high rates of viral mutation and the short half-life of the drugs (10). Condra et al. (11,12,13) reported that resistance might develop after missing as little as one dose in five. The genetic mutations that result in drug resistance often confer resistance to an entire class of protease inhibitors or non-nucleosides. Thus, in failing one regimen, a non-adherent patient may severely limit future antiviral options (10). Secondly, the level of adherence that must be achieved and maintained for maximal effectiveness exceeds that needed for effective therapy in many other chronic conditions. With HAART, patients must maintain near-perfect adherence to maintain an undetectable viral load (10). A study by Paterson et al. reported only 50% patients with 80%-90% adherence achieved undetectable viral loads. Patients required better than 95% adherence to achieve highest rate of undetectable viral load ( 14). The development of resistant strains is also a significant public health concern because of the possibility of transferring the resistant strain to others. Evidence was reported documenting the sexual transmission of virus resistant to all known classes of anti-retroviral drugs including protease inhibitors (15). In a prospective study of 93 patients, self reported adherence was independently associated with undetectable seminal HIV RNA level after six months of therapy (16).
One hundred percent adherence to HAART is not easy to achieve. Studies of HIV/AIDS patients have reported low adherence rates. In a cross sectional study by Mostashari et al (17), involving 102 HIV infected females, 62% females reported taking all medications for >= 6 days a week, and were classified as adherent. In an observational cohort study called the ATHENA study (18), adherence to HAART was obtained by self-report and validated by blood assays.
Of the 224 patients, 53. l % reported taking all the medications on time, and also followed dietary requirements for the last week. The rest reported missing doses or not taking them on time and were classified as non-adherent. In a study on 46 patients with HN, Singh et al. (19) reviewing monthly prescription fill records assessed adherence to antiretroviral therapy. All patients filling >=80% of their medications were defined as being adherent. With this criteria, 63% patients were adherent. In a study that of 180 patients randomized into either MEMS, diaries and no surveillance groups, the adherence in the past four weeks was 80.6%, 92% and 93% respectively (20). In a retrospective study of pharmacy claims data regarding prescription fills to assess adherence to HIV medications, only 26% patients had more than 80% adherence. Adherence was defined as proportion of days on which drugs were taken during the first 365 days on therapy. The mean adherence was 53% (21). Thus, the adherence has been found to be less than adequate.

B] Assessment of adherence
A major problem in studying adherence is the lack of a standard measure (8).
There is no "gold standard" for measuring adherence. The four methods used most commonly to measure adherence are self reported (questionnaire/ interview/ diary), pill counts, drug assay, and electronic monitoring.
Self Reported Questionnaires: It is the most common, inexpensive and simple method of determining adherence (22). Advantages of this method include low costs, easily obtainable results and flexibility to tailor the method to the language and reading competency of the subjects (23). Patient self reports are often the only available method. However, the validity of this measure is questionable. In general, self-reports tend to overestimate adherence compared with other methods of determining adherence, like pill counts or electronic monitoring.
Recall bias is another concern (23). Among HIV-infected patients however, there tends to be a strong correlation between self-reports and virologic outcomes.
Though this method may not be as accurate as desired, there may be reason to believe that it is useful because patients reporting non-adherence are usually at least as non-adherent as indicated by interview (24).
Pill Counts: Having a physician, nurse or other health care practitioner count pills remaining in a bottle is another way to measure adherence (8). This method involves a comparison of the medicine left in the bottle and the quantity that should have been left if the medication had been taken. The advantage of this method is that they are potentially affected less than the other methods by subjective patient response (8). Adherence assessed by this method correlate better with that measured from electronic bottle caps than does self-reported adherence (25). However, this method had several limitations. Patients may forget to bring their bottles to the clinic when instructed. It is very time-consuming.
Patients may empty the bottle or may take all the remaining pills before the visit to the clinic (26).
Drug assay: Plasma and urinary blood levels provide useful objective assessment of adherence (27). The accuracy of this method depends in part on the half-life of the drug (26), which is the time required for the potency of the drug to fall to half or to be eliminated from the body. This means that it depends on how soon the drug reaches the systemic circulation so as to be detected in a drug assay. These studies are very inconvenient and expensive. Some patients may object to having their blood drawn, regarding this as unnecessary and intrusive. Also, patient-to-patient variability is a drawback (28). In addition, results may be confounded by pharmacokinetic factors, such as poor drug absorption or drug -drug interactions, which may mimic poor adherence (8).
Electronic monitoring: Bottles fitted with caps harboring electronic chips that register each time a pill bottle is opened or closed constitute the most sophisticated method currently available for measuring adherence. Two systems are available: Medication Event Monitoring System (MEMS) and the eDEM monitor (8). Data from the MEMS allows calculation of 1) the adherence rate , 2) prescribed frequency, and 3) prescribed interval. This measure also does not directly measure whether the patient took the medication; hence the accuracy of this method is suspect (26).

