We will conduct a 5-year, prospective, cluster-randomized multi-center clinical trial in 27 clinics from 13 states in the US. All clinics employ clinical pharmacists and 47% of the patients are under-represented minorities.12
Clinics were stratified (defined below) and then randomized to a: 1) 9-month PPCM arm (n=219), 2) 24-month PPCM arm (n=219), or 3) control group (n=219) that also includes a distracter intervention for asthma ( see Appendix for Methods). A study coordinator in each clinic will enroll 24 subjects with hypertension who will be followed for 24 months (n=648). In addition, we will perform a retrospective chart review for another 486 subjects (18 per clinic) with hypertension which will serve as an observational cohort (pure control). The purpose of the observational cohort is to determine if the intervention diffuses into the rest of the practice, even for patients not enrolled into the trial.
The primary hypothesis for the study is that BP control at 9 months will be significantly greater in subjects from clinics randomized to the two intervention groups compared to the control group. The secondary aims will compare mean BP between study arms, mean BP in minority groups and those from lower socioeconomic groups and evaluate the association between provider attitudes to deliver PPCM with mean BP and BP control rates.
Sites, Patients and Providers
The study will be conducted in the National Interdisciplinary Primary Care Practice Based Research Network medical offices. Features of the medical offices were previously published and showed that 85.4% are located in family medicine residencies, 10.4% in internal medicine residencies and 4.2% in faculty practices.12
Sixty-seven percent of the offices utilized electronic medical records when surveyed in 2006.12
The mean number of providers was: 10 attending physicians, 23 resident physicians, 6 nurses, 5 medical assistants, 0.4 nurse practitioners, 0.4 physician assistants, 0.7 psychologists, 0.3 social workers, 0.4 dieticians, and 0.4 patient educators.
Only two of the 27 offices have pharmacies. All 27 offices have clinical pharmacists (mean 1.9 per clinic) who are faculty in the office and are primarily employed to provide education for physicians and patient care. They spend an average of 2 days per week (0.4 cFTE) in direct patient care and cover clinic hours an average of 75% of the time. Most have a Pharm.D. degree (96%) and a postdoctoral residency or fellowship (78%), and 43% are board certified pharmacotherapy specialists. Seventy percent of clinics have had clinical pharmacy services for over 5 years.12
Patient revenue is used to cover pharmacist salaries in 22% of practices.
These clinics have a mean of 26% African Americans and 21% Hispanics. We expect to recruit at least 40% of patients from these minority groups. Payers include Medicaid/government (28%), Medicare (22%), private insurance (34%), uninsured (12%) and other (4%).10
We used a validated survey instrument13
to score clinical pharmacy services based on the extent of direct patient care services provided in the medical office. Clinic scores fell into two distinct levels similar to previous findings.13, 14
Clinics were stratified based on the pharmacy structure scores (low and high) and percent minority patients (<44% vs. ≥44%).
Data and Safety Monitoring
A Data and Safety Monitoring Board (DSMB) was appointed following NHLBI guidelines and includes internationally recognized experts in hypertension and/or team-based care models. The DSMB will meet at least once a year to review the rate of patient recruitment, evaluate the safety of the study and make recommendations to the investigators to improve the conduct of the study.
Providers in clinics randomized to the intervention arms received training from one investigator (BLC) on BP guidelines15,16
strategies to overcome clinical inertia and suggested methods of communication between physicians and pharmacists. Eight regional training sessions were conducted between September 2009 and January 2010 and were delivered to one pharmacist and one physician investigator from each medical office in “train the trainer” sessions. Separate training programs occurred for providers from clinics randomized to the BP intervention or usual care (alternative asthma intervention). The trainers also included a physician/pharmacist team from one community-based Family Medicine program that successfully implemented the intervention model in a previous study.8
These individuals encouraged physicians to improve participation of providers, instilled confidence and enthusiasm, addressed strategies to overcame barriers BP control and discussed methods to effectively overcome clinical inertia, adverse drug reactions and poor medication adherence.7, 8
One to three months after the “train-the-trainer” session, each physician-pharmacist pair will deliver the same training program for all providers in their own clinic. Pharmacists in clinics randomized to the 9-month BP intervention will provide training sessions with their physicians twice a year for two years while those randomized to the 24-month intervention will provide training sessions for three years.
