Abstract
Background
Medication adherence, or the extent to which patients take prescribed medications, is an important determinant of successful medical treatment.1,2 One meta-analysis found an average treatment adherence rate of 75.2%, with rates ranging from 4.6% to 100% across studies. 3 Nonadherence to clinically indicated medications can result in high costs to the health care system1,2 and increase subsequent emergency department visits and hospitalizations.4-7
Given the consequences of poor adherence, it is important to determine the prevalence of and reasons for self-reported nonadherence in marginalized populations that already experience poor health outcomes. In Canada, people who reported poorer health, lower incomes, and no prescription drug coverage were significantly more likely to report nonadherence for cost-related reasons. 8 Several Canadian provinces include a premium or deductible in their drug benefit programs, 9 which could result in poor medication adherence and an increase in hospital-based service utilization. 10 For individuals who are homeless or vulnerably housed, additional structural barriers such as lack of privacy, lack of a place to store medications, and lost or stolen medications may reduce adherence. 11 In a study of 966 homeless single adults in the United States, 36% of participants reported an unmet need for prescription medications. 12 A study of 368 homeless adults in Toronto, Canada, found that 32% of respondents were unable to obtain their prescription medications, with 63% of these individuals attributing this to a lack of drug benefits. 13
This study examines the prevalence of and factors associated with self-reported nonadherence to medications among homeless and vulnerably housed adults in 3 large Canadian cities. This study also determines the association between self-reported medication nonadherence and frequent emergency department visits during a 1-year follow-up period. An important feature of this study is the inclusion of individuals who are vulnerably housed and at high risk of becoming homeless, a group that has been less well studied than homeless individuals.
Methods
Data were obtained from the Health and Housing in Transition (HHiT) study, a prospective cohort study of homeless and vulnerably housed single adults in Vancouver, Toronto, and Ottawa. The design of the study has been described previously. 14
Study Setting
This study was conducted in the provinces of Ontario and British Columbia, both of which have publicly funded universal health insurance systems that provide coverage for hospital and physician services and drug benefit programs for residents who are receiving welfare or disability benefits. 9 Drug benefit coverage is limited to drugs listed on provincial formularies.15,16 Additional programs sometimes provide additional coverage for certain individuals whose prescription drug costs are high relative to their income, but these programs include a deductible and copayments.16,17
Participants and Recruitment
Between January and December 2009, 1191 single adults (18 years or older) were enrolled, of whom 595 were homeless and 596 were vulnerably housed. Homelessness was defined as living within the last 7 days at a shelter, public place, abandoned building, vehicle, or someone else’s place, and not having one’s own place. Recruitment of homeless adults occurred at shelters and meal programs using sampling methods designed for this population. 18
Individuals were considered to be vulnerably housed if they were living in their own room or apartment, but had been homeless or had 2 or more moves in the past 12 months. Vulnerably housed individuals were recruited at rooming houses, single room occupancy hotels, and meal programs. All study participants provided informed consent and were reimbursed $20 CDN for each interview. This study was approved by the Research Ethics Board at St Michael’s Hospital, the University of Ottawa, and the University of British Columbia.
Survey Instrument
Data were obtained through 60- to 90-minute structured in-person interviews conducted at the time of recruitment and at a follow-up interview approximately 12 months later. The survey instrument obtained information on demographic characteristics, housing status, health status and health conditions, quality of life, health care utilization, substance use, and risk behaviors.19,20 Past or present mental health diagnoses were based on self-report. Eligibility for a drug benefit program was determined on the basis of receiving welfare or disability benefits.
Alcohol abuse was assessed using the Alcohol Use Disorders Identification Test (AUDIT). A positive screen is defined as a score of ≥8 and is indicative of hazardous or harmful drinking. 21 Drug abuse and drug-related problems were assessed using the Drug Abuse Screening Test (DAST-10). 22 A positive screen is defined as a score of ≥3 and is indicative of moderate, substantial, or severe drug use problems.
