Persistence with antihypertensive medication in the first 6 months among 13 205 incident patients newly starting treatment.
Reduction in relative risk of nonpersistence in the first 6 months by medical management (clinical decision-making) and communication (data collection: history and physical examination skills) ability. Standardized examination scores were converted from a standard mean of 500 to the raw score mean of 66.8% overall for all examination administrations and the mean standard deviation of 5.73. Adjusted odds ratios (Table 4) were used to plot the change in relative risk of nonpersistence for each 1-SD change in examination score within the score range observed in the study population, from the lowest score category (−3 SDs) to the highest (+3 SDs).
Tamblyn R, Abrahamowicz M, Dauphinee D, Wenghofer E, Jacques A, Klass D, Smee S, Eguale T, Winslade N, Girard N, Bartman I, Buckeridge DL, Hanley JA. Influence of Physicians' Management and Communication Ability on Patients' Persistence With Antihypertensive Medication. Arch Intern Med. 2010;170(12):1064-1072. doi:10.1001/archinternmed.2010.167
Copyright 2010 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2010
Less than 75% of people prescribed antihypertensive medication are still using treatment after 6 months. Physicians determine treatment, educate patients, manage side effects, and influence patient knowledge and motivation. Although physician communication ability likely influences persistence, little is known about the importance of medical management skills, even though these abilities can be enhanced through educational and practice interventions. The purpose of this study was to determine whether a physician's medical management and communication ability influence persistence with antihypertensive treatment.
This was a population-based study of 13 205 hypertensive patients who started antihypertensive medication prescribed by a cohort of 645 physicians entering practice in Quebec, Canada, between 1993 and 2007. Medical Council of Canada licensing examination scores were used to assess medical management and communication ability. Population-based prescription and medical services databases were used to assess starting therapy, treatment changes, comorbidity, and persistence with antihypertensive treatment in the first 6 months.
Within 6 months after starting treatment, 2926 patients (22.2%) had discontinued all antihypertensive medication. The risk of nonpersistence was reduced for patients who were treated by physicians with better medical management (odds ratio per 2-SD increase in score, 0.74; 95% confidence interval, 0.63-0.87) and communication (0.88; 0.78-1.00) ability and with early therapy changes (odds ratio, 0.45; 95% confidence interval, 0.37-0.54), more follow-up visits, and nondiuretics as the initial choice of therapy. Medical management ability was responsible for preventing 15.8% (95% confidence interval, 7.5%-23.3%) of nonpersistence.
Better clinical decision-making and data collection skills and early modifications in therapy improve persistence with antihypertensive therapy.
Effective management of chronic disease is a major challenge in health care.1 Among chronic conditions, cardiovascular disease is a leading cause of morbidity and mortality.2 Drug treatment can substantially reduce morbidity,3 but long-term therapy adherence is required to realize these benefits.4 Population-based studies of treatment persistence consistently show disappointing results, particularly for hypertension treatment.5- 9 Less than 75% of people prescribed drugs for hypertension are still using them after 6 months, and nonpersistence is associated with an increased risk of hospitalization for cardiovascular problems.5,6,8- 12
Few potentially modifiable determinants of persistence have emerged despite decades of research.4,6,10,13,14 Consequently, efforts to improve persistence target a variety of potentially important factors, each with a small contribution, such as patient understanding, motivation, forgetting, reinforcement, and cost.15 Results have been mixed, and these multifaceted complex programs are difficult to implement and sustain in regular practice.15
What is often overlooked is the role that physicians play in influencing adherence. Physicians determine treatment, educate patients, manage side effects, and influence patient knowledge and motivation.12 The physician's ability to manage these aspects of treatment may be particularly relevant for hypertension, in which discontinuation of treatment and poor adherence occur early in the course of therapy.6,8,10,16 The physician's choice of antihypertensive therapy appears to be one important determinant of treatment adherence. Although diuretics are recommended as first-line treatment,17 they have lower persistence rates than angiotensin II receptor blockers and angiotensin-converting enzyme inhibitors, possibly because of differences in adverse effects.5,6,8,10 Physicians vary in the extent to which they discuss the rationale and expected effects of new medications,16,18,19 especially possible adverse effects.16,18 Because 22% to 62% of patients experience adverse effects from antihypertensive drugs in the initial period of use,10,16,20 effective physician-patient communication and therapeutic management of these problems is critical. Approximately 15% to 25% of patients will not spontaneously report adverse effects to their physicians, and, among those who do, physicians act in only 63% of cases.16,21,22
Although cross-sectional studies have provided compelling evidence about the importance of physician communication in improving adherence,14,22- 28 until recently there has been no mechanism for systematically evaluating these skills in the training or licensure process. Moreover, medical management skills may also be important, particularly in treatment initiation and effective monitoring of treatment effects. These skill sets are amenable to improvement with structured education and practice support29,30 but have not been investigated.