C] Determinants of Adherence
Given the importance of adherence with medication regimen in the success of HAART, most research in medication adherence in HIV infection has focused on predictors of adherence and factors affecting adherence. These factors can be classified as patient characteristics, clinical characteristics, treatment regimen characteristics, clinician and clinician-patient relationship and psychological and emotional characteristics.
Patient Characteristics: The literature on adherence strongly and consistently demonstrates that adherence cannot be predicted solely on the basis of gender, age, race or educational status (29). Factors that affect the initiation and adherence to anti-retroviral therapies are knowledge and beliefs about the disease and medication, social support, co-morbid conditions, substance abuse, cognitive impairment, depression and other mental illnesses (5). Thus adherence is may not related to income, social class, occupation or educational background and nor can it be accurately predicted by physicians (30). In a pilot study to test the effect of behavioral medical management of adherence, self reported adherence in the past four days improved from a mean of 80% to 98% in the group receiving behavioral based intervention of education about the therapy, positive reinforcements and encouragement, counseling and life style assessments (31). Thus knowledge about therapy and positive reinforcement enhance adherence Clinical Characteristics: After a critical literature review, Haynes (29) commented that there are few associations between disease features and adherence. The only exception being that when patients get better from any illness they are less likely to adhere to treatment regimen (32,33,34).
Medications are more likely to be taken for short term, symptomatic illnesses, where there is a more easily appreciated direct connection between medication and therapeutic effect. (35).
Treatment regimen characteristics: It has been well documented that the likelihood of adherence declines with an increase in the number of medications, frequency of dosing, severity of side effects, and complexity and anticipated duration of side effects. The more the regimen requires alterations or disruptions in daily routines and lifestyle, the less likely will be excellent adherence (36).
Unfortunately, these negative characteristics are associated with the current, complex anti-retroviral medication regimen. Combination anti-retroviral medication regimen involve large number of pills with varying dosing schedules, ( food requirements, lifestyle rearrangements and lifelong administration (5).
Furthermore, there are numerous side effects associated with the therapy. These include nausea, vomiting, anemia, granulocytopenia, pancreatitis, peripheral neuropathy, oral stomatitis, malaise, skin rash, and fever, to name a few (3).

Psychological and emotional characteristics:
Mood status is an important predictor of adherence. A level of anxiety either too high or too low may be related to non-adherence (37 Reaction of others to the patients' diagnosis constitutes a significant concern (Ross and Rosser, 1988

B. Data Collection
A standardized questionnaire was administered to patients meeting the eligibility criteria who visited one of the three sites. The patients were told that the questionnaire was about how they think and feel about their HN related medications, and about different strategies that people use to take their medications. Research assistants explained the questionnaire to the patients in a private location on each site, and were available to answer questions while the respondents were filling out the questionnaire.
Some patients did not complete the questionnaire at the clinic and were allowed to fill out the questionnaire at home and mail it to the clinic. They were told that they would each receive a $20 gift certificate after they had turned the questionnaire in. The data was collected during the year 1996-97.
The survey questionnaire administered to the patients included questions to gather data on demographics, living arrangements, education, employment, income, insurance coverage, social support, side effects, and psychological measurement It was a self-reported questionnaire. All the questionnaires were checked for completeness before the incentives were awarded.

