Medication nonadherence is a common precipitant of heart failure (HF) hospitalization and is associated with poor outcomes. Recent analyses of national data focus on long-term medication adherence. Little is known about adherence of patients with HF immediately after hospitalization. Hospitalized patients with HF were identified from the Atherosclerosis Risk in Communities study. Atherosclerosis Risk in Communities data were linked to Medicare inpatient and part D claims from 2006 to 2009. Inclusion criteria were a chart-adjudicated diagnosis of acute decompensated or chronic HF; documentation of angiotensin-converting enzyme inhibitor/angiotensin receptor blocker (ACEI/ARB), β blocker (BB), or diuretic prescription at discharge; and Medicare part D coverage. Proportion of ambulatory days covered was calculated for up to twelve 30-day periods after discharge. Adherence was defined as ≥80% proportion of ambulatory days covered. We identified 402 participants with Medicare part D: mean age 75, 30% men, and 41% black. Adherence at 1, 3, and 12 months was 70%, 61%, and 53% for ACEI/ARB; 76%, 66%, and 62% for BB; and 75%, 68%, and 59% for diuretic. Adherence to any single drug class was positively correlated with being adherent to other classes. Adherence varied by geographic site/race for ACEI/ARB and BB but not diuretics. In conclusion, despite having part D coverage, medication adherence after discharge for all 3 medication classes decreases over 2 to 4 months after discharge, followed by a plateau over the subsequent year. Interventions should focus on early and sustained adherence.
Most studies of medication adherence in patients with heart failure (HF) have focused on long-term adherence. Little is known about the temporal trend of medication adherence immediately after hospitalization in patients with documented discharge medications. Previous studies often required a filled prescription for study inclusion, which may overestimate adherence. In the few studies that have used Medicare part D data, adherence has been described in patients with either an inpatient or outpatient HF claim. However, no study using Medicare part D data has examined adherence to HF-specific medications immediately after hospitalization. This issue is of significant policy interest because the Centers for Medicare and Medicaid Services is now tying payment to readmission rates for some chronic diseases, including HF. These initiatives have led to increased emphasis on interventions to reduce readmissions. Previous work has demonstrated improved rates of guideline-concordant medication prescribed at discharge, but we know relatively little about adherence and its determinants after discharge. To determine whether medication adherence changes over time, we examined monthly medication adherence for angiotensin-converting enzyme inhibitor/angiotensin receptor blocker (ACEI/ARB), β blocker (BB), and diuretic therapies for up to 1 year after hospitalization using part D data available for participants of the Atherosclerosis Risk in the Communities (ARIC) study. We included ARIC participants who had an adjudicated diagnosis of hospitalized acute decompensated HF (ADHF) or chronic HF in 2006 to 2009 and documentation of discharge medications from chart abstraction.
Methods
The ARIC study is an on-going predominantly biracial cohort of 15,792 men and women from 4 US communities (Forsyth County, North Carolina; Minneapolis, Minnesota; Jackson, Mississippi; and Washington County, Maryland) and followed since 1987 to 1989. The ARIC study began detailed abstraction of hospital discharge records for cohort members hospitalized with HF in 2005, as previously described. In brief, inclusion criteria for detailed abstraction included an International Classification of Diseases, Ninth Revision, Clinical Modification discharge diagnosis code for HF or a related condition or symptom (398.91, 402.01, 402.11, 402.91, 404.01, 404.03, 404.11, 404.13, 404.91, 404.93, 415.0, 416.9, 425.4, 428.x, 518.4, and 786.0x). Discharge diagnosis codes could be in any position for inclusion. Study participants’ hospitalization records were reviewed for evidence of signs and symptoms of HF, including new onset or worsening shortness of breath, peripheral edema, paroxysmal dyspnea, orthopnea, and hypoxia. In the presence of such evidence, a detailed abstraction of the medical record was completed. HF was classified as definite or possible ADHF or as chronic stable HF by independent physician reviewers. The ability to distinguish between ADHF and chronic stable HF is a strength of the ARIC study.
