Measuring and predicting the factors associated with persistence to antidepressant therapy in patients with diabetes

Objective: The objective of this study was to measure the rate of persistence to antidepressants and to identify the factors influencing persistence to these medications in patients with diabetes. Methods: We conducted a retrospective study among patients with diabetes enrolled in a commercial health plan between 2009 – 2012. The study population includes patients who were at least 18 years of age and diagnosed with major depression during this period. The patients were eligible for acute phase treatment and continuation phase treatment if they were enrolled at least 90 days and 180 days after the Index antidepressant Prescription Start Date (IPSD) respectively. The patients were eligible for the study if (1) there was no history of diagnosis of major depression for at least 120 days prior to the first episode of major depression (Index Episode Start DateIESD) and (2) there was no history of an antidepressant dispensing for at least 90 days prior to the IPSD. Results: The mean age of the patients in both the phases was approximately 60 years. A majority were prescribed SSRI (Selective Serotonin Reuptake Inhibitors) antidepressants in acute (71.5%) and continuation (74.9% ) treatment phases , 81.8% of the patients in acute phase and 72.8% in continuation phase had monotherapy, 210 patients in acute phase and 112 patients in continuation phase had no follow up visits. Only 60.1 % and 43.5% of patients were found to be persistent to acute and continuation treatment phases respectively. Acute Phase Treatment: The odds of non-persistence were higher for patients in age group 18-40 compared to patients aged 40 above (OR 0.46 P=0.0036). Across the class of antidepressants patients utilizing trazadone or mirtazapine (OR=2.35 P=0.02) were more likely to non-persist. Patients who had 1 to 3 (OR=0.19 P<0.0001) or more than 3 (OR= 0.63 P<0.0001)follow up visits were found to have lower odds for non persistence compared to patients with no follow up visits during the treatment. Patients who had a combination treatment with either buproprion or tricyclic antidepressants (TCA) were found to be more likely to non-persist (OR 2.85 P=0.003). Continuation Phase Treatment: The odds of non-persistence were higher for patients in age group 18-40 compared to patients aged 40 above (OR 0.52 P=0.03). Patients who had 1 to 3 (OR 0.1 P<0.0001) or more than 3 (OR 0.13 P<0.0001)follow up visits were found to have lower odds for non-persistence compared to patients with no follow up visits during the treatment period Conclusion: In this population of commercially-insured patients having diabetes, acute phase persistence with antidepressant therapy was found to be associated with age, antidepressant class, type of therapy and intensity of follow up visits where as continuation treatment persistence was associated with age and intensity of follow up visits.


