Objectives: To evaluate whether provider adoption and use of the formulary decision support element of e-prescribing were directly associated with significant prescriber and patient behavior change.
Study Design: Retrospective, observational prepost study comparing a treatment group with a control group.
Methods: Pharmacy fill information and provider enrollment records from January 2009 to December 2011 were used to compare fill volumes—primary fill and subsequent refill volume by retail and mail order fulfillment channels—of prescribers who had adopted e-prescribing (treatment group) with those who had not (control group). Propensity scores were used to match treatment and control groups by baseline prescribing patterns. Primary outcomes of interest were prepost differences for 90-day retail fills, initial fills (primary adherence), and subsequent refills.
Results: E-prescribers demonstrated significantly higher average increases in 90-day retail fills (+2.8 fills per provider, P <.001), primary adherence (+12.0 fills per provider, P <.001), and subsequent adherence (second fill: +4.5 fills per provider, P <.001; and third fill: +2.4 fills per provider, P <.001) compared with non–e-prescribers.
Conclusions: Provider adoption and use of e-prescribing applications with formulary decision support were associated with significant prescriber and patient behavior change in a large, nationally representative population. E-prescribing adoption was positively associated with prescribers’ selection of longer supply prescriptions and improvements in adherence. E-prescribing infrastructure may allow for additional provider messaging opportunities designed to improve quality of care and patient outcomes.
Am J Pharm Benefits. 2016;8(4):133-140

Pharmacy benefit managers (PBMs) are critical stakeholders in the healthcare industry, responsible for administering prescription benefit programs for the vast majority of insured Americans. Prescription drug spending represents 10% of total healthcare expenditures, or roughly 1.8% of US gross domestic product.1

Importantly, prescription drugs are first- or second-line treatments for many chronic diseases associated with an outsized portion of healthcare spend.2,3 Appropriate and persistent use of prescription drugs to treat chronic diseases has also been demonstrated to reduce downstream medical expenditures,4 achieving substantial cost savings.5,6

In addition, pharmacies are the most frequently utilized healthcare service, with the average patient filling nearly 4 times as many prescriptions as the number of their visits to office-based physicians.7,8 Given these facts and additional demographic context—an aging population with increasing prescription drug use on a per capita basis9,10—it becomes clear that PBMs can play a significant role in the current healthcare reform landscape.11,12

PBMs are adopting benefit design strategies that reflect the broader healthcare trend of aligning payment to value and outcomes. A common theme among the evolving strategies is an effort to address and improve patient adherence to medication. Due to their position of processing prescription claims, PBMs are in a unique position to have the most comprehensive view of a patient’s prescription treatment regimen.
PBMs can leverage claims, eligibility, pharmacy data, and provider data to understand prescribing patterns and patient behavior by identifying gaps in care and suboptimal adherence, and potentially to effect positive behavior change relating to medication adherence through the timely delivery of information to providers, patients, and pharmacies.
Surescripts is a health information network that connects PBMs, pharmacies, and providers using electronic health records (EHRs) for the purpose of enabling real-time electronic prescribing. Through Surescripts, PBMs can electronically deliver prescription formulary, eligibility, and co-pay information to providers using a certified EHR. MedImpact Healthcare Systems, Inc, is one such PBM, managing pharmacy benefits for more than 47 million individuals nationwide, connected to the Surescripts network. Provider use of the benefit information through Surescripts is associated with increased levels of preferred brand and generic prescribing, resulting in lower co-pays that are in turn linked to improved adherence.13,14
Surescripts also provides PBM-supplied decision support information about availability of retail 90-day supply and mail-order benefits, both of which have been linked to improved adherence.15,16 The ability of providers to route prescriptions electronically provides an auditable trail between what the provider prescribed and whether the patient filled the prescription at the pharmacy, addressing a current challenge with quality measures relating to primary adherence.17,18
Additionally, provider use of electronic prescribing (e-prescribing) is a core measure in the EHR Incentive Program, a federal incentive program to promote provider adoption and meaningful use of EHRs.19 The e-prescribing core measure requirement is currently limited to the routing component of e-prescribing; however, e-prescribing infrastructure allows numerous opportunities for formulary decision support and clinical messaging.
