Adherence and Healthcare Utilization among Older Adults with COPD and Depression

Abstract

Background and Objective

Adherence to chronic obstructive pulmonary disease (COPD) maintenance medications and antidepressants may reduce healthcare utilization among multimorbid individuals with COPD and depression. We quantified the independent effects of adherence to antidepressants and COPD maintenance medications on healthcare utilization among individuals co-diagnosed with COPD and depression.

Procedures

We conducted a retrospective cohort study using a 2006–2012 5% random sample of Medicare beneficiaries co-diagnosed with COPD and depression who had two or more prescription fills of both COPD maintenance medications and antidepressants. We measured adherence to medications using the proportion of days covered per 30-day period. The primary outcomes were all-cause emergency department (ED) visits and hospitalizations. Beneficiaries were followed over a minimum 12 month follow-up period.

Results

Of the 16,075 beneficiaries meeting inclusion criteria, 21% achieved adherence ≥80% to COPD maintenance medications and 55% achieved adherence ≥80% to antidepressants. Compared to no use and controlling for antidepressant adherence and potential confounders, higher (≥80%) levels of adherence to COPD maintenance medications were associated with decreased risk of ED visits (hazard ratio (HR) 0.79; 95% CI 0.74, 0.83) and hospitalizations (HR 0.82; 95% CI 0.78, 0.87). Similarly, higher levels (≥80%) of adherence to antidepressants resulted in decreased risk of ED visits (HR 0.74; 95% CI 0.70, 0.78) and hospitalizations (HR 0.77; 95% CI 0.73, 0.81) compared to no use.

Conclusions

Clinicians can assist in the improved management of their multimorbid patients’ health by treating depression among patients with COPD and monitoring and encouraging adherence to the regimens they prescribe.

Introduction

Multiple chronic conditions are common in chronic obstructive pulmonary disease (COPD) and influence its natural history and management.^1,2 In particular, depression is a common comorbidity among COPD patients, with prevalence ranging from 17% to 44%.^3–7 Comorbidities such as depression can complicate medication regimens, negatively affect adherence to maintenance medications, and lead to increased healthcare utilization.^3–11 Healthcare utilization and costs among individuals with COPD are two to three times higher compared to individuals without COPD.^12,13 It is likely that some of these increased utilization and costs are attributable to comorbid illness burden.

Use of COPD maintenance medications results in decreased hospitalizations, mortality, and costs; however adherence is low, ranging from 29%–56%.^14–21 Treatment of comorbid illness could impact adherence and outcomes among individuals with COPD; however, few studies have examined this question. Among Medicare beneficiaries diagnosed with both COPD and depression, antidepressant treatment was associated with reduced mortality; however, receipt of depression care was not associated with fewer COPD-related hospitalizations among a similarly diagnosed cohort of veterans.^22,23 This discrepant finding may be due to a very strict definition of guideline-concordant depression care that enriched the control cohort with individuals who received some depression treatment.

A limitation of these prior studies is the failure to account for concomitant treatment of COPD and depression. Use of COPD maintenance medications is an independent predictor of decreased healthcare utilization and may be associated with adherence to antidepressants.^14–16 Hence, when assessing healthcare utilization among individuals co-diagnosed with COPD and depression, it is critical to account for adherence to treatment for both conditions. Thus, the objective of this study was to quantify the independent effects of adherence to antidepressant treatment and COPD maintenance medications on emergency department (ED) visits and hospitalizations among Medicare beneficiaries co-diagnosed with COPD and depression. Multi-morbidity has an impact across disease states; hence, we focused specifically on all-cause rather than disease-specific healthcare utilization.²⁴

Methods

Study Population

Medicare administrative claims data from the Centers for Medicare & Medicaid Services (CMS) Chronic Condition Data Warehouse (CCW) were the data source for this study. We used a 5% random sample of Medicare beneficiaries to identify beneficiaries with at least 1 inpatient or outpatient claim for COPD and related respiratory disorders containing an International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes (490, 491.x, 492.x, 494.x, or 496) during 2006–2011. These ICD-9-CM codes are used by CMS to identify beneficiaries with COPD and related disorders and have been used in several published analyses of administrative claims data.^{8,14,15,22,25}