ACKNOWLEDGMENTS
I would like to earnestly thank my major professor Dr. Larrat, whose support is the main reason I could complete this project.I thank him for all the guidance, resources and opportunities he has provided me since I started my masters program.I thank him for the knowledge I have gained, for his prompt responses and feed backs given to me whenever needed.His kindness and disposition will always be an inspiration to me.I would also like to heartfully thank Dr. Kogut for his immense support during this project.I am always grateful for his guidance which enabled me to think better and work efficiently on my project.Working with him not only enhanced my research and technical skills but also my personal qualities such as being positive, dedication to work and being optimistic.I also thank him for providing the data for this project and allowing me to work on other research projects.
I would also like to thank Dr. Willey, whose classes equipped me with basic and fundamental research skills.The two courses she taught were the best classes I have taken here.I also thank her for time and feedback on my class work which I was able to apply to my thesis work.I would also like to sincerely thank Dr. Rosenbaum and Professor Marcoux for being on my committee and providing suggestions to further improve my work.
I am also very thankful to my parents and my sister who have been a source of infinite love and encouragement to me.It would not have been possible for me to complete this project without their support.
Once again I would like to express my deepest gratitude to Dr. Larrat and Dr. Kogut.v  for patients with diabetes to adhere to their medication to have glycemic control 2 and also to decrease hospitalizations and health care costs. 3proximately 15 million adults in the United States are affected by major depression which is highly recurrent. 4,5It is estimated that 9.5% of the adult population suffer from depressive illness every year. 6[12] The increased prevalence of depressive symptoms and major depressive disorder in patients with diabetes has been documented by many studies. 13The prevalence of depression in diabetic population is approximately double the prevalence of depression in the general population. 14Symptoms of depression are present in approximately 30% of people with diabetes 15 and approximately 10% of people with diabetes have major depression. 16Li et al found that 45% of patients with diabetes have undiagnosed depression which suggests even higher rates of prevalence of both these conditions together. 17Comorbid depression in diabetes patients is severe and persistent. 18Lustman et al found that in patients with comorbid major depression and diabetes the rates of relapse of depression were as high as 79% over 5 year period with a mean of 4 or more episodes during that period. 19,20tients with comorbid diabetes and depression are more likely to be non-adherent to the medication regimen 21 and also show poor diabetes management compared to patients without depression. 225][26][27] It is also associated with greater symptom burden, 28 functional impairment, 25,26 micro and macro vascular complications, 29 higher health care costs 26 and mortality. 29For example a study conducted by Ciechanowski et al on patients enrolled in a health maintenance organization having diabetes with higher severity of depression symptoms, subjects had worse physical and mental functioning, higher non-adherence to oral hypoglycemic regimens (15% vs. 7%) 51% higher primary , 75% higher ambulatory and 86% higher total health care costs compared to patients with low severe symptoms of depression. 26These facts suggest the importance of effective treatment of depression in patients with diabetes to improve the health outcomes.
When compared to non-depressed patients, depressed patients are three times more likely to be non-adherent to treatment recommendations. 30In one study conducted amongst primary care patients about one-third discontinued their antidepressant therapy within one month of the initiation of the treatment and about half discontinued within three months. 31Lowest rates of adherence were found in patients with diabetes (67.5%) compared to patients with other chronic conditions (pulmonary-68.8% and cardiovascular diseases-76.6%) in a study conducted by DaMatteo et al. 32 These facts together suggest even lower adherence rates in patients with comorbid depression and diabetes.A major barrier to improving care in people with comorbid depression and diabetes is their poor adherence to the treatment. 30e objective of this study was to measure the rate of persistence to antidepressants and to identify the factors influencing persistence to these medications in patients with diabetes.
We The patients were eligible for the study if (1) there was no history of diagnosis of major depression for at least 120 days prior to the first episode of major depression (Index Episode Start Date-IESD) and ( 2) there was no history of an antidepressant dispensing for at least 90 days prior to the IPSD.Since persistence with particular drug therapy was a variable of our interest, we excluded the patients who switched the drug therapy during the treatment period.the overall measure of disease burden and predicts the mortality by weighing different comorbid conditions of the patient; this score was calculated for each patient for all the disease conditions in the time frame studied and was used for assessing the comorbidity of the patients in this study.The score was coded in categories with scores of 1 to 3 and more than 3. Follow up visits for any purpose during the treatment phase were counted for both the phases.The visits were coded as 0, 1-3 visits and more than 3 visits.
Statistical Analysis: Descriptive statistics for each variable were used to summarize the study population.Bivariate analyses were conducted to assess the association between the dependent variable and each independent variable.Univariate logistic regression was performed to select the variables with significant P-Values (0.2) for final model (Table 7 and Table 8 in Appendix).Interactions and colinearity were assessed between the independent variables.Multivariate logistic regression was performed to examine the multivariate associations of the independent variables with the dependent variable (persistence to antidepressant therapy during acute phase treatment and continuous phase treatment) and odds ratios were used to measure the association.Separate analyses were performed for the acute and continuous phases.
The significance level was set at 0.05 and 95% Confidence Intervals were examined.
All the statistical analyses were performed using SAS statistical software version 9.In the bivariate analysis (Table 2) age, antidepressant class, type of therapy and follow up visits were found to have a significant association with persistence to antidepressant medication during the acute phase treatment.Patients adherent during the acute phase were aged 41 or above (61.8%)vs. 18-40 years (47.5%),prescribed an SSRI antidepressant (61.7%) or SSNRI (63.3%) or bupropion (64.1%) and were on monotherapy (62.6%).Higher rates of persistence were found in patients having 1-3 (79.4%) or more than 3 follow up visits (53.9%).
In the multivariate analysis ( phase were also enrolled in a non-HMO health plan.A higher percentage of people had monotherapy (72.8%).No follow up visits were observed in 18.6% of patients during the treatment period.Most of the patients were prescribed (83.6%) generic antidepressants similar to the patients in acute phase treatment.A majority of them were prescribed SSRI's (74.9%) followed by SSNRI's (9.0%).
In the bivariate analysis (Table 5) age and follow up visits were found to have a significant association with persistence to antidepressant medication during the continuous phase treatment.Patients adherent during the continuous phase were aged 41 or above (45.1%)vs. 18-40 years (30.3%).Higher rates of persistence were found in patients having 1-3 (56.2%) or more than 3 (48.9%)follow up visits compared to patients with no follow up visits (11.6%).
In the multivariate analysis (Table 6) age and follow up visits were found to be significantly associated with persistence to continuous phase treatment after adjusting for other variables.The odds of non-persistence were higher for patients in age group 18-40 compared to patients aged 40 above (OR=0.52P=0.03).Patients who had 1 to 3 (OR=0.1 P<0.0001) or more than 3 (OR=0.13P<0.0001)follow up visits were found to have lower odds for non-persistence compared to patients with no follow up visits during the treatment period.

