Measuring Adherence with Antidepressant Medication: Comparison of HEDIS and PDC Methodologies

Depression is a significant problem for the managed care system. Antidepressant medication helps ameliorate the symptoms of depression yet adherence to medication is known to be poor. The current approach to adherence measurement (i.e HEDIS) is limited or lacking. Other methods are used (e.g Proportion of Days Covered –PDC) in other chronic diseases to measure adherence. Medication adherence is a growing concern for clinicians and other health care stakeholders (e.g. payer) because of the increasing evidence that non-adherence is prevalent and places patients at an increased risk for adverse health outcomes and higher cost of care. We conducted a retrospective cohort study of patients enrolled in a Medicaid plan. For inclusion in the study population patients had to meet HEDIS inclusion criteria and PDC criteria respectively. Patients included in the HEDIS study’s cohort were adults at least 18 years of age with a new diagnosis of depression confirmed by outpatient medication use and an ICD-9 diagnostic code. The upper limit age was set at 75 years old in order to maintain the confidential information about the patients. Patients had to meet certain enrollment eligibility criteria as well. For the PDC study population patients met the same age requirement as for the HEDIS measurement inclusion criteria. Patients included in the study for the PDC cohort were not required to have a new diagnosis of depression certified by a diagnostic code; they only had to be antidepressant medication users during the study period. We evaluated antidepressant medication adherence by applying the HEDIS measures and PDC measures. The measure of effect was the odds ratio in separate models. We also applied HEDIS criteria to the PDS cohort to provide a head-to-head comparison of the rates of adherence. Adherence was assessed with seven medication classes as recommended by HEDIS: Miscellaneous Antidepressants. Monoamine Oxidase Inhibitors (MAOIs). Phenylpiperazine Antidepressants. Selective Serotonin-Norepinephrine Reuptake Inhibitors Antidepressants (SSNRIs). Selective Serotonin Reuptake Inhibitors Antidepressants (SSRIs). Tetracyclic Antidepressants (TeCAs). and Tricyclic Antidepressants (TCAs). Differences in baseline characteristics and the odds of adherence were assessed between the groups for each methodology separately as well as patient demographic and health related variables. We constructed multivariate logistic regression models to measure the odds of adherence with antidepressant medication for each methodology while controlling for potential confounders and assessing for interaction terms. The level of significance and the corresponding 95% confidence intervals of the odds ratios were presented as well. A total of 626 eligible antidepressant users were identified according to the HEDIS criteria, and 22,351 eligible antidepressant users were identified according to PDC criteria and were evaluated for adherence with antidepressant medication. In both study samples patients 50 years and older were significantly more likely to be adherent with antidepressant medication than the younger group (<35 years ) patients. In both groups patients that had respiratory disease had an increased odds of adherence with antidepressant medication relative to patients that were not classified having a respiratory disease. Patients that had other mental health diagnoses in addition to depression had a statistically significant increased odds of adherence with antidepressant medication relative to patients that did not have such diagnoses. The beta coefficient representing the relationship between the antidepressant medication adherence and the therapy regimen was positive and statistically significant for both samples. Patients that were using more than one drug were significantly more likely to be adherent to antidepressant medication regimen than patients that were using only one type of antidepressant drugs. Our results implicate older age and comorbid diseases such as respiratory and other mental health diseases and polymedication as risk factors associated with better adherence with antidepressant medication therapy among Medicaid insured patients. Interventions that strive to improve adherence with antidepressant medication therapy should continue to be implemented and evaluated.

evidence that non-adherence is prevalent and places patients at an increased risk for adverse health outcomes and higher cost of care.
We conducted a retrospective cohort study of patients enrolled in a Medicaid plan. For inclusion in the study population patients had to meet HEDIS inclusion criteria and PDC criteria respectively. Patients included in the HEDIS study's cohort were adults at least 18 years of age with a new diagnosis of depression confirmed by outpatient medication use and an ICD-9 diagnostic code. The upper limit age was set at 75 years old in order to maintain the confidential information about the patients. Patients had to meet certain enrollment eligibility criteria as well. For the PDC study population patients met the same age requirement as for the HEDIS measurement inclusion criteria. Patients included in the study for the PDC cohort were not required to have a new diagnosis of depression certified by a diagnostic code; they only had to be antidepressant medication users during the study period. We evaluated antidepressant medication adherence by applying the HEDIS measures and PDC measures. The measure of effect was the odds ratio in separate models. We also applied HEDIS criteria to the PDS cohort to provide a head-to-head comparison of the rates of adherence. Adherence was assessed with seven medication classes as recommended by HEDIS: Miscellaneous Antidepressants. Monoamine Oxidase Inhibitors (MAOIs).

