Using Aggregated Pharmacy Claims to Identify Primary Nonadherence

ABSTRACT

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

Take-Away Points

This study demonstrates the ability of aggregated pharmacy claims data available through the native electronic health record (EHR) to identify patients with primary nonadherence. The availability of multi-payer claims data within the EHR may serve as a foundation for ongoing medication monitoring and improving adherence in a nonintegrated primary care network.

Hypertension is the foremost modifiable risk factor for cardiovascular disease in US adults¹; however, despite its importance, achieving blood pressure control remains challenging in routine clinical practice.² Medication nonadherence is common and strongly associated with poor blood pressure control and adverse clinical outcomes.^3-6 Traditionally, claims-based measures have focused on adherence after filling at least 1 prescription, or secondary adherence.⁷ More recently, links between prescribing information and claims records indicate that up to 30% of patients do not fill an initial antihypertensive prescription, which is referred to as primary nonadherence.^8-12 Identifying patients who do not fill an initial prescription may be an important first step in addressing nonadherence and improving outcomes in clinical practice, particularly given recent evidence that outreach to patients with primary nonadherence can increase the proportion of patients who fill a prescribed medicine.⁷

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.

METHODS

Study Design

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.

Patient Population

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.

Analysis

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).

RESULTS

Figure

The shows the development of our cohort. Of 15,781 patients who received a new prescription for an antihypertensive medication, 791 patients met our study definition. The majority of patients (11,207; 71%) were excluded because they did not have available pharmacy fill records to address our study question. Of those excluded, 8721 had no imported pharmacy fill history, 1998 had no pharmacy claims recorded, and 488 had pharmacy records that had fewer than 30 days of observation following the index visit.

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).

DISCUSSION

Primary nonadherence is an increasingly identified barrier to optimal hypertension management. However, the ability to intervene has been limited by the lack of pharmacy claims records available in clinical practice in nonintegrated delivery systems. We used aggregated pharmacy claims data available through the primary care EHR to determine that one-third of patients prescribed a new antihypertensive medication in our cohort lacked evidence of filling the medication within 30 days. This is consistent with Fischer et al who identified a 28.4% incidence of primary nonadherence in a similar cohort.¹⁰ Lower rates of primary nonadherence have been found in studies within integrated delivery systems,^9,17 perhaps reflecting the ability of an integrated delivery system to capture more complete medication refill history. Although the Surescripts network aggregates the majority of payers in the United States, it may not have complete capture in our multi-payer environment. Differences in settings and methods of data capture may explain these differences in primary nonadherence rates.

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.

Limitations

The results of our study need to be taken in the context of its limitations. First, only 791 of 3284 patients who were prescribed an antihypertensive met our study definition. This was largely due to the fact that the pharmacy fill history was not available because the provider had not accessed the medication history after the new prescription. Second, indication is not attached to e-prescriptions. Although we only studied medications with a primary indication of hypertension, it is possible that some antihypertensives in our data set were also used for additional indications. Next, the completeness of our data is uncertain, as prescriptions paid for with cash or coupons, or those filled by pharmacies or pharmacy benefit managers who do not contribute to our source database, may not be available. We did find, however, that patients with more discrepancies at baseline were less likely to have evidence of a fill. This may indicate missing data, but it could also be that nonadherence to baseline medications at the time of an index prescription is associated with future nonadherence. However, the consistency of our rate of primary nonadherence with previous literature is reassuring. Finally, we do not have information regarding co-payments or out-of-pocket costs, which are known to be associated with nonadherence.⁹

Despite these limitations, our findings suggest that aggregated pharmacy claims available within a provider EHR may be useful in identifying patients with primary nonadherence in routine clinical practice. However, if interventions are meant to impact clinical care, the data must be sufficiently complete, accurate, and accessible in real time to clinicians with minimal interruptions to work flow. Ideally, identification and data sharing could be automated and presented to providers in a standardized and actionable format. For example, the EHR could generate a prompt for a follow-up call or a letter to be mailed if there is no evidence of a fill within a certain time period. Our results showed that the majority of patients who do fill their medications do so on the day it is prescribed, suggesting that interventions could be applied in the first few days following prescription. This approach has been used to improve the proportion of patients who fill statin prescriptions and could be broadened to other medication classes with appropriate supporting technology.⁷

Addressing nonadherence will undoubtedly require interventions that address a range of contributing factors including cost burden and access to medications; patient understanding, motivation, and behaviors; and the lack of coordinated care.^20,21 However, aggregated claims data within the native EHR could serve as the foundation to more appropriately identify patients demonstrating nonadherence in real time in clinical practice.

CONCLUSIONS

Primary nonadherence is associated with adverse clinical outcomes, yet can be difficult to measure in a multi-payer environment. Our study used aggregated pharmacy fill data to identify that nearly one-third of patients prescribed a new antihypertensive medication in our primary care cohort did not fill that medication within 30 days. Our findings suggest that the increased availability of medication fill histories in clinical practice can provide objective insight into a patient’s medication adherence, and may provide a foundation for targeted interventions to improve primary nonadherence.

Author Affiliations: Thomas Jefferson University, Jefferson College of Population Health (DC, JC), Philadelphia, PA; Christiana Care Value Institute (RA, PW, DE), Newark, DE; Department of Medicine, Christiana Care Health System (DE), Newark DE.

