Implementing real-time prescription benefit tools: Early experiences from 5 academic medical centers

Abstract

Background:

Medication price transparency tools are increasingly available, but data on their use, and their potential effects on prescribing behavior, patient out of pocket (OOP) costs, and clinician workflow integration, is limited.

Objective:

To describe the implementation experiences with real-time prescription benefit (RTPB) tools at 5 large academic medical centers and their early impact on prescription ordering.

Design: and Participants:

In this cross-sectional study, we systematically collected information on the characteristics of RTPB tools through discussions with key stakeholders at each of the five organizations. Quantitative encounter data, prescriptions written, and RTPB alerts/estimates and prescription adjustment rates were obtained at each organization in the first three months after “go-live” of the RTPB system(s) between 2019 and 2020.

Main measures:

Implementation characteristics, prescription orders, cost estimate retrieval rates, and prescription adjustment rates.

Key results:

Differences were noted with respect to implementation characteristics related to RTPB tools. All of the organizations with the exception of one chose to display OOP cost estimates and suggested alternative prescriptions automatically. Differences were also noted with respect to a patient cost threshold for automatic display. In the first three months after “go-live,” RTPB estimate retrieval rates varied greatly across the five organizations, ranging from 8% to 60% of outpatient prescriptions. The prescription adjustment rate was lower, ranging from 0.1% to 4.9% of all prescriptions ordered.

Conclusions:

In this study reporting on the early experiences with RTPB tools across five academic medical centers, we found that variability in implementation characteristics and population coverage. In addition RTPB estimate retrieval rates were highly variable across the five organizations, while rates of prescription adjustment ranged from low to modest.

1. Introduction

The problem of unaffordable prescription drugs in the U.S. has been highlighted in many disease areas including cancer,^1–4 diabetes,^5–8 obstructive lung diseases,^9,10 and multiple sclerosis.¹¹ Although proposals have been introduced in Congress, no comprehensive legislation addressing affordable drug pricing has passed, resulting in highly variable patient costs depending on formularies, insurance plans, and costs previously incurred within the benefit year.¹² Prescribers are unable to make meaningful shared decisions with patients about treatment if the relative affordability of medications options is not known. While many electronic health records (EHRs) have tried to improve price transparency, for example by incorporating patient-level drug plan and formulary coverage information, an informed estimate of the patient’s out-of-pocket (OOP) cost at the time of prescribing has not been systematically available to help guide medication treatment decisions.

To address this lack of drug price transparency, the Centers for Medicare & Medicaid Services (CMS) requiring all sponsors of prescription drug plans in Medicare (i.e. Part D) to implement at least one electronic real-time prescription benefit (RTPB) tool by January 1, 2023.^15,16 Several RTPB tools integrate with commercial EHRs to provide prescribers with patient cost estimates at the time of prescribing a medication. RTPB systems source their data from Pharmacy Benefits Managers (PBMs) and retail pharmacies, and they rely on the same networks originally developed to verify insurance eligibility and/or adjudicate prescription drug claims in real time. The companies that provide the RTPB data, contract with individual Pharmacy Benefits Managers (PBMs) or other sources. An organization would need to understand their PBM mix when deciding upon one or more RTPB vendor(s).

When an RTPB tool is activated, the prescriber is able to see patient-specific medication costs on their EHR display as well as any available less expensive alternatives from the same drug class or at alternative pharmacies. These tools may be deployed in the ambulatory care setting only, the hospital setting for outpatient discharge only (e.g. when a hospitalist or an Emergency Department (ED) prescriber orders a prescription on discharge), or both, depending on the implementation decisions made by the organization. One stated goal of these tools is to empower patients and prescribers to compare patient costs and choose prescriptions representing better value. Two real-world examples of RTPB alerts are shown in Fig. 1A and B.

Fig. 1A.

Example of a real-time prescription benefit alert using Electronic Health Record Vendor X.

Fig. 1B.

