Abstract
Keywords
Introduction
The quality of the patient–provider relationship is an important factor in improvement of patient outcomes (Kaplan, Greenfield, & Ware, 1989; Roter & Hall, 2009; Von Korff, Gruman, Schaefer, Curry, & Wagner, 1997). Among older adults, the patient–physician relationship has also been identified as a potential barrier to medication adherence, and negative perceptions of physicians can influence a patient’s willingness to follow a physician’s advice and/or seek care (Beach, Keruly, & Moore, 2006; Bova et al., 2012). It has been reported that patients who perceive their physician as a poor communicator are at higher risk of nonadherence to treatment than patients whose physician is perceived to communicate well (Zolnierek & Dimatteo, 2009).
Literature examining patient perceptions of physician communication and its association with medication adherence is limited and the findings are mixed. Some studies have reported that poor physician communication about medication was associated with decreased likelihood of adhering to medications (Cohen, Rogers, Burke, & Beilin, 1998; Muntner et al., 2011; Schoenthaler, Allegrante, Chaplin, & Ogedegbe, 2012; Schoenthaler et al., 2009; Wilson et al., 2007), whereas others have reported no association between patient perceptions of physician communication and medication adherence (Daniels, René, & Daniels, 1994; Turner, Hollenbeak, Weiner, Ten Have, & Roberts, 2009). A patient’s perception of his or her physician’s knowledge and competence may influence his or her decision to adhere to medication therapy (Beach et al., 2006; Goff, Mazor, Meterko, Dodd, & Sabin, 2008; Ledford et al., 2010), and for patients with chronic disease (e.g., hypertension), the overall physician–patient relationship may have an impact on the patient’s response to appropriately taking medication regimens or other treatments after their experience with their physician (Diette & Rand, 2007; Ha, Anat, & Longnecker, 2010). Although the findings of prior studies are mixed, potential explanations for these differences include small sample size (Turner et al., 2009), short study follow-up periods for examining adherence (e.g., 30 days after an event; Muntner et al., 2011), and potential confounding factors (Cohen et al., 1998). The current study is strengthened by the use of a national sample of older adults, the ability to measure adherence for a longer period of time (up to 1 year), and the ability to control for various factors such as mortality risk, other existing health conditions, prescription insurance, and physician visits, all of which are important to address among patients with a chronic condition, such as hypertension.
Hypertension is a major contributor to morbidity and mortality due to cardiovascular disease (Fields et al., 2004; Hajjar & Kotchen, 2003), and approximately 67% of older adults aged 60 or older report having hypertension with highest rates reported among non-Whites and women (Aslam, Haque, Agostini, Wang, & Foody, 2010; Yoon, Burt, Louis, & Carroll, 2012). Among patients with such chronic diseases, the patient–provider relationship has been identified as a potential barrier to optimal treatment (Stewart, 1995).
Measures examining patient perception of physicians are limited, and no patient perceptions of physician measures developed from large claims databases with a nationally representative sample of older adults were identified. In addition, no prior studies were found that examined association between patient perceptions of physicians and medication adherence using data from a nationally representative sample of older adults and large administrative claims data. Therefore, the objective of this study was to assess association between patient perceptions of physicians and adherence to antihypertensive medication using data from the Medicare Current Beneficiary Survey (MCBS).
Method
Data Source
The MCBS is a nationally representative sample of Medicare beneficiaries maintained by the Center for Medicare and Medicaid Services (CMS). The 2007 MCBS survey items that asked about patient perceptions of physicians and the 2008 Medicare Part D claims for MCBS respondents were utilized for this study. The study was approved by the institutional review board at Purdue University.
Study Sample
Individuals were included in the study sample if they were 65 years or older in 2007, dwelling in the community, had a self-reported diagnosis of hypertension, were enrolled in Medicare Part D for 12 months in 2008, and had Medicare Part D claims for antihypertensive medication in 2008. Individuals were excluded if they had end-stage renal disease or disability, resided in a long-term care facility, had a diagnosis of Alzheimer’s disease or dementia, or had a proxy responder.