Following the initial training sessions, two investigators will conduct a telephone conference with each physician-pharmacist pair to discuss any questions or issues that the providers were not able to answer regarding the intervention. The investigators will continue to support all offices quarterly the first year, and twice a year in year two and year three (for clinics randomized to the long-term BP intervention).
Study Coordinator Training
A study coordinator (RN, LPN or MA) employed in each medical office will enroll patients, collect study data and abstract medical records for the observational cohort. All coordinators completed human subjects education and were trained in Iowa City in December 2009 on the web-based case report forms that will be used on the secure website operated by study Data Coordinating Center (DCC).
One investigator (BLC) provided training and certification on proper BP measurement technique using an automated Omron HEM 907-XL device.17
Study monitors from the DCC will make site visits to each medical office at least yearly to evaluate data fidelity and will re-certify each research nurse in BP measurement at each monitoring visit.
The Pharmacist Intervention
The suggested PPCM model will specify recommended visit frequency and activities. However, since this is an effectiveness study, strict adherence to the model and intervention schedule will not be required. Instead, we will encourage the use of the model and then measure the extent to which it is implemented. The pharmacist will be asked to document all visits, medication recommendations made to the physician, recommendations accepted by the physician and the time required to complete various steps in the study visit.
Patients in the two intervention groups will receive the intervention for 9 months. The therapeutic strategies will be based on JNC-7 and the BP goal will be:15
1) BP <140/90 mm Hg for patients with uncomplicated hypertension; 2) BP <130/80 mm Hg for patients with diabetes or chronic kidney disease. (We will modify these goals if JNC-8 changes the goals).
At baseline, the pharmacist will review the medical record and perform a structured interview with the patient, including: 1) a detailed medication history; 2) assessment of patient knowledge of BP medications, purpose of each medication, goals of therapy, medication dosages and timing, and potential medication side effects; 3) potential contraindications to specific BP medications; and 3) expectations for future dosage changes, monitoring and issues that may become future barriers to BP control (e.g., side effects, non-adherence, patient self efficacy). The pharmacist will supply a wallet card listing all medications and doses, contact phone numbers and BP goals.
The pharmacist will create a care plan with treatment recommendations for the physician at the baseline visit so that an immediate change in medication can be made.7, 8, 18
If the physician agrees with the care plan, or makes a modification in the plan, the pharmacist will implement the plan. The study case report forms will capture whether or not the physician accepted the pharmacists’ recommendations. The suggested PPCM model includes structured face-to-face visits with the patient at baseline, 1, 2, 4, 6 and 8 months, a telephone call at 2 weeks and additional visits if BP remains uncontrolled.
If BP is controlled, the recommended action will be for the pharmacist to schedule the patient for routine follow-up every 3–6 months.15
If BP control is lost, the pharmacist is encouraged to increase visit frequency similar to the baseline schedule.
Pharmacists in control sites will not provide the intervention for patients with hypertension but will continue to provide “curbside consultations” if physicians specifically ask questions about patients with hypertension. Instead, pharmacists in the control group will be providing an alternative intervention for asthma patients (Appendix).
Subjects will be eligible if they are English or Spanish speaking males or females, over 18 years of age with a diagnosis of hypertension and have uncontrolled BP defined as ≥ 140 mm Hg SBP or ≥ 90 mm Hg DBP for uncomplicated hypertension; or ≥ 130 mm Hg SBP or ≥ 80 mm Hg DBP for patients with diabetes or chronic kidney disease. Qualification will be based on a seated BP (average of the second and third reading) as measured in the office by the study coordinator.
Subjects will be excluded with: 1) current signs of hypertensive emergency (acute angina, stroke, or renal failure; 2) systolic BP >200 or diastolic BP > 114 mm Hg); 3) history of MI, stroke, or unstable angina in the prior 6 months; 4) systolic dysfunction with a LV ejection fraction < 35% documented by echocardiography, nuclear medicine study, or ventriculography; 5) glomerular filtration rate less than 20 ml/min or proteinuria > 1 gram per day; 6) diagnoses of cirrhosis, Hepatitis B or C infection, or laboratory abnormalities (serum ALT or AST > 2 times control or total bilirubin > 1.5 mg/dl) in the prior 6 months; 7) pregnancy; 8) pulmonary hypertension or sleep apnea (unless treated by continuous positive pressure ventilation); 9) life expectancy estimated less than 2 years; 10) residence in a nursing home or diagnosis of dementia; and 11) inability to give informed consent or impaired cognitive function`.
