Patient-reported outcomes and medication adherence in patients with heart failure

Abstract

Aims

Patient-reported outcome measures (PROMs) may predict poor clinical outcome in patients with heart failure (HF). It remains unclear whether PROMs are associated with subsequent adherence to HF medication. We aimed to determine whether health-related quality of life, anxiety, and depression were associated with long-term medication adherence in these patients.

Methods and results

A national cohort study of Danish patients with HF with 3-year follow-up (n = 1464). PROMs included the EuroQol five-dimensional, five-level questionnaire (EQ-5D-5L), the HeartQoL and the Hospital Anxiety and Depression Scale (HADS). Patient-reported outcomes (PRO) data were linked to demographic and clinical data at baseline, and data on all redeemed prescriptions for angiotensin-converting enzyme inhibitors/angiotensin II receptor blockers/angiotensin receptor neprilysin inhibitors (ACEI/ARB/ARNI), β-blockers, and mineralocorticoid receptor antagonists during follow-up. Medication non-adherence was defined as <80% of proportion of days covered. In adjusted regression analyses, lower health-related quality of life (EQ-5D and HeartQoL) and symptoms of depression (HADS-D) at discharge were associated with non-adherence. After 3 years of follow-up, lower health-related quality of life (EQ-5D) was associated with non-adherence for ACEI/ARB/ARNI [adjusted OR 2.78, 95% confidence interval (CI): 1.19–6.49], β-blockers (adjusted OR 2.35, 95% CI: 1.04–5.29), whereas HADS-D was associated with non-adherence for ACEI/ARB/ARNI (adjusted OR 1.07, 95% CI: 1.03–1.11) and β-blockers (adjusted OR 1.06, 95% CI: 1.02–1.10).

Conclusion

Lower health-related quality of life and symptoms of depression were associated with non-adherence across HF medications at 1 and 3 years of follow-up. Person-centred care using PROMs may carry a potential for identifying patients at increased risk of future medication non-adherence.

Introduction

Patients with heart failure (HF) have a poor prognosis^1,2 and are afflicted by a low quality of life including depressive symptoms, and frequent contacts with the healthcare system.^3,4 Evidence-based HF medication is associated with improved survival and adherence to medical treatment is thus essential to improve prognosis.² Non-adherence to HF medication may lead to exacerbations, decline in physical functioning, readmission, death, and high healthcare costs.^5,6

Medication adherence has been estimated to range between 37% and 71% in a large sample of healthcare beneficiary patients with HF.⁷

A number of patient characteristics have been reported to be associated with non-adherence to HF medication, including age, sex, low health status, low health literacy, and medication side-effects. Better tools are required to identify and target patients at high risk of non-adherence to HF medication.^8,9

Patient-reported outcome measures (PROMs) are questionnaires with self-reported information by patients regarding subjective health, such as health-related quality of life, symptoms of anxiety and depression or symptom burden. Patient-reported outcome (PRO) data may facilitate a more systematic person-centred approach to care.^4,10 A few small-scale studies have found depression and anxiety to be associated with medication non-adherence.^11,12 In addition, a systematic literature review on determinants of non-adherence to HF medication, identified depression as one of the several factors.⁹ Thus, using PROMs in the clinical setting may have the potential to identify patients at risk of lower adherence to medication, to target individualized care, and to ultimately lower the risk of adverse outcomes.

However, PROMs have not been investigated in relation to medication adherence in patients with HF using different domains of mental and physical health. We aimed to investigate the association between health-related quality of life, anxiety and depression, and long-term medication adherence among patients with HF.