Current widely used methods of measuring medication adherence (such as proportion of days covered or medication possession ratio) rely on the presence of a claim for the first prescription fill, which is available only if the patient successfully picked up the initial prescription from the pharmacy. A challenge with these methods is that as many as 28% of first-fill prescriptions are abandoned before they reach the pharmacy, and therefore, such measures may be overstating true levels of medication adherence.20
Our study is novel in that we sought to understand whether provider adoption and use of the formulary decision support element of e-prescribing was directly associated with improved medication adherence as indicated by increases in the number of claims for 90 days’ supply, first fill, and subsequent refills. We did this by pairing data from Surescripts’ provider network and MedImpact’s pharmacy information.
We investigated 3 core questions: first, we assessed whether e-prescribing resulted in prescribing behavior change, as indicated by selection of 90-day prescriptions; second, whether e-prescribing is associated with an increase in primary adherence, as indicated by the number of claims for first-fill prescriptions; and third, whether there was a discernible subsequent impact on second and third medication fills.
Study Design
A prepost, retrospective analysis was performed by linking summary prescriber-level fill information from MedImpact and prescriber enrollment data from Surescripts. MedImpact’s pharmacy fill information from January 2009 to December 2011 was summarized to monthly prescriber-level data.
More than 75 variables were aggregated by prescriber, using National Provider Identifier (NPI) codes, to the following levels: a) fulfillment channel and days’ supply amount: retail 30-day supply, retail 90-day supply, and mail 90-day supply; and b) adherence defined as the number of prescriptions filled for members’ first, second, and third fills. We measured the raw volume of these claims associated with each prescriber (treatment and control) for 1-year pre- and post e-prescribing adoption.
Total prescription claims was defined as the sum of retail 30-day supply, retail 90-day supply, and mail pharmacy claims for primary and subsequent fills. For the purpose of this analysis, comparisons of prescribing behavior were conducted based on a treatment versus control group research design. Prescribers adopting e-prescribing were included in the treatment group.
Prescribers not using e-prescribing technology were included in the control group. The pre-period was defined as the 12 months prior to a provider’s first use of the e-prescribing formulary decision support element, and the subsequent 12 months was defined as the post period. Differences among provider-level variables over the pre- and post periods were assessed to measure the impact of electronic prescribing on the study population. The primary outcomes of interest were prepost differences in claims for 90-day retail fills, initial fill (primary adherence), and subsequent refills.
Study Cohort Selection
A 2-stage matching algorithm was employed to enhance statistical rigor and to include adjustments for baseline differences in prescribing behavior between treatment and control groups. The goal of the 2-stage matching approach was to improve the comparability of the treatment and control groups while limiting potential confounding.21
Robust matching procedures combined simulation-based adjustments (1-stage matched methods), and propensity score matching (2-stage matched methods) was used to minimize the effect of bias that may have occurred in the absence of such advanced statistical techniques. Both 1- and 2-stage matched results are presented to describe the impact of advanced matching methods between e-prescribers and non–e-prescribers, and to validate findings of 1-stage matching methods.
1-Stage Matched Methods
First, an adoption month and year were randomly assigned to each control group prescriber to simulate the introduction of e-prescribing into the population and to control for rapid changes in adoption of e-prescribing during the study period. This approach was used to avoid measuring pre- and post prescribing rates for a control prescriber in January 2010 to a treatment prescriber in December 2010.
Random assignment of proxy exposure dates to unexposed populations is a research design used to identify the onset of risk for exposure to the unexposed control group.22 Frequency distributions were calculated based on eligibility transaction dates from providers in the treatment group. Each control group prescriber was assigned a proxy eligibility date to replicate the observed distribution of the treatment group. Establishing this proxy or simulated eligibility transaction date in the control group created a basis for comparison over the study period between the treatment and control groups.
2-Stage Matched Methods
Second, propensity score methods were used to balance treatment and control groups by baseline prescribing patterns.23 Propensity scores were based on estimated probabilities from a multivariable logistic regression model used to predict the probability of prescribers’ membership in the treatment group. Covariates placed in the generalized linear model with logit link function included baseline prescribing patterns for retail 90-day, retail 30-day, mail 90-day, first fill, second fill, and third fill.