DISCUSSIONS, LIMITATIONS AND CONCLUSION
In this study we assessed the antidepressant treatment persistence rates and the factors associated with persistence to these medications in the acute phase and continuation phase treatment periods.Results of our study indicate that only 60.12 % and 43.52 % of patients were persistent to acute and continuation phase respectively.
Persistence was significantly influenced by age, class of antidepressant, type of therapy and intensity of follow up visits during the acute phase treatment where as only age and intensity of follow up visits had an effect on persistence to the therapy during continuation phase.
[35] Older age was found to be associated with persistence to the antidepressant medication in both acute and continuous phases.This finding was consistent with the previous research [36][37][38] that older people were more likely to be adherent to the therapy compared to younger people.For example, in a study conducted by Akincigil et al on privately insured patients, patients who were age 50 or above were found to be 2.48 times more likely to be adherent compared to patients with 18-25 years.This might be due to the possibility that older people are more experienced in managing their medication regimen to various disease conditions which makes it easier for them to manage antidepressant medication as well.Another possible explanation might be they are more worried about mortality compared to younger population.
The class of antidepressant utilized was significantly associated with persistence during the acute phase but had no influence during the continuation phase.This might be due to the side effects or adverse events due to these drugs during the initial treatment.It could be assumed that patients who had no side effects with these drugs in acute phase treatment persisted with their medication in continuation phase and therefore the class of drug had no influence on persistence in this phase .Patients who were prescribed SSRI's were more likely to be persistent to acute phase treatment which was a similar result in previous studies 39 .There was no difference in the rates of persistence to the therapy among SSRI's SSNRI's, TCA and bupropion.Patients receiving other antidepressants (trazadone and mirtazapine) were less likely to persist with effective acute phase treatment.One reason for this might be that these drugs were prescribed for insomnia while the primary treatment was psychotherapy.Another reason might be due to the adverse effects of these drugs such as weight gain, dizziness etc.
It was also found that patients who had a combination therapy with SSRI/SSNRI and bupropion or TCA were less likely to persist.This might be due to the side effects due to the combination of these drugs such as remarkably lower blood pressure with combination of TCA and SSRI. 40However a meta-analysis by Seetal et al found that combination of SSRI/SSNRI with bupropion was well tolerated. 41Another reason might be these patients have severe depression resulting in less motivation to take the medicines.
We also found that patients who have either 1 to 3 or more than 3 follow up visits are more likely to be persistent to the therapy compared to patients with no follow up visits.It was found in the previous research that patients with 3 or more follow up visits were more likely to persist. 36,37,42One reason for this might be that the physician could possibly educate the patient about the importance of taking the medications regularly.Another reason might be these patients are more cautious about their health and therefore have frequent follow up visits with the physician and take their medications regularly.
Our study has several limitations.In this population of commercially insured patients having diabetes, acute phase persistence with antidepressant therapy was found to be associated with age, antidepressant class, type of therapy and intensity of follow up visits where as continuation treatment persistence was associated with age and intensity of follow up visits.The variance inflation factors in the model with no interaction terms were close to 1 and the condition index was less than 30 confirming no colinearity.The model fit was assessed using Hosmer-Lemeshow Goodness of fit (GOF) and ROC curves.The GOF P-value (0.9) was greater than 0.05 and AUC (0.7) was greater than 0.5indicating an adequate model fit