A unique opportunity to address these issues arose in Canada when medical regulatory authorities requested the development of a national clinical skills examination to qualify for licensure. This examination assesses the physician's ability to communicate with patients, conduct a clinical assessment, and develop an appropriate management plan.31 Previous research has shown that lower scores on this examination predict complaints about poor communication and quality of care in the first decade of practice, as well as unsatisfactory peer review assessment.32- 34 The possibility that physicians who possess greater proficiency in communication and/or medical management will achieve better medication adherence among their patients has not been examined. By linking physician examination results with information about newly treated hypertensive patients in the Quebec health care system, we were able to test the hypothesis that the newly treated hypertensive patients of physicians with superior communication and medical management abilities would achieve better adherence.
The study was conducted in Canada, where universal health coverage is provided through provincial health insurance plans. Each province maintains a population-based registry of insured persons, discharge abstracts of all hospitalizations (diagnoses and procedures), claims for all physician visits remunerated on a fee-for-service basis, and prescriptions provided through the public drug insurance plan. Claims for physician visits include information on the diagnosis, procedure, date, location, and cost of service. Each prescription claim includes the drug, quantity, duration, and date of dispensing.35 All claims also record unique physician and patient identifiers that can be used to create longitudinal histories of utilization. In Quebec, 99% of residents have provincial health insurance, 85% to 95% of medical visits are remunerated on a fee-for-service basis, and 50% of persons (all elderly, welfare recipients, and persons not insured through their employer) have public drug insurance.35
Starting in 1992, to receive a license to practice in Canada, all physicians had to complete the Medical Council of Canada (MCC) national clinical skills examination (MCCQE2) in addition to the traditional written examination that assesses knowledge and clinical decision making (MCCQE1). The clinical skills examination assessed communication, history, and physical examination skills and clinical management by direct observation of performance by physician examiners in 18 to 20 standardized patient cases.31
A prospective cohort of Quebec physicians who completed the MCCQE2 between 1993 and 1996 was assembled. All hypertensive patients in Quebec who were started on an antihypertensive medication regimen by these physicians between 1993 and 2007 were followed up for the first 6 months after the initial prescription to assess persistence. Claims data were available for 100% of patients. Newly treated patients with hypertension were defined as having a dispensed prescription for an antihypertensive medication (diuretic, β-blocker, calcium channel blocker, angiotensin-converting enzyme inhibitor, angiotensin II receptor blocker, or α-receptor modifier), no previous prescription for an antihypertensive medication in the past year, and a diagnosis of hypertension (International Classification of Diseases, Ninth Revision, code 401) recorded in a physician visit in the past 12 months. Only patients whose treatment was initiated by physicians in the Quebec physician cohort were included. Patients who started treatment with more than 1 antihypertensive drug or who had insufficient follow-up because they died in the first 6 months after the start of treatment were excluded.
To protect confidentiality, the MCC and the Quebec medical regulatory authority assembled the physician cohort and replaced nominal physician data with a study number. The provincial health insurance agency (Régie de l’assurance maladie du Québec [RAMQ]) retrieved records for the 5.8 million patients who were seen by study physicians between 1993 and 2007, including all medical claims, prescriptions, and hospitalizations for these patients irrespective of the treating physician. The RAMQ encrypted all physician and patient identifiers, added the physician study number, and sent de-identified data for analysis. The MCC, Quebec privacy commission, RAMQ, and McGill University institutional review board provided ethics and legal approval.