A 1-to-1 matching algorithm was performed using the nearest available neighbor without replacement method to match control group prescribers to prescribers in the treatment group based on the propensity scores of the treatment group.24
Statistical Analysis
Outcomes were calculated for both 1-stage and 2-stage matching scenarios. The treatment group was compared with 1- and 2-stage matched control groups. Changes for each outcome, day supply, and adherence were calculated to determine the relative percent difference in claim volume for each group, and the absolute difference of percent differences between treatment and control groups. Relative percent difference for each group was determined as %▲ = (Post –Pre)/Pre. For example, the treatment group was measured as follows: %▲T = (PostTreatment – PreTreatment)/PreTreatment.
Absolute percent differences between treatment and control groups were defined as e-prescribing valuen = %▲T  – %▲Cn, where n = 1 for 1-stage matching and n = 2 for 2-stage matching. E-prescribing value serves as a proxy to measure the impact of e-prescribing on each measure, as the differences among the treatment and control groups were estimated to be attributed to e-prescribing. Postperiod data were compared with baseline results by group using parametric t tests. Statistical significance was determined at P <.05 level. All analyses were conducted with SAS version 9.4 (SAS Institute, Inc, Cary, North Carolina).
Derivation of Study Cohorts
Selection of the treatment (e-prescribers) and control (non–e-prescribers) groups included a multistep exclusion process (Figure 1). Approximately 20,372 prescribers who had adopted e-prescribing during the 2010 calendar year were identified using the Surescripts provider network. E-prescribing adoption was evidenced by a prescriber’s registration, activation, and first e-prescribing transaction. To be selected, prescribers must have completed at least 1 eligibility transaction with MedImpact through the Surescripts network.
An eligibility transaction is a request for drug coverage and formulary information generated at the point of care and routed electronically to the corresponding PBM through the Surescripts network. The eligible provider must have been successfully matched in the Surescripts directory and must have had at least 25 claims, decreasing the sample by 3404 providers.
Eligibility transactions dates were divided into 2 periods: the 12 months prior to the first eligibility transaction date (pre-period), and the 12 months following the first eligibility transaction date (post period). Prescribers with claim totals of 0 in either the pre- or post periods (426) were excluded.
As both treatment and control groups were expected to have no e-prescription claims in the pre-period, such providers were removed from the population. Omitting e-prescription claims derived from integrated healthcare delivery systems led to an additional reduction in prescriber count by 10,247 providers. The final treatment group consisted of 6295 providers.
The control group was selected from providers who had not adopted e-prescribing prior to and during the study period by using the nationwide Surescripts provider directory. 63,305 prescribers were removed in the exclusion process using a combination of the provider’s Drug Enforcement Administration number, NPI code, and last name to remove all prescribers registered on the Surescripts network, in addition to removing prescribers with fewer than 25 claims. 
This resulted in retaining only the providers who had not adopted e-prescribing during the study period, with more than 25 claims. The final control group consisted of 18,934 providers after removing 810 providers due to 0 total claims in the pre- or post period, and 16,591 providers due to affiliations with integrated healthcare systems.
Prepost Fill Volume
Average prepost volume of retail 90-day fills increased by 3.3 fills per provider (Rx/provider) in the treatment group versus 0.5 Rx/provider in the 1-stage matched control group, an absolute difference of +2.8 Rx/provider (P <.001) (Table). The absolute difference between treatment and 2-stage matched control group was +3.2 Rx/provider (P <.001). Increases in retail 30-day fills were greater in the treatment group. The overall absolute difference in retail 30-day fills indicated greater increases among e-prescribers, 16.4 Rx/provider (P <.001) for 1-stage matched comparisons and a 21.2 Rx/provider (P <.001) change for 2-stage matched comparisons.
As shown in Figure 2, estimated e-prescribing value, as measured by differences in retail 90-day fills, ranged from 17.2% to 29.9%, indicating that prepost differences were substantially higher for the treatment group compared with both control groups. This trend was consistent for both mail 90-day and retail 30-day measures as well.
