PREDICTORS OF PATIENT REPORTED MEDICATION ADHERENCE TO ANTIRETROVIRAL THERAPY

The acquired immunodeficiency syndrome (AIDS) caused by human immuno deficiency virus (HIV) is still a significant public health problem. In the United States approximately 1,000,000 people live with HIV/AIDS infection.The recommended treatment is HAART (highly active antiretroviral therapy) . The HAART treatment is very complex, because of a large number of prescribed drugs in a regimen, frequent dosing and also a number of side effects. In HIV management precise adherence• 26 to the prescribed medication regimen is the key for the maximal viral suppression and improved health status, and the only way to turn this deadly disease into a manageable chronic disease. The importance of many factors associated with medication adherence, including patient characteristics, disease characteristics, medication regimen characteristics, and the patient-provider relationship has been documented. One of the aspects of the complexity of the prescribed regimen is the number of medications in the regimen. Results: The number of medications did not show association with medication adherence. Patients on a more complex antiretroviral medication regimen did not miss higher percent of prescribed medication, and we may not discriminate between adherent and non-adherent patients only based on a number of prescribed antiretroviral medication. Also the variables number of side effects and how long ago HIV positive were negatively associated with medication adherence.

Estimated prevalence tells that currently around the world about 36 million people live with HIV I AIDS infection. The WHO also has reported that AIDS caused 22 million deaths . 2 In the United States approximately 1,000,000 people live with HIV/AIDS infection. 3 The incidence of HIV infection is about 40,000 cases each year. 3 In this country more than 420,000 people have already died from AIDS. 4

Adherence-key factor in HIV/AIDS treatment
Current therapeutic options for HNI AIDS patients include more than 15 antiretroviral drugs, commonly grouped in the three classes as: nucleoside reverse transciptase inhibitors (NRTD, nonnucleoside reverse transcriptase inhibitors (NNRTD and protease inhibitors (PI). 5 ' 6 ' 7 ' 8 From 1996 the highly active antiretroviral therapy (HAAR T) has been used as a standard of care for symptomatic as well asymptomatic HN positive patients. 5 • 9 • 10 The HAART treatment is very complex, because of a large number of prescribed drugs in a regimen and frequent dosing. 9 • 10 Also a number of ( side effects, drug interactions, toxicities and also very specific food restrictions make this treatment not only complex but also very demanding for the patients and consequently difficult for lifetime precise adherence. 5 Adherence to a prescribed medication regrmen reqmres taking medication as prescribed consistently. 22 Adherence or compliance was defined by Fisher as 'a concept used to measure' 23 positive patients' behavior 'in meeting their therapeutic goals. ' 23 In HIV management precise adherence 7 ' 9 ' 12 ' 24 to the prescribed antiretroviral medication regimen is the key for the maximal viral suppression and improved health status, longer and better quality of life, and the only way to turn this deadly disease into a manageable chronic disease. 7 ' 12 ' 22 ' 24 ' 25  Poor adherence to the prescribed antiretroviral medication regimen of 60% and 42% were reported by Eldred JL,et al. 31 and Muma et al. 32 Medication non-adherence also causes higher costs of HIV I AIDS treatment because of increased number of complications and development of opportunistic infections, hospitalizations, and finally increased mortality of the HIV infected patients. 13 Consequences of medication non-adherence are the most obvious in HIV population.  16, 19,20,2 1,22,24,29,30,34  The disadvantage of these methods is that they do not give information about Jong-term medication use behavior 28 . Although direct methods probably have a higher sensitivity and specificity than indirect methods, they are more expensive and more inconvenient for the patients. 49 Indirect methods include pill counts, physician assessment, and self-reported adherence to a prescribed medication regimen, prescription refill records, and electronic monitoring. 11 ,12, 15 ,38,48,49 A. The pill count method is based on the count of returned pills. This method was very frequently in use in older research literature. 50 A self-report is more convenient for the patients, and less expensive than other measures. 24 With cooperative patients, who are willing to disclose their behavior, self-report has the potential to be an accurate measure of patient adherence. 24 D. The prescription refill records method is based on the use of pharmacy records, and gives unique opportunity to explore adherence usually in a larger population, and usually over longer periods of time. The disadvantage of this method is that we know that prescription was filled, but we do not know did actually patient use the medication as prescribed.
E. Electronic monitors are method based on use of the computerized drug containers.
Electronic monitors provide data with dates as well as time intervals of each opening of the bottle. Electronic monitors use may also overestimate the adherence rate because we know that bottle was opened but we do not know whether the prescribed medication was used. 48 They are also expensive and impractical for large populations as well as multi-drug regimens . For some forms of medication electronic monitors cannot be used. Also because of awareness that their behavior is monitored patients may be more adherent. Because there is no gold standard perhaps the most accurate estimate of medication adherence may be obtained by using several methods in the same study, which also gives an opportunity for reliability and validity comparison between the measures. 11