Methods

Setting and design

This nationwide, cohort study was based on a population of Danish patients with HF. We used data from the DenHeart Survey, a collaborative study between the five heart centres in Denmark, collecting PRO data and ancillary information on lifestyle and general health in patients with cardiac disease.¹³ Healthcare in Denmark is tax-financed with free access for all citizens and out of hospital medicine expenses are partly reimbursed. The Danish Civil Registration System keeps records of vital status and habitation, using the unique civil registration number, given to each citizen at birth or immigration. This enables linkage of individual level data across all public registries.¹⁴

The investigation conforms with the principles outlined in the Declaration of Helsinki. Approval for this study was given by the DenHeart Steering Committee (DenHeart registered at ClinicalTrials.gov: NCT01926145) and the Danish Data Protection Agency (no: 2012-58-006). Patients provided written consent when filling in the questionnaire and in addition, the study was approved by the Danish Patient Safety Authority, authorizing access to information from medical records (no: 3-3013-1691).

Study population

Patients discharged from one of the five heart centres in Denmark between 15 April 2013 and 15 April 2014 with a HF diagnosis, either as a primary (main reason for hospitalization) or as a secondary discharge diagnosis (secondary reason for hospitalization) and participating in the DenHeart Survey were eligible for inclusion in this study (Supplementary material online, Table S1).

Patients were invited to fill in the DenHeart Survey, consisting of six PROMs at discharge, at transfer to another hospital or within 3 days after discharge.

Patients <18 years, not able to understand written or oral Danish, with no civil registration number, or unable to participate due to severe illness, such as unconscious or terminal illness were excluded.

Patient-reported outcomes measures

The PROMs included the EuroQol five-dimensional, five level questionnaire (EQ-5D-5L), the HeartQoL, and the Hospital Anxiety and Depression Scale (HADS). The EQ-5D-5L is a generic questionnaire measuring five domains of current health-related quality of life, where a higher index score indicates higher health-related quality of life.¹⁵ The HeartQoL is a disease-specific 14-item questionnaire. It measures health-related quality of life in patients with heart disease within the past 4 weeks. An emotional and a physical score provide a global score ranging from 0 to 3 points, and a higher score indicates a better state.^16,17 The HADS is a 14-item generic questionnaire measuring symptoms of anxiety and depression within the last week. It is summarized into a total score between 0 and 21 points in each of the two subscales, where a higher score indicates a possible mood disorder. The HADS is not a diagnostic tool and will be referred to as symptoms of anxiety (HADS-A) and symptoms of depression (HADS-D).^18–20

Outcome

Outcome was HF medication non-adherence after discharge from a cardiac-related hospitalization. Data on HF medication after discharge were retrieved from the Register of Medicinal Product Statistics, which holds information on all prescriptions redeemed at Danish pharmacies, including date of dispensing, strength, and package size.²¹

The first 90 days following discharge from index hospitalization were considered as a blanking period, to allow up-titration, breaks in therapy, or change of drug. In patients alive after Day 90, we traced all prescriptions for angiotensin-converting enzyme inhibitors (ACEI)/angiotensin II receptor blockers (ARB), β-blockers, and mineralocorticoid receptor antagonists (MRAs) using the Anatomical Therapeutic Chemical (ATC) coding system (Supplementary material online, Table S2). In 2016, the Angiotensin receptor neprilysin inhibitor (ARNI) was introduced.² For patients dispensing ACEI/ARB, substitution to ARNI during follow-up was allowed and substitution of a within-group pharmaceutical agent was also allowed. Combination drugs defined by medication with ≥2 active drugs were excluded.

Medication non-adherence to ACEI/ARB/ARNI, β-blockers, and MRAs was defined as having <80% of proportion of days covered (PDC)²² in 1 and 3 years of follow-up in patients surviving the blanking period, according to an individual gold standard. PDC was calculated as the percentage of time from end of blanking period and until end of keeping the individual gold standard within 1 and 3 years.

We defined the gold standard in each patient as follows: the first and second redeemed prescription after Day 90 in patients alive was traced and estimated the mean daily dose (MDD). The MDD was calculated as the number of pills dispensed at the first redemption divided with the number of days between redemption one and two, multiplied with the strength of the pills. This MDD was defined as the maximum tolerable dosage and the gold standard for each patient.