Phenylpiperazine
Antidepressants. Selective Serotonin-Norepinephrine Reuptake Inhibitors Antidepressants (SSNRIs). Selective Serotonin Reuptake Inhibitors Antidepressants (SSRIs). Tetracyclic Antidepressants (TeCAs). and Tricyclic Antidepressants (TCAs). Differences in baseline characteristics and the odds of adherence were assessed between the groups for each methodology separately as well as patient demographic and health related variables. We constructed multivariate logistic regression models to measure the odds of adherence with antidepressant medication for each methodology while controlling for potential confounders and assessing for interaction terms. The level of significance and the corresponding 95% confidence intervals of the odds ratios were presented as well.
A total of 626 eligible antidepressant users were identified according to the HEDIS criteria, and 22,351 eligible antidepressant users were identified according to PDC criteria and were evaluated for adherence with antidepressant medication. In both study samples patients 50 years and older were significantly more likely to be adherent with antidepressant medication than the younger group (<35 years ) patients. In both groups patients that had respiratory disease had an increased odds of adherence with antidepressant medication relative to patients that were not classified having a respiratory disease. Patients that had other mental health diagnoses in addition to depression had a statistically significant increased odds of adherence with antidepressant medication relative to patients that did not have such diagnoses. The beta coefficient representing the relationship between the antidepressant medication adherence and the therapy regimen was positive and statistically significant for both samples. Patients that were using more than one drug were significantly more likely to be adherent to antidepressant medication regimen than patients that were using only one type of antidepressant drugs.
Our results implicate older age and comorbid diseases such as respiratory and other mental health vii Likelihood of antidepressant medication adherence according to selected characteristics ………31 Table 6. Univariate logistic regression reduced model for the PDC methodology. Medication adherence is a growing concern for clinicians, and other health system stakeholders (e.g. payer) because of the increasing evidence that non-adherence is prevalent and associated with adverse outcomes and higher cost of care. 1 Adherence to a medication regimen is generally defined as the extent to which patients take medication as prescribed by their health care providers. 2  Depression is one of the most common disorders and a leading cause for disability worldwide.

LIST OF TABLES
Almost 50% of the U.S population has experienced at least one psychiatric disorder in their lifetime. 5 The lifetime prevalence of major depressive disorder is reported to be as high as 17% and the 12-month prevalence is 5%-9%. 6,7 The World Health Organization (WHO) considers major depressive disorder one of the most debilitating diseases to society. Its negative outcomes include suicide, substantial impairment, lower quality of life, increased health care utilization and cost, and adverse impact on employment productivity. 8 It is anticipated that major depressive disorder (MDD) will be the second leading cause of disability worldwide by 2020 9 with a lifetime risk of 7% -12% for men and 20%-25% for women. 9 Depression is a serious public problem, particularly among people with low income. Low socioeconomic status is associated with a higher prevalence of depression. Moreover, the duration of new episodes of depression are longer in people with a low socio-economic level. 10 According to data from the National Health and Nutrition Examination Survey (2009-2012) persons living below poverty level are nearly 2½ times more likely to have depression than those at or below poverty level. 11 The incidence of depression is rising and the costs are escalating. People that have depression cannot enjoy a fulfilled life because they experience sadness, a sense of isolation and feeling like they are a burden. Depressed people are 30 times more likely to commit suicide and each year in U.S. approximately 41.000 individuals complete suicide. 12 Depression impacts the academic development of a person, the marriage perspective and an average loss of $10,400/year in income by the time the person reaches 50 years of age. 13 Depressed people are 7 times more likely to be unemployed. 8 According to Smith JP et al (2010) a person that suffers from depression has a loss of 20% in potential income, and a lifetime loss for each family who has a depressed family member of $300,000. 14 From the employer point of view depression is a big concern also. The employer is affected by the missed days of work (absenteeism) and the reduced productivity (presenteeism) of individuals with MDD.
According to data from the National Health and Nutrition Examination Survey (2009-2012) almost 43% of individuals with severe depressive symptoms reported serious difficulties in work, home and social activities. The rate of difficulty in work, home or social related activities, increased with the severity of depression. 11 The consequences of untreated depression or inadequately treated depression are significant, therefore adherence to antidepressant medication is crucial. Response to drug therapy occurs predominantly in patients who are strictly adherent to antidepressant medication regimens. Most patients experience good outcomes with appropriate antidepressants taken for the appropriate period of time. Patients must be closely monitored during the acute phase (first three months) and the initial continuation phase (the first six to nine months) of treatment for necessary drug dosage adjustments. Measuring adherence is important to make sure that patients comply with provider recommendations for a good outcome and a reduced economic burden on health care system. The antidepressant therapy must be coupled with the appropriate form of psychological therapy.
Antidepressant medications are the standard approach for treating depression. According to the Sequenced Treatment Alternatives to Relieve Depression (STAR*D) trial, depressive disorder is challenging to treat. 14 STAR*D is the largest study of its kind. The results demonstrated that after exposure to four different levels of medication intervention options, approximately one-third of the patients in this study failed to achieve remission. 14 Also, approximately 50% of the patients in this trial prematurely discontinued antidepressant therapy for a number of different reasons, including patient-related (e.g. side effects, misperception about medication), and clinician-related factors (e.g. poor instruction by the clinician about the medication, lack of follow-up care). Depression itself is a condition that causes patients to have difficulty following the medication regimen, yet they can have the greatest potential to benefit from treatment adherence. 15 Currently, there are more than 20 antidepressant medications available worldwide, and they are used singularly or in combination with other antidepressants to treat the disabling affects of depression for many patients.
According to IMS Health National Prescription Audit PLUS, in 2010 antidepressants were the second most commonly prescribed medication, second to drugs to lower cholesterol. Approximately 254 million antidepressant prescriptions were written in 2010, resulting in nearly $ 10 billion in costs. 16 Eleven percent of Americans aged 12 years and over take antidepressant medication and more than 60% of patients have taken therapy for 2 years or longer, with 14% having taken the medication for 10 years or more. 11 For most patients, the course of major depressive disorder (MDD) is recurrent or chronic. The goal of antidepressant treatment is the full remission of symptoms and prevention of relapse. Remission of symptoms is related to improved functioning and prognosis. 17 Antidepressants are recommended to be continued for at least 4 months beyond the initial symptom resolution. However, reported levels of nonadherence have been consistently high and this remains a serious concern. 18 Thus, a majority of antidepressant users are chronic users of these medications, and as in other chronic diseases such as diabetes and hypertension, the health system must direct resources towards promoting patient adherence with prescribed chronic therapies.