Medicare has some beneficiaries who are below 65 years old but have Medicare eligibility due to having disability or end-stage renal disease. Beneficiaries with end-stage renal disease or disability comprise of approximately 16% of the Medicare population (MedPAC, 2018), and were excluded from this study because beneficiaries with end-stage renal disease or disability are more likely to have assistance with taking their medication due to the severity of their condition. Also, beneficiaries residing in long-term care facilities were excluded because they are not likely to be administering their own medications while residing in the facility. Beneficiaries with Alzheimer’s disease/dementia were excluded due to a greater likelihood of inaccurate responses due to memory loss. Diagnosis of Alzheimer’s and/or dementia was determined by self-reported diagnosis from the following two questions included in the survey: (a) “Have you ever been told that you have Alzheimer’s” and (b) “Have you ever been diagnosed with dementia”? Finally, beneficiaries with proxy responders were excluded due to prior research indicating that proxies may not be useful in providing an accurate assessment of perceptions on behalf of patients (Blazeby, Williams, Alderson, & Farndon, 1995; Pierre, Wood-Dauphinee, Korner-Bitensky, Gayton, & Hanley, 1998; Slevin, Plant, Lynch, Drinkwater, & Gregory, 1988).
Hypertension diagnosis was identified through self-report on MCBS questions and through examination of claims data. Beneficiaries had to have both a self-reported diagnosis of hypertension and have prescription claims for antihypertensive medications. That was done in an effort to exclude persons without hypertension that may have been taking antihypertensive meditation for other indications, such as kidney protection in persons with diabetes. Beneficiaries were identified as having hypertension if they answered “yes” to at least one of two survey items: (a) “Has your doctor ever told you that you have/had hypertension, sometimes called high blood pressure?” and (b) “Since a year ago, did your doctor tell you that you still had hypertension or high blood pressure?” Antihypertensive medication use was identified from the Medicare Part D prescription claims, using First Databank Generic Name, for any antihypertension medication that was included in a list created, by the researchers, using Facts and Comparison eAnswers and Micromedex 2.0 (Facts and Comparison eAnswers, 2012; Micromedex 2.0, 2012). The list included 95 antihypertensive medications and combination therapies from 20 therapeutic drug classes with an indication to treat hypertension. A list of the medications are included in the supplemental material.
Measures
Patient Perception of Physicians Scale
The Patient Perception of Physicians Scale was constructed using items from the 2007 MCBS. The MCBS data have been collected since 1991, and there have been many studies utilizing these data to assess long-term care (Lee et al., 2016; Moyo, Huang, Simoni-Wastila, & Harrington, 2018; Shen, Zuckerman, Palmer, & Stuart, 2015), health care cost and utilization (Shen et al., 2015), disability (Ben-Shalom & Stapleton, 2016; Hennessy et al., 2015), Medicare plans (Henning-Smith, Casey, & Moscovice, 2017; Jacobs & Buntin, 2015), and patient satisfaction (Bogner et al., 2015). MCBS survey items include a variety of questions asking about sociodemographic characteristics, clinical characteristics, health status and functioning, access to care, health care resource use, and health care costs. Items selected for consideration to be included in the patient perceptions of physician scale were based on prior literature addressing the patient–provider relationship. Survey items that asked patients about the respondents’ perceptions of the physician from whom they regularly sought care were identified by the authors in the 2007 MCBS questionnaires and data. The identified items were utilized to construct the Patient Perception of Physicians Scale. Some examples of items identified included the following: “Your doctor answers all of your questions,” “Your doctor tells you all you want to know about your condition and treatment,” “Your doctor is competent and well trained,” “Your doctor is very careful to check everything when examining you,” and “You have great confidence in your doctor.” For each item, responses were coded so that 1 indicated strongly disagree, 2 indicated disagree, 3 indicated agree, and 4 indicated strongly agree. Items that were negatively worded, “Your doctor often does not explain your medical problems to you,” “You often have health problems that should be discussed but are not,” “Your doctor often acts as though he or she was doing you a favor by talking to you,” and “Your doctor seems to be in a hurry” were reverse coded so that “4” indicated strongly disagree, “3” indicated disagree, “2” indicated agree, and “1” indicated strongly agree. All values from the items were summed to create an overall Patient Perception of Physicians Scale score. Higher scale scores indicated more favorable perceptions of the physician. The Patient Perception of Physicians Scale was assessed for reliability to determine the accuracy and precision of scale, and was also assessed for convergent validity to determine whether similar constructs corresponded with one another. A list of the items included in the Patient Perception of Physicians Scale are included in the supplemental material.