Adherence was also assessed as dispensed HF medication in patients alive across three time intervals: 3–6, 9–12, and 33–36 months after discharge from the index hospitalization. Here, we estimated the dispensing of zero, one, two, or three different drugs (ACEI/ARB/ARNI, β-blockers, and MRAs).

Patient characteristics

A range of demographic and clinical patient characteristics were a priori identified as potential confounding factors. The characteristics were identified based on extensive regression analyses of three PROMs in the DenHeart Survey, where all selected characteristics were associated with a worse score across PROMs. One question in relation to self-perceived level of social support from the DenHeart Survey was used and the response options were dichotomized. Demographics and data from the index hospitalization were retrieved from the Danish National Patient Registry (DNPR), including information on all contacts to the Danish healthcare system.²³

Length of hospital stay covered the total number of bed days at index hospitalization, including any days after transfer from a heart centre, and dichotomized at median (>2) days. Comorbidity was indexed according to the weighted Charlson comorbidity index (CCI) using the ICD-10 coding system (Supplementary material online, Table S3), based on every primary or secondary discharge diagnosis 10 years prior to the index hospitalization and categorized into: no comorbidity, moderate comorbidity level, and high comorbidity level.²⁴ A device-related procedure during index hospitalization included a pacemaker or an implantable converter defibrillator implantation or replacement. Information on HF-related variables was retrieved from medical records. Left ventricular ejection fraction was the last measured LVEF, and if missing at index hospitalization the last measured LVEF was used if referred to as unchanged. Systolic blood pressure (mmHg) was the last observed value during hospitalization.

Statistical analysis

After the blanking period of 90 days from index hospitalization, all patients were followed until end of keeping their individual gold standard, emigration, death, or end of follow-up after 3 years. Patients dying in the blanking period were excluded from further analyses.

The PROM scores were presented with mean and standard deviation (SD) and analysed on a continuous scale. As an additional analysis, the scores were also dichotomized by the worst quartile, except in HADS, where a cut-off at ≥8 points was used.¹⁸ The EQ-5D and the HeartQoL response scales were reversed in the regression analyses, thus a higher score indicated a worse state in all three questionnaires in the regression analyses.

To account for short breaks in therapy, 30 grace days between every two redeemed prescriptions were applied. We used multivariable logistic regression presented with OR estimates and 95% confidence intervals (CIs) and adjusted for the selected patient characteristics including age, sex, length of hospital stay, LVEF, comorbidity, systolic blood pressure, incident HF, device-related procedure, and self-reported social support.

In a secondary analysis, multinomial regression analysis was used to compute relative risk ratios (RRRs) and 95% CI of number of dispensed HF medications in three periods after discharge. Dispensing of three different drugs was considered as reference value.

Multiple chained imputations (Markov chain method) were performed to handle missing data (1.5–4.0%) and imputed on the following patient characteristics and PROMs: information on social support, LVEF, systolic blood pressure, the HeartQol, the EQ-5D, and the HADS. We imputed 50 datasets, using Rubin's Rule, under the assumption of data being missing at random.²⁵ All assumptions behind the applied statistical tests were tested before analysis.

We performed sensitivity analyses and tested for interaction prior to the regression analyses. In the sensitivity analyses, 14 grace days were applied to assess the robustness of our primary analysis and tested for interaction by age and sex. We also performed an additional analysis stratified according to the type of index HF diagnosis (primary or secondary) to assess medication adherence in patients with a primary HF diagnosis compared with patients with a secondary HF diagnosis at discharge from index hospitalization.

Analyses were performed using STATA version 14.0 (StataCorp).

Results

Participants

Of 3114 eligible patients with HF, a total of 1537 patients completed the questionnaire. We excluded a total of 73 patients due to congenital heart disease (n = 8) and acute life-threating conditions (n = 23) including cardiac arrest, ventricular fibrillation and acute thoracic surgery. Finally, patients dying during the 90-day blanking period after discharge from index hospitalization (n = 42) were excluded. A total of 1464 patients were included in the study (Figure 1).

