by Dr. Kathryn H. Bowles, Senior Editor & Associate Professor
Sheryl Potashnik, PhD, MPH, Project Manager
Nai-Wei Shih, MPhil, IMBA, Research Associate
School of Nursing, Philadelphia, PA
University of Pennsylvania
This column was made possible by an educational grant from
Chamberlain College of Nursing
Bowles, K. H., Potashnik, S. & Shih, N. (October, 2011). Barriers to Meaningful Use: A Case for Sticking to the Standards. Achieving Meaningful Use in Research with Information Technology Column. Online Journal of Nursing Informatics (OJNI),15 (3). Available at http://ojni.org/issues/?p=876
An interdisciplinary research team at the University of Pennsylvania, School of Nursing is further developing its existing decision support system for discharge referral decision making. This five-year project, led by Dr. Bowles and funded by the National Institute of Nursing Research (NINR) (NRO1-007674), uses selected content from the Allscripts (formerly Eclipsys) clinical information system. This paper will describe the barriers encountered and solutions created as we attempted to draw data from four hospitals with the “same” clinical information system vendor.
The refined decision support system is being designed and built using case studies generated from the de-identified hospital records of discharged older adult patients. The case studies will be evaluated by geographically dispersed, interdisciplinary panels of health care experts (physicians, nurses, social workers, and physical therapists) who will review the case studies and tell us whether or not they would decide to refer the patients for post-acute care, where to, and why. This methodology was successfully used in our previous NINR funded study, “Factors to Support Discharge Decision Making” (Bowles et al., 2008; Bowles et al., 2009).
The majority of data used to develop the case studies came from the Adult Patient Profile within Allscripts. The Adult Patient Profile is a comprehensive patient assessment process created by the Clinical Practice Model Resource Center (CPMRC) now owned by Elsevier. It is completed upon admission by the admitting nurse and other interdisciplinary team members and updated throughout the hospital stay. The content of the Adult Patient Profile provides a robust opportunity to use data for a meaningful use because other sites or facilities that have the same clinical information system can collaborate on research using the elements within the Adult Patient Profile.
The project involves four hospitals all with the Allscripts CPMRC Adult Patient Profile and other components collectively called Knowledge Based Charting (KBC). The project involves four hospitals to gain a clinically, geographically, and socioeconomically diverse sample of patients all with the same clinical information.
To begin, the team created a detailed data matrix of the variables of interest outlining where they are found in the system, what they are named, and how they are measured. The matrix variables were compiled based upon prior research guided by the Orem Self Care Deficit theory (Orem, 1995); feedback from grant reviewers; in-person and conference call meetings with study co-investigators, and technical and statistical consultants; and the literature. Key staff members were identified at each of the participating hospitals including nurse informaticians, administrators, and other information technology personnel. Early in the process we established relationships with each hospital by at least one site visit, as well as multiple telephone conference calls/work sessions during which the study goals and procedures were reviewed and the data matrix introduced. The nurse informatician at each site reviewed the data elements and affirmed either an exact match with the original Adult Patient Profile elements or listed how they currently name and measure their items for which some variation was noted. This detailed procedure enabled the research team to know exactly how each site named and measured their data elements, which ones were missing and needed to be added, and what needed to be recoded upon receipt of the data download.
As we began to specify the data elements needed, we found the various hospitals had differing versions of CPMRC Adult Patient Profile and in addition, we were surprised to find customization of the systems at the local level had occurred and created a barrier to easily obtaining the desired large, diverse dataset. This paper addresses the latter issue, system customization which creates problems aggregating data across sites, thus diminishing the ability to clearly describe data elements.
Although vendors often create flexibility within their systems to allow their customers the freedom to “customize” their systems, from a research or quality improvement view this is not a desirable path unless such customization is constructed in a way that can be easily converted back to the standard format. Standards provide many advantages including providing a clear description of the data elements and matching data elements across systems. (Watzlaf, Zeng, & Jarymowycz, 2004) For our purposes, they support transferability, interoperability, and ultimately meaningful use.
To illustrate some examples, Table 1 shows a sample of a few data elements of interest, the standard set by the vendor, and the customization that occurred at the sites. Notice how the elements no longer exactly match and how a number of them will need to be renamed or recoded once the data are received. Several elements have been graciously rebuilt by our hospital partners to improve the matches. This labor intensive process would not be needed if electronic health record customers would “stick to the standards”.
We have worked with the hospitals to reconcile differences as much as possible without undue burden. To manage the remaining discrepancies we have a team comprised of a database developer/query writer, statistician, clinicians, and informaticians to configure a database to accept the data and provide the expertise to recode and merge the four datasets.
The plan to use hospital sites for our study with the same clinical information system should enable efficient collection of existing de-identified, standardized data for research purposes. This is the recommended direction for informatics researchers to make meaningful use of the rich data generated from clinical information systems. However, as in our experience, if we want to be truly efficient and avoid the time, cost, and effort it takes to match, rebuild, re-code and merge datasets the information system users must “stick to the standards” and avoid customizing their systems in ways that detract from transferability and cross-site collaboration. If customization is necessary, do it in a thoughtful way that will allow for reconciling or collapsing variables to some common denominator representing data that can be standardized to enable data analysis to proceed. Further, sites must stay up to date with the latest versions of software.
The CPRMC has a consortium of users who meet on an annual basis. We recommend a discussion of how to support cross-site research and quality measurement using uniform data be a recurrent agenda item. Keeping this issue on the forefront of discussion might deter the temptation to freely change the standard terms or to fall behind in valuable upgrades.
Bowles, K. H., Holmes, J. H., Ratcliffe, S. J., Liberatore, M., Nydick, R., & Naylor, M. D. (2009). Factors identified by experts to support decision making for post-acute referral. Nursing Research, 58(2), 115-122.
Bowles, K. H., Ratcliffe, S. J., Holmes, J. H., Liberatore, M., Nydick, R., & Naylor, M. D. (2008). Post-acute referral decisions made by multidisciplinary experts compared to hospital clinicians and the patients’ 12-week outcomes. Medical Care, 46(2), 158-166.
Orem, D. E. (1995). Nursing: Concepts of practice. (5th ed.). St. Louis, MO: Mosby.
Watzlaf, V. J. M., Zeng, A., & Jarymowycz, C. P.A. (2004). Standards for the content of the electronic health record. Perspectives in Health Information Management, 1, 1.
Proofed by Monica Key