Diabetes, Hypercholesterolemia, Hypertension
Conditions
Keywords
medication adherence, choice architecture, pharmacy home
Brief summary
This study is a pilot test of an intervention that delivers timely diagnostic information about medication nonadherence to doctors, and then offers the services of clinical pharmacists to treat these nonadherence problems. Participating doctors will be notified when a patient is 10 days late refilling a medication for diabetes, hypertension, or hypercholesterolemia. In one randomization arm the pharmacist will contact the patient as the default option (with no action required by the doctor), and in the other the pharmacist will contact the patient only if the doctor actively chooses that the pharmacist take action. Patients of participating doctors will be randomized to 1) one of these two pharmacist options, 2) an information only control arm in which the doctor gets adherence information but does not have access to a pharmacist for that patient, and 3) a no information control arm. The investigators' central hypothesis is that the pharmacist will be consulted more often when intervention by the pharmacist is the default outcome and that the default pharmacist intervention will be the most beneficial for adherence outcomes.
Detailed description
Poor adherence with prescription medications is ubiquitous, regardless of the disease, medication, patient population, or country studied. It is also expensive - annual costs of poor adherence in the United States were recently estimated at $290 billion. This problem has two components: diagnosis and treatment. Regarding diagnosis, doctors' assessments of patients' adherence are inaccurate, and doctors often do not discuss adherence problems with their patients. This makes it attractive to use pharmacy claims to identify nonadherence. While diagnostic data is necessary to solve the non-adherence problem, it is not sufficient. Once diagnosed, doctors must take action to treat nonadherence. Research shows that simply giving doctors claims data about nonadherence is ineffective, probably because it is not clear what action to take, and because the costs in time and energy of taking action are too great. What is currently lacking is a practical way to effectively integrate this diagnostic information and treatment expertise into work flows in primary care doctors' offices, and an effective method of inducing doctors to act on it. Behavioral economics suggests that barriers to doctors' action may be overcome in a cost effective way by altering the architecture of choices doctors face. The long term goal of this research is to develop systems that effectively connect pharmacy benefits managers (PBMs), primary care doctors, clinical pharmacists, and patients in ways that improve medication adherence and patients' health outcomes. The overall objective of this application, which is the next step toward attainment of the investigators long term goal, is to conduct a pilot test of an intervention that delivers timely diagnostic information about nonadherence to doctors, and then offers the services of clinical pharmacists to treat these nonadherence problems. Participating doctors will be notified when a patient is 10 days late refilling a medication for diabetes, hypertension, or hypercholesterolemia. Taking advantage of the principle of intelligent choice architecture from behavioral economics, in one arm the pharmacist will contact the patient as the default option (with no action required by the doctor), and in the other the pharmacist will contact the patient only if the doctor actively chooses that the pharmacist take action. Patients of participating doctors will be randomized to 1) one of these two pharmacist options, 2) an information only control arm in which the doctor gets adherence information but does not have access to a pharmacist for that patient, and 3) a no information control arm. The investigators central hypothesis, which is strongly supported by work in other fields, is that the pharmacist will be consulted more often when intervention by the pharmacist is the default outcome and that the default pharmacist intervention will be the most beneficial for adherence outcomes. This study is a collaboration between researchers at Brown University, Tufts University, Harvard University, and Johns Hopkins University; Express Scripts; a large regional commercial insurer; and a network of primary care doctors in Eastern Massachusetts. The team is led by Dr. Ira Wilson, an experienced adherence researcher, and includes behavioral and health economists, and a statistician experienced in adherence issues. The investigators will accomplish the investigators overall objectives by pursuing the following Specific Aims: 1. Establish and test the technical and communications infrastructure required for the conduct of this clinical trial. The following steps must occur in a secure environment: a) Express Scripts notifies the study that a patient is late filling a prescription, b) the study notifies the doctor, c) the doctor makes a choice about how to respond, and d) a pharmacist, in some cases, contacts the patient. 2. Conduct and evaluate a clinical trial of an intervention comparing methods of offering pharmacist services to primary care doctors. Eligible doctors and patients will be randomized to a) pharmacist services under one of two choice architecture conditions (default or choice), b) adherence information only, or c) no information; further randomization for patients in the experimental arms will occur where the patient's HMO/PPO status will be revealed to the physician, or not. Outcomes include medication adherence, duration of nonadherence event, and physician participant behavioral outcomes.