Medication adherence (proportion of days covered [PDC])
Medication adherence with antihypertensive medications was measured using 2008 Medicare Part D administrative claims data. PDC was calculated as the number of days with medication on hand divided by the number of days in an interval (Benner et al., 2002; Martin et al., 2009; Nau, 2012; A. M. Peterson et al., 2007). The PDC calculation can have values between 0 and 1, and may be converted to a percentage. PDC was calculated for all antihypertensive medications taken by each beneficiary in 2008. If a medication refill was dispensed prior to the end of the previous medication being completed, it was assumed that the refill was not started until the previous medication was completed. For beneficiaries taking more than one antihypertensive medication throughout the entire study interval, an average PDC was calculated by calculating the PDC for each medication, summing each PDC, and dividing by the total number of antihypertensive medications. Calculations of PDC were adjusted for those who switched between antihypertensive medications, added an additional antihypertensive medication, or did both during the year. A weighted average was calculated for those who switched medications, added medications, or both. The study interval in which PDC was examined was the calendar year 2008.
Demographic and clinical variables
Demographic variables for each beneficiary were based on their status in December 2007 and included age, gender, race, education, income, and marital status. Clinical variables included number of medications, having private insurance, perceived health status, number of doctor visits, Medicare Part D prescription coverage gap, Charlson comorbidity index scores created from 2007 Medicare inpatient and outpatient claims, and 10 clinical conditions. The 10 clinical conditions were heart disease, heart failure, stroke, nonskin cancers, diabetes, rheumatoid arthritis, depression, osteoporosis, Parkinson’s disease, and respiratory diseases (emphysema, chronic obstructive pulmonary disease [COPD], or asthma). Presence or absence of each of the 10 clinical conditions was identified using self-reported responses to survey items in the 2007 MCBS data for the 10 clinical conditions. For each disease/condition, beneficiaries were asked: (a) whether they had ever been told by their doctor that they had a diagnosis of the specific disease/condition and (b) in the past year, had they been told by their doctor that they had a diagnosis of the specific disease/condition. If the beneficiary answered yes to either of the two questions, they were identified as having that disease/condition. These items were used to determine whether the patient had the clinical conditions prior to 2007. This resulted in 10 variables, one for each disease/condition, indicating whether a beneficiary had the disease/condition or not.
The Charlson comorbidity index is used to assess the likelihood of mortality for patients with comorbid conditions. In calculating the Charlson comorbidity index, various conditions receive different weights based on their adjusted risk of mortality. The sum of the weights results in a single comorbidity score for each patient. The higher the score, the more likely the predicted outcome would result in mortality. For this study, the Charlson comorbidity index scores were created from 2007 Medicare inpatient and outpatient claims using an algorithm by Romano and colleagues (Romano, Roos, & Jollis, 1993).
Statistical Analysis
Patient Perception of Physicians Scale
Principal component analysis with varimax rotation was conducted on 12 potential items for inclusion in the Patient Perception of Physicians Scale. Varimax rotation was used, so that each item would be associated with only one factor and provide simplicity in interpretation of results (Nunnally, 1978). Factors with eigenvalues greater than or equal to 1 were retained. In addition, the scree plot of eigenvalues plotted against the number of factors was examined.
Item-total correlations and change in Cronbach’s alpha if each item was deleted were used to determine item retention. If item-total correlation was less than .5 for an item, the item was deleted (Nunnally, 1978). Each factor was also considered to represent a subscale within the Patient Perception of Physician Scale, and subscale scores were calculated by summing the responses for items in each subscale. All retained patient perception of physician items were summed to create an overall Patient Perception of Physicians Scale. Cronbach’s alpha was used to assess internal consistency reliability of the scale and subscales. According to R. Peterson (1994), a Cronbach’s alpha level below .60 is considered unacceptable, .70 is low, .80 is considered moderate, and .90 is considered high; therefore, if Cronbach’s alpha was greater than .80, the scale was considered to have good internal consistency reliability (R. Peterson, 1994).