Although the absolute difference from pre- to post periods was not statistically significant and indicated a decrease in mail 90-day fills of 0.1 Rx/provider (P <.846) for the treatment group, significant differences using 2-stage matched methods showed an overall absolute increase for the treatment group in mail 90-day fills of +1.8 Rx/provider (P <.001). Calculations of percent difference for mail 90-day indicated a decrease of 10.0% for the treatment group, a decrease of 23.4% for the control group 1-stage, and a decrease of 29.1% for control group 2-stage.
Decreases in mail fills for the treatment group were significantly less than decreases for the control groups (1- and 2-stage), yielding estimated e-prescribing values ranging from 13.4% to 19.1% for 1- and 2-stage matched results, respectively. E-prescribing value for retail 30-day ranged from 6.2% to 17.4%, indicating greater increases in retail for the treatment group compared with both control groups (1- and 2-stage).
E-prescribing providers experienced significantly higher increases in both primary and subsequent fills compared with the 1-stage and 2-stage control groups (Table). For primary adherence, prepost differences between treatment and control groups were 12.0 Rx/provider (P <.001) and 14.9 Rx/provider (P <.001), respectively, for 1-stage and 2-stage matched groups. This was observed for subsequent (second and third) fills as well. The treatment group exceeded the 1-stage control group in second fills by 4.5 Rx/provider (P <.001) and by 5.8 Rx/provider (P <.001) for 2-stage matched groups.
For third fills, increases in the treatment group were significantly higher than in the control groups. Differences between treatment group and 1-stage matched groups were 2.4 Rx/provider (P <.001) and greater differences occurred for 2-stage matched groups (3.0 Rx/provider P <.001). Overall, the estimated value of e-prescribing on primary adherence ranged from 14.3% to 20.5%, indicating greater increases in primary fills for e-prescribers (Figure 2). E-prescribing value on subsequent fills ranged from 10.9% to 28.3% for second fill and 5.9% to 23.8% for third fills, indicating significantly greater results for the treatment group when compared with the control groups.
Nonadherence to medications results in substantial downstream medical costs, and therefore warrants the significant attention and accompanying investments that payers and care providers have created with programs and resources to improve patient adherence. Our findings suggest that provider adoption and use of e-prescribing applications with formulary decision support were associated with improvements in medication adherence in a nationally representative population.
Specifically, we found that e-prescribing adoption was positively associated with providers’ selection of longer-supply prescriptions, which in turn has been linked to increased medication adherence, and improvements in adherence to first fill, as well as subsequent fills. Considering the growth in e-prescribing adoption in 2013, with close to 640,000 prescribers routing roughly half of all prescriptions electronically, the potential improvements in medication adherence and health outcomes are significant.
We observed a marked increase among e-prescribers in the selection of 90-day supply prescriptions when compared with the control group. Importantly, the increase in 90-day supply prescriptions—both to mail (+13.4% to +19.1%) and retail pharmacies (+17.2% to +29.9%)— among e-prescribers was comparatively greater than that of 30-day supply prescriptions. When providers adopt EHRs and are made aware of 90-day channel (mail and at retail) benefit information conveyed through e-prescribing, they may be more likely to: a) pay attention to this information when presented, and b) shift their overall prescribing habits and choose 90-day rather than 30-day prescriptions for maintenance medications.
Our findings show that providers who adopted e-prescribing during the study period saw a substantial increase in primary medication adherence (+14.3% to +20.5% in first fills), suggesting that electronically routed prescriptions increased the likelihood of patients starting therapy. This finding is consistent with other studies observing that between 22% and 31% of paper prescriptions (ie, those printed out and handed to patients) do not make it to the pharmacy.20,26,27
E-prescription routing addresses this issue of “leakage,” which is a contributing factor to the increase in primary adherence. Many pharmacies have developed outreach programs to remind patients that their prescriptions are ready, programs that are likely benefited by the increase in prescriptions submitted electronically. We suspect that prescriber use of the benefit information from e-prescribing, and patient knowledge that the prescription is likely to have lower co-pays (due to increased use of generic prescriptions and 90-day supply prescriptions), were also factors in the improved primary adherence numbers.