HEDIS methodology inclusion and exclusion criteria.
Patients were at least 18 years of age as of July 1, 2013 and continuously enrolled for at least 180 days of a 231 day period between the period of July 1, 2013 to September 30, 2014. Patients older than 75 years of age were excluded from the study in order to maintain the confidentiality of subjects. All patients were confirmed to have a diagnostic code for depression. International classification of disease ninth edition (ICD-9) codes from the 2014 Healthcare Effectiveness Data and Information Set (HEDIS) 19 were used to identify the presence of any code indicative of depression or a depression related complication (Appendix D). Also patients had to have a diagnosis of depression in an inpatient, outpatient, emergency department, intensive outpatient or partial hospitalization setting during the 60 days prior of the starting of medication through 60 days after the Index Prescription Start Date (IPSD). All patients were confirmed to have used a medication for treatment of depression or a depression related disease, defined as the presence of a claim for any oral or injectable antidepressant agent during each patient's continuous enrollment period. Medication use was evaluated 105 days prior to the IPSD with seven medication classes (Appendix E) in order to test the negative medication history (NMH). Patients that had a prescription during the NMH period were excluded from the study, thus providing a sample of newly treated patients.

PDC methodology inclusion and exclusion criteria.
Patients were at least 18 years of age as of July 1, 2013 and continuously enrolled for 365 days during the study period. Patients older than 75 years of age were excluded from the study in order to maintain the confidentiality of subjects. No ICD-9 codes for a diagnostic of depression were required for the PDC methodology inclusion criteria. Patients that used any of the antidepressant medications from the seven classes (Appendix E) were included in the study. Patients were also required to have at least 91 days between the index medication and the end of the study period.