C. Measures and Variables assessed
The questionnaire included questions regarding the following: • Demographics: age, gender, ethnicity, years of education, family income, health insurance coverage, number of people in the household, and employment status.
• Current health status and mood status.
• Social support: emotional, financial, physical support from family and friends . • Medical status: self reported disease and medication history, number of doses missed in the past one month, number of doses missed in the past three months.
• Coping: ways in which people cope with HIV and its treatment. The Ways of coping scale as modified by Dunkel-Schetter, Feinstein, Taylor, Flake (53) to suit their study on cancer patients was included.

D. Assessment of Medication Adherence
Medication adherence with anti-retroviral and protease inhibitor medications was assessed separately using data on two scales. The adherence was calculated separately for antiretroviral and protease inhibitor medications because the respondents had answered questions for each type of medication separately. The two scales used to measure adherence are: • During the last 3 months, have you ever taken less of your antiretroviral medicine than your doctor prescribed because you felt better?
• During the last 3 months, have you ever taken less of your antiretroviral medicine than your doctor prescribed because you felt worse?
• Since you began taking protease inhibitor/antiretroviral medication, have you ever purposely taken more/less of the medicine than your physician prescribed or discontinued your medication?
On the response options, a "Yes" was coded as "2" and "No" was coded as "l ".
The score for the scale was obtained by summing the response codes on each item on the scale. The range for the scales could thus be 6 to 12. Any respondents who had not responded to more than one item were dropped from the analyses. The scores were calculated separately for antiretroviral drugs and protease inhibitor drugs. The MAS score for each anti-retroviral drug was calculated. Further, the average score for all the anti-retroviral drugs was calculated and used in the analyses. Similarly, the scores for all protease inhibitor drugs were calculated and averaged.

Percentage Adherence:
Percentage adherence in the past one month was calculated using the answers to the questions "During the past month, about how many times did you miss a dose of the medication?" and "How often do you take this medication?" The responses to the question "How often do you take this medication" were used to determine the total doses prescribed for each medication. From this question, the number of doses the respondent should take for one month was calculated.
Percentage adherence in the past one month was calculated using the formula: Percentage adherence = Number of doses of 1-medication missed in the past one month x 100 Total number of doses in the past one month The percentage adherence in the past one-month was determined separately for antiretroviral drugs and protease inhibitors. Protease inhibitor medications were newly introduced at the time of the study, and hence it was thought interesting to explore adherence to these drugs separately. The percentage adherence was calculated for all the antiretroviral drugs and was averaged to get an average percentage adherence in the past one month to all antiretroviral medications the patient was on. Similarly, the percentage adherence was determined for all protease inhibitors and was averaged. Thus, the range of values for percentage adherence can be from 0 to 100.
Two definitions were followed to classify respondents as adherent or nonadherent. A respondent was classified as "adherent" if his percentage adherence was 100, i.e. he reported not missing any dose in the past one month. This stringent cut off was chosen to offset the likely overestimation of adherence by respondents. A big drawback of self-reported adherence is that the patients tend to overestimate the adherence (23,24). All the respondents having less than 100 % adherence were classified as "non-adherent". But since in the real world, it would ( be almost impossible to attain 100% adherence, an alternative cut off of 95% was also chosen.
For the second cut-off, all patients showing 2:95% adherence were classified as "adherent" and those showing <95% adherence were classified as "non-adherent".
This cut off was chosen based on a study by Paterson D et al. (14) that reported that even with adherence as high as 95%, only 80% of patients had undetectable viral loads. The coding system followed was: "1" for adherence, and "O" for nonadherence.

Coping:
Coping was assessed using the responses to a 50-item scale. The scale gives the frequency of use of each coping style by the respondents. The questions were of the type: In the last month, how often did you think, feel or do each item?
Description of these items is in appendix ill.
The final score for each factor was obtained by summing the responses on the items constituting that factor. Lazarus and Folkman (44) described this method of raw sconng.
E. Variables Used: The following variables were determined to be of interest and were included in the analyses. 3. Percentage adherence: to protease inhibitor drugs in the past one-month.