Example of a real-time prescription benefit alert using electronic health record vendor Y.

Since commercial payors often adopt CMS rules, it is likely that these RTPB tools will become more prominent in the next few years. However, the availability, usability, and utility of these tools is not well understood. Data on the use of these tools, how they are implemented, as well as their potential effects on prescribing behavior, patient OOP costs and clinician workflow integration is needed.

To date, little has been published about the impact of these tools or how these tools are implemented in the peer reviewed literature.¹⁷ We sought to describe the early experiences with implementing RTPB tools at a convenience sample of 5 large academic medical centers/health systems: University of Pittsburgh Medical Center (UPMC), Yale New Haven Health (Yale), Stony Brook Medicine (Stony Brook), Froedtert & the Medical College of Wisconsin (F&MCW) and Johns Hopkins Health System (JHHS).

2. Methods

2.1. Data sources and study sample

In this cross-sectional study, we systematically collected information on the implementation characteristics of RTPB tools through discussions with key stakeholders at each of the five organizations (UPMC, Yale, Stony Brook, F&MCW and JHHS). These stakeholders included Chief Medical Information Officers, leaders in clinical informatics, health plan or health system pharmacy leadership, medical directors for electronic health records, and representatives from EHR and RTPB tool vendors. Representatives from each of the five organizations met virtually every month to refine the data collection process and discuss interim results.

Quantitative data on the number of encounters, providers and RTPB estimates retrieved or prescription orders that were adjusted was obtained from in-house analyst teams at UPMC, Yale, F&MCW and JHHS or from the EHR vendor, in the case of Stony Brook. The RTPB estimate retrieved at UPMC was sourced from monthly reports compiled by the RTPB vendor (A). This study was approved or determined to be not human subjects research by the institutional review boards and/or quality improvement study boards at all five organizations.

2.2. RTPB implementation characteristics

Through an iterative process, we collected information on the RTPB vendor(s), the types of services each vendor provided (i.e. their functionality), go-live dates, type of EHR used, and deployment setting (ambulatory, at hospital discharge or both) for RTPB implementation in each medical center. Because the RTPB could be configured differently across sites (Box. 1), we also collected the following information: 1) RTPB estimate notification process (i.e. whether the patient cost estimate was automatically generated upon signing an e-prescription order versus requested “on-demand” through a user action by a prescriber), 2) prescriber opt-out options, and 3) cost differential settings that trigger notifications through the RTPB tool.

Box. 1.

Real-time Prescription Benefits Tools Implementation Parameters (i.e. Configurations)

RTPB Implementation Parameters	Available Options	Example
Deployment setting	Inpatient, outpatient or both	A health system may choose to restrict availability and use of RTPB tools to their outpatient EHR.
Patient coverage	Contract with multiple RTPB vendors versus only one vendor	A health system contracts with more than 1 RTPB vendor to enhance population coverage.
Trigger for displaying results of a RTPB data transaction on a clinician’s screen (also known as a RTPB alert)	Automatic display versus “on-demand”	The EHR may choose to only display RTPB alerts after the prescriber chooses an “Estimate” button as one additional step of the existing e-prescribing workflow.
Cost differential display thresholds	A specific dollar amount versus no threshold (i.e. all RTPB transactions will display, regardless of the difference in OOP cost between the original medication ordered and the suggested alternative)	With the goal of reducing hard stops that may interrupt a prescriber’s workflow, a health system may choose to only display prescription alternatives when the difference between the original medication ordered and the suggested alternative exceeds a prespecified dollar amount (e.g. $3 per month).
Prescriber opt-out	Yes versus no	When displayed, the RTPB alert may have an option allowing prescribers to opt-out from viewing future RTPB alerts.

2.3. Quantitative data collection

We obtained data on the total number of encounters where a medication was ordered, the total number of unique providers associated with those encounters, the total number of outpatient medication orders, the number of RTPB estimates retrieved and the number of times a prescription order was changed as a result of an RTPB alert. These data were collected in the first three months after “go-live” of the RTPB system(s) at each of the five organizations, which occurred in 2019 or 2020 (Table 1). When possible, we selected observation windows to avoid any overlap with the COVID-19 pandemic (i.e. starting on or after March 2020).