The study has been designed to minimize selection bias. Potential subjects will be identified from a list generated from each clinic and then randomized for inclusion by the DCC. Coordinators will receive the ordered lists and will then review the medical records in this order and invite a patient to participate if the study criteria are met. The study coordinator will continue this process until 24 patients are consented and enrolled. We will record the reasons patients declined, did not meet criteria or dropped from the study in order to evaluate the generalizability of the selected subjects to the entire population.
If a potential subject only speaks Spanish, we will utilize bilingual coordinators or translators within the office to explain the study and assist with obtaining informed consent. The consent form and scripts have been translated into Spanish. All patient questionnaires are scripted to ensure consistent administration.
The study coordinator will measure BP in the sitting position after appropriate rest17
baseline, 6, 9, 12, 18 and 24-months.7, 8
The study coordinator will collect the following at baseline: height, weight, and pulse, the duration of HTN, presence of other cardiovascular risk factors, symptoms and adverse drug reactions, medication adherence, socio-demographics, co-morbidities, current medications and dose, how the patient actually took the medication and an evaluation of medication adherence using validated instruments.19, 20
The timing of collection of these variables are shown in .
Patients enrolled in the control group will receive usual care from their physicians. Patients in intervention clinics will be referred to the intervention pharmacist.
Study monitors from the DCC will perform site visits to compare the completeness of the case report forms with the medical records and the web-based database. Each site will be visited at least once a year to evaluate data fidelity.
Observational Cohort (pure control group)
The purpose of the retrospective observational cohort will be to evaluate whether the effect of the intervention diffuses throughout the practice for patients not actively enrolled in the study using our previous methods.8, 21–23
Once 24 patients have been enrolled into the prospective interventional study at a given site, the study coordinator will identify another 18 patients who meet the same inclusion criteria. Qualification in the retrospective observational cohort will be an index visit that occurred during the study period and based on a qualifying BP on the last two clinic visits. Patients in the observational cohort will not be consented or have any study visits, and all data will be collected retrospectively by medical record audit.
We recognize the hazard of comparing research BP values with clinic-measured BPs so these will not be compared to the prospective data in the analyses. However, by comparing the observational cohort across all three study groups, we will be able to determine if PPCM diffuses to non-study patients to evaluate BP in the absence of the Hawthorne effect.
Predictors of implementation
The theoretical model we have used to predict physician and pharmacist behavior to adopt PPCM is the Theory of Planned Behavior (TPB).24
We will determine the provider’s attitude (do they favor it), their social pressure (subjective norm) and whether they feel in control of the action (perceived control).24, 25
The 16-item instrument will be administered to physicians in all clinics prior to patient enrollment and again at the end of the study.25
CPT codes for pharmacist medication management have been developed but are not routinely recognized by insurers.26
At baseline and at the end of the study, we will ask the pharmacists to provide, for the most recent year, the: 1) total amount billed and collected for all clinical services, and 2) total amount billed and collected strictly for services related to hypertension. The amounts collected will be used as covariates to predict BP control.
We will evaluate and control for self-reported adherence at baseline, 9 and 24 months using a strategy recently published.19, 20
We will use a symptom survey from our previous studies and evaluate symptoms between the three study arms at baseline, 9 and 24 months.7
Costs will be assigned to each BP medication at baseline, 9 and 24 months. All clinic visits (including pharmacist), telephone follow-up, emergency room visits, hospitalizations and laboratory procedures will have costs assigned and the three groups will be compared using methodologies previously described.14
Incremental costs as a function of differences in BP will be calculated at baseline, 9 months and 24 months. These findings will be expressed as dollars per incremental reduction in BP (mm Hg). We will compare the costs associated with this intervention to of other studies that estimated the value to society following controlled BP.
Formative evaluations will be conducted with physicians, pharmacists, study coordinators and office administrators once the study has been completed. A detailed description of the formative evaluations appears in the appendix.