4.
Percentage adherence: to protease inhibitor drugs in the past one-month.

Medication Adherence Scale for antiretroviral drugs: Dichotomous
measure of adherence-abbreviated as MAS A.V.
6. Medication Adherence Scale for protease inhibitor drugs: Dichotomous measure of adherence-abbreviated as MAS P.I.

Independent Variables:
The IV's of primary interest were the coping styles. These were used as continuous variables for univariate and bivariate analysis, but had to be categorized for use in final logistic regression analysis. The coping styles are: Seeking social support -abbreviated as 'sss'.

Demographic Variables
Age: The variable age was categorized into three groups of< 35 years, 35-41 years and 2: 42 years. The first category was coded as 0, the 35-41 years age category was coded as 1, while the > 42 years age group was coded as 2 for the analysis.
Gender: For the purpose of analysis, males were coded as '1' and females as 'O'.

Annual income:
The respondents were dichotomized as having income of less than $15,000 (coded as' 1 '),or more than$ 15,000 (coded as 'O').

Years of Education:
Respondents having attained more than 12 years of education were coded as 'O' where as those with less than 12 years of education were coded as ' 1 '.

Insurance:
The respondents which reported having any form of insurance were coded as 'O', where as those without any insurance were coded as '1 '.

Clinical variables
Bodily pain: Respondents who reported moderate to severe bodily pain were grouped into one category and were coded as '1 ', where as those which reported none to mild pain were coded as 'O'.

Times since diagnosis with HIV:
Patients who had been diagnosed with HIV before less than 2 years were coded as '1 ', those diagnosed before 3-4 years were coded as '2', whereas those who were diagnosed before 5 years were coded as '3'.

CD4 count:
The patients with CD4 cell count between 50-200 were coded as 'O', whereas those with count between 201-500 were coded as '1 '.

Injection drug use:
Occasional and regular drug users were coded as '1 ', while those who were not drug users were coded as 'O'.
There were 13 7 respondents who were prescribed antiretroviral medications.
There were 77 respondents who were prescribed protease inhibitor medications.

Data Analysis:
The above-mentioned variables constitute the independent and dependent variables as described. The associations between the independent variables and the dependent variables were examined using bivariate and multivariate statistics.
The data was analyzed using the Statistical Analysis System (SAS) version 8.00 on the computers of Department of Applied Pharmaceutical Sciences, University of Rhode Island.
The data was screened for normality, linearity and homoscedasticity. The variable "adherence using MAS" was markedly negatively skewed for both antiretroviral medications and protease inhibitor medications. Several transformations including square root, exponential, log, were tried to make the variable normal. The variable was dichotomized due to a markedly skewed distribution. All respondents with a score of 6 were categorized as being "adherent" and those with score of 7 and above were categorized as being "non-adherent", that is, any respond who responded 'yes' to even a single question were categorized as being non-adherent.
Further, bivariate analyses were run between the primary independent variables and all other variables and also between the primary dependent variables and all other variables to check for the potential confounding variables. Bivariate statistics were used to determine the association between each dependent variable and each independent variable, excluding the independent variables of primary interest i.e. copmg styles. Similarly, association between each primary independent variable and other independent variables was determined. The associations between each dependent variable (100% A.V., 95% A.V., 100% P.I., 95% P.I., MAS A.V., MAS P.I.) and the independent variables (age, gender, race, income, years of education, insurance, bodily pain, time since diagnosis, CD4 count, injection drug use) excluding the primary independent variables, i.e. the coping styles, were explored using chi-square tests. The associations between each primary independent variable (sss, dis, fop, bea, cea) and other independent variables (gender, race, income, years of education, insurance, bodily pain, CD4 count, injection drug use) were explored using multiple T-tests. ANOVA's were run to explore the association between the coping styles (ss, dis, fop, bea, cea) and the variables "time since diagnosis" and "age".
Further, each primary independent variable (coping styles) was categorized into three categories so that preliminary logistic regression models could be run between each coping style and each dependent variable to assess the parametric form. The primary independent variables were transformed into categorical variables as they did not show a linear relationship with the dependent variables and hence could not be used as continuous variables in the final logistic models.
Each coping style was categorized into three level variables based on the frequency distribution. Each coping style was categorized as "seldom used", "used often" and "used very often". Further, these categorical independent variables were transformed into dummy variables as follows: Seeking Social Support: ssshigh-'using seeking social support very often' sssmed-'using seeking social support often reference category-'using seeking social support seldom' Distancing: dishigh-'using distancing very often' dismed-'using distancing often reference category-'using distancing seldom'