We excluded beneficiaries with a history of respiratory cancer, tuberculosis, asbestosis, and sarcoidosis because their medication use may differ from that of other individuals with COPD. We required continuous Medicare Parts A, B, D, and no Part C (Medicare Advantage) coverage in the 12 months following initial diagnosis of COPD (index date). Once entered into the cohort, beneficiaries continued to contribute follow-up time until 12/31/2012 unless they: 1) were deceased; 2) lost Part D coverage; or 3) enrolled in a Part C plan. All Medicare beneficiaries who met criteria since 1999 have the date of first diagnosis of COPD reported in the CMS Master Beneficiary Summary File (MBSF). We used this date to exclude beneficiaries whose first diagnosis of COPD occurred prior to 2006, ensuring that we were focusing on newly-diagnosed COPD cases.

We used ICD-9-CM codes 296.2x, 296.3x, and 311.xx to identify beneficiaries with depression. These codes have been used previously and exclude bipolar disorder, schizoaffective disorder, and dysthymic disorder.^{8,11,22,26–28} Presence of depression was defined as at least one diagnosis code on at least 1 inpatient claim or 2 outpatient claims during the study period. To be included in our study cohort, we required a diagnosis of depression at least once following initial COPD diagnosis.

Exposure

The primary exposures in this study were adherence to inhaled COPD maintenance medications and antidepressant medications. We searched for all inhaled maintenance medications (including inhaled corticosteroids, long-acting β-agonists, and long-acting anticholinergics, alone or in combination with each other) in the Part D prescription drug events file. We excluded oral methylxanthines because they can be used as either acute or maintenance medications and are limited to severe COPD.²⁹ We also searched for and included Part B claims for nebulized medications. We divided follow-up time post-COPD index date into 30-day months. We required at least 2 fills of both COPD maintenance and antidepressant medications so that we could measure adherence over the duration of follow-up.

Adherence was measured using Proportion of Days Covered (PDC) (number of daily doses in the prescription/30), which ranges from 0 to 1, where 0 = no medication adherence and 1 = total medication adherence. PDC is a well- and widely- accepted approach to assessing adherence in administrative claims data and has been used in studies examining adherence to COPD medications.^{11,14,15,22,30–33} PDC ≥0.8 is considered the standard threshold for ‘good’ adherence.^30–32 Adherence was measured monthly from the date of the first fill of a COPD maintenance medication following COPD diagnosis through the end of the study period. We created a rolling 3-month average adherence to reduce variability in monthly adherence measures.³⁴ Even though we required at least 2 fills of COPD maintenance medications and started measuring adherence during the month of the first fill, many beneficiaries still had zero values for adherence during some months of observation. This could occur during follow-up when a beneficiary did not have a fill for three or more consecutive months. Distribution average adherence was highly skewed, so we created adherence categories: 0, <0.2, ≥0.2 to <0.4, ≥0.4 to < 0.6, ≥0.6 to <0.8, and ≥0.8. We also examined a narrower categorization: 0, <0.8, ≥0.8. To ensure temporality, we lagged the rolling adherence variable by one month. Similarly, we searched for all antidepressants and created lagged 3-month rolling adherence categories.

Outcome

The primary outcomes of interest were all-cause ED visits and hospitalizations. We searched inpatient and outpatient claims to create a count of ED visits occurring each month and searched inpatient claims to create a count of hospitalizations occurring each month. We dichotomized this as ≥1 ED visit or hospitalization per month.

Covariates

Comorbidities at COPD diagnosis were determined using CCW’s 27 flagged comorbid conditions.²⁵ If the date of first diagnosis of a chronic condition was prior to the date of COPD diagnosis, the patient was considered to have that chronic condition at baseline. We also controlled for diagnosis of asthma (ICD-9-CM codes 493.xx on 1 or more inpatient claims or 2 or more outpatient claims). To measure multi-morbidity, we summed indicator variables for nine chronic conditions (Alzheimer’s disease and related dementias, atrial fibrillation, chronic kidney disease, heart failure, diabetes, ischemic heart disease, osteoarthritis, stroke, and asthma). Our multi-morbidity measure ranged from 0–9 and was dichotomized at the mean (mean=2.0) based on its distribution. We created time-varying comorbidity diagnoses to control for onset of new morbidities in our regression model by comparing the first date of diagnosis with the first day of each month following diagnosis of COPD.