Interventions
BEHAVIORALPharmacist calls patient unless physician cancels call
BEHAVIORALPatient nonadherence information sent to physician
BEHAVIORALPharmacist calls patient if physician requests call
BEHAVIORALDoctor receives information and may be allowed certain actions
Sponsors
National Institute on Aging (NIA)
Tufts Medical Center
Harvard University
Johns Hopkins University
Brown University
Study design
Allocation
RANDOMIZED
Intervention model
PARALLEL
Primary purpose
HEALTH_SERVICES_RESEARCH
Masking
SINGLE (Subject)
Eligibility
Sex/Gender
ALL
Age
18 Years to No maximum
Healthy volunteers
No
Inclusion criteria
Physician Inclusion Criteria: * New England Quality Care Alliance (NEQCA) primary care physicians of adult patients insured through large commercial insurer partner Patient Inclusion Criteria: * Adult patients of consented New England Quality Care Alliance (NEQCA) primary care physicians * Insured through large commercial insurer partner * Prescribed chronic medications for one or more of the three study conditions in the past six months Patient Exclusion Criterion: * On the insurer's do not contact list
Design outcomes
Primary
| Measure | Time frame | Description |
|---|---|---|
| Probability of Resolution of Nonadherence Within 30 Days | Outcome measure examines fills within 30 days of a nonadherence event. Participants were followed over a total of 6 months. | Patients who were more than 10 days late refilling a chronic medication prescription were in the analytic sample frame and were targeted for intervention according to how they were randomized. This outcome is the rate at which these patients have filled a prescription by 30 days. Outcome is 1 if the patient fills the prescription by 30 days (considered resolution of nonadherence); otherwise it is 0. Outcome measures reported are the means of the per-person proportions of nonadherence (NAE) events resolved within 30 days across all patients in each particular arm. |
| Duration of Nonadherence Event | Participants were followed over a total of 6 months | Patients who were more than 10 days late refilling a chronic medication prescription were in the analytic sample frame and were targeted for intervention according to how they were randomized. This outcome is the duration of nonadherence event (the length of time the patient took to refill a prescription if the refill had been late), in days. |
Secondary
| Measure | Time frame | Description |
|---|---|---|
| Probability of Physician Viewing Nonadherence Event Information | Participants were followed over a total of 6 months | Patients who were more than 10 days late refilling a chronic medication prescription were in the analytic sample frame and were targeted for intervention according to how they were randomized. This outcome is the rate at which physicians viewed nonadherence event information. Outcome measures reported are the means of the per-person proportions of NAE event notices viewed by the physician across all patients in each particular arm. |
| Probability of Pharmacist Action Triggered | Participants were followed over a total of 6 months | Patients who were more than 10 days late refilling a chronic medication prescription were in the analytic sample frame and were targeted for intervention according to how they were randomized. This outcome is the rate at which pharmacist action was triggered to resolve nonadherence. Outcome measures reported are the means of the per-person proportions of NAE events which triggered pharmacist action across all patients in each particular arm. |
Countries
United States
Participant flow
Recruitment details
Physician recruitment with signed consent forms was in-person. 91 physicians were consented and enrolled at launch. All eligible patients were automatically enrolled, and mailed an opt-out card to be returned if participation was refused.
Pre-assignment details
Both patients and doctors were enrolled in the study. Separate Periods represent the sequential nature of the study design: Period 1 includes physicians only. Only those patients with a nonadherence event were randomized to a control or intervention arm, so the number consented at launch (2,606) is higher than the number included in the study results (1,474). Patients were removed if insurance coverage expired or if their enrolled physician withdrew.