Data on participant demographics and hospitalizations came from the ARIC study. Validated hospitalizations for ADHF or chronic stable HF were identified and merged with Medicare Provider Analysis and Review inpatient stay records, Medicare hospice claims, and Medicare part D claims using unique ARIC study participant identifiers and information concerning dates of service. The Medicare Provider Analysis and Review records were used to obtain information on days in skilled nursing facility and all-cause hospital admissions that were not abstracted. For this analysis, claims for Medicare part D prescriptions were identified using National Drug Codes and examined by medication class. Because the ARIC study collects information on all hospitalizations through active surveillance, we were able to include Medicare Advantage and fee-for-service enrollees. To ensure capture of all part D medication fills (e.g., 90-day fills), ARIC study participants were included in this retrospective analysis if they were enrolled in Medicare part D at the time of discharge and for at least 3 months before hospitalization. Only the first HF hospitalization observed from April 2006 to December 2009 was used for each participant to depict adherence over a 12-month period and enable comparability of assessment to other studies looking at annual rates of adherence. Participant characteristics including co-morbidities, history of HF hospitalization, blood pressure, heart rate, laboratory values, and discharge medications were abstracted from the chart. Assessments of left ventricular function and left ventricular ejection fraction (LVEF) were documented in the current or previously available medical records or in reports of cardiac imaging (echocardiography, cardiac catheterization, and nuclear or other cardiac imaging) available to ARIC.
The study population included 834 ARIC cohort participants who had a hospitalization for ADHF or chronic stable HF from April 2006 to December 2009. We excluded 5 participants who were veterans and had no part D claims filed for the 3 medications at any point during the study period. A total of 426 participants had documentation regarding medications prescribed at discharge and part D enrollment for at least 3 months before discharge. We excluded 24 participants who were not discharged on any of the 3 classes of medications. The final study sample was 402.
Prescriptions at discharge for ACEI/ARB, BB, or diuretic therapy were documented by chart abstraction. Proportion of ambulatory days covered (PADC) was determined using fill date and days supplied and was calculated as proportion of days the drug was available during the study period divided by the number of days the patient was not in a hospital, skilled nursing facility, or hospice. To account for switches within a class or changes in dose, PADC was not calculated in excess of 1 for any time period. The PADC was determined for up to twelve 30-day periods after hospital discharge for each medication; the 30-day periods are subsequently referred to as “months.” Medication adherence was defined as ≥80% PADC over the specified time period. Observations were censored by death or end of the study period (December 31, 2009); observations were only included in the calculation of adherence each month if the patient had at least some days in an ambulatory setting (i.e., days not in a hospital, skilled nursing facility, or on the Medicare hospice benefit). Bivariate associations of medication adherence class over the first 3 months after discharge were assessed using the chi-square tests (for dichotomous participant characteristics) or t tests (continuous measures) for each drug. We examined the following participant characteristics from the hospitalization in relation to adherence: demographics, co-morbidities, history of HF hospitalization, ADHF versus chronic HF, HF type, LVEF, assessment of LVEF during hospitalization, blood pressure, heart rate, serum sodium, serum creatinine, length of stay, and adherence to the other 2 medication classes.
Results
Overall, we studied 402 ARIC cohort participants who were hospitalized for ADHF or chronic stable HF from April 2006 through December 2009, had medications documented at discharge, and were enrolled in part D for at least 3 months before discharge. Table 1 provides characteristics of the sample of 402 participants with documentation of at least one of 3 medication classes at discharge. Participants were 75 ± 6 years old, 30% were men, and 40% were black. Most participants, 70%, were hospitalized for ADHF. Documentation of LVEF was available for 83% of the 402 participants either by assessment during hospitalization, from information available in ARIC study records for a previous hospitalization, or imaging studies before the study period. Based on available documentation, 55% of participants with an assessment of left ventricular function had an LVEF <50%. Table 2 lists the proportion of the 402 participants prescribed various combinations of the 3 medication classes. In total, 62% (n = 248) were discharged on ACEI/ARB, 76% (n = 305) on BB, and 80% (n = 321) on a diuretic. Only 39% of participants were discharged on all 3 medication classes.