Persistence with antihypertensive treatment was measured in the first 6 months because the greatest attrition in adherence to antihypertensive therapy occurs in the initial treatment period and is likely influenced by physician-related factors.5,6,8,10,11,20 Persistence was measured by means of records of prescription claims from all community-based pharmacies and was defined as having a continuous supply of at least 1 antihypertensive medication, with gaps no greater than 90 days, in the first 6 months of treatment. Prescription refills provide a good proxy measure of medication use because refill rates are strongly correlated with targeted clinical outcomes such as blood pressure.36- 38 Nonpersistence with antihypertensive treatment was operationally defined as not refilling prescriptions for any antihypertensive treatment for at least 3 months and was assessed by checking 90 days after the end date of each prescription. If there were no further antihypertensive prescriptions dispensed, the end date on the last prescription was considered the start date of nonpersistence.
Communication ability was assessed by means of standardized scores achieved in the communication, history, and physical examination components of the MCCQE2. Physician-patient communication skills are assessed in 3 or 4 cases, selected to represent challenging situations in which effective communication is required for management (eg, discuss refusal of treatment for a terminal illness, counsel an adolescent about birth control) by trained physician observers using case-specific checklists.31 Data collection skills by history and physical examination are assessed in a 5- or 10-minute interaction with 16 or 17 cases selected to represent common and important problems in practice (eg, conduct a focused history and physical examination for a patient presenting with abdominal pain).31,33,39 We also assessed overall scores achieved on the clinical skills examination, taken in the second postgraduate year, based on the mean of scores achieved in each standardized patient case.31
Medical management ability was assessed by means of the standardized medical knowledge and clinical decision-making scores on the MCCQE140 and clinical management scores on the MCCQE2.31 Medical knowledge is assessed by 450 multiple-choice questions that test medical knowledge in medicine, surgery, obstetrics-gynecology, psychiatry, pediatrics, and preventive medicine.40 Key feature problems are used to assess clinical decision making.41 Examinees are asked to respond to critical data collection, diagnosis, and/or management features of 36 to 40 clinical problems.41 Clinical management is assessed on the MCCQE2 by post–standardized patient encounter responses to short-answer questions about the case. We also assessed overall scores achieved on the MCCQE1, taken at the end of medical school, based on the weighted standardized score achieved on the multiple-choice questions and clinical decision-making components.40
Additional physician-related factors were examined as possible mediators between physician ability and persistence. Choice of initial therapeutic class for hypertension treatment has been shown to be associated with the risk of treatment discontinuation.6,8,10 Initial treatment was classified into 5 mutually exclusive categories: diuretics, β-blockers, calcium channel blockers, angiotensin-converting enzyme inhibitors, and angiotensin II receptor blockers.
Therapy change, defined as a change in drug or dosage of initial treatment within the first 2 weeks, was assessed as an indicator of physician response to potential adverse effects or lack of therapeutic effect, identified by the patient or physician. We limited our assessment to the first 2 weeks after the first prescription was dispensed because most adverse effects for new antihypertensive treatment will occur in the first few weeks and to avoid reverse causality bias. Specifically, reverse causality bias can occur because therapy change necessitates the filling of a second prescription, which extends the period of use, and thus apparent persistence, unless there is sufficient follow-up after the expected end date of the prescription to accurately classify persistence. Because all prescriptions in Quebec are dispensed for 30 days or less,42 patients who had a new prescription for a drug or dosage change within the first 2 weeks after starting treatment, but stopped therapy after the second prescription, would still terminate their supply of medication before the end of the second month of follow-up. By the end of the 6-month follow-up period, such patients would have had no new prescriptions for more than 90 days and, thus, would be appropriately classified as nonpersistent.
Follow-up monitoring was defined in 2 ways: (1) as the number of visits to the prescribing physician in the 6 months after starting treatment and (2) as at least 1 early follow-up visit in the first 2 months when most adverse drug effects are expected. The first measure assessed the number of opportunities for the prescribing physician to reinforce the importance of therapy, whereas the latter measured the opportunity to assess initial treatment effects.