Participants by arm
| Arm | Count |
|---|---|
| Default Patient Patient nonadherence information sent to physician; Pharmacist calls patient unless physician cancels call | 244 |
| Information Patient Patient nonadherence information sent to physician | 477 |
| Control Patient Control - no intervention | 479 |
| Choice Patient Patient nonadherence information sent to physician; Pharmacist calls patient if physician requests call | 274 |
| Information Doctor Physician randomized to information only arm | 29 |
| Choice Doctor Physician randomized to choice arm | 31 |
| Default Doctor Physician randomized to default arm | 31 |
| Total | 1,565 |
Withdrawals & dropouts
| Period | Reason | FG000 | FG001 | FG002 | FG003 | FG004 | FG005 | FG006 | FG007 | FG008 | FG009 | FG010 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Initial Physician Randomization | Physician Decision | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 1 | 0 | 0 |
Baseline characteristics
| Characteristic | Default Patient | Information Patient | Control Patient | Choice Patient | Information Doctor | Choice Doctor | Default Doctor | Total |
|---|---|---|---|---|---|---|---|---|
| Age, Continuous | 56.80 years STANDARD_DEVIATION 0.49 | 55.99 years STANDARD_DEVIATION 0.51 | 56.58 years STANDARD_DEVIATION 0.47 | 56.29 years STANDARD_DEVIATION 0.68 | 53.86 years STANDARD_DEVIATION 1.67 | 56.87 years STANDARD_DEVIATION 1.53 | 56.16 years STANDARD_DEVIATION 1.78 | 56.33 years STANDARD_DEVIATION 0.45 |
| Region of Enrollment United States | 244 participants | 477 participants | 479 participants | 274 participants | 29 participants | 31 participants | 31 participants | 1565 participants |
| Sex: Female, Male Female | 78 Participants | 167 Participants | 163 Participants | 107 Participants | 16 Participants | 11 Participants | 12 Participants | 554 Participants |
| Sex: Female, Male Male | 166 Participants | 310 Participants | 316 Participants | 167 Participants | 13 Participants | 20 Participants | 19 Participants | 1011 Participants |
Adverse events
| Event type | EG000 affected / at risk | EG001 affected / at risk | EG002 affected / at risk | EG003 affected / at risk | EG004 affected / at risk | EG005 affected / at risk | EG006 affected / at risk | EG007 affected / at risk | EG008 affected / at risk | EG009 affected / at risk | EG010 affected / at risk |
|---|---|---|---|---|---|---|---|---|---|---|---|
| deaths Total, all-cause mortality | 0 / 244 | 0 / 123 | 0 / 121 | 0 / 274 | 0 / 134 | 0 / 137 | 0 / 220 | 0 / 221 | 0 / 29 | 0 / 31 | 0 / 31 |
| other Total, other adverse events | 0 / 244 | 0 / 123 | 0 / 121 | 0 / 274 | 0 / 134 | 0 / 137 | 0 / 220 | 0 / 221 | 0 / 29 | 0 / 31 | 0 / 31 |
| serious Total, serious adverse events | 0 / 244 | 0 / 123 | 0 / 121 | 0 / 274 | 0 / 134 | 0 / 137 | 0 / 220 | 0 / 221 | 0 / 29 | 0 / 31 | 0 / 31 |
Outcome results
Primary
Duration of Nonadherence Event
Patients who were more than 10 days late refilling a chronic medication prescription were in the analytic sample frame and were targeted for intervention according to how they were randomized. This outcome is the duration of nonadherence event (the length of time the patient took to refill a prescription if the refill had been late), in days.