We observed a positive spillover adherence effect, as subsequent increases in second fills (+10.9% to +28.3%) and third fills (+5.9% to +23.8%) were also higher among patients whose providers e-prescribed. This spillover effect makes intuitive sense, as any significant increase in primary adherence is likely to result in increases in longer-term adherence, which is consistent with discontinuation patterns of medication adherence.28,29
E-prescribing increases the number of prescriptions making it to the pharmacy. Providing greater visibility into the prescription fulfillment process, in particular the discrepancy between what is prescribed versus what is picked up by the patient, has the paradoxical effect of revealing higher abandonment rates for primary, or first fill, prescriptions. Quality organizations should consider the policy implications of advocating for adherence measures that may inadvertently harm or reduce performance scores from e-prescribers.
This issue is becoming more salient given 3 additional trends: First, that roughly three-quarters of all office-based prescribers now actively e-prescribe, and this proportion is likely to continue increasing given federal incentive programs to stimulate meaningful use of EHRs30,31; second, that e-prescribing use is increasing among those who have adopted it; and third, that many payer organizations, including Medicare, are instituting pay-for-performance programs that rely on adherence quality measures.32
In a period of significant change within the healthcare system, stakeholders—including the federal government, private health insurers, providers, pharmacies, patients, and PBMs—are all investing in health information technology. EHRs can be leveraged to connect providers in new ways with richer information that can facilitate improved care coordination and enable population health management.
PBMs and providers are leveraging these connections through e-prescribing to enhance prescribed pharmacotherapy. The ability to connect providers and PBMs to enable real-time electronic messaging via the e-prescribing channel could be leveraged to improve prior authorization processes and medication adherence interventions.
Further investigation on how PBMs and providers can better use e-prescribing communication channels is critical. Our results present additional questions worthy of further investigation, including: a) whether some EHR systems are inherently more functional than others, and therefore may drive better outcomes; b) whether other advanced functionalities within e-prescribing, such as medication history or electronic prior authorization, are associated with positive clinical outcomes; and c) whether previous findings of e-prescribing reducing costs can be substantiated, given that e-prescribing adoption has reached maturity.
First, the prepost study design used in this analysis assumes that prescribers in the treatment and control groups would have demonstrated similar prescribing behavior in the absence of electronic prescribing. To address this assumption, our approach used simulation-based adjustments and propensity scoring methods to produce comparable pre-period comparison groups; however, we acknowledge residual confounding due to selection bias may have remained.
Second, although covariates derived from a combination of provider demographics are often used to match comparison groups, such demographic data were unavailable in this study. However, baseline utilization and prescriber patterns were used to maximize matching of providers between groups. Third, we recognize that patients’ actual medication adherence may differ from estimated rates using pharmacy claims data; however, adherence estimates by way of pharmacy claims has been well validated in integrated health systems.33,34
We linked nationally representative prescription information from a PBM with an e-prescribing network to examine changes in prescribing patterns and patient adherence after e-prescribing adoption. We found greater increases in initial fill rates, refill rates, and 90-day supply prescribing among prescribers adopting technology, compared with prescribers not adopting e-prescribing. E-prescribing infrastructure allows for additional provider messaging opportunities intended to increase efficiencies and improve patient outcomes.
Author Affiliations: Surescripts, LLC (JYS, SBJ, MS), Arlington, VA; MedImpact Healthcare Systems, Inc (RSL, BVP, CK), San Diego, CA.
Source of Funding: No funding was received for conducting this research. Author Disclosures: Ms Smith, Mr Joseph, and Mr Sow are employees of Surescripts, LLC. Dr Patel, Dr Leslie, and Mr Kucharski are employees of MedImpact Healthcare Systems, Inc, which provides e-prescribing–related supportive services.
Authorship Information: Concept and design (RSL, BVP, CK, MS, JYS, SBJ); acquisition of data (RSL, BVP, CK, MS, SBJ); analysis and interpretation of data (RSL, BVP, CK, JYS); drafting of the manuscript (MS, JYS, SBJ); critical revision of the manuscript for important intellectual content (RSL, BVP, MS, JYS, SBJ); statistical analysis (JYS); provision of patients or study materials (RSL, BVP, CK); administrative, technical, or logistic support (MS, SBJ); and supervision (RSL, BVP, MS, SBJ).