In all, 74.1% of the patients were males and 81.3% had an LVEF ≤40%. A total of 61.7% of the patients had a moderate or a high comorbidity level and 37% of the patients had incident HF (Table 1). Table 2 shows the PROM scores at discharge.

Medication non-adherence

We found that 31.7% of the patients using ACEI/ARB/ARNI, 31.6% of the patients using β-blockers and 24.2% of the patients using MRAs were non-adherent at 1-year follow-up (Figure 2). After 3 years of follow-up, 68.7% of the patients were non-adherent in the use of ACEI/ARB/ARNI, 64.9% of the patients using β-blockers, and 66.1% of the patients using MRAs (Figure 2). The mean follow-up was 2.7 years (SD: 0.73) from entry in the study to death or end of follow-up after 3 years, and patients had a mean follow-up of 2.1 years (SD: 0.97) before not keeping their individual gold standard, regardless of drug.

Association between PROMs and medication non-adherence

Using multivariable logistic regression, we demonstrated associations between the PRO data at discharge from a cardiac-related hospitalization and subsequent medication adherence.

We found that patients reporting a lower health-related quality of life (EQ-5D), were more likely to be non-adherent in the use of MRAs (adjusted OR 3.49, 95% CI: 1.10–11.1) at 1-year follow-up (Table 3). Lower health-related quality of life was also associated with higher odds of being non-adherent after 3-year follow-up when using ACEI/ARB/ARNI and β-blockers (adjusted ORs 2.78, 95% CI: 1.19–6.49 and 2.35, 95% CI: 1.04–5.29, respectively). Patients reporting scores within the worst quartile of the EQ-5D were more likely to be non-adherent in β-blockers after 3-year follow-up (Table 4).

Lower cardiac health-related quality of life (HeartQoL) was associated with non-adherence in the use of β-blockers at 1-year follow-up, when analysing the HeartQoL on a continuous scale (adjusted OR 1.26, 95% CI: 1.06–1.49) and at 3-year follow-up when dichotomizing the HeartQoL (adjusted OR 1.37, 95% CI: 1.01–1.87) (Tables 3 and 4).

Finally, patients with a higher score on the HADS-D, indicating symptoms of depression, were more likely to be non-adherent at 1 year for all three drug classes: ACEI/ARB/ARNI (adjusted OR 1.04, 95% CI: 1.01–1.07), β-blockers (adjusted OR 1.05, 95% CI: 1.02–1.09), and MRAs (adjusted OR 1.06, 95% CI: 1.01–1.11) (Table 3). At 3-year follow-up, we also found an association between symptoms of depression (HADS-D) and non-adherence in patients using ACEI/ARB/ARNI (adjusted OR 1.07, 95% CI: 1.03–1.11) and β-blockers (adjusted OR 1.06, 95% CI: 1.02–1.10), but not in patients using MRAs (Table 4). When dichotomizing the PROM scores at 1- and 3-year follow-up, a score ≥8 points indicating symptoms of depression (HADS-D) remained associated with non-adherence in all estimates across drugs, except in the use of ACEI/ARB/ARNI at 1-year follow-up and MRAs at 3-year follow-up (Tables 3 and 4). Symptoms of anxiety according to the HADS did not show any association with non-adherence in any analyses.

We performed an additional analysis stratified according to the diagnosis given at discharge from index hospitalization (primary or secondary HF diagnosis) and found no systematically differences in the point estimates. Hence, the associations between PROs at discharge and subsequent medication adherence appeared to be comparable among patients with a primary and secondary HF diagnosis at index discharge (Supplementary material online, Tables S4 and S5).

Association between PROMs and dispensed HF medication

The association between PRO data at discharge from a cardiac hospitalization and number of subsequent dispensed HF medications were examined in multinomial adjusted regression analyses (Supplementary material online, Tables S6–S8).

A lower cardiac health-related quality of life (HeartQoL) score was associated with the dispensing of one drug compared with three drugs 9–12 months after discharge (adjusted RRR 1.32, 95% CI: 1.05–1.65). A score in the worst quartile of the HeartQoL was associated with a 1.5- to 1.9-fold increase risk of dispensing two or one drug compared with three drugs 3- 6 and 9–12 months after discharge (Supplementary material online, Tables S6 and S7).