STATEMENT OF PURPOSE
A review of the literature shows the importance of precise and strict adherence to complex antiretroviral regimens in the HN population. The importance of many factors associated with adherence, including patient characteristics, disease characteristics, medication regimen characteristics, and the patient-provider relationship characteristics have been documented. The primary objective of this study is to analyze and estimate the point prevalence of HN positive patients' adherence to complex antiretroviral medication regimens and to examine the association of the following factors with adherence: • Medication regimen complexity as a predictor of medication adherence while controlling for patient characteristics, disease characteristics, and medication regimen characteristics.
The secondary objective is to explore other potential predictors of medication adherence. The hypothesis is that a number of medications may influence patients' medication adherence. Patients on a more complex multi-drug regimen very likely will be less adherent to the prescribed medication regimen than patients on a less complex drug regimen. The patients on a more complex antiretroviral regimen, including a larger number of prescribed medications very likely will miss more doses of prescribed medication and demonstrate the poorer adherence to the prescribed antiretroviral medication regimen.

METHODOLOGY Study Design
This is a descriptive, cross-sectional study of self-reported medication adherence to prescribed HIV related medication regimen.

Study Sample
The study sample consists of 145 participants, all diagnosed as HIV positive patients.
A patients' eligibility to participate in this study was 1) over 18 years of age, 2) ability to read and write English, and 3) currently using an antiretroviral medication, or medication for HIV related complications. The study sample was not random sample.
All study participants were patients attending clinics. For participation, each person received $20.

Measures:
The questionnaire collected information about the following: A. provider, and questions about medication use.

Method
The total number of prescribed HN related medications was calculated for each participant; a larger number of prescribed medications corresponds to a more complex regimen. The participants of this study were considered absolutely adherent (100%) to prescribed antiretroviral therapy if they did not report missing any doses of prescribed regimen in the previous month. Dependent variable medication adherence was calculated using the following formula: {(total# of doses prescribed I total# of medication) -(total #of doses missed) I total# medication)} I (total# of doses prescribed I total# of medication)* 100. Consequently, a higher percent of doses missed in the past month correspond to a higher level of medication non-adherence.
With multiple ANOV As the participants who reported 100% adherence were separated as a sample and were used for comparison with the non-adherent sample.

Statistical Analvsis:
Data was analyzed using Multiple Regression Analysis with SAS program (Statistical Analysis Software), Version 8.0. The computer facilities and the library at the University of Rhode Island was used for the data analysis as well as for the all necessary research work. Data were checked for the accuracy of the data entry and for the basic assumptions of linearity, normality, and homoscedasticity. SAS procedures proc mean and proc plot were used to check for outliers and linearity assumption of all. Collinearity and singularity was checked also. The dependent variable, selfreported medication adherence (percent of doses taken), in the past month was used in analysis because of the least skewness and kurtosis compared to self-reported medication adherence in the past week or self-reported adherence in the past three months. Also one month was considered as better time period for the evaluation of medication taking behavior compared to one-week period. Recall bias was less likely in a period of a month compared to three months period. Because of severe nonnormal distribution logarithmic transformation was performed on the dependent variable medication adherence. Variables were considered significant predictors for the p-value below 0.05. If independent variables were categorical then independent variables were dummy coded.
The Questionnaire was designed to ask about the number of doses missed in the previous month for each medication in patients' regimen; from the first drug in the regimen to the sixth drug in a regimen. But the questionnaire was not designed to ask about the number of doses missed for the regimen of 7 or more drugs. Seven study ( participants, (5%) were on the regimen of 7 or 8 drugs. Because most study participants, (71 %) were on regimen of four or less than four drugs they did not respond on the questions about the number of doses missed in the previous month for the fifth or sixth drug in a regimen. Because information on the number of doses missed in the previous month was missing for approximately 80% of variables for the fifth or sixth drug in the regimen was not used in the analysis. This is illustrated in Table 1. In accordance with this all computations were performed for the regimen of four or less than four drugs. The variables to be evaluated included:

Dependent variable:
Two dependent variables were created. The first dependent variable medication adherence was calculated for all study participants using the following formula:

A. Medication adherence, (percent of doses taken in the previous month):
0-100% (continuous), was defined and calculated as: Medication adherence was calculated for the regimen of four or less than four drugs in the regimen.
The second dependent variable was calculated only for the study participants who reported that they had missed their medication. Consequently study participants who reported that they never missed their medication were not included. Medication adherence was defined and calculated as: {(total #of doses prescribed I total# of medication) -(total #of doses missed) I total # medication)} I (total # of doses prescribed I total # of medication) * 100.
Medication adherence was calculated for the regimen of four or less than four drugs in the regimen.

A: Patient Characteristics
Detailed demographics of this study population are given in  In general, a majority of the sample population, (73%) reported excellent, very good, or good health status. Also 63% reported that they had never been hospitalized. Only 13% of study participants reported severe or very severe bodily pain. Almost half, 46%, confirm that they never feel so weak that they need to spend a day in a bed. The majority, 63%, reported long duration of disease; they had been diagnosed 5 years or longer ago. In this study population basically 43% had developed AIDS, because 13% reported a CD4 cell count less than 50, and 30% reported a CD4 cell count between 50 and 200. Table 3 shows the frequency of indicators of health status.

C: Medication regimen characteristics
The mean number of prescribed medications was 3.7 drugs. Prescribed HIV related medications regimen range from monotherapy to 8 drugs therapy. 6 study participants, or 4%, were on monotherapy, 30 study participants, or 21 % were on 2-drug therapy, and 41, or 28%, were on 3-drug therapy. 24 study participants or 17% were on a more complex regimen of 4 drugs. 44 study participants or 30% were on a regimen of 5 and more drugs. Antiviral medications were prescribed for almost all study participants (99%); 50% had protease inhibitors in their regimen, and 56% additionally used antinfective medication for HIV related complications. The most commonly prescribed antiviral medications in this study population were: lamivudine or epivir (3TC), prescribed for 121 patient or 83%; then stavudine or zerit (D4T), prescribed for 69 patients or 49%; and zidovudine or retrovir (AZT), prescribed for 65 participants or I 45%. Among antiinfectives, bactrirn or septra (Trimethoprirn) was the most commonly prescribed drug, for 67 patients or 46%. Among protease inhibitors, the most commonly prescribed drug was Indinavir (Crixivan), prescribed for 58 participants or 40%. The most common frequency of dosing was two or three times a day for each drug in a regimen. The prescribed regimens did not cause any side effects in 26% of study population. The rest confirm that they experienced from one to a number of side effects. Table 4 shows the frequency of indicators of medication regimen.

D: Patient-Provider relationship characteristics
Physicians were described as the most helpful health care providers. If study participants have questions about their medications, 91 % ask their physicians, 31 % ask a pharmacist, and 30% a nurse. Patient provider relationship characteristics are shown in Table 5.

E: Medication adherence
Almost half of the study sample reported that they had never missed a dose of prescribed medication. There appears to be a significant difference in reported strict, 46% reported forgetfulness as a reason for missing a dose of prescribed medication.
Also a high number reported intentional non-adherence 29 ; 25% reported sometimes being careless about taking prescribed antiretroviral medication regimen. 19% admit that sometimes they stop the use of prescribed medication. 14% reported drug holidays for 3 or more days.
The mean number of prescribed doses missed in the previous month was 5 doses. Selfreported mean medication adherence in the previous month was high and was calculated to be 97%, with a minimum of 58% and maximum of 100%. The mean medication adherence in the past week was 95%, and the mean medication adherence in the past three months was 98%.

F: Multiple Regression Analvsis and Stepwise multiple Regression Analysis
Multiple regression analyses were performed for the following models: predictors' group patient characteristics, predictors' group disease characteristics, predictors' group medication regimen characteristics and Stepwise multiple regression analysis for the final model.               ANNUAL INCOME was coded as: 1-<24,000 0->25,000.
(    It is difficult to define a real degree of non-adherence in the study sample because we do not have precise and accurate methods to measure patients' adherence to prescribed medication regimen. Multiple regression analyses as well as Stepwise multiple regression analysis was carried out on the total sample as well as on the smaller subs"cunple of the study participants who reported missing medication. This method did not help to define more potential predictors of medication adherence.
When adherent study participants were compared with multiple ANOVA's to nonadherent study participants, they were similar. No significant differences were found between the two groups, although the non-adherent sample had slightly more women, more hospitalizations, longer duration of disease, and worse CD4 cell count.
Adherence was high in this study population. Perhaps, adherence rates were overestimated, or this population was highly motivated, and, with high social support, they were adherent to the prescribed medication regimen.

