research
Evaluation of e-prescribing in chain community pharmacy: Best-practice recommendations Michael T. Rupp and Terri L. Warholak
Received March 9, 2007, and in revised form June 2, 2007. Accepted for publication August 22, 2007.
Abstract Objectives: To measure the attitudes and beliefs of community-based pharmacists and technicians toward electronic prescribing (e-prescribing) and the processing of e-prescriptions and to generate best-practice recommendations for changes to improve e-prescribing in the community setting. Design: Descriptive, nonexperimental, cross-sectional study. Setting: 422 chain community pharmacies in six states that met a minimum dispensing volume of five e-prescriptions per day. Data were collected between April and July 2006. Participants: Pharmacists, technicians, and student interns. Intervention: Receiving, processing, and dispensing of e-prescriptions assessed via self-administered survey and follow-up interviews of key pharmacy operations and information technology management in each participating chain pharmacy organization. Main outcome measures: Attitudes, beliefs, and satisfaction of pharmacy personnel regarding e-prescribing, compared with conventional prescribing, and recommendations for improving e-prescribing in the community practice setting. Results: 1,094 surveys were returned from pharmacy personnel practicing in 276 chain community pharmacies. Pharmacy personnel preferred e-prescriptions over conventional prescriptions on each of seven desired outcomes of care. Pharmacists were found to view e-prescribing more positively than technicians (P < 0.05) for its net effect on three key outcomes: patient safety, effectiveness of care, and efficiency of care. A total of 2,235 written comments were received on the returned surveys. Of these, 57% (1,277) mentioned negative features of e-prescribing, while 43% (958) noted positive features. Improved clarity and/or legibility of prescriptions was the most frequently cited advantage of e-prescribing, followed closely by improved speed or efficiency of processing. Prescribing errors were the most commonly cited negative feature of e-prescribing, particularly those stating a wrong drug or wrong directions. Conclusion: Pharmacy personnel were generally satisfied with the current status of e-prescribing, but they also perceive key weaknesses in how it has been implemented in physicians’ practices and their own organizations. A total of 11 best-practice recommendations are offered to improve e-prescribing in the chain community pharmacy setting. Keywords: E-prescribing, community pharmacy, practice guidelines. J Am Pharm Assoc. 2008;48:364–370. doi: 10.1331/JAPhA.2008.07031
364 • JAPhA • 4 8 : 3 • M a y / J u n e 2 0 0 8
w w w.j a p h a . o r g
Michael T. Rupp, PhD, BPharm is Professor, College of Pharmacy, Midwestern University, Glendale, Ariz. Terri L. Warholak, PhD, BPharm, is Assistant Professor, College of Pharmacy, University of Arizona, Tucson. Correspondence: Michael T. Rupp, PhD, Midwestern University, 19555 N. 59th Ave., Glendale, AZ 85308. Fax: 623-572-3549. Email:
[email protected] Disclosure: The authors declare no conflicts of interest regarding products or services discussed in this manuscript. The authors declare no conflicts of interest or financial interests in any product or service mentioned in this article, including grants, employment, gifts, stock holdings, or honoraria. Acknowledgments: To Christine Marsh, BPharm, for valuable assistance interpreting the comments of pharmacy staff. Funding: Supported by grant 1 U18 HS016394 from the Agency for Healthcare Research and Quality. Previous presentation: Portions of this research were presented at the American Society of Health-System Pharmacists Midyear Clinical Meeting, Anaheim, Calif., December 6, 2006.
Journal of the American Pharmacists Association
Evaluation of e-prescribing in chain community pharmacy
A
rate-limiting step in the diffusion of electronic prescribing (e-prescribing) in the community practice setting is the willingness of physicians and pharmacies to adopt and sustain its use in their practices. Their willingness to do so is determined in part by the perceived advantages that e-prescribing offers compared with conventional prescribing and prescription-processing modalities. Although most research on e-prescribing has focused on physicians, several studies directed at pharmacists have appeared in the literature. A study by Anderson and Malone1 was conducted at a meeting of community pharmacists