Convergent validity was evaluated using Spearman correlation to assess associations between the scale scores and selected convergent validity variables. Correlation coefficients less than .25 indicate little or no relationship, coefficients from .25 to .50 indicate a fair relationship, .50 to .75 indicate moderate to good, and above .75 indicate good to excellent relationship (Portney & Watkins, 2000). Four variables from the MCBS, private insurance, perceived health status, satisfaction with information provided, and satisfaction with doctor’s concern, were selected to evaluate convergent validity of the Patient Perception of Physicians Scale based on prior literature (DiMatteo & Hays, 1980; Rutten, Augustson, & Wanke, 2006; Wanzer, Booth-Butterfield, & Gruber, 2004). The Patient Perception of Physicians Scale was expected to have positive association with each of the selected variables examined for convergent validity.
Medication adherence
Means and standard deviations were calculated for PDC, and frequency tabulations were developed for medication adherence status. Beneficiaries were considered to be adherent if PDC was 80% or higher. If PDC was less than 80%, the beneficiary was considered to be nonadherent. A PDC of 80% has been used in prior outcomes research (Hansen et al., 2009; Wu et al., 2009; Zedler, Joyce, Murrelle, Kakad, & Harpe, 2011), and is the cutoff selected by the Pharmacy Quality Alliance as the commonly used threshold for performance measures for most classes of drugs used to treat chronic diseases, such as antihypertensive medications (Nau, 2012).
Association between patient perception and medication adherence
Multiple logistic regression using stepwise selection was used to assess association between scores on the Patient Perception of Physicians Scale and likelihood of being adherent to antihypertensive medication. All the demographic and clinical characteristics listed above were all candidates for inclusion as covariates in the logistic regression model using stepwise selection.
SAS® version 9.3 for the Unix environment was used for data analysis. An a priori alpha level of .05 was used to evaluate significance for all analyses.
Results
Sample
After applying inclusion and exclusion criteria, the study sample included 2,510 beneficiaries. Figure 1 shows the sample selection results. Table 1 summarizes the sample demographic characteristics. Mean age of individuals in sample was 76.4 years with a standard deviation of 6.88 years. The majority of the sample was female (65%), White (83%), and married (53%); had a high school education or less (63%); and had income of US$30,000 or less per year (62%).
Sample selection.
Sample Characteristics.
Non-White includes African American, Asian, Native Hawaiian or Pacific Islander, American Indian or Alaska Native, Other race, and more than one race.
Not married includes never married, widowed, divorced, or separated.
Number of individuals missing response on education was 11 or less and values were removed as required by data use agreement.
Number of individuals missing income did not provide an income level in an above listed category. Income was reported as less than US$25,000, US$25,000 or more, or did not know.
Patient Perception of Physicians Scale
There were 262 beneficiaries missing responses to items utilized in the development of the Patient Perception of Physician Scale, resulting in a sample of 2,248 beneficiaries. Principal component analysis with varimax rotation yielded two factors; so, the “Patient Perception of Physicians Scale” had a total of 12 items, and consisted of two factors (subscales). Factor 1 had eight items, was primarily a measure of perceived physician knowledge, and was named the “Perceived Physician Knowledge” subscale. Factor 2 had four items, was primarily a measure of perceived communication, and was named the “Perceived Concern” subscale. The proportion of total variance explained by the Patient Perception of Physicians Scale was 67%. The Cronbach’s alpha coefficient for the Patient Perception of Physicians Scale was .92, indicating good reliability. The overall Patient Perception of Physician Scale and each of the two subscales had Cronbach’s alpha coefficients greater than .80. After assessing convergent validity, the Patient Perception of Physicians Scale had positive correlations with “private insurance,” “perceived health status,” “satisfaction with information provided about your health,” and “satisfaction with doctor’s concern about your health” as hypothesized. The highest correlations were with satisfaction with information provided about health (r = .28, p ⩽ .0001) and satisfaction with their doctor’s concern for their health (r = .31, p ⩽ .0001). The theoretical range for the Patient Perception of Physicians Scale is 12 to 48 with higher scores representing more favorable perceptions of their physicians. The actual range for the sample was 15 to 48 with a mean of 38.79 (SD = 5.013). See Table 2.