A: Patient Characteristics
Potential predictors of medication adherence grouped as patient characteristics include: age, education, race, employement, annual income, health status, and number of people for physical assistance or place to stay. Model for the total sample was not significant. Model which include sub-sample of non-adherent study participants, (0 doses missed were deleted), was significant. Only variables, race (Non-White) and education showed positive association with medication adherence in this model.

B: Disease characteristics and C: Medication regimen characteristics
Predictors grouped as disease characteristics, and medication regimen characteristics did not show statistically significant association with variable medication adherence.  21 Christensen et al.,40 and Sung et al. 42 For the better-fit variable total number of medication was transformed by squaring it, but the final result was almost the same as before transformation. Also results for the sub-sample was not significantly different from the results for the final model, including total sample.

D: Final model; Stepwise Multiple Regression Analysis
One of the aspects of the complexity of a prescribed regimen is the number of medications in a regimen, as well as the frequency of dosing, number of side effects, food requirements, taste, and cost. All this is true for the medication regimen prescribed for HIV I AIDS patients. This study findings suggest that more complex regimen, with larger nun1ber of medication does not necessarily lead to medication non-adherence. This is illustrated in Figures 1 and 2. Previous research also reported that it is most likely that it is not only the number of medication, but probably other aspects of the regimen complexity, such as the frequency of medication use, number of side effects, or perhaps nun1ber of psychological variables that contributes significantly to the patients medication non-adherence. 17 In the final model two variables showed negative statistically significant association with medication adherence: number of side effects and number of children.
Previous research reported stable family situation as important and positive factor for patients' medication adherence, but association between the number of children and medication adherence was not reported. This study found negative association between number of children and medication adherence. Perhaps people with children had more duties including childcare and skip prescribed medication more frequently than people without children did.
Number of side effects showed negative statistically significant association with medication adherence. The higher number of side effects in the regimen the less likely patients will adhere to such a regimen. This finding is consistent with the previous research. Catz et al. 57  The study has the following benefits: The advantages of a self-report, such as low cost, and fast and easy distribution, probably make it a good choice to study and collect data about larger as well as smaller populations. The relatively large study population, 145 patients, and the precise and nonjudgmental questioning gives researchers a unique opportunity to explore adherence issues in depth and to assess important and valuable information about adherence behavior as well as a nun1ber of other important factors associated with adherence in this population. Although the self-report is an indirect adherence measure, it has been reported by some authors as a measure with higher sensitivity and specificity over other measures. 25 ' 35 A better understanding of adherence to a prescribed antiretroviral medication regimen in this HN population may help us to develop intervention strategies, which would be valuable for this and other HIV patients, and a number of other chronic illnesses that require lifelong treatment. The results of this research will provide information about the patients' needs for health care providers as well as for pharmaceutical companies and manufacturers.
The study has the following limitations: Because these are self-reported data estimate of the number of doses missed in the past month because of forgetfulness may not be exactly the number of doses missed.
Some patients simply do not want to report non-adherence; recall bias and overestimation of adherence rate are real possibility in this type of data collection. 22 . 2 4 Although numerous methods for measuring adherence exist; no instrument satisfied all the necessary criteria to be accepted as gold standard to measure adherence rate. Data was collected during '96-'97, when combined therapies were being just introduced as a standard of care and at that time knowledge and infomrntion about the treatment options of HIV positive individuals was limited. However, the recommended standard of care is (HAART), a multiple drug therapy including potent protease inhibitors drugs, other than drug combinations used previously. Study sample size was a limitation for more in depth research about adherence rates in some subpopulation groups, such as incarcerated HIV positive participants, or HIV positive women. Study participants were mostly unemployed, without health insurance, and with long duration of disease. Race distributions were not typical for HIV population, and middle age White males were over sampled. Adherence was probably overestimated or ( study participants were maybe more adherent to prescribed medication regimen than other, average HIV patients because they accepted to participate, and they already were patients in clinic. Because they were patients on the clinics maybe they received counseling about importance of medication adherence in HIV I AIDS management.
Study design was limitation for an estimate of the prevalence of non-adherence over longer period of time.