The primary endpoint is BP control at 9 months between study arms. Secondary endpoints will include: 1) BP control and mean BP differences between groups at 12, 18 and 24 months, 2) between group comparison of BP control and mean BP in patients from under-represented minorities, education level, household income, 3) medication intensification, 4) provider-level variables as predictors of mean BP and BP control and 5) pharmacist-level variables (billing and collections) as predictors of mean BP and BP control. A separate analysis for items 1–5 will be conducted across the three arms for patients in the observational cohort.
Sample size calculations assumed that there will be 1 to 1 to 1 randomization to two intervention arms and a control arm and that the primary comparison is the comparison of the BP control rate in the two intervention arms combined to the BP control rate in the control arm. Effect sizes were determined from our previous study that found an effect size of 25% vs 60% at 6 months.8
Sample sizes were computed to ensure that there would be sufficient minority group subjects for a 5% level test (alpha) of proportions to achieve 90% power to detect a difference of 60% in the combined intervention arms versus 35% in the control arm using the method described by Donner et al.27
The method computes a sample size for independent observations then inflates the sample size 15% to account for the correlation among subjects in the same clinic. Briefly, the sample size calculations inflate the number of subjects needed for independent samples by a multiplicative factor to account for the correlation among subjects at the same clinic. The required sample size is n'=m*(1+(n-1)*κ) where m is the sample size computed assuming independent samples, n is the number of subjects in each clinic, κ is an estimate of the dependence among subjects in the same clinic, and n' is the sample size adjusted to take into account the correlation among subjects in a clinic. For continuous responses (SBP and DBP), likelihood-based mixed models with random patient effects will be fit in SAS Proc Mixed to incorporate all available data from baseline through 9 months in an intention-to-treat analysis. For BP control, a Generalized Estimating Equation (GEE) model using the binomial distribution and the logit link was fit in SAS Proc Genmod, accommodating the correlations across patients. This analysis accounts for the correlation among subjects from the same clinical center. The model will contain a term for treatment assignment (PPCM versus control), baseline systolic blood pressure, baseline diastolic blood pressure, and age. The analysis will provide an estimate and 95% confidence interval for the odds ratio of achieving BP control in the combined intervention arms compared to the control arm adjusted for any differences in baseline blood pressures and age. These calculations suggested that we would need to enroll 648 subjects in 27 clinics.
The primary comparison at 9 months will have a minimal detectable difference in BP control of 35% versus 50% with an alpha of 5% and 90% power and 35% versus 48% with 80% power. We assumed approximately 40% of subjects in the minority group, the minimal detectable difference for either of the two minority subgroups would be 35% versus 60% with 90% power and 35% versus 58% with 80% power. The model will contain a term for treatment assignment (intervention versus control), baseline BP and age. If a patient’s BP control status is missing at 9 months, that subject will be considered to have uncontrolled BP. We will perform a sensitivity analysis to determine the potential dependence of the results of the primary analysis on the missing values. This will include using multiple methods to impute the missing values to determine the effect of the choice of imputation method on the results. Methods will include worst case (patients with missing BP not controlled in the intervention group but patients with missing data have controlled BP in the control group), best case (missing assumed controlled in the intervention group and assumed not controlled in the control group), random assignment of outcome, and using regression methods to predict the outcome for each subject with a missing value based on their baseline values of known predictors of BP. Because randomization is by center there may be differences among the treatment groups with respect to other potent predictors of outcome including patient-, physician-, or clinic-specific factors. While the primary analysis will only include baseline BP and age, we will explore the effects of other potential covariates in a separate secondary analysis including: gender, race, education, insurance status, household income, marital status, smoking status, alcohol intake, body mass index, number of co-existing conditions at baseline, number of baseline antihypertensive medications, baseline medication adherence, and total number of clinic visits. We will first determine if there are significant (clinical or statistical) differences between the treatment groups with respect to any of these potential covariates. We will then repeat the GEE analysis described above but with the additional covariates identified in the previous step. In order to avoid multicolinearity, we will first fit univariate models to identify covariates that appear to have some influence on the effect of the treatment. We will then include those covariates whose univariate p-values are less than or equal to 0.2 in multivariate models. These analyses will provide estimates of the effect size for the intervention adjusted for differences among the treatment groups with respect to important potential confounders.