Mean, Standard Deviation, Cronbach’s Alpha Reliability, and Item-Total Correlation Range for the Patient Perception of Physicians Scale, Perceived Physician Knowledge Subscale (Factor 1), and Perceived Concern Subscale (Factor 2).
PDC
There were 59 beneficiaries who had only one prescription for unique antihypertensive medications in 2008, which probably represented trials of medication and, therefore, did not permit calculation of PDC. Those 59 individuals were dropped from the analysis, and this resulted in a sample of 2,189 beneficiaries.
PDC values ranged from 0.13 to 1. The mean PDC was 0.82 with a standard deviation of 0.236. PDC was converted to a binary variable to indicate whether a beneficiary was adherent or not. If PDC was 0.80 (i.e., 80%) or higher, the beneficiary was considered to be adherent. If PDC was less than 0.80 (i.e., 80%), the beneficiary was considered nonadherent. Based on that classification only, 65% of the sample was adherent in filling their antihypertensive medication.
Association Between Patient Perception of Physicians and Medication Adherence
A multiple logistic regression model using stepwise selection was developed to assess association between the patient perception of physicians and medication adherence to antihypertensive medication. The sample size for the regression model was reduced from 2,189 beneficiaries to 1,935 beneficiaries after an additional 254 beneficiaries were excluded from the multiple logistic regression model due to missing data on education and/or income.
The response variable in the model was medication adherence status, with “1” indicating adherence and with “0” indicating nonadherence based on the 80% cutoff as described above. The predictor variable, in the initial model was the Patient Perception of Physicians Scale as a continuous variable and did not enter the regression model, indicating no significant association between patient perception of physicians and adherence to antihypertensive medications. The significant covariates included age, respiratory conditions, nonskin cancers, and number of unique medications (results not reported here).
In an effort to categorize the patient perceptions as more favorable and less favorable, the predictor variable, patient perception of physicians score, was examined as a categorical variable in the logistic regression model due to the skewness of the variable. The frequency distribution of the Patient Perception of Physicians Scale scores was evaluated to determine a cutoff point for more favorable perceptions versus less favorable perceptions. With scores ranging from 15 to 48, the median was 37. Logistic regression with the receiver operator characteristics (ROC) option was used to determine a cutoff for the perception scores. The cutoff identified from the ROC curve was 37, with scores of 37 or higher representing more favorable perceptions and scores less than 37 representing less favorable perceptions. Sensitivity analysis was conducted by using various cutoffs ranging from 36 to 38 to determine the robustness of the model using the cutoff selected. The sensitivity analysis indicated that the cutoff point from the ROC curve was acceptable, which was the point at which sensitivity and specificity were maximal. Therefore, the ROC curve cutoff of 37 was utilized to distinguish between more favorable perceptions and less favorable perceptions.
The logistic regression model adjusted for demographic and clinical characteristics and all two-way interactions. The following variables remained in the model after stepwise selection: patient perception of physicians, age, having a respiratory condition, having a nonskin cancer, and number of medications. Beneficiaries with patient perception of physician scale scores of 37 or higher were more likely to be adherent to antihypertensive medications than beneficiaries with scores less than 37, odds ratio (OR) = 1.341, 95% confidence interval (CI) = [1.101, 1.632], p = .0035. Other significant variables in the model were age, having a respiratory condition, having a nonskin cancer, and number of medications. Beneficiaries with a respiratory condition were more likely to be adherent in comparison with beneficiaries who did not have a respiratory condition (OR = 1.356, 95% CI = [1.037, 1.772], p = .0258), and beneficiaries who had a nonskin cancer were more likely to be adherent than beneficiaries who did not have a nonskin cancer (OR = 1.377, 95% CI = [1.064, 1.783], p = .0152). Beneficiaries aged 80 to 84 years old were less likely to be adherent to antihypertensive medications in comparison with beneficiaries aged 65 to 69 years (OR = 0.673, 95% CI = [0.497, 0.911], p = .0103). Beneficiaries who filled 21 or more unique medications during the study period were less likely to be adherent to antihypertensive medications in comparison with beneficiaries who filled one to five unique medications during the study period (OR = 0.111, 95% CI = [0.070, 0.175], p < .0001). Table 3 presents the results of the multiple logistic regression model.