Nine to twelve months after discharge, patients with a higher score on the HADS-D at discharge, indicating symptoms of depression, were more likely to dispense only one or zero drugs compared with three drugs (adjusted RRRs 1.05, 95% CI: 1.01–1.10 and 1.08, 95% CI: 1.03–1.14, respectively) (Supplementary material online, Table S7).

A score in the worst quartile on the respondent scale of the EQ-5D indicating lower health-related quality of life was associated with the dispensing of one drug compared with three drugs in 9–12 months after discharge (Supplementary material online, Table S7).

Symptoms of anxiety (HADS-A) were associated with a higher risk of only dispensing one drug compared with three drugs, 3–6 months after discharge (adjusted RRR 1.06, 95% CI: 1.01–1.10) (Supplementary material online, Table S6). When dichotomizing the PROM scores, having ≥8 points on the anxiety subscale of HADS, 9–12 months after discharge and ≥8 points on the depression subscale of HADS 33–36 months after discharge, was associated with a higher risk of dispensing no drugs compared with three drugs (Supplementary material online, Tables S7 and S8).

Finally, none of the PROMs analysed on a continuous scale were associated with dispensing of drugs in the last period of 33–36 months after discharge in adjusted analyses (Supplementary material online, Table S8).

Discussion

The results from this large nationwide study suggest that lower health-related quality of life and symptoms of depression at discharge from a cardiac hospitalization were associated with HF medication non-adherence after follow-up at 1 and 3 years.

PROMs and HF medication non-adherence

To our knowledge, this is one of the first nationwide cohort studies investigating the association between a combination of PROMs following discharge from a cardiac-related hospitalization, including the EQ-5D, the HeartQoL, and the HADS and subsequent medication non-adherence in patients with HF.

Medication adherence in HF has been widely investigated, but few studies have addressed the role of health-related quality of life or anxiety and depression. No studies of the EQ-5D or the HeartQoL in relation to medication adherence in patients with HF were identified. A systematic review from 2011 included 11 studies on determinants of medication adherence in HF and identified three studies presenting conflicting results regarding depression as a risk factor of medication non-adherence.⁹ None of the three studies used PROMs assessed by the HADS.^26–28 Two studies found a non-significant association between depression and medication non-adherence,^27,28 whereas one study found ‘carelessness’ from the patients about medication adherence in depressed patients.²⁶ The lack of evidence of a clear association in these studies may reflect small population sizes ranging from 51 to 134 patients,^26–28 or that medication adherence was assessed by self-reported measurements rather than more objective methods.^26,27

In the present study, we did not find any association between symptoms of anxiety and non-adherence. However, in the analyses of number of dispensed HF medications, symptoms of anxiety were associated with dispensing of only one drug in 3–6 months and no drugs in 9–12 months. These findings correspond to a randomized trial of a 12-week collaborative care intervention of health behaviours in 134 patients with cardiac disease, including HF. Patients were randomized to a care manager coordinating physiatrists recommendations or usual care and one of the items of assessed health behaviour was self-reported medication adherence.²⁹ Patients allocated to the intervention had significantly less symptoms of anxiety after 6 weeks and showed improvement in self-reported adherence. In contrast to our findings of dispensed medication, the association did not remain at 6-month follow-up.²⁹ The lack of association over time may be a result of a small sample size, short follow-up, only one item of adherence covered medication or study design. High-intensity intervention has been shown to lead to immediate improvements, but maintenance of achieved health behaviours over time may be more challenging.²⁷

In our study, we did not identify an association between symptoms of anxiety and non-adherence to HF medication. However, symptoms of depression were associated with non-adherence across HF medication in our analyses. A possible explanation could be how these symptoms affect the individual person. Key symptoms of depression might be characterized by impaired motivation and loss of initiative, potentially leading to risk of lower medication adherence as opposed to symptoms of anxiety, where patients might be more focused on adherence to healthcare recommendations, including medication.