The main outcome variable for our analysis was self-reported nonadherence to medication, assessed using the questions “Are you currently supposed to be taking medication that was prescribed by a doctor?” and “Are you actually taking this medication?” 23 Nonadherence was defined as responding “yes” to the former question and “no” to the latter question. Reasons for nonadherence were identified by asking “If you are NOT taking the medication prescribed to you, why not?” Response options included the following: the medication is too expensive; difficulties storing medication; unable to take the medication as recommended; disliking side effects; not believing in taking medication; and other reasons. Other reasons were categorized by 2 independent reviewers.
The second outcome measure used in this analysis was self-reported emergency department visits. Participants were re-interviewed approximately one year after the baseline interview and asked “In the past 12 months, have you gotten health care from a hospital emergency room?” and if so, “How many times in the past 12 months?” Frequent emergency department utilization was defined as ≥3 visits in the past year.
Statistical Analysis
Chi-square test and Fisher’s exact test were used to compare the characteristics of participants who reported medication nonadherence with those who reported adherence. Logistic regression was used to identify factors associated with medication nonadherence, as well as to evaluate whether there was an association between medication nonadherence at the baseline interview and frequent emergency department use at follow-up.
Multivariate models were constructed using a purposeful selection technique. 24 After generating univariate logistic regression models for each covariate against each outcome of interest, variables that met the inclusion criteria (P ≤ .25) were added into a multivariate model. Variables were then removed from the model if they did not meet inclusion criteria for retention (P ≤ .10) and removal of the variable did not alter the parameter estimate of any remaining variable by ≥15%. After all parameters were tested iteratively, variables that were not significant in the univariate models were added back in one at a time to determine if they were significant in the final model. Models were tested for best fit using the Hosmer–Lemeshow test. All analyses were performed using SAS 9.3 software (SAS Institute Inc, Cary, NC).
Results
Of 1191 study participants, 4 were excluded because they did not indicate if they were prescribed a medication. A total of 716 (60%) participants reported that they were supposed to be taking a prescribed medication, and 189 (26% of those prescribed a medication) reported that they were not taking the medication. One-year follow-up data were available for 970 (82.6%) participants.
Table 1 compares the characteristics of participants who reported medication nonadherence and adherence. Participants reporting nonadherence were significantly more likely to be younger than 40 years, to have been employed in the past 12 months, and to have a higher AUDIT risk level. They were significantly less likely to have a primary care provider.
Characteristics of Participants at Baseline.
Abbreviations: DAST, Drug Abuse Screening Test; AUDIT, Alcohol Use Disorders Identification Test.
Table 2 displays the results of the bivariate and multivariate logistic regression models. In the multivariate model, being 40 years or older was associated with decreased likelihood of nonadherence (adjusted odds ratio [AOR] = 0.59; 95% confidence interval [CI] = 0.41-0.84), as was having a primary care provider (AOR = 0.49; 95% CI = 0.34-0.71). A positive AUDIT screen was associated with increased likelihood of nonadherence (AOR = 1.86; 95% CI = 1.31-2.63). Table 3 summarizes participants’ reasons for not taking a prescribed medication. The most common reasons included medication side effects, the cost of medications, and lack of access to a physician.
Factors Associated With Nonadherence Among Homeless and Vulnerably Housed Participants Who Were Prescribed a Medication by a Physician.
Abbreviations: OR, odds ratio; AOR, adjusted odds ratio; CI, confidence interval; DAST, Drug Abuse Screening Test; AUDIT, Alcohol Use Disorders Identification Test.
Reasons Given for Nonadherence to Prescribed Medications, Among 177 a Participants Reporting Nonadherence at Baseline.
Of the 189 participants reporting nonadherence at baseline, 177 provided a reason for not taking prescribed medications. A total of 191 reasons were given, as participants were able to provide multiple reasons.
Of the 970 participants who were reinterviewed 1 year after enrollment, 175 (18%) reported ≥3 visits to the emergency department during the 1-year follow-up period. Self-reported nonadherence at baseline was associated with frequent visits to the emergency department at the bivariate level (OR = 1.55; 95% CI = 1.02-2.35). However, this relationship did not retain statistical significance (AOR = 1.49; 95% CI = 0.96-2.32) in a multivariate model that included the following variables: housing status, mental health diagnoses, having a positive AUDIT screen, and being a victim of physical violence in the past 12 months.