Multiple Logistic Regression Examining Association Between the Overall Patient Perception of Physicians Scale and Medication Adherence.
The following variables were eligible to enter regression model: patient perception of physician scale, age, gender, race, education, income, marital status, private insurance, Medicaid, reaching the Medicare Part D prescription coverage gap, perceived health status, number of unique medications, number of doctor visits, Charlson comorbidity index, and clinical conditions (heart disease, heart failure, diabetes, stroke, nonskin cancers, rheumatoid arthritis, depression, Parkinson’s disease, osteoporosis, and respiratory conditions).
p ⩽ .05 indicates significance.
Respiratory conditions include emphysema, chronic obstructive pulmonary disease, and asthma.
Discussion
In prior literature, reported rates of nonadherence to antihypertensive medications have ranged from 25% to 61% (Shaya et al., 2009; Tong, Chu, Fang, Wall, & Ayala, 2016). Among older adults with hypertension, adherence rates have ranged from 44% to 63% (Dickson & Plauschinat, 2008; Lo, Chau, Woo, Thompson, & Choi, 2016; Monane et al., 1996). We found 65% of Medicare beneficiaries enrolled in a Medicare Part D plan were adherent to their antihypertensive medications. Monane and colleagues (1996) examined medication adherence to antihypertensive medication among older adults in the New Jersey Medicaid program in the 1980s, and reported that only 49% were adherent. The adherence rate, in the current study, 65%, was slightly higher than the 63% reported by Dickson and Plauschinat (2008), for older adults taking antihypertensive medications. Whereas our study utilized the MCBS and Medicare claims data, the Dickson and Plauschinat (2008) study utilized South Carolina Medicaid administrative claims data.
A 12-item Patient Perception of Physicians Scale, consisting of two subscales that assessed perceived physician knowledge and perceived concern, was constructed from MCBS items. The scale showed evidence of internal consistency reliability and convergent validity. In comparison with the DiMatteo and Hays’ (1980) and Safran et al.’s (1998) scales, the Patient Perception of Physicians Scale created from MCBS items has fewer items and fewer subscales. However, this scale had three more items than the patient physician relationship index developed by Ostacoli et al. (2007). Overall, our Patient Perception of Physicians Scale had a Cronbach’s alpha of .92. Internal consistency reliability of the Patient Perception of Physicians Scale was consistent with that of other scales, which had Cronbach’s alphas ranging from .92 to .95 (DiMatteo & Hays, 1980; Safran et al., 1998), except the scale developed by Ostacoli and colleagues (2007), which had a Cronbach’s alpha of .81. In addition, this scale showed some evidence of convergent validity, having positive correlations with having private insurance, perceived health status, patient satisfaction with doctor’s concern, and patient satisfaction with information provided (DiMatteo & Hays, 1980; Rutten et al., 2006; Wanzer et al., 2004).
In examining association between patient perception of physicians and adherence to antihypertensive medications, beneficiaries with more favorable perceptions of their physicians (scores 37 or higher) were more likely to be adherent to antihypertensive medications than beneficiaries with less favorable perceptions (scores less than 37). Prior work by Wilson and colleagues (2007) examined the prevalence of the physician–patient dialogue about medication adherence among Medicare beneficiaries. They reported that more than 30% of Medicare beneficiaries did not have a discussion with their doctor about their medications in the last 12 months. Among those who skipped doses or stopped taking a medication due to side effects, more than one quarter of them did not discuss these changes with their physician. Wilson and colleagues (2007) concluded there was a need for better physician–patient communication, particularly among older adults (65 years or older), because many of them had multiple chronic diseases and were taking multiple prescription medications. In a qualitative study examining patient barriers to medication adherence, Polinski and colleagues (2014) reported that having “a trusting patient-provider relationship, shared decision-making support, full disclosure of side effects and cost sensitivity were attributes that might enhance primary adherence” (p. 755). In addition, the significant result from the logistic regression and the sensitivity analysis may indicate there may be a threshold effect, that is, that providers may need to achieve some minimum level of favorable perception from their patients to have a positive influence on proportions of medication adherence. Our findings provide some quantitative evidence that patient perceptions of their physician indeed have some association with patients adhering to medication therapy.