Table 1.

Characteristics of RTPB Implementation at 5 academic medical centers/health systems.

	UPMC	Yale	F&MCW	JHHS	Stony Brook
Measurement Dates	7–10/2020	4–7/2019	10/2019–1/2020	4–7/2019	9–12/2019
Total number of encounters where a medication was ordered	241,450	277,272	94,972	97,651	79,894
Number of unique providers	3291	3583	1625	1218	1666
RTPB tool Vendor(s)	A	B	A, B	B	B, C
Deployment setting (IP, OP, both)	Both	OP	Both	Both	Both
EHR Vendor	Epic; Cerner	Epic	Epic	Epic	Cerner
Automatic Display	Yes	No	Yes	Yes	Yes
Opt-out Option	Yes	Yes	Yes	Yes	No
Display Threshold	None	None	>$3/mo or $0.10/d	None	None

UPMC = University of Pittsburgh Medical Center.

F&MCW = Froedert & Medical College of Wisconsin.

JHHS = Johns Hopkins Health System.

RTPB = real time prescription benefit.

IP = inpatient.

OP = outpatient.

EHR = electronic health record.

2.4. Definition of RTPB estimate retrieval rate and prescription adjustment rate

We additionally sought to calculate the RTPB estimate retrieval rate and prescription adjustment rate, both collected in the first three months after “go-live.” We define RTPB estimate retrieval rate as the number of electronic prescriptions where estimates were electronically received over the total number of prescriptions in the measurement period. The RTPB estimate retrieval rate may be less than 1 in cases where there is missing data – for example, if a patient’s pharmacy benefits manager does not have an existing contract or data sharing agreement with the RTPB tool vendor. In some health systems, the RTPB estimate may be suppressed in cases where alternative covered medications have a higher OOP cost then the originally ordered medication. We define the prescription adjustment rate as the number of prescriptions where a medication that had been ordered was modified by selecting an alternative proposed by the RTPB tool over the total number of RTPB data retrievals available.

3. Results

3.1. RTPB services

Three RTPB vendor services were used by the health systems and all functioned similarly. For scheduled ambulatory services, the EHR performed an insurance verification the night prior to the appointment, confirming pharmacy benefits. When a clinician wrote a prescription, the EHR would use the RTPB service to query the PBM linked to that patient through their pharmacy benefit information, and it would store an estimate. That estimate could either be surfaced through a user action or automatically upon signing of the prescription order, depending on the local configuration. For same-day appointments or ‘on-the-fly’ encounters, pharmacy benefit information would not be verified without a staff member triggering that process. If data were not successfully returned, no automatic alerts would be triggered and clicking the ‘estimate’ button within the EHR would note that data were not available.

3.2. Implementation characteristics

The characteristics of each of the five organizations are shown in Table 1. Implementations occurred between April 2019 and July 2020. Yale, F&MCW and JHHS all used a single EHR vendor (X), while Stony Brook used a different EHR vendor (Y). UPMC had a hybrid system, using vendor X for its outpatient encounters and Y for inpatient encounters. Organizations varied in their choice of vendors and their approach to implementation, but all implemented the tool within the EHR and prescribing workflow. Two organizations used a combination of two vendors to increase the coverage of PBMs, while three used a single vendor. All of the organizations except one chose to display patient cost estimates and suggested alternative prescriptions automatically if less expensive alternatives were available and returned. Yale chose to make its system on-demand, meaning that each prescriber must affirmatively find and click on an “Estimate” button in order to view the data generated by the RTPB tool.