Defining Adherence
Eligibility for the HEDIS methodology required that all patients have at least one antidepressant claim during the index period associated with an ICD-9 diagnostic of depression 60 days prior or after the first prescription. This will ensure that the patients has a new diagnosis of depression according to HEDIS requirements. Patient must not have an antidepressant claim during a 105 days period prior to the index date. Patient with a claim during the negative medication history were excluded from the study population.
The continuation phase of the treatment lasts 231 days, during which a patient needs to fill a sufficient number of antidepressant prescriptions to provide medication for at least 180 days. Medication gap due to washout or refill can total a maximum of 51 days during the 231-day period. Adherence was determined as the proportion between the days supply of dispensing divided by the days of follow up during the chronic phase (231 days) . Patients were classified as adherent if they remained on antidepressant medication for at least 180 days (6 months) of the 231 days in the period.
For the PDC methodology, all patients were required to have at least one antidepressant claim during the index period. The index date should occur at least 91 days prior to the end of the study period.
Adherence was also evaluated as a dichotomous variable and calculated as the proportion between the days supply of dispensing during the study period divided by the number of days between the fill days and the end of the study period. According to the Pharmacy Quality Alliance (PQA) patients will be considered adherent to their medication if the PDC is equal to or exceeds 80% and non-adherent if the PDC is less than 80%. Adherence with antidepressant medication was evaluated separately for the two methodologies as the inclusion criteria for the measures were different, yielding different samples.

Independent Variables
Age: We created three age groups, approximately the youngest, middle aged, and older patients in the population. Most older adults 65 years old or older receive their prescription drugs benefit through Medicare Part D plans, and were not included in this study.
Gender: Analyzed as a dichotomous variable.

Statistical Analysis
The analysis of the data was performed using Statistical Analysis Software Version 9.4 (SAS® A third sample was created by applying the inclusion and exclusion criteria for PDC methodology to patients that met the HEDIS inclusion and exclusion criteria. The sample was analyzed to determine if there is a correlation between measuring adherence with antidepressant medication using HEDIS and PDC methodology using a continuous measure of adherence.

Resources required
The journal articles, clinical guidelines and other important information necessary for the successful completion of this research were obtained using the University library or interlibrary exchange services, PubMed through the College of Pharmacy, or other professional services available on World Wide Web such as government information policy on HEDIS and PDC measures. Data used in this study were provided by Major Professor Stephen Kogut. This study was approved by URI Institutional Review Board. Analyses of the data were conducted using SAS® Version 9.4 -University of Rhode Island.

RESULTS
The two separate samples were created after applying the inclusion/exclusion criteria for both HEDIS and PDC methods that were subsequently statistically analyzed. The main difference between the samples created is that patients included in the HEDIS sample have a new diagnosis of depression documented by an ICD-9 code, whereas the patients in the PDC sample are simply antidepressant users irrespective of diagnosis.

Application of inclusion and exclusion criteria for HEDIS methodology
The final cohort comprised a total of 626 Medicaid patients that met all the inclusion and exclusion criteria for HEDIS methodology (Figure 1). The initial cohort was comprised of 32,052 patients initially from which a number of 15,056 patients were excluded because they did not met the enrollment criteria and had no prescription for an antidepressant drug during the study period. Further 22,370 patients were excluded from the cohort for (1) not having a diagnosis of depression according to a ICD-9 code, (2) not being prescribed an antidepressant agent during the intake period and (3) not meeting the eligibility criteria according to HEDIS requirements.

Application of inclusion and exclusion criteria for PDC methodology
The final cohort was comprised of 22,351 patients ( Figure 2). Initially a total number of 34,481 patients were identified as using an antidepressant medication agent. After selecting only the patients that had an antidepressant prescription during the study period, 1,352 patients were excluded. After applying the continuous enrollment criteria another 1,591 patients were excluded from the sample. Further, 5,190 patients were excluded from the sample because they (1) did not use an antidepressant agent included in the study criteria and (2) their index medication was not greater than 91 days prior to the end of the study period.

HEDIS Baseline Characteristics
The analytic cohort for the HEDIS measure was comprised of 626 patients ( antidepressants were prescribed to 4.95% of the patients and 4.63%, respectively. Patients that were prescribed mirtazapine accounted for 3.35% of the sample.

PDC Baseline Characteristics
The analytic cohort for PDC measure was comprised of 22,351 patients ( were more frequently adherent than users of SSRIs. TCAs, and mirtazapine (P-value<0.001).