We created variables to assess COPD complexity, including any use of acute COPD medications (e.g., short acting beta-agonists and short-acting anticholinergics) and use of supplemental oxygen within the month. Supplemental oxygen has been used to assess severe COPD and to proxy COPD severity in multiple studies using administrative claims data.^15,33,35,36 Acute COPD medications are used to treat exacerbations, which are associated with moderate to severe disease.³⁷ We controlled for nursing home stay during the month by searching for the presence of healthcare common procedure coding system (HCPCS), or Place of Service codes on a skilled nursing facility (SNF) claim. These variables were lagged by one month to ensure temporality.

Individuals who engage in healthy behaviors may be less likely to be hospitalized.^38,39 We created a healthy behavior measure based on the following preventive health services utilization measures: influenza vaccination, colorectal cancer screening, prostate cancer screening, mammography and Papanicolaou smears and created an annual count as an indicator of healthy behavior.³⁹ This variable ranged from 0–3 (mammography or Papanicolaou test counted as 1) and was dichotomized at 1 or more based on its distribution.

Data Analysis

We examined the distributions of demographic and clinical variables during the month of COPD diagnosis and examined the average 3-month adherence to COPD maintenance medications and antidepressants during follow-up. We used a discrete time approach to model risk of the primary outcomes (two separate models) as a function of lagged 3-month rolling adherence to both COPD maintenance medications and antidepressants and covariates per 30-day period following the first fill of a COPD maintenance medication and an antidepressant. To accomplish this, we used generalized linear models with a binomial distribution and a complementary log-log link.⁴⁰ We censored individuals after the first event; hence, our results are interpretable for the first occurrence of an ED visit or hospitalization.

First, we examined unadjusted models of adherence to both COPD maintenance medications and antidepressants. We added age, sex, and race to the models. Next, we added covariates to the model, keeping those whose p-value was <0.001 or whose inclusion resulted in greater than a 10% change in the effect estimates. We tested effect modification by pre-COPD diagnosis history of ED visits and use of oxygen during the first 6 months following diagnosis. Hazard ratios (HR) and 95% confidence intervals (CI) are reported.

Our final discrete time regression models contained key independent variables: adherence to COPD maintenance medications and antidepressants, and covariates: months since COPD diagnosis, age, sex, race, percent of the county population with a high school diploma, median household income (county), time-varying comorbidity indicators (acute myocardial infarction, Alzheimer’s disease and related dementias, asthma, atrial fibrillation, chronic kidney disease, cancer, heart failure, hip fracture, history of depression, ischemic heart disease, stroke, more than one chronic condition), an indicator variable for preventive health measures, polypharmacy, and lagged measures of oxygen use, acute inhaler use, and nursing home residence. Models were run with both categorizations of adherence.

This study was approved by the Institutional Review Board at the University of Maryland, Baltimore (HP-00055737). All analyses were conducted with SAS v9.4 (SAS Institute, Cary, NC). The p-value for statistical significance was <0.05 unless otherwise noted.

Results

There were 305,997 beneficiaries newly diagnosed with COPD between 2006 and 2012 and of these, 77,444 (25%) were diagnosed with depression following COPD diagnosis. Of the 30,695 (40% of 77,444) with at least 12 months of continuous Medicare Parts A,B,D and no Part C coverage following COPD diagnosis, 16,075 (52% of 30,695) had at least 2 prescription fills each of COPD maintenance medications and antidepressants; this group formed our study sample.