Differences were also noted with respect to a patient cost threshold for automatic display. Only one out of five organizations (F&MCW) set a limit under which an RTPB display would be suppressed. At F&MCW, if the estimated difference in OOP cost between the original prescription and the alternative prescription was less than $3 per month, the tool would not automatically alert clinicians. In the four other organizations, no threshold was set, meaning that the tool would always trigger an alert, even if the difference in the OOP cost between the original prescription and the suggested alternative was $0.

3.3. RTPB estimate retrieval and prescription adjustment rate

In the three months after each system’s “go-live” date, Yale had the largest number of unique encounters (n = 277,272), while Stony Brook had the fewest (n = 79,894). Similarly, Yale had the largest number of unique prescribers associated with these encounters (largest: Yale (n = 3583); smallest: JHHS (n = 1218)). In the first three months after “go-live,” RTPB estimate retrieval rates varied greatly across the five organizations, ranging from a low of 8% at Yale (with a single RTPB and data only returned upon clinician request) to a high of 60% at F&MCW (with two vendors and automatic estimate retrieval at prescription signing) (Fig. 2). The prescription adjustment rate were lower than estimate retrieval rates ranged from 0.1% of all prescriptions ordered at Stony Brook to 4.9% at UPMC.

Fig. 2.

Real-time prescription benefit (RTPB) data retrieval and prescription adjustment rate at five healthcare organizations in the first three months after go-live. RTPB Retrieval Rate is defined as the number of electronic prescriptions where estimates were electronically received divided by the total number of prescriptions in the measurement period. The Rx (Prescription) Adjustment Rate is the number of prescriptions where a medication that had been ordered was modified by selecting an alternative proposed by the RTPB tool divided by the total number of RTPB data retrievals available in the measurement period.

4. Discussion

In this study reporting on the early experiences with RTPB tools across five academic medical centers, we found that organizations implemented one or more vendors with different configurations, a high degree of variability in coverage of the tools with respect to RTPB estimates being returned, and low to modest rates of prescription adjustment.

Our results show that there was a high degree of variability in RTPB estimate retrievals. For example, during the study interval, less than one in five prescription orders at Yale and Stony Brook resulted in successful retrieval of RTPB estimates. In contrast, half or more of all prescription orders at UPMC and F&MCW successfully retrieved RTPB estimates. These differences may have been due to: (1) implementation design choices adopted by each center (e.g. Yale only displayed patient cost information on-demand) or (2) mismatches between the patients served by the medical center and the population covered by each RTPB vendor(s). For example, many of the patients that received care at UPMC were covered under a PBM that was contracted with the chosen RTPB vendor. In contrast, Yale’s RTPB vendor (B) contracted with one PBM that only covered a small percentage of Yale’s patients. As a result, the majority of e-prescription orders could not generate a RTPB transaction. This situation has since improved at several of the organizations examined in this study. For example, Stony Brook has now established contracts with 2 other RTPB vendors while Yale now has 3 RTPB vendors providing coverage for the majority of its patients.

Although the estimate retrieval rates were variable across the five organizations, our results show that actual prescription adjustment rates were low or modest (~5%). These rates may be due to a combination of factors: low RTPB estimate retrieval rates, implementation design choices in creating meaningful cost thresholds to prompt adjustments, and implementation challenges such as lack of provider awareness or confidence in these new EHR-integrated tools. It could also be the case that the RTPB tool suggested prescription alternatives were not clinically appropriate (e.g. patient has already tried and not responded to the suggested alternative, or is allergic to that medication). Alert fatigue could also explain potential differences in data retrieval to prescription adjustment rates. Despite these early challenges, rates of prescription adjustment may increase as RTPB tools themselves improve and as more widespread adoption of these tools occur.