Figure 1 .Index 6 Figure 2 .Figure. 3
Figure 1.Time frame for the enrollment of subjects in the study (1)Reliance on claims data, which might have coding errors due to insufficient information provided by the physician about ICD codes, CPT codes etc.Also the deliberate miscoding of major depression with other diagnosis codes by physicians43 further increases the chances of coding errors in identifying patients with major depression.(2)If the patient receives prescription or care outside the health plan network or if the patient receives samples by the provider in the office it will not be shown in the claims data.(3)We assumed that patients who had prescription for the antidepressants as persistent which might not actually reflect the actual utilization of these drugs by the patients.(4)In the HEDIS (Healthcare Effectiveness Data and Information Set) algorithm patients who switched the therapies are also included in the calculation of rates of persistence but in this study since we excluded them so the rates of persistence in this study might be underestimated if the patients who switched their therapy were persistent.(5)Thegeneralization of these results might be limited to commercial health plans and patients with diabetes.

Figure 4 .
Figure 4. Receiver Operating Characteristic (ROC) Curve for acute phase model

Figure 5 .
Figure 5. ROC Curve for continuation phase model

Table 1
provides the base line demographic and clinical characteristics of the study population.The mean age of this population was 60.5 years and a majority of the patients were females (56.8%).The mean comorbid score in this study population was 3.4.Additionally 44.69% of the patients received psychotherapy during the acute phase treatment.Insulin utilization was observed in 31% of these patients.Forty-four (6.4%) patients were hospitalized.A majority (65.2%) of them were enrolled in non-HMO (Health Maintenance Organization) plans.Most of these patients (81.8%) had antidepressant monotherapy.No follow up visits were in 30.6% of patients during the treatment period.Generic antidepressant drugs were prescribed to 84.6% of the patients.A majority of the population (71.5%) were prescribed SSRI's followed by SSNRI's (8.7%).

Table 3 )
age, class of antidepressant, type of therapy and follow up visits were found to be significantly associated with persistence to acute phase treatment after adjusting for other variables.The odds of non-persistence were higher for patients in age group 18-40 compared to patients aged 40 above (OR 0.46 P=0.0036).Across the class of antidepressants patients utilizing other (trazadone and mirtazapine) antidepressants (OR=2.35P=0.02) were more likely to not persist compared to patients utilizing SSRI's, whereas no difference was observed in the rates of persistence in patients utilizing SSNRI's, TCA, bupropion and SSRI's.Patients who had 1 to 3 (OR=0.19P<0.0001) or more than 3 (OR= 0.63 P<0.0001)follow up visits were found to have lower odds for non-persistence compared to patients with no follow up visits during the treatment.Patients who had a combination treatment with either buproprion or TCA were found to be more likely to non-persist (OR 2.85 P=0.003) where as there was no difference in rates of persistence in patients who had a combination treatment either with trazadone or mirtazapine.

Table 1 .
Demographic and clinical characteristics of a population of commercially insured patients: Acute phase persistence with antidepressant medications (N=687)

Table 3 .
Estimated odds ratios and 95% confidence intervals of antidepressant treatment persistence during acute phase treatment

Table 4 .
Demographic and clinical characteristics of a population of commercially insured patients: Continuation phase treatment persistence with antidepressant medications

Table 5 .
Bivariate Analysis: Factors associated with persistence to continuation phase antidepressant treatment in commercially insured patients

Table 7 .
Results of univariate logistic regression analysis for acute phase treatment persistence

Table 8 .
Result of univariate logistic regression analysis for continuation phase treatment persistence.

Table 9 .
Goodness of Fit test for acute phase model