Physician sex, age, and specialty (family medicine/general practice, medical specialist, or surgical specialist) were retrieved from MCC and medical regulatory files. Physician practice experience was measured as a time-dependent variable that reflected the number of practice years since the completion of training when the physician prescribed hypertension treatment for a given patient. All physician-related variables were included in the analysis as potential confounders.
Patient age, sex, and household income were measured by means of the RAMQ beneficiary database.43 Using hospitalization and medical services claims diagnostic codes for each patient in the year before the initiation of treatment, we assessed hypertension status (complicated or uncomplicated) and comorbidity. Complicated hypertension was defined as having a previous diagnosis of heart disease, diabetes, chronic kidney disease, atherosclerotic disease, cerebrovascular disease, peripheral arterial disease, asthma, chronic obstructive pulmonary disease, or benign prostatic hypertrophy. Comorbidity was measured by means of the Charlson Comorbidity Index,44 a weighted index of conditions that increase the risk of mortality. The setting in which the initial drug was prescribed, categorized as community, hospital, or emergency department, was also used as an indicator of patient acuity because patients whose antihypertensive treatment was initiated in acute care settings likely have more severe disease and would also have greater challenges in coordinating adequate community follow-up. Medication complexity was assessed by the number of different medications taken in the year before starting antihypertensive treatment.
Nonpersistence during follow-up was descriptively summarized by life-table analysis. We tested the hypothesis that physician communication and management ability would be associated with persistence with hypertension treatment by means of the generalized estimating equation extension of multivariate logistic regression with an exchangeable covariance structure to account for clustering of patients within physicians (SAS, version 9.2; SAS Institute Inc, Cary, North Carolina).45 The patient was the unit of analysis, and nonpersistence at the end of the 6-month follow-up period was the binary outcome variable. We first fit the generalized estimating equation model with physician characteristics only and then adjusted their effects for patient characteristics and other physician-mediated determinants of persistence. We modeled each licensing examination score separately to avoid multicolinearity. We assessed nonlinearity by testing a quadratic term for examination score in each model. To determine whether the association between licensure examination scores and persistence with treatment was attenuated with increasing practice experience, we tested the statistical significance of interactions between ability scores and years of practice experience. In secondary analyses, we also assessed, in separate generalized estimating equation models with adjustment for physician and patient characteristics, the associations between licensure examination scores and the following intermediate physician-mediated outcomes of treatment: (1) antihypertensive drug choice (diuretics vs other classes), (2) therapy change in the first 2 weeks, and (3) at least 1 follow-up visit in the first 2 months. With a population of 13 205 patients, there was 88% statistical power to detect a relative increase of 10% per 1 SD in score, assuming a type 1 error of .05 and 20% nonpersistence.
To assess the clinical relevance of statistically significant relationships between physician communication and management ability and persistence, we calculated the prevented fraction by using the case-based approach.46 For this calculation, we used the reduction in risk of nonpersistence for patients of physicians in the 3 upper score quartiles relative to the bottom quartile because improving skills for the bottom 25% of physicians is a feasible goal for future intervention. The prevented fraction represents the proportion of expected cases of nonpersistence that were averted by better physician management and communication abilities and the potential reduction in nonpersistence that may be achieved by improving the skills of physicians in the bottom quartile.46
A total of 645 of the 1161 physicians in the Quebec examination cohort who entered practice between 1993 and 2007 initiated hypertension treatment for at least 1 patient in their first 1 to 15 years of practice. The mean (SD) number of years of practice for these physicians was 5.6 (3.0) years; 54.3% were female, 57.8% were family physicians, and 32.9% were medical specialists (Table 1). The medical management and communication scores of study physicians were slightly above the mean score of 500 for all Canadian-trained physicians but diverged from the mean by up to 4 to 7 SDs.
Overall, physicians prescribed hypertensive treatment for 172 118 patients, of whom 13 205 (7.7%) met the inclusion criteria. Overall, 129 410 (75.2%) were excluded because of previous hypertensive therapy, 21 626 (12.6%) because there was no diagnosis of hypertension, 6794 (3.9%) because they started treatment with more than 1 drug, 263 (0.2%) because treatment duration was less than 5 days, and 820 (0.5%) because they died within the first 6 months of treatment. The mean age of newly treated hypertensive patients was 61.4 years, 57.7% were female, and the average household income was $46 010 (Table 2). Before starting antihypertensive treatment, 40.4% of patients were classified as having complicated hypertension, 43.1% were taking 2 to 5 other medications, and more than 30% had significant comorbidity on the Charlson index.