Time frame: Participants were followed over a total of 6 months
Population: All Information Only patients are grouped together in these results (regardless of physician arm), because the treatment for all Information Only patients is identical across all physician arms.
| Arm | Measure | Value (MEAN) | Dispersion |
|---|---|---|---|
| Default Patient | Duration of Nonadherence Event | 40.91 Days | Standard Deviation 37.89 |
| Choice Patient | Duration of Nonadherence Event | 39.46 Days | Standard Deviation 35.02 |
| Information Patient | Duration of Nonadherence Event | 41.35 Days | Standard Deviation 43.51 |
| Control Patient | Duration of Nonadherence Event | 41.04 Days | Standard Deviation 41.8 |
p-value: <0.01Regression, Linear
p-value: <0.01Regression, Linear
p-value: <0.01Regression, Linear
Primary
Probability of Resolution of Nonadherence Within 30 Days
Patients who were more than 10 days late refilling a chronic medication prescription were in the analytic sample frame and were targeted for intervention according to how they were randomized. This outcome is the rate at which these patients have filled a prescription by 30 days. Outcome is 1 if the patient fills the prescription by 30 days (considered resolution of nonadherence); otherwise it is 0. Outcome measures reported are the means of the per-person proportions of nonadherence (NAE) events resolved within 30 days across all patients in each particular arm.
Time frame: Outcome measure examines fills within 30 days of a nonadherence event. Participants were followed over a total of 6 months.
Population: All Information Only patients are grouped together in these results (regardless of physician arm), because the treatment for all Information Only patients is identical across all physician arms.
| Arm | Measure | Value (MEAN) | Dispersion |
|---|---|---|---|
| Default Patient | Probability of Resolution of Nonadherence Within 30 Days | .4286 Proportion of resolved NAEs within 30 | Standard Deviation 0.4953 |
| Choice Patient | Probability of Resolution of Nonadherence Within 30 Days | .4290 Proportion of resolved NAEs within 30 | Standard Deviation 0.4953 |
| Information Patient | Probability of Resolution of Nonadherence Within 30 Days | .4554 Proportion of resolved NAEs within 30 | Standard Deviation 0.4982 |
| Control Patient | Probability of Resolution of Nonadherence Within 30 Days | .4437 Proportion of resolved NAEs within 30 | Standard Deviation 0.4971 |
p-value: <0.01Regression, Logistic
p-value: <0.01Regression, Logistic
p-value: <0.01Regression, Logistic
Secondary
Probability of Pharmacist Action Triggered
Patients who were more than 10 days late refilling a chronic medication prescription were in the analytic sample frame and were targeted for intervention according to how they were randomized. This outcome is the rate at which pharmacist action was triggered to resolve nonadherence. Outcome measures reported are the means of the per-person proportions of NAE events which triggered pharmacist action across all patients in each particular arm.
Time frame: Participants were followed over a total of 6 months
Population: We considered only claims for patients who had been randomly assigned to intervention arms where the pharmacist was available - Default and Choice.
| Arm | Measure | Value (MEAN) | Dispersion |
|---|---|---|---|
| Default Patient | Probability of Pharmacist Action Triggered | 0.6102 Proportion with pharmacist calls | Standard Deviation 0.4881 |
| Choice Patient | Probability of Pharmacist Action Triggered | 0.1903 Proportion with pharmacist calls | Standard Deviation 0.3929 |
p-value: <0.01Regression, Logistic
Secondary
Probability of Physician Viewing Nonadherence Event Information
Patients who were more than 10 days late refilling a chronic medication prescription were in the analytic sample frame and were targeted for intervention according to how they were randomized. This outcome is the rate at which physicians viewed nonadherence event information. Outcome measures reported are the means of the per-person proportions of NAE event notices viewed by the physician across all patients in each particular arm.
Time frame: Participants were followed over a total of 6 months
Population: We excluded the Control group because we considered only claims where physicians assigned to a treatment arm would have been notified by email of nonadherence.
| Arm | Measure | Value (MEAN) | Dispersion |
|---|---|---|---|
| Default Patient | Probability of Physician Viewing Nonadherence Event Information | 0.2822 Proportion of NAEs viewed | Standard Deviation 0.4505 |
| Choice Patient | Probability of Physician Viewing Nonadherence Event Information | 0.3856 Proportion of NAEs viewed | Standard Deviation 0.4872 |
| Information Patient | Probability of Physician Viewing Nonadherence Event Information | 0.3257 Proportion of NAEs viewed | Standard Deviation 0.4688 |
p-value: <0.01Regression, Logistic
p-value: <0.01Regression, Logistic