Source of Funding: This project was funded by the Delaware Health Sciences Alliance pilot award, project order #7. Dr Comer was individually funded by the PhRMA Foundation Postdoctoral Fellowship for Health Outcomes.

Author Disclosures: The authors report no relationship or financial interest with any entity that would pose a conflict of interest with the subject matter of this article.

Authorship Information: Concept and design (DC, DE, JC, RA); acquisition of data (DE, RA); analysis and interpretation of data (DC, PW, DE, RA); drafting of the manuscript (DC, PW, DE); critical revision of the manuscript for important intellectual content (DC, DE, JC, RA); statistical analysis (PW, DE); provision of patients or study materials (DE); obtaining funding (DE); administrative, technical, or logistic support (DE, JC); and supervision (DE, JC).

Address correspondence to: Daniel Elliott, MD, MSCE, Associate Chair for Research, Department of Medicine, Christiana Care Health System, 4755 Ogletown-Stanton Rd, Newark, DE 19718. E-mail: delliott@christianacare.org.

REFERENCES

1. Lloyd-Jones D, Adams RJ, Brown TM, et al; American Heart Association Statistics Committee and Stroke Statistics Subcommittee. Heart disease and stroke statistics—2010 update: a report from the American Heart Association. Circulation. 2010;121(7):e46-e215.

2. Vital signs: awareness and treatment of uncontrolled hypertension among adults—United States, 2003—2010. CDC website. http://www.cdc.gov/mmwr/preview/mmwrhtml/mm6135a3.htm. Published September 7, 2012. Accessed April 18, 2014.

3. Schroeder K, Fahey T, Ebrahim S. How can we improve adherence to blood pressure-lowering medication in ambulatory care? systematic review of randomized controlled trials. Arch Intern Med. 2004;164(7):722-732.

4. Vrijens B, Vincze G, Kristanto P, Urquhart J, Burnier M. Adherence to prescribed antihypertensive drug treatments: longitudinal study of electronically compiled dosing histories. BMJ. 2008;336(7653):1114-1117.

5. Dragomir A, Côté R, Roy L, et al. Impact of adherence to antihypertensive agents on clinical outcomes and hospitalization costs. Med Care. 2010;48(5):418-425.

6. Jackevicius CA, Li P, Tu JV. Prevalence, predictors, and outcomes of primary nonadherence after acute myocardial infarction. Circulation. 2008;117(8):1028-1036.

7. Derose SF, Green K, Marrett E, et al. Automated outreach to increase primary adherence to cholesterol-lowering medications. JAMA Intern Med. 2013;173(1):38-43.

8. Lagu T, Weiner MG, Eachus S, Tang SS, Schwartz JS, Turner BJ. Effect of patient comorbidities on filling of antihypertensive prescriptions. Am J Manag Care. 2009;15(1):24-30.

9. Shah NR, Hirsch AG, Zacker C, et al. Predictors of first-fill adherence for patients with hypertension. Am J Hypertens. 2009;22(4):392-396.

10. Fischer MA, Stedman MR, Lii J, et al. Primary medication non-adherence: analysis of 195,930 electronic prescriptions. J Gen Intern Med. 2010;25(4):284-290.

11. Tamblyn R, Eguale T, Huang A, Winslade N, Doran P. The incidence and determinants of primary nonadherence with prescribed medication in primary care: a cohort study. Ann Intern Med. 2014;160(7):441-450.

12. Storm A, Andersen SE, Benfeldt E, Serup J. One in 3 prescriptions are never redeemed: primary nonadherence in an outpatient clinic. J Am Acad Dermatol. 2008;59(1):27-33.

13. Grant RW, Singer DE, Meigs JB. Medication adherence before an increase in antihypertensive therapy: a cohort study using pharmacy claims data. Clin Ther. 2005;27(6):773-781.

14. Fischer MA, Choudhry NK, Brill G, et al. Trouble getting started: predictors of primary medication nonadherence. Am J Med. 2011;124(11):1081.e9-e22.

15. Elixhauser A, Steiner C, Harris DR, Coffey RM. Comorbidity measures for use with administrative data. Med Care. 1998;36(1):8-27.

16. Comer DM, Couto J, Aguiar R, Wu P, Elliott D. Usefulness of pharmacy claims for medication reconciliation in primary care. Am J Manag Care. 2015;21(7):486-493.

17. Raebel MA, Ellis JL, Carroll NM, et al. Characteristics of patients with primary non-adherence to medications for hypertension, diabetes, and lipid disorders. J Gen Intern Med. 2012;27(1):57-64.

18. Unni E, Farris KB. Determinants of different types of medication non-adherence in cholesterol lowering and asthma maintenance medications: a theoretical approach. Patient Educ Couns. 2011;83(3):382-390.

19. Choudhry NK, Fischer MA, Avorn J, et al. The implications of therapeutic complexity on adherence to cardiovascular medications. Arch Intern Med. 2011;171(9):814-822.

20. Steiner JF. Rethinking adherence. Ann Intern Med. 2012;

157(8):580-585.

21. Baroletti S, Dell’Orfano H. Medication adherence in cardiovascular disease. Circulation. 2010;121(12):1455-1458.