The most commonly used first-line treatments were angiotensin-converting enzyme inhibitors, diuretics, and calcium channel blockers (Table 2, eTable 1). Most antihypertensive treatment was started in a community-based setting, and the majority of patients had a follow-up visit with their prescribing physician in the first 2 months. The number of visits in the 6-month follow-up period was skewed (mean, 7.1; SD, 8.1) and therefore was reclassified as an ordinal variable in further analyses. After starting treatment, 8.0% of patients had a change in antihypertensive therapy in the first 2 weeks, the majority of which were drug changes.
Within the first 6 months, 2926 patients (22.2%) had discontinued all antihypertensive medication. More than half of nonpersistent patients (1717 [58.7%]) discontinued treatment in the first 60 days (Figure 1). Younger patients, males, and those who started treatment in hospital and emergency department settings were more likely to be nonpersistent, as were patients with lower income (Table 3).
Higher scores in medical management were protective against nonpersistence (Table 4). Clinical decision-making ability was the strongest predictor, for which the risk of nonpersistence was reduced by 23% (odds ratio [OR], 0.74; 95% confidence interval [CI], 0.63-0.87) per 2-SD increase in examination score and by 65% (0.35; 0.20-0.61) when comparing physicians who were 3 SDs above vs below the mean score (Figure 2). The relationship between clinical decision-making ability and persistence was not attenuated with longer practice experience, at least for the first 15 years of practice (score × years in practice interaction: P = .72). Better communication skills were also associated with a reduced risk of nonpersistence, specifically data collection skills, for which there was a 12% reduction in the risk of nonpersistence per 2-SD increase in score (OR, 0.88; 95% CI, 0.78-1.00) (Table 4, Figure 2). Higher overall scores in both the knowledge and clinical skills examinations were also associated with a lower risk of nonpersistence (Table 4).
Physician-mediated treatment characteristics were also associated with nonpersistence. Compared with patients taking diuretics, patients started on regimens of other classes of antihypertensive therapy were less likely to be nonpersistent (Table 5). Regardless of the class of antihypertensive treatment, early therapy change significantly reduced the risk of nonpersistence by 55% (OR, 0.45; 95% CI, 0.37-0.54). Both the number of follow-up visits and early follow-up were independently protective, the latter reducing the risk of nonpersistence by 14% (OR, 0.86; 95% CI, 0.75-0.98). Although there was no association between physician ability and choice of starting therapy, physicians with higher overall scores in knowledge and decision making were more likely to make therapy changes in the first 2 weeks (OR, 1.25; 95% CI, 1.01-1.53; per 2-SD increase in score), and physicians with higher scores in communication were more likely to have their patients come in for follow-up visits in the first 2 months (1.14; 1.01-1.29; per 2-SD increase in score) (eTable 2).
After adjusting for patient and treatment characteristics, 15.8% (95% CI, 7.5%-23.3%) of nonpersistence was prevented by physicians in the top 3 quartiles of clinical decision-making skills compared with physicians in the bottom quartile. For data collection, only physicians in the top quartile were significantly better than those in the bottom quartile, accounting for 4.8% (95% CI, 1.1%-7.9%) of nonpersistence.
We found that a substantial proportion of newly treated patients with hypertension discontinued all antihypertensive treatment within the first 6 months. Medical management ability, choice of therapy, and early changes in dosage and drug improved persistence. Physician-patient communication was also important, confirming results reported in previous cross-sectional studies,14,22- 28 particularly more advanced skills in history taking and physical examination.