The sample was primarily female (76.5%) and white (86.4%) with an average age of 67.6 years (standard deviation 12.8) (Table 1). Comorbid illness burden was high, with 71.2% having more than one chronic condition. During the month of COPD diagnosis, 8.9% used oxygen and 26.2% used acute COPD medications. There were an average of 8.1 (sd 11.1) ED visits and 4.4 (sd 5.2) hospitalizations over the 58.2 (sd 21.3) average months of follow-up post-COPD diagnosis. Over the course of the study, 32.5% of the sample visited the ED at least once, 23.8% visited the hospital at least once, and 35.2% had either an ED visit or a hospitalization at least once (data not shown).

Table 1.

Baseline Characteristics of Medicare Beneficiaries diagnosed with Chronic Obstructive Pulmonary Disease (COPD) and Depression between 2006 and 2011 with ≥2 prescription fills for COPD maintenance medications and ≥ 2 prescription fills for antidepressants, n=16,075

Characteristic	Distribution
Age in years, mean (sd)	67.6 (12.8)
Age in years, n(%)
< 65	5,507 (34.3)
65–74	5,416 (33.7)
75–84	3,824 (23.8)
> 84	1,328 (8.3)
Female, n(%)	12,298 (76.5)
Race/Ethnicity, n(%)
White	13,889 (86.4)
Black	1,204 (7.5)
Hispanic	5588 (3.5)
Other	424 (2.6)
Original reason for Medicare entitlement, n(%)
Age	8,865 (55.2)
Disability/ESRD	7,210 (44.8)
Low-income subsidy, n(%)	10,342 (64.3)
Chronic conditions, n(%)
AMI	660 (4)
Alzheimer’s Disease or Related Dementias	2,106 (13.1)
Asthma	5,469 (34.0)
Chronic Kidney Disease	2,699 (16.8)
History of depression pre-COPD diagnosis	8,951 (55.7)
Diabetes	5,954 (37.0)
Heart Failure	5,685 (35.4)
Ischemic Heart Disease	8,489 (52.8)
Rheumatoid Arthritis/ Osteoarthritis	9,208 (57.3)
Stroke/Transient Ischemic Attack	2,591 (16.1)
>1 Comorbid conditions	11,465 (71.2)
Oxygen use month of COPD diagnosis, n(%)	1,432 (8.9)
COPD acute medication use month of COPD diagnosis, n(%)	4,208 (26.2)
≥1 Preventive health measures*, n(%)	1,397 (8.7)
Evidence of nursing home residence, n(%)	1,092 (6.8)

^*Includes influenza vaccination, colorectal cancer screening, prostate cancer screening, mammography and Papanicolaou smears

During the study period, 21% of the sample achieved average three-month rolling adherence to COPD maintenance medications ≥80% (good adherence) and 39% had an average adherence of zero (Table 2). Adherence to antidepressants was higher, with 55% achieving adherence ≥80%. Among individuals with ≥80% adherence to COPD maintenance medications, 67% had good adherence to antidepressants. However, among individuals with ≥80% adherence to antidepressants, only 26% had good adherence to COPD maintenance medications (data not shown).

Table 2.

Three-month Adherence Categories Averaged Over Follow-up among Medicare Beneficiaries diagnosed with Chronic Obstructive Pulmonary Disease (COPD) and Depression between 2006 and 2011 with ≥2 prescription fills for COPD maintenance medications and ≥ 2 prescription fills for antidepressants, n=16,075

Categories	Person months of follow-up (%)
COPD Maintenance Meds
0	311,760 (39.2)
<0.2	90,951 (11.4)
≥0.2 and <0.4	89,099 (11.2)
≥0.4 and <0.6	59,780 (7.5)
≥0.6 and <0.8	75,693 (9.5)
≥0.8	167,635 (21.1)
Antidepressants
0	154,408 (19.2)
<0.2	19,457 (2.4)
≥0.2 and <0.4	48,519 (6.0)
≥0.4 and <0.6	46,964 (5.8)
≥0.6 and <0.8	95,240 (11.8)
≥0.8	440,306 (54.7)

In adjusted analysis, adherence to COPD maintenance medications was associated with a significantly decreased hazard of an emergency department visit when compared with a zero level of adherence (no use) and controlling for adherence to antidepressants and other covariates (Tables 3 and 4). Similarly, adherence to antidepressant medications was associated with decreased hazard of an emergency department visit when compared with a zero level of adherence and controlling for adherence to COPD maintenance medications and other covariates. Similar results were observed for all-cause hospitalizations (Tables 3 and 4). There was no effect modification by pre-COPD diagnosis history of ED visits or by use of oxygen during the first six months following diagnosis.