Focusing on Positive:
fophigh-'using focusing on positive very often' fopmed-'using focusing on positive often reference category-'using focusing on positive seldom'

Behavioral Escape Avoidance:
beahigh-'using behavioral escape avoidance very often' beamed-'using behavioral escape avoidance often reference category-'using behavioral escape avoidance seldom'
The variable 'time since diagnosis' was dummy coded as follows: longtime->= 5 years.
Finally, logistic regression analysis was run to assess the effect of each coping style on each dependent variable. Logistic models were run separately for each independent variable with each dependent variable. Logistic regression models to assess the effect of each primary independent variable on each dependent variable were tested following the strategy described by David Kleinbaum. The 'chunk' tests were performed to detect any interactions. The Maximum Likelihood ratio tests were used to check for the significance of the interaction terms in the model.
The likelihood ratio test is a chi-square test that makes use of maximum likelihood values. The full model with the interaction terms included and the reduced model (without interaction terms) were compared using the difference between the log likelihood statistics for the two models. Checking the effect of adding each variable to the model separately assessed confounding. Confounding assessment followed the interaction assessment. The confounding assessment was guided by considerations of validity and precision as described by Klienbaum.
Starting with the 'gold model', i.e. the model with all Independent variables included, variables were sequentially dropped to check the effect on the odds ratios and 95% confidence intervals. Only the variables whose deletion did not caused a change in the odds ratio and C.I. were dropped. Separate models were run for each primary independent variable due to high correlation between them.
Each primary independent variable was conceptually very different and separate models were run to assess the effect of each primary l.V. on each D.V.
The logistic regression models are listed below: For Anti-retroviral drugs:  reported that they had never at all or never in the past 6 months used intravenous drugs. Table 3. Adherence with Anti-retroviral (A.V.) and Protease Inhibitor (P.I.)

Medications (Dependent Variables):
For patients on A.V Medications: With a 95% cut off (patients whose adherence was above 95% were categorized as adherent, while those below 95% were categorized as non-adherent), 85.61 % (n = 113) of the patients were found to be adherent, whereas 14.39% (n = 19) were found to be non-adherent.
With a 100% cut off (patients whose adherence was 100% were categorized as adherent, while those below 100% were categorized as non-adherent), 47.73% (n = 63) of the patients were found to be adherent whereas 52.27% (n = 69) were found to be non-adherent.
Using the MAS, 45.26% (62) patients were found to be adherent, while 54.74% (75) patients were found to be non-adherent.