Finally, we could have expected a lower adherence to MRAs, since this has previously been reported in other studies of medication adherence in HF.³⁰ However, due to the large proportion of patients with prevalent HF and the fact that 35% of the patients in this cohort received MRAs at discharge from index hospitalization, this might explain our finding in relation to MRAs.

Clinical implications

It is well established that adherence to evidence-based HF medication is pivotal when trying to reduce mortality and risk of adverse outcomes, and knowledge about modifiable factors in relation to medication adherence is essential.

Our findings support the utility of using PROMs in patients with HF. Use of PROMs does not only map the subjective health but appears also to provide valuable information on patients at risk of non-adherence to HF medication.

We demonstrated a consistent pattern of symptoms of lower health-related quality of life and symptoms of depression at discharge and subsequent risk of medication non-adherence. This underlines the potential value of implementing PROMs in routine care. Hence, screening patients for symptoms of depression and mapping health-related quality of life may enable targeted individualized efforts to assist vulnerable patients, which could well be an efficient way of optimizing and improving HF care in settings with limited resources.

Strengths and limitations

A major strength of this study was the combination of self-reported information, data from registries, and clinical information from medical records, ensuring detailed data on all patients and enabling thorough adjustment for confounding. The nationwide design enhanced the generalizability of results.

Data from Danish registries have a high completeness and are validated for epidemiological use.²³ Using data from the Register of Medicinal Product Statistics to assess medication non-adherence enabled us to track every redeemed prescription over time, with no risk of recall bias from patients. However, although repeated redemption of prescriptions of patient co-paid HF medication would indicate that the patients also took the medication, we have no objective proof of that.

Using a blanking period and beginning the register-based follow-up with data from the Register of Medicinal Product Statistics approximately 3 months after completing the questionnaire could have influenced the given answers over time. It is not straightforward to predict whether the time passed would reflect a better or worse score.

Moreover, the PROMs used were survey data at discharge and hence no repeated measurements of the PRO data were available. We did not account for hospitalizations during follow-up, although this might be associated with adherence, and a correlation between low adherence and increased risk of readmission most likely exists.

Non-response was 48%, but sensitivity analyses of demographics showed no differences between responders and non-responders, limiting the risk of selection bias. However, we were not able to account for clinically relevant factors with potential impact on the mental and physical health in the non-responders. Non-responders are more likely to suffer from severe illness in relation to their HF and the population included have a higher quality of life, leading to a potential underestimation of the association between mental and physical health and medication non-adherence. Furthermore, we were unable to distinguish between non-adherence and patients not prescribed the HF medication of interest in this study.

Conclusion

Lower health-related quality of life and symptoms of depression were associated with HF medication non-adherence over time. These findings were independent of other well-established prognostic factors. This knowledge stresses the potential of measuring health-related quality of life and symptoms of depression to target differentiated treatment and care and subsequently improve the prognosis in patients with HF.

Data availability

The data in this article cannot be shared publicly for ethical/privacy reasons. Due to Danish law in relation to data protection, all data from Danish health registries and medical records cannot be shared publicly.

Supplementary material

Supplementary material is available at European Heart Journal – Cardiovascular Pharmacotherapy online.

Acknowledgements

We wish to thank all the participants for taking time to complete the questionnaire, the staff at the participating departments for engagement in the project and Anne Vinggaard Christensen at Rigshospitalet—Copenhagen University Hospital for her practical support on retrieving data from health-registries and retrieving data from the DenHeart Survey.

Funding

This work was supported by The Danish Nurses’ Organization and the Danish Heart Foundation [grant number: 18-R124-A8454-22099].

Conflict of interest: S.P.J. has been paid as a consultant/speaker by Bayer, Bristol-Myers Squibb, Pfizer, and Sanofi and received research grants from Bristol-Myers Squibb and Pfizer (unrelated to the current study). None of the other authors have any conflicts to declare.

References