The strengths of this study are that we investigated novel approaches to assessing physician-related determinants of medication adherence using comprehensive population-based data on new users of antihypertensive treatment and national examination measures of physician ability. Limitations are that we used an indirect measure of medication persistence36- 38 and we may have underestimated complications and comorbidity. Because administrative data have high specificity for diagnoses but only moderate sensitivity,47 these problems will produce random errors in measurement and attenuate associations. Similar to other studies, we found that diuretics were associated with higher risks of nonpersistence than other antihypertensive classes.6,8,10 The hypothesis that these differences are due to class-specific side effects is supported by our findings because the effect of therapeutic class was independent of patient and physician characteristics.
One of the most significant determinants of persistence was treatment change—specifically, the modification of the dosage or drug before the end of the first prescription. This is a new, but not unexpected, finding because 22% to 62% of new users of antihypertensive medications experience adverse effects that lead them to reduce or discontinue their use of prescription medication.10,16,20,48 We found that more knowledgeable physicians were more likely to make treatment changes, either because they followed up their patients more effectively or their patients were more motivated to report problems. Even after adjustment for therapy change, medical management ability, particularly clinical decision making, provided additional benefits, accounting for 15.8% of persistence in the first 6 months. Patients expect physicians to explain the rationale and benefits of hypertension treatment and adverse effects.49 More knowledgeable physicians may be more likely to provide relevant information to patients and to monitor and intervene for side effects. It is also possible that physicians with better clinical decision-making ability incorporate patient and family preferences into the treatment decision-making process—aspects of treatment decision making that have been shown to improve persistence.22,25,26,50 Indeed, during the past decade, many medical schools and specialty training programs have incorporated patient-centered case-based learning into the curriculum to enhance clinical decision-making abilities and the ability to involve patients in the decision-making process.51,52 This also may be one reason why physicians with better communication skills, particularly in history taking, had a lower risk of nonpersistence; they would be able to elicit patient preferences more effectively and incorporate this information into treatment decision making.
This study highlights several important avenues for future research. First, the hypothesis that physicians with better clinical decision-making and data collection skills are more likely to incorporate patient preferences into the treatment decision-making process needs to be tested. If true, training programs could improve the ability of physicians to manage chronic conditions by ensuring that these aspects of management are systematically incorporated into undergraduate and postgraduate medical training. For practicing physicians, continuing medical education interventions aimed at enhancing communication and clinical decision making should be evaluated to determine whether improving these skill sets will lead to better patient treatment adherence. However, the development and evaluation of tools to support clinical decision making and medication monitoring in the practice environment will likely be the most fruitful. In particular, new information technologies that can be integrated with the electronic medical record may have the greatest promise for enhancing drug- and patient-specific follow-up of adverse effects and treatment effectiveness. Technologies such as computerized decision support, community drug profiles from retail pharmacies, and follow-up monitoring tools such as interactive voice recording approaches have been shown to be effective in management of some chronic conditions,53- 57 and their increased use in hypertension treatment should improve patient adherence and, ultimately, clinical outcome.
Correspondence: Robyn Tamblyn, PhD, Department of Medicine, McGill University, 1140 Pine Ave W, Montreal, QC, H3A 1A3 Canada (email@example.com).
Accepted for Publication: November 30, 2009.
Author Contributions:Study concept and design: Tamblyn, Abrahamowicz, Dauphinee, Jacques, Klass, Winslade, and Buckeridge. Acquisition of data: Tamblyn, Wenghofer, Jacques, Smee, Girard, and Buckeridge. Analysis and interpretation of data: Tamblyn, Abrahamowicz, Jacques, Eguale, Girard, Bartman, Buckeridge, and Hanley. Drafting of the manuscript: Tamblyn, Dauphinee, and Bartman. Critical revision of the manuscript for important intellectual content: Tamblyn, Abrahamowicz, Dauphinee, Wenghofer, Jacques, Klass, Smee, Eguale, Winslade, Girard, Buckeridge, and Hanley. Statistical analysis: Tamblyn, Abrahamowicz, Eguale, Girard, and Hanley. Obtained funding: Tamblyn and Dauphinee. Administrative, technical, and material support: Tamblyn, Wenghofer, Jacques, Klass, Smee, Winslade, Bartman, and Buckeridge. Study supervision: Tamblyn, Dauphinee, and Buckeridge.
Financial Disclosure: None reported.
Funding/Support: This study was supported by the Medical Council of Canada, Canadian Institutes for Health Research.