For patients on P.I Medications:
With a 95% cut off (patients whose adherence was above 95% were categorized as adherent, while those below 95% were categorized as non-adherent), 86.67% (n = 65) of the patients were found to be adherent whereas 13.33% (n = 10) were found to be non-adherent.
With a 100% cut off (patients whose adherence was 100% were categorized as adherent, while those below 100% were categorized as non-adherent), 49.33% (n ( = 37) of the patients were found to be adherent whereas 51.95% (n = 40) were found to be non-adherent.
Using the MAS, 48.05% (37) patients were found to be adherent, while 51.95% ( 40) patients were found to be non-adherent.    None of the independent variables was significantly associated with adherence in the chi square test. None of the variables showed a significant association with medication adherence. The variables "insurance" and "time since diagnosis" did not have enough sample size per cell and hence the chi-square was not a valid test to check for the differences in the proportions of respondents who were adherent and those who were non-adherent. The variable "gender" (p= 0.02) was found to be significantly different between the adherent and non-adherent patients. Greater proportion of males were adherent. The variables "annual family income"(fisher's p value= 0.16), "T-cell count" (pvalue = 0.03) were found to be significantly different between the adherent and non-adherent patients. Respondents with annual income more than $15,000 and those with T-cell count of less than 200 were found to be more adherent than those with income less than $15,000 and those with T-cell count greater than 200. The mean score on the variable "Seeking Social Support" was significantly different between the patients with insurance and patients with no insurance (p-value=0.0003). The mean score was also significantly different between patients living alone and patients living with someone (p-value=0.005). The patients who were insured and those who lived alone had greater mean score on "Seeking Social Support" as compared to those who were uninsured and those who did not live alone. The mean score on the variable "Distancing" was significantly different between the patients living alone and patients not living alone (p-value=0.03). The patients living alone had a greater mean score on the variable "Distancing" as compared to those who did not live alone. The mean score on the variable "Focusing on Positive" was significantly different between the patients with excellent/good health and patients with fair/poor health (p-value=0.03). The score was also significantly different between the patients living alone and those not living alone (p-value=0.05). Also, the mean score was significantly different between patients with income <15,000 and those with income 2: 15,000 (p-value=0.02). The mean score on the variable "Behavioral Escape Avoidance" was significantly different between the patients living alone and those not living alone (p-value=0.03). The score was also significantly different between the patients with none/mild pain and patients with moderate/severe pain (p-value=0.01). The patients living alone and those with moderate to severe pain reported more behavioral escape avoidance as compared to patients not living alone and those with none to mild pain. The mean score on the variable "Cognitive Escape Avoidance" was significantly different between the white patients and the non-white patients (p-value=0.01).
Whites had a significantly greater score on cognitive escape avoidance scale as compared to the non-whites. None of the continuous primary I.V's showed significant differences across the groups of "time since diagnosis" or "age". The mean score on the variable "Seeking Social Support" was significantly different between the patients with insurance and those without insurance (p-value=0.002). The scores were also different between patients with none/mild body pain and those with moderate/severe pain (p-value=0.03). The patients with some insurance and those with moderate to severe pain reported using more seeking social support as compared with those with no insurance and those with none to mild pain.  The mean score on the variable "focusing on positive" was significantly different between the patients with insurance and those without insurance (p-value=0.04).
The patients with some insurance reported using focusing on positive as compared to those with no insurance. The mean score on the variable "behavioral escape avoidance" was significantly different between the patients living alone and those not living alone (p-value=0.008).  There was significant difference in the means of the variable "Cognitive Escape Avoidance" between the groups of variable "time since diagnosis" (p-value=0.02) for people prescribed A.V. medications.
The results for patients on P.I. medications were non significant. Table 24 to Table 47 summarize the final logistic regression models run between each of the independent variables (coping styles) with each of the dependent variables (100% A.V. , 95 % A.V., MAS A.V., 100% P.I. , 95 % P.I., MAS P.I.) controlling for the potential confounding variables. In the bivariate tests, some demographic and clinical variables were found to be significantly associated with either some independent variable or some dependent variable, but none was found to be significantly associated with both the independent variable and dependent variable, and hence did not qualify to be a confounder. Introducing the variables in ascending order as well as descending order assessed the effect of each independent variable on the dependent variables. In the final model, 'behavioral escape avoidance' was significantly associated with medication adherence as ( assessed by medication adherence scale for patients on antiretroviral medications.
The people who used behavioral escape avoidance very often are 60% less likely to be adherent as compared to those who use this coping style seldom. Similarly, people who used behavioral escape avoidance often are 70% less likely to be adherent as compared to those who use this coping style seldom. BEA was also significantly associated with adherence to antiretroviral medications using a 100% cut off definition. The respondents using BEA very often are 70% less likely and those using BEA often are 90% less likely to be adherent as compared to their counterparts who use BEA seldom. All other logistic regression models were nonsignificant. Also, the final models revealed some interesting associations.
Education and living arrangement were also found to be significantly associated with adherent to protease inhibitors as defined by 95% cut off. The respondents with less than 12 years of education and those living alone were 90% less likely to be adherent as compared with those with more than 12 years of education and not living alone.