Monane and colleagues (1996) reported that older adults, aged 85 and older, were more likely to be adherent to their antihypertensive medications in comparison with younger seniors. However, we found that older adults, aged 80 to 84 years, were less likely to be adherent to their antihypertensive medications in comparison with younger seniors (65-69 years). Shaya and colleagues (2009) reported that antihypertensive patients with a lower Charlson comorbidity index score were more likely to be adherent in comparison with those with a higher Charlson comorbidity index score. Our findings indicated that patients with respiratory conditions and nonskin cancers were more likely to be adherent to their medication. This may be due to the severity of long-term respiratory conditions such as asthma and COPD, as well as the importance of consistently taking cancer medications when being treated.
Limitations
There are several limitations to this study. The study was limited to the use of Medicare Part D claims to estimate medication adherence due to full claim information being only available for Part D prescriptions. In calculating PDC from the administrative claims data, if a patient refilled the medication prior to the end of the previous medication being completed, we assumed that the refill was not started until the previous obtained medication was completed. Also, if patients were taking multiple antihypertensive medications, an average PDC was calculated. Although it is reasonable to assume that the medication refill was not started until the previous obtained medication was completed, it is still important to mention this as a study limitation with regard to our assessment of medication adherence.
The sample was limited to persons with hypertension because more than half of older adults have been diagnosed with this chronic disease. However, we have no a priori reason to think that the beneficiaries in the sample would differ greatly from other MCBS beneficiaries on patient perceptions of physicians.
Data utilized for this study were the 2007 and 2008 MCBS data. Although these data may be older, they were the most current data available to us when conducting the research. We believe that the findings are still informative as they relates to the patient–provider relationship and adherence to medications among older adults.
The Patient Perception of Physicians Scale was derived from self-reported responses from the beneficiaries up to 1 year prior to the administrative claims that were assessed for medication adherence. Although there is no external reference, such as patient interviews, for the Patient Perception of Physicians Scale, it was assessed for reliability and construct validity. Also, the time frame for which the data on patient perceptions of the provider were collected and the antihypertensive medication claims assessed were over a 2-year period.
Finally, we are unable to determine whether the prescriber of the patient’s antihypertensive medication was the primary provider for which the patient provided his or her self-reported perception responses. However, it is our assumption that this was likely to be true in the majority of cases. However, if that was not the case for a significant number of cases, it would be expected to reduce correlation between the perception measure and adherence, and as such, likely make the findings regarding the magnitude of the relationship between perceptions and adherence conservative.
Conclusion
Older adults’ perceptions of their physician may influence their adherence to antihypertensive medications. In this study, the more favorable the perception of the provider, the more adherent older adults were to their antihypertensive medications. To our knowledge, this is the first study to use a nationally representative sample of older adults to assess patient perceptions and medication adherence to antihypertensive medication among older adults. This work provides evidence of the importance of the patient–provider relationship and its potential influence on medication adherence.
Supplemental Material
Supplemental_Material_1 – Supplemental material for Patient Perception of Physicians and Medication Adherence Among Older Adults With Hypertension
Supplemental material, Supplemental_Material_1 for Patient Perception of Physicians and Medication Adherence Among Older Adults With Hypertension by Lori M. Ward and Joseph Thomas in Journal of Aging and Health
Footnotes
Authors’ Note
Ward and Thomas have conceptualized the project, and acquired and analyzed the data. Ward and Thomas also drafted, revised, and approved the final version of this article.
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.
Supplemental Material
Supplemental material is available for this article online.
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
