The American Journal of Managed Care December 2015
Using Aggregated Pharmacy Claims to Identify Primary Nonadherence
Objectives: Aggregate pharmacy claims available within an electronic health record (EHR) provide an opportunity to understand primary nonadherence in real time. The objective of this study was to use pharmacy claims data available within the EHR to identify the prevalence and predictors of primary nonadherence to antihypertensive drug therapy in a multi-payer primary care network.
Study Design: We conducted a retrospective cohort study of patients prescribed a new antihypertensive medication in a large primary care practice network between January 2011 and September 2012.
Methods: We matched prescriptions for the new antihypertensive to pharmacy claims listed in the EHR. The primary outcome was the presence of a fill for the new medication within 30 days of the prescription.
Results: Of 791 patients in our study cohort, two-thirds (522; 66%) filled their prescription within 30 days. The majority (409; 78.4%) of that group filled the prescription on the day it was issued. Lower diastolic blood pressure and Medicare coverage increased the probability of nonadherence.
Conclusions: Medication fill data within the provider EHR can identify primary nonadherence in clinical practice. As adoption of this technology increases, it provides an opportunity to identify nonadherence, allowing for the effective design of interventions to improve adherence to therapy.
Am J Manag Care. 2015;21(12):e655-e660
Primary nonadherence research has relied largely on pharmacy claims within integrated delivery systems or health plans.9,13,14 Historically, providers in most primary care practices, particularly those in multi-payer environments, do not have access to these data to identify or monitor for nonadherence. The recent adoption of electronic-prescribing (e-prescribing) systems has made prescription fill information increasingly available to providers within their native electronic health record (EHR). This access to aggregated, multi-payer pharmacy data creates an opportunity to identify and address primary nonadherence in clinical practice, possibly even in real time.
The objective of this study was to use aggregated pharmacy claims available within the practice-based EHR to identify the prevalence and predictors of primary nonadherence to antihypertensive therapy in primary care practice in a nonintegrated delivery system. Our findings will help to determine the potential value of aggregated pharmacy data in identifying primary nonadherence in routine clinical practice. This information could be the foundation for clinically meaningful and timely interventions to improve medication adherence.
We conducted a retrospective cohort study of patients prescribed a new antihypertensive medication in a primary care practice. We documented evidence of the first prescription fill from pharmacy claims available through the e-prescribing system within the provider EHR. The study was approved by the Institutional Review Board of Christiana Care Health System.
Study Setting and Data Sources
We conducted this study in a large, multi-specialty medical practice that includes 14 primary care sites providing care for more than 100,000 patients in northern Delaware and the surrounding communities. These practices share an EHR (Centricity, GE Healthcare), used for all clinical encounters, which contains a complete list of problems, prescriptions, allergies, office notes, and diagnostic test results. Additionally, all prescriptions in the practices are generated through the EHR, with e-prescribing in place since 2010. Through the e-prescribing platform, physicians or designees within the practice can request up-to-date pharmacy claim histories through Surescripts, the largest health information network in the United States. Accessing medication histories requires patient consent, which was incorporated into the patient registration process. The Surescripts medication history query provides aggregate pharmacy claims from the majority of retail and mail-order pharmacies, as well as from pharmacy benefit managers for private and public insurers. Once accessed, the medication history is available in the EHR.
We identified established patients with hypertension who were prescribed a new antihypertensive agent from January 2011 to September 2012. Patients included in the study needed to have a formal diagnosis of hypertension (International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM] code 401.xx) or elevated blood pressure (ICD-9-CM code 796.2) before the time of the index prescription. We required patients to have at least 1 primary care visit recorded in the EHR in the 18 months prior to the index visit date in order to avoid including patients whose prescriptions appeared to be new but actually represented a patient transitioning to a new provider or a practice transitioning from paper records to electronic records.