Address correspondence to: Jaime Y. Smith, MAE, Surescripts, LLC, 2800 Crystal Dr, Ste 800, Arlington, VA 22202. E-mail: jaime.smith@surescripts.com.
  1. National Center for Health Statistics. Health, United States 2012: with special feature on emergency care. Hyattsville, MD: Centers for Disease Control; 2013.  http://www.cdc.gov/nchs/data/hus/hus13.pdf
  2. Stanton MW. The high concentration of US health care expenditures [AHRQ Pub. No. 06-0060]. Agency for Healthcare Research and Quality website. http://meps.ahrq.gov/mepsweb/data_files/publications/ra19/ra19.pdf. Published June 2006. Accessed February 29, 2016.
  3. Salmun, LM. Medco therapeutic resource centers: driving growth through clinically compelling, specialist services. Presented at: Xavier University of Louisiana College of Pharmacy’s Annual Health Disparities Conferences; April 19, 2009; New Orleans, LA. http://xula09.the1joshuagroup.com/XULA09%20-%20Salmun,%20Luis.pdf. Accessed February 29, 2016.
  4. Sokol MC, McGuigan KA, Verbrugge RR, Epstein RS. Impact of medication adherence on hospitalization risk and healthcare cost. Med Care. 2005;43(6):521-530.
  5. Roebuck MC, Liberman JN, Gemmill-Toyama M, Brennan TA. Medication adherence leads to lower health care use and costs despite increased drug spending. Health Aff (Millwood).  2011;30(1):91-99. doi: 10.1377/hlthaff.2009.1087.
  6. Stuart B, Loh FE, Roberto P, Miller LM. Increasing Medicare part D enrollment in medication therapy management could improve health and lower costs. Health Aff (Millwood). 2013;32(7): 1212-1220. doi: 10.1377/hlthaff.2012.0848.
  7. Schappert SM, Rechtsteiner EA. Ambulatory medical care utilization estimates for 2007. Vital Health Stat 13. 2011;(169):1-38.
  8. Retail prescription drugs filled at pharmacies (annual per capita by age). Kaiser Family Foundation website. http://kff.org/other/state-indicator/retail-rx-drugs-by-age/. Published 2013. Accessed April 15, 2014.
  9. Kaufman DW, Kelly JP, Rosenberg L, Anderson TE, Mitchell AA. Recent patterns of medication use in the ambulatory adult population of the United States: the Slone survey. JAMA.  2002;287(3):337-344.
  10. Centers for Disease Control and Prevention. Trends in aging—United States and worldwide.  MMWR Morb Mortal Wkly Rep. 2003;52(6):101-104,106.
  11. Morgan SG. Prescription drug expenditures and population demographics. Health Serv Res.  2006;41(2):411-428.
  12. Merlis M. Explaining the growth in prescription drug spending: a review of recent studies.  Presented at: Pharmaceutical Pricing Practices, Utilization, and Costs Conference; August 9, 2000; Washington, DC.  Accessed June 2014.
  13. Fischer MA, Vogeli C, Stedman M, Ferris T, Brookhart MA, Weissman JS. Effect of electronic prescribing with formulary decision support on medication use and cost. Arch Intern Med. 2008;168(22):2433-2439. doi:10.1001/archinte.168.22.2433.
  14. Shrank WH, Choudhry NK, Fischer MA, et al. The epidemiology of prescriptions abandoned at the pharmacy. Ann Intern Med. 2010;153(10):633-640. doi: 10.7326/0003-4819-153-10-201011160-00005.
  15. Iyengar R, Henderson R, Visaria J, Glave Frazee S. Dispensing channel and medication adherence: evidence across 3 therapy classes. Am J Manag Care. 2013;19(10):798-804.
  16. CVS Caremark. Consumer benefits of receiving medication through the mail. The Washington Post website. http://www.washingtonpost.com/sf/brand-connect/wp/2014/03/17/consumer-benefits-of-receiving-medication-through-the-mail/. Published March 17, 2014. Accessed November 13, 2014.