Discussion
Of the 716 participants who were prescribed a medication, 26% reported nonadherence. This rate is almost identical to the average rate of nonadherence in a meta-analysis of studies conducted in a broad range of patient populations. 3 The level of medication nonadherence in our study did not vary significantly by housing status. In a multivariate model, only age, positive AUDIT screen, and access to a primary care provider were significantly associated with medication nonadherence. This association between nonadherence and younger age is consistent with other studies, including a 7-country comparison of medicine underuse due to cost 25 and a US study examining adherence to antiretroviral therapy among homeless or unstably housed adults living with HIV/AIDS. 26 Clinicians should therefore be particularly alert to nonadherence among younger adults who are homeless or vulnerably housed. While less than 3% of our nonadherent participants attributed their nonadherence to substance use, a positive screen for hazardous or harmful drinking was significantly associated with nonadherence, highlighting the importance of recognizing this factor as contributing to nonadherence.
In our multivariate model, having a primary care provider was associated with a significantly lower likelihood of nonadherence. This finding likely reflects both better access to a prescriber and support from a primary care provider that may enable individuals to achieve higher adherence. Previous studies have shown that patient-centered primary care, collaborative physician–patient interactions, and patient and provider symmetry in health locus of control beliefs are associated with better adherence.27-29 In contrast, homeless and vulnerably housed individuals who do not have a primary care provider may obtain care in settings such as emergency departments or walk-in clinics that are not structured to encourage long-term medication adherence.
Participants provided a variety of reasons for not taking prescribed medications. Avoiding unpleasant side effects or denying the severity of a health problem are reasons commonly cited in the general population.1,2 However, other reasons may indicate barriers related to housing status, such as not having regular access to a doctor, lacking a safe place to store medications, or having competing priorities. Previous studies have found that unstable housing and homelessness are associated with antiretroviral medication nonadherence among HIV-infected persons.30-33
A statistically significant association between nonadherence and subsequent self-reported frequent emergency department utilization was observed in univariate but not multivariate models. We speculate that this finding may be due to a lack of adequate power rather than the absence of a true association. Future studies should correlate medication adherence among homeless and vulnerably housed individuals with administrative health data, as previous studies have demonstrated a relationship between medication nonadherence and subsequent health care utilization.4-6
This study has certain limitations. Medication adherence was measured using self-report. While other methods of assessing adherence have been used in clinical studies, these methods are difficult to apply to large population-based samples. Another limitation was a lack of information on the specific medications for which participants were adherent or nonadherent. Data on emergency department visits were also based on self-report. Finally, the generalizability of our findings to other health care systems is uncertain.
Conclusion
Homeless and vulnerably housed adults face significant barriers to achieving medication adherence. Some of these barriers may be alleviated by having safe, secure, and stable housing. Our findings confirm the importance of having a source of regular primary care to enhance medication adherence. In addition, health care providers should be attuned to the increased risk of nonadherence among younger homeless and vulnerably housed adults and those with a positive AUDIT screen.
Footnotes
Acknowledgements
We would like to acknowledge the following individuals from our community partner organizations: Laura Cowan, Liz Evans, Stephanie Gee, Clare Hacksel, Erika Khandor, and Wendy Muckle. The authors also thank the study coordinators and interviewers in each of the 3 cities as well as the shelter, drop-in, and municipal and provincial staff for their assistance with participant recruitment and follow-up.
Authors’ Note
The views expressed here are the views of the authors and do not necessarily reflect the views of any of the funding organizations.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Funding for the Health and Housing in Transition (HHiT) study was received from an operating grant (MOP-86765) and an Interdisciplinary Capacity Enhancement Grant on Homelessness, Housing and Health (HOA-80066) from the Canadian Institutes of Health Research. The Centre for Research on Inner City Health in the Li Ka Shing Knowledge Institute at St. Michael’s Hospital gratefully acknowledges the support of the Ontario Ministry of Health and Long-Term Care.