Univariate Logistic Regression Analyses for HEDIS and PDC methodologies
Univariate logistic regression analyses were performed for both methodologies in order to test the association of each explanatory variable separately with the outcome (adherence with antidepressant medication) in order to identify variables for inclusion in the multivariate analyses. We also excluded the variables from the further analysis that did not show significant association with the outcome (adherence with antidepressant medication) on their own as they are not likely to be associated with the outcome after adjusting for other variables. The results of the univariate logistic regression for all the variables considered in the analysis are presented in Table 3. For both methodologies, the middle and older age groups had statistically significant positive beta coefficients that showed the association of age with the outcome (adherence with antidepressant medication

Likelihood of antidepressant medication adherence according to HEDIS selected characteristics
The results of the multivariate logistic regression analysis of the effect of antidepressant therapy on medication adherence according to HEDIS methodology are presented in Table 5. Collinearity was not found between any of the independent variables assessed for the inclusion into the model. Interaction between variables was assessed as well to decide whether any interaction terms should be included into the

Likelihood of antidepressant medication adherence according to PDC selected characteristics
The results of the multivariate logistic regression analysis of the effect of antidepressant therapy on medication adherence according to PDC methodology are presented in Table 6. Collinearity was not found between any of the independent variables assessed for the inclusion into the model. Interaction between variables was assessed as well to decide whether any interaction terms should be included into the final model. The interactions found between the independent variables are discussed in the limitations section of the paper, as the interactions terms were not suited to be included in the final model. The likelihood ratio chi-square was used to decide on the best model and was calculated by subtracting the -

Correlation between adherence with antidepressant medication using HEDIS and adherence with antidepressant medication using PDC methodology
The final cohort comprised a total of 626 Medicaid patients that met all the inclusion and exclusion criteria for HEDIS methodology applied to the PDC sample. The sample was analyzed to determine if there is a correlation between measuring adherence with antidepressant medication using HEDIS and PDC methodology. We calculated the correlation coefficient of the two variables and we found that the two methodologies are highly positively correlated with a correlation coefficient of 0.95. This means that the two variables are closely correlated and that if we measure adherence with antidepressant medication according to HEDIS methodology for patients that meet the PDC criteria, in 95% of the cases they will be found adherent. A correlation matrix plot is presented in Figure 3.

DISCUSSION
Major depressive disorder is a prevalent and debilitating disease among the U.S adult population.
Effective treatment and adherence with antidepressant medication regimen is essential for optimal therapeutic outcome and for lower economic implications. Depression is a chronic disease, and many patients will be taking medication for long periods of time, maybe lifelong. According to the World Health PDC we considered in our analysis that no more than two drugs were prescribed for same medical condition and coded polytherapy by looking at the drugs filled in the same day. We excluded from our analysis one of the drugs refilled in the same day during our study period. A limitation of the HEDIS method is that it does not adequately discern between combination therapy and medication switch. Thus, patients using more than more type of antidepressant are more likely to be classified as adherent. Factors associated with medication adherence are dependent upon the characteristics of the patient population, the medication class and data source evaluated. According to the World Health Organization influence patient adherence to therapy. 24 Correlation between the two methodologies proved that adherence with antidepressant medication can be measured using the PDC method used for measuring adherence with medication for chronic diseases.

LIMITATIONS AND CONCLUSIONS
The present study utilized a retrospective cohort design with variables derived from data based claims of insured patients in a Medicaid state program. Because of the nature of the data we lack information on race/ethnicity, disease severity, social support of perceived stigma of patients with depression. Adherence is a problem that touches more than one level and poor adherence can be influenced by factors such as knowledge, attitudes, skills, environment of the patient and providers' practices. Due to the retrospective, not-randomized study design and the use of claims data we couldn't analyze all the variables that might have influenced adherence with antidepressant medication regimen. Because we captured only comorbid disease states through ICD-9 coding only during the study period, we might have underestimated the comorbid disease prevalence. There is a potential for misclassification surrounding the outcome (adherence) considering that patients that were prescribed two drugs on the same day were classified as using polytherapy. Some interactions were found between age and therapeutic regimen. We derived measures of adherence from claims data rather than observing actual medication use. It is possible that patients picked up medication but then did not proceed to take it, resulting in a misclassification of baseline adherence.
Our study is believed to be the first evaluating the correlation between the measuring adherence with antidepressant medication according to HEDIS methodology and PDC methodology, so there was no evidence in the literature to support our findings. The generalization of our results is limited and confirmation of our results is to be warranted.
In conclusion, in both cohorts age (over 35 years) was found to be a risk factor for adherence with antidepressant medication for both groups. Patients that had respiratory disease had an increased odds of adherence with antidepressant medication relative to patients that did not have a respiratory disease.
Patients that had a mental health disease had increased odds of adherence with antidepressant medication relative to patients that did not have these comorbid conditions. Patients that were using more than one drug were significantly more likely to be adherent to antidepressant medication regimen than patients that were using only one type of antidepressant drugs.
The methodologies of measuring adherence are closely correlated and that if we measure adherence with antidepressant medication according to HEDIS methodology for patients that meet the PDC criteria, in 95% of the cases they will be found adherent.