  17. Brennan TA, Dollear TJ, Hu M, et al. An integrated pharmacy-based program improved medication prescription and adherence rates in diabetes patients. Health Aff (Millwood).  2012;31(1):120-129. doi: 10.1377/hlthaff.2011.0931.
  18. Khandelwal N, Duncan I, Rubenstein E, Ahmed T, Pegus C. Community pharmacy and mail order cost and utilization for 90-day maintenance medication prescriptions. J Manag Care Pharm. 2012;18(3):247-255.
  19. Eligible Professionals Meaningful Use Core Measures: measure 4 of 13—stage 1 (2014 definition). CMS website. http://www.cms.gov/Regulations-and-Guidance/Legislation/EHRIncentivePrograms/downloads/4_e-prescribing.pdf. Updated May 2014. Accessed November 2014.
  20. 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. doi: 10.1007/s11606-010-1253-9.
  21. Dugoff EH, Schuler M, Stuart EA. Generalizing observational study results: applying propensity score methods to complex surveys. Health Serv Res. 2014;49(1):284-303. doi: 10.1111/1475-6773.12090.
  22. Harvey R, Jankus DD, Mosley D; UnitedHealthcare. Random assignment of proxy event dates to unexposed individuals in observational studies: an automated technique using SAS. Midwest SAS Users Group website. http://www.mwsug.org/proceedings/2012/PH/MWSUG-2012-PH02.pdf. Published 2012. Accessed March 1, 2014.
  23. Leslie RS, Ghomrawi H. The use of propensity scores and instrument variable methods to adjust for treatment selection bias. Presented at: SAS Global Forum; March 16-19, 2008; San Antonio, TX. http://www2.sas.com/proceedings/forum2008/366-2008.pdf. Accessed June 2014.
  24. Coca-Perraillon M. Local and Global Optimal Propensity Score Matching. Presented at: SAS Global Forum; April 16-19, 2007; Orlando, FL. http://www2.sas.com/proceedings/forum2007/185-2007.pdf. Accessed March 2014.
  25. Tamblyn R, Equale 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. doi: 10.7326/M13-1705.
  26. 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. doi: 10.1016/j.amjmed.2011.05.028.
  27. Vrijens B, Vincze G, Kristanto P, Urquhart P, Burnier M. Adherence to prescribed antihypertensive drug treatments: longitudinal study of electronically compiled dosing histories.  BMJ. 2008;336(7653):1114-1117. doi: 10.1136/bmj.39553.670231.25.
  28. Vrijens B, De Geest S, Hughes DA, et al; ABC Project Team. A new taxonomy for describing and defining adherence to medications. Br J Clin Pharmacol. 2012;73(5):691-705. doi: 10.1111/j.1365-2125.2012.04167.x.
  29. Hsiao CJ, Jha AK, King J, Patel V, Furukawa MF, Mostashari F. Office-based physicians are responding to incentives and assistance by adopting and using electronic health records. Health Aff (Millwood). 2013;32(8):1470-1477. doi: 10.1377/hlthaff.2013.0323.
  30. Joseph SB, Sow MJ, Furukawa MF, Posnack S, Daniel JG. E-prescribing adoption and use increased substantially following the start of a federal incentive program. Health Aff (Millwood).  2013;32(7):1221-1227. doi:10.1377/hlthaff.2012.1197.
  31. Medicare health & drug plan quality and performance rating 2013 Part C & Part D technical notes. CMS website. http://www.cms.gov/Medicare/Prescription-Drug-Coverage/PrescriptionDrugCovGenIn/Downloads/Technical-Notes-2013-.pdf. Updated August 9, 2012. Accessed June 2014.
  32. Lau HS, de Boer A, Beuning KS, Porsius A. Validation of pharmacy records in drug exposure assessment. J Clin Epidemiol. 1997;50(5):619-625.
  33. Steiner JF, Prochazka AV. The assessment of refill compliance using pharmacy records: methods, validity, and applications. J Clin Epidemiol. 1997;50(1):105-116.