Table 3.

Adjusted^* Association of lagged 3-month rolling adherence to COPD Maintenance Medications and Antidepressants with All-Cause Emergency Department (ED) Visits and Hospitalizations Using Multiple Adherence Levels relative to Zero Adherence among Medicare Beneficiaries diagnosed with Chronic Obstructive Pulmonary Disease (COPD) and Depression between 2006 and 2011 with ≥2 prescription fills for COPD maintenance medications and ≥2 prescription fills for antidepressants, n=16,075

	ED Visits	Hospitalizations
	Hazard Ratio (95% Confidence Interval)	Hazard Ratio (95% Confidence Interval)
COPD Maintenance Meds
0	Reference	Reference
<0.2	0.72 (0.67, 0.77)	0.70 (0.66, 0.75)
≥0.2 and <0.4	0.71 (0.67, 0.76)	0.77 (0.71, 0.82)
≥0.4 and <0.6	0.65 (0.60, 0.70)	0.71 (0.65, 0.76)
≥0.6 and <0.8	0.73 (0.68, 0.78)	0.75 (0.69, 0.80)
≥0.8	0.79 (0.75, 0.83)	0.82 (0.78, 0.87)
Antidepressants
0	Reference	Reference
<0.2	0.68 (0.60, 0.79)	0.77 (0.67, 0.88)
≥0.2 and <0.4	0.78 (0.71, 0.86)	0.80 (0.73, 0.88)
≥0.4 and <0.6	0.66 (0.60, 0.73)	0.71 (0.64, 0.78)
≥0.6 and <0.8	0.73 (0.67, 0.78)	0.79 (0.73, 0.85)
≥0.8	0.74 (0.70, 0.78)	0.77 (0.73, 0.81)

^*Models adjusted for months since COPD diagnosis, age, sex, race, percent of the county population with a high school diploma, median household income (county), acute myocardial infarction, Alzheimer’s disease and related dementias, asthma, atrial fibrillation, chronic kidney disease, cancer, heart failure, hip fracture, history of depression, ischemic heart disease, stroke, more than one chronic condition, an indicator variable for preventive health measures, polypharmacy, and lagged measures of oxygen use, acute inhaler use, and nursing home residence.

Table 4.

Adjusted^* Association between lagged 3-month rolling adherence to COPD Maintenance Medications and Antidepressants and Emergency Department Visits or All-Cause Hospitalizations Using Adherence Levels of <0.8 and ≥0.80 Relative to Zero Adherence among Medicare Beneficiaries diagnosed with Chronic Obstructive Pulmonary Disease (COPD) and Depression between 2006 and 2011 with ≥2 prescription fills for COPD maintenance medications and ≥ 2 prescription fills for antidepressants, n=16,075

	Emergency Department Visits	Hospitalizations
	Hazard Ratio (95% Confidence Interval)	Hazard Ratio (95% Confidence Interval)
COPD Maintenance
0	Reference	Reference
<0.8	0.70 (0.67, 0.74)	0.73 (0.70, 0.77)
≥0.8	0.79 (0.74, 0.83)	0.82 (0.78, 0.87)
Antidepressants
0	Reference	Reference
<0.8	0.72 (0.68, 0.76)	0.77 (0.72, 0.82)
≥0.8	0.74 (0.70, 0.78)	0.77 (0.73, 0.81)

^*Models adjusted for months since COPD diagnosis, age, sex, race, percent of the county population with a high school diploma, median household income (county), acute myocardial infarction, Alzheimer’s disease and related dementias, asthma, atrial fibrillation, chronic kidney disease, cancer, heart failure, hip fracture, history of depression, ischemic heart disease, stroke, more than one chronic condition, an indicator variable for preventive health measures, polypharmacy, and lagged measures of oxygen use, acute inhaler use, and nursing home residence.