| Characteristic | Mean (±SD) or % |
|---|---|
| Age, years | 75.4 ± 5.6 |
| Male | 30% |
| ARIC geographic region and race | |
| Washington County white | 18% |
| Forsyth County white | 21% |
| Forsyth County black | 3% |
| Minneapolis white | 21% |
| Jackson black | 37% |
| Hypertension | 87% |
| Diabetes mellitus | 54% |
| Lung Disease | 39% |
| Myocardial infarction | 24% |
| Atrial fibrillation | 26% |
| Stroke/Transient Ischemic Attack | 20% |
| Depression | 10% |
| Dialysis | 6% |
| Smoker | 12% |
| Previous HF hospitalization | 42% |
| Acute decompensated HF | 70% |
| Systolic blood pressure (mmHg) | 145 ± 33 |
| Heart rate (bpm) | 87 ± 23 |
| Serum sodium (mmol/L) | 136 ± 4.4 |
| Serum blood urea nitrogen (mg/dl) | 38 ± 23 |
| Serum creatinine (mg/dl) | 2 ± 1.6 |
| Glomerular filtration rate (ml/min/1.73m 2 ) | 42 ± 21 |
| LV assessment during hospitalization | 51% |
| LVEF | 43% ± 17.4 |
| LVEF ≤ 40% | 40% |
| LVEF < 50% | 55% |
| Hospital length of stay (days) | 7.6 ± 15.5 |
| Medication Classes (n=402) | |
|---|---|
| ACEI/ARB + BB + Diuretic | 155 (39%) |
| BB + Diuretic | 76 (19%) |
| Diuretic | 46 (11%) |
| ACEI/ARB + Diuretic | 44 (11%) |
| ACEI/ARB + BB | 42 (10%) |
| BB | 32 (8%) |
| ACEI/ARB | 27 (2%) |
Medication adherence as measured by PADC was suboptimal (<80%) and decreased after discharge for each medication class ( Figures 1 to 3 ). Overall, adherence at 1, 3, and 12 months was 70%, 61%, and 53% for ACEI/ARB, 76%, 66%, and 62% for BB, and 75%, 68%, and 59% for diuretics. Participants with a medication fill during the first 30-day period after discharge, PADC >0, demonstrated a higher adherence during subsequent periods. If the PADC was >0 in the first 30-day period, adherence at 1, 3, and 12 months was 86%, 71%, and 60% for ACEI/ARB; 88%, 73%, and 63% for BB; and 86%, 75%, and 64% for diuretics. Adherence decreased the most over the first 2 to 4 months after hospital discharge for all 3 medication classes. For participants who did not have a medication fill within 30 days of discharge (PADC = 0 in the first period), medication adherence was poor (i.e., <80% PADC). Adherence to ACEI/ARB and diuretics never exceeded 40% over the study period. Adherence to BB improved over time but remained low. We also calculated the proportion of participants with nonpersistence, defined as those who did not have a medication fill for ≥90 consecutive days and those with discontinuation defined as nonpersistence with no subsequent medication claims. At least 1 episode of nonpersistence was observed in 25% of participants prescribed an ACEI/ARB and 16% for both BB and diuretics. Discontinuation was documented in 21%, 12%, and 11% of participants prescribed an ACEI/ARB, BB, and diuretic, respectively.
We explored bivariate associations between adherence to individual drug classes over the first 3 months after hospitalization and participant characteristics listed in Table 1 . We found a limited number of associations that were significant at p ≤0.05 ( Table 3 ). Blacks in Jackson and Forsyth had lower adherence than whites for ACEI/ARB and BB. Relative to participants in other sites, participants in the Washington County, Maryland, were more adherent to ACEI/ARB and BB. Participants with lung disease had worse adherence to BB. Higher serum creatinine was associated with worse adherence to ACEI/ARB therapy. Adherence to any single drug class was positively associated with the likelihood of being adherent to the other discharge medication classes. Age, gender, other co-morbidities, hospitalization because of ADHF compared with chronic HF, assessment of LVEF during hospitalization, any documentation of LVEF, LVEF <50% or ≤40%, heart rate, or length of stay were not associated with medication adherence.
| Dichotomous Characteristics | ACEI/ARB | Beta Blockers | Diuretics | |||
|---|---|---|---|---|---|---|
| Percent with PADC≥80% | P value | Percent with PADC≥80% | P value | Percent with PADC≥80% | P value | |
| Race | ||||||
| Black (Jackson, Forsyth) | 51% | 0.05 | 50% | <0.01 | 62% | 0.58 |
| White | 63% | 69% | 65% | |||
| ARIC Site ∗ | ||||||
| Jackson | 49% | 0.02 | 49% | <0.01 | 60% | 0.26 |
| Forsyth County | 54% | 0.48 | 62% | 0.83 | 72% | 0.06 |
| Minneapolis | 64% | 0.34 | 71% | 0.11 | 51% | 0.34 |
| Washington County | 54% | 0.01 | 58% | 0.02 | 63% | 0.76 |
| Prior HF Hospitalization | ||||||
| Yes | 50% | 0.05 | 61% | 0.46 | 66% | 0.59 |
| No | 65% | 56% | 69% | |||
| Lung Disease | ||||||
| Yes | 56% | 0.71 | 52% | 0.01 | 65% | 0.64 |
| No | 59% | 67% | 62% | |||
| ACEIARB PADC≥80% | ||||||
| Yes | 75% | <0.01 | 79% | <0.01 | ||
| No | 38% | 43% | ||||
| Beta Blocker PADC≥80% | ||||||
| Yes | 74% | <0.01 | 74% | <0.01 | ||
| No | 37% | 48% | ||||
| Diuretics PADC≥80% | ||||||
| Yes | 71% | <0.01 | 72% | <0.01 | ||
| No | 33% | 46% | ||||
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