We excluded prescriptions if there was a previous prescription for a medication within the same class to avoid changes that were made in response to formulary requirements. To ensure sufficient observation time, we excluded patients whose pharmacy fill history was obtained less than 30 days after the index prescription, unless the pharmacy fill history showed evidence of filling the index prescription. We required patients to have evidence of at least 1 medication fill at any time in the entire pharmacy fill history, in order to minimize incorrectly labeling patients as nonadherent when the aggregated data source may not have included the patient’s pharmacy or payer information.
Outcome and Key Covariates
Our primary outcome was primary nonadherence, which we defined as having no evidence of filling the index prescription within 30 days. Previous studies have used 1 to 9 months as a cut point to classify primary nonadherence,8,11,14 but we specified 30 days as our time frame to maximize the clinical relevance of the information. The EHR provided patient demographics (age, gender, race, and insurance type) and clinical data (blood pressure, the total medication count, and the number of active antihypertensive medications at the time of the index visit). We measured comorbidity burden using the Elixhauser comorbidity index based on diagnoses related to previous encounters in our system.15 We measured previous healthcare utilization within our system as the count and time to prior-year hospitalizations, emergency department visits, and primary care provider visits. From a previous related study, we included a count of medication discrepancies between the medication lists recorded in the EHR and pharmacy claims at the time of the index prescription for a subgroup of our study cohort.16 In the previous study, we defined medication discrepancies as either a medication that was listed on the EHR prescriber list but did not have a corresponding fill or a filled medication that was not listed on the EHR prescriber list.
We used descriptive statistics to explain population characteristics associated with primary nonadherence at 30 days. We developed univariate and multivariate logistic regression models to examine the relationship of selected covariates with primary nonadherence. We used the same approach in the subset of patients for whom the number of medication discrepancies at the time of the index visit was available to determine sensitivity of results to potential EHR errors. All analyses were done with SAS version 9.3 (SAS Institute, Cary, North Carolina).
Table 1 describes our cohort. Two-thirds of the study population (522/791; 66%) filled prescriptions within the first 30 days. The majority of patients (409/522; 78.4%) who filled their prescriptions within 30 days filled it on the same day the prescription was issued.
Table 2 shows the results of regression analyses for predicting first-fill nonadherence. In the univariate analysis, patients without evidence of a fill were more likely to be older (odds ratio [OR], 1.01; 95% CI, 1.00-1.02), have Medicare (OR, 1.61; 95% CI, 1.16-2.23), and have a higher number of medications (OR, 1.04; 95% CI, 1.00-1.07). Having a higher diastolic blood pressure at the index visit was associated with a lower rate of nonadherence (OR, 0.98; 95% CI, 0.97-1.00). In the multivariate model, increasing diastolic blood pressure was found to be significantly associated with decreased nonadherence (OR, 0.98; 95% CI, 0.97-1.00) and having Medicare was found to be significantly associated with increased nonadherence (OR, 1.57; 95% CI, 1.02-2.42). The results of our subgroup analysis contained 502 patients for whom the number of medication discrepancies was available. Patients with more medication discrepancies at baseline were more likely to meet our definition of nonadherence (OR, 1.30; 95% CI, 1.15-1.46).
We identified the association of primary nonadherence with increasing medication burden, age, and noncardiovascular comorbidities. This is consistent with previous literature suggesting that medication nonadherence may increase as competing comorbidities, particularly with active symptoms, take precedence over asymptomatic conditions such as hypertension.18 Although the evidence for association of primary nonadherence with comorbidity is mixed, with some evidence suggesting increased primary nonadherence with a higher medication burden,19 our study and others suggest the opposite.11
Interestingly, we identified that patients with lower diastolic blood pressure were more often nonadherent. The association may indicate a perceived lack of urgency, particularly given the asymptomatic nature of hypertension, although this is not clear from our data. Specific interventions for those with lower grades of hypertension may need to include education on the importance of adherence to prevent worsening hypertension.