Compared to zero adherence and controlling for potential confounders and antidepressant adherence, both lower (<0.8) and higher (≥0.8) levels of adherence to COPD maintenance medications were associated with decreased risk of ED visits (HR 0.70; 95% CI 0.67, 0.74 and HR 0.79; 95% CI 0.74, 0.83, respectively) and hospitalizations (HR 0.73; 95% CI 0.70, 0.77 and HR 0.82; 95% CI 0.78, 0.87, respectively)(Table 4). Any level of antidepressant use resulted in decreased risk of ED visits and hospitalizations compared with zero adherence, but the difference between adherence levels was not significant.

Discussion

In this national study of Medicare beneficiaries co-diagnosed with COPD and depression who used both COPD maintenance medications and antidepressants, concomitant use of these medications resulted in significantly reduced ED visits and hospitalizations. Differences between adherence levels had a negligible impact. What appears important is that individuals use these medications consistently, as those who did not use antidepressants or COPD maintenance medications consistently during follow-up (i.e., zero adherence) were up to 30% more likely than those with even minimal adherence to experience an ED visit or hospitalization. This study provides some of the first evidence that treatment of both comorbid COPD and depression results in better health outcomes.

Other studies support our finding of low adherence to COPD maintenance medications.^{14,17–19,21,33} Prior research that measured adherence using PDC or medication possession ratio (similar to PDC but measured over 6 months to 1 year), reported that 21%–47% of those taking COPD medications had average adherence ≥80%.^14,17,21,33 In our study, only 21% of beneficiaries achieved average adherence of ≥80%, suggesting that depression may interact with COPD to further reduce adherence. This is supported by results from a prior study that reported decreased adherence to COPD maintenance medications among disabled Medicare beneficiaries diagnosed with depression.¹¹

Surprisingly, we did not observe a dose-response association between adherence and reduced hazard of ED visits and hospitalization. Individuals who have the highest adherence may require more medication due to symptom breakthrough and/or worsening of disease. Findings also may indicate that there are adherence thresholds (minimum and maximum) needed to reduce hospitalization and ED visits. As well, we did not control for duration of medication use. Results from one study suggest that the combination of high adherence with increased duration may provide the best protection against hospitalization.¹⁴

Adherence to antidepressant therapy was higher than that observed for COPD maintenance medications, with 55% achieving adherence of ≥80%. Prior studies reported a wide range of antidepressant adherence estimates (30–97%) but used a variety of methods to measure adherence and were conducted in mixed age groups.^41,42 A more recent study conducted among members of a large health insurance plan reported that 57% of individuals aged 65 and older diagnosed with depression had adherence of ≥75% to antidepressants, consistent with our results.⁴³

Results should be interpreted in light of certain limitations. While the use of administrative claims data facilitates research on the association between medication adherence and health care utilization, it is limited by lack of laboratory results, including spirometric confirmation of COPD diagnosis and severity of airway obstruction. Consequently, it is possible that there is residual confounding of our results by COPD complexity, despite our strict inclusion criteria and controlling for oxygen and COPD acute medication use. Future studies should consider how COPD maintenance medication and antidepressant adherence influence COPD exacerbations. Challenges also include measuring adherence to inhaled medications using administrative claims data.^44,45 The gold standard for inhaler adherence assessment, weighing the canisters for displaced medication, can’t be used with administrativeclaims data, nor are patient-reported adherence measures available.¹⁶ While not directly comparable to other studies which used different cohorts, lengths of follow-up, and measures of outcomes (e.g., exacerbations versus all-cause or respiratory-related hospitalizations), our findings on the frequency of ED visits and hospitalizations fall within the ranges reported in the literature.^14,46–50 Finally, generalizability is limited to Medicare beneficiaries with Part D and FFS coverage.

Conclusions

This study provides some of the earliest evidence that understanding the interplay between medication regimens for COPD and depression is vital to preventing potentially avoidable ED visits and hospitalizations. Clinicians can assist in the improved management of their multimorbid patients’ health by treating depression among patients with COPD and monitoring and encouraging adherence to the regimens they prescribe.