Our study had several limitations. First, it only included quantitative data from five academic organizations during the first three months after go-live of each RTPB tool. It is possible that with greater use, familiarity and more provider education/training, rates of RTPB tool use and prescription adjustment rates may have changed. Second, because of configuration options and prescriber opt-outs, not all electronic estimates would have been displayed to the prescriber at the time of order signing. Unfortunately, reliable data on the rates of display of these estimates were not obtainable. Future research should seek to understand factors that affect retrieval of RTPB information. Third, the low rates of prescription adjustments we report may be due to undercounts or missing data. For example, after viewing an RTPB alert, a prescriber may ultimately select an alternate medication than he or she originally intended, but sent the prescription using a different workflow. These types of prescription adjustments would not be captured based upon our current outcome definition(s). The prescription adjustment rate at Yale was relatively low despite the fact that the system required prescribers to click an “Estimate” button to generate a RTPB request. This could be due, in part, to a lack of alternatives (with lower OOP cost) to the originally ordered medication. Fourth, the lack of standardization across health systems, payer-mix, RTPB vendors, and EHRs makes it difficult to generalize our findings to other systems across the country. Our report, however, is able to provide a range of experiences that many large systems may encounter. Despite these limitations, our study provides much-needed data across 5 health systems that will help move the field forward for future research.

4.1. Lessons learned

Our results show three common themes: first, despite their potential to help address the problem of unaffordable prescription drugs, the utility of RTPB tools to date may be limited in use by data fragmentation based upon PBM, payor, and RTPB vendor contracts. Currently, PBMs are not incentivized to share data with all RTPB vendors. Since no single RTPB vendor can obtain benefit information for all of the patients seen at any health system that serves a large diverse population, use of more than one RTPB vendor may be needed to ensure a high degree of population coverage. In our study, F&MCW contracted with two RTPB tool vendors while JHHS contracted with one vendor. It is possible that the higher RTPB estimate retrieval rate for F&MCW (60%) when compared to JHHS (38%) was due to differences in coverage of the patient populations by RTPB vendors. However, choosing to enter into contracts with multiple RTPB vendors may add additional financial or operational burden to organizations. Second, RTPB tools are less likely to be used by clinicians when they are configured to only display upon a user’s action, such as clicking an “Estimate” button. Chief Medical Information Officers or other decision makers should consider automatic display of RTPB estimates as a default setting. Third, even when faced with a hard stop in the EHR that suggests an alternative prescription or pharmacy that may save their patients money, prescribers did not frequently change their original prescriptions. This may be due to a variety of factors including: alert fatigue, a suggested alternative that is not clinically appropriate (e.g. due to allergy/intolerance), or where the pharmacy alternative is not appropriate (e.g. mail order for an acute medication), patient preferences, lack of provider education, and misunderstandings about the purpose of these tools.^18–22 If clinicians are unsure of the purpose of these tools, or how they deliver value to them and their patients, they may be less likely to engage with the tools or may even opt-out. It is also possible that clinicians also hesitate to use tools they do not fully trust. For example, if a prescriber believes that the patient estimates shown in RTPB alerts are not 100% reliable, they may avoid having a discussion about medication costs with their patients. Future studies, potentially using qualitative methods should explore which of the listed potential factors are most important to prescribers. In addition more specific provider education in advance of launching these tools may help to correct misunderstandings and thereby make the tool more useful clinically. Quantifying the impact on savings for patients, and when the tool may lead to the most benefit could also support further iterative improvement in the tools themselves, as well as adoption. One final lesson is that current implementation settings for RTPB tools do not allow any granular adjustments, such as enabling RTPB alerts only for certain medications, such as specialty medications, or to certain prescribers. These adjustments may help limit the number of low value messages in a clinician’s workflow thereby avoiding over-alerting.

In conclusion, although this landscape is evolving rapidly, our data on the implementation experiences across five academic medical centers show that RTPB tools can be successfully deployed and integrated with existing EHRs. However, we found that the use and clinical utility of RTPB tools may be limited in the first few months after implementation. Given the novelty of these tools and their potential to inform debates about drug price transparency and value-based prescribing, we argue that substantially more research is needed to understand factors that affect use and impact of these RTPB tools to achieve the ultimate goal of improved health outcomes.

Data availability

Data will be made available on request.