The very careful reader will have noted that I met a very smart (and, to a rapidly becoming old fogey like me, disturbingly young!) THCB reader, Jonathan Pearlstein at HIMSS. Jon works for the academic survey firm NORC and has been heavily involved in the AHRQ/CMS assessment of ePrescribing that I’ve featured on THCB here, here, and here. John has written a commentary specially for THCB on the studies just done. He tells me that these are of course his opinions not those of the organizations he works for or with—but he has to say that! Those of you involved in eRx should particularly pay attention to the issue around re-keying data into pharmacy systems in the section End to End Transmissions but overall this is a fascinating read.
From September 2006 to March 2007, I participated in an evaluation of five electronic prescribing (eRx) pilot projects, sponsored by AHRQ and CMS, and mandated as part of the Medicare Modernization Act. All told, the evaluation cost the government over $7 million and involved the heaviest hitters in eRx and health IT evaluation—SureScripts, RAND, and Brigham & Women’s Hospital, to name a few.
The product of our labors is available here. My commentary on the study follows:
Our primary charge in evaluating the pilot projects was to investigate the effectiveness and interoperability of proposed standards for eRx messaging. These proposed standards enable some of the more advanced features of eRx, such as sending automatic notifications when a patient has filled a prescription, or allowing prescribers to access patient medication histories and formulary information. Other advanced aspects include digital methods for requesting prior authorization, representing drug dictionaries, and writing sigs (patient instructions for taking medications, such as “by mouth”).
A lot of what we found falls in line with what Matthew commented on early this month regarding the HSC article on eRx in Health Affairs. Although our study probably is more different than it is similar to the one in HA, I think a comparison is worthwhile. Their methods mainly involved qualitative interviews of doctors using eRx in 2005. Ours, on the other hand, took place in 2006 and involved a panoply of methods including interviews, surveys, expert panels, observational studies, medical chart reviews, and technical testing of the standards. Here is how some of our findings stack up:
One year after HSC conducted their eRx study, Medication Histories are still not complete. RxHub and SureScripts represent two competing solutions for compiling Med History information, gathering data from payers/PBMs and pharmacies, respectively. Neither one works (i.e. neither generates a comprehensive list). Further interoperability among standards is needed to integrate Med History from a number of different sources. And, as was discussed in the HSC paper, until physicians are confident that the medication history they get through eRx is complete, they will not use it.
Unlike the HSC study, we did find that there is a glimmer of hope for Med History. There were a few devoted prescribers who frequently used the function and expressed high satisfaction. Even so, a new problem we encountered was that most of the physicians in our pilot testing did not even realize that their eRx system had Med History available. Can someone say “user training”?
Formulary and Benefit Information
Here we have a similar story to the Med History. Due to the complexity of formulary information and the wide variety of formularies, information presented by the eRx systems may not be complete. As a result, physicians in our study, as in HSC’s, doubt the information’s reliability and do not use it.
A noteworthy challenge in this area is the dynamic nature of benefit information, which makes it difficult to know the extent to which a specific patient will be covered. Think Medicare Part D. Ideally, a prescriber would know whether the patient were approaching the “donut hole,” in order to avoid racking up some major bills. We still need to investigate further how the Formulary & Benefit Standard will incorporate real-time, patient-centered data sources.
End-to-End Transmissions Still a Challenge
The HSC study found that most sites were not actually engaging in e-prescribing, but rather, e-faxing. What we found is that one year later, things look similarly bleak. Of the five pilot sites we evaluated, only one achieved full end-to-end transmission of new prescription information. The problem was not the technical standard itself (NCPDP SCRIPT New v8.1 worked fine, for those who care). The problem was manipulating the data to be usable on the pharmacy information systems.
At most of the pilot sites, pharmacists had to re-key, or re-enter at least some of the information they received through the eRx system. Their pharmacy information systems simply could not use the data to auto-populate forms for dispensing and filling prescriptions. Although this type of transmission—Electronic Data Interchange + Re-Keying—may be a step up from e-faxing, it still seems a far cry from the ideal of true end-to-end eRx. We still don’t know the extent to which re-keying information contributes to medication errors or lost efficiencies. My suspicion, however, is that the losses will be enough to merit getting eRx to work on the back-end.
The use of the eRx system by nurses, MAs, and PAs rather than by physicians—described as “surrogate prescribing”—was endemic across all our pilot sites. The HSC study vaguely alluded to this finding in their paper; however, we found it to be a ubiquitous practice. Some of our sites recorded that as many as 77% of all prescriptions were entered into the system by surrogate prescribers, not physicians. Moreover, this workflow pattern did not appear to be a transitional stage; rather, it was the preferred workflow across all five sites.
The implications of surrogate prescribing are ambiguous. On the one hand, it seems to undermine eRx’s much-touted decision support functionalities. Prescribers won’t react to a drug interaction alert if they never see it. On the other hand, the efficiencies it creates may drive adoption and help create a business case for eRx.
The key to success is in careful implementation. Tech vendors and health systems can choose to embrace surrogate prescribing and may be rewarded for it. One of our pilot sites built surrogate prescribing into their system, constructing a queue for physicians to sign off on prescriptions originally written by nurses. The potential for eRx to shift work away from the physician—and at the same time to create efficiencies—represents a good example of “commoditizing the caregiver” through technological innovation, discussed recently by Clayton Christenson.
All pessimism aside, it’s worth noting that in our study we endorsed the technical standards for three types of eRx messages: Medication History, Formulary & Benefit, and Prescription Fill Status Notification. Now, with the likely promotion and promulgation of these standards by Congress and HHS, it’s probable that we will see accelerated efforts in this arena. Widespread use, coupled with our recommendations for improvement, will facilitate more comprehensive Med History and Formulary & Benefit information.
In addition, the outlook for physician adoption looks good. Supposing that we decide surrogate prescribing can be safe and effective, the real need now is to get eRx working on the pharmacy back-end. For if we want eRx to be the “killer app” that drives health IT adoption, rather than just plain “killer,” we must cautiously approach all new types of potential errors.