Discussion
The purpose of this research was to assess coping with HIV as a predictor of medication adherence. Two measures of medication adherence, MAS and percentage adherence were used to assess the adherence to anti-retroviral and with the styles of coping used. Patients with some insurance and those with moderate to sever pain reported using ' seeking social support' with significantly ( greater frequency as compared with those having no insurance and with mild pain. This could be because patients with insurance could have more meaningful social interactions and those in pain and suffering have a tendency to seek external support. Patients with some insurance were also found to use 'focusing on positive' to a significantly greater extent than those with no insurance. In the final logistic regression models, the variable gender was significantly associated with percentage adherence using the 100% cut off (OR, 0.2; 95% Cls, 0.08-0.8). Males were 80% less likely to be adherent to protease inhibitors as compared with females. In the previous researches, gender has been inconsistently associated with medication adherence.
For patients prescribed antiretroviral medications, those with some insurance coverage and those living alone used 'seeking social support' with significantly greater frequency as compared with those with no insurance and not living alone.
Also, patients living alone reported significantly more use of 'distancing' and 'behavioral escape avoidance' as compared with patients not living alone. This could be because people who use passive coping strategies tend to be more depressed and withdrawn. Patients having some insurance, those with good health, living alone and having income greater than $15,000 reported using 'focusing on positive' significantly more than those without these attributes. With the knowledge of these associations, people with the attributes associated with non-adherent can be identified and targeted for interventions.
Copingthe results of this study indicate the coping style 'behavioral escape avoidance', was significantly associated with adherence to antiretroviral   19), utilized the 'Ways of coping questionnaire' to assess coping styles and their association with medication refill adherence at 6 months. The authors reported that adherence was significantly associated with better adaptive coping and non adherence with poor coping. Although in this research no association was found between adaptive coping styles like seeking social support or focusing on positive, non-adherence was significantly associated with behavioral escape avoidance. Although the previous researchers have used different coping scales for assessing coping as predictor of medication adherence, the coping behaviors involving escape avoidance strategies have consistently been associated with nonadherence.
This indicates that patients who exhibit behavioral escape avoidance tend to be less adherent and is a group, which should be focused for behavioral intervention.
The conceptualization of adherence as the extent to which the patient follows medical instructions is an oversimplification of a multidimensional complex construct. It is now agreed upon that adherence is the extent to which a patient's behavior-taking medications, following a diet, and/or executing lifestyle changes, corresponds with agreed recommendations from health care provider (58).
Adherence involves a motivated behavioral change. This broader view of adherence highlights the importance of psychological constructs and predictors of adherence. This study is an addendum to and compelling support to the previous researches that have investigated psychological predictors to adherence. The results obtained are within some limitations, which despite diligence in statistical I methodology, could have contributed significantly to the results. These limitations are discussed in the following paragraph.

Limitations:
The limitations, in part were due to the nature of the data and also the lack of reliable and foolproof measures for the variables such as medication adherence.
The sample size was only 145. Many variable distributions were skewed.
Statistical techniques were used to rectify this shortcoming, but the skewness may indicate selection bias.
Self-reported data: A major shortcoming was the fact that the data was selfreported. The validity of self-reports is questionable as it lends itself to patient's recollection of the past. Recall bias plagues self-reports. There is a degree of subjectivity that seeps in as the responses depend on the situational mood of the respondent, education, social desirability concerns. These might influence the patient's ability and willingness to give accurate responses.

Measurement:
The lack of any standard for the measurement and quantification of medication adherence is another limitation. The varied assessment techniques and definitions used for adherence and the varied results in adherence studies stand evidence to this.
This study was an effort to work within these limitations and tries to assess coping as a predictor of medication adherence. More research with more objective measures needs to be carried out on a larger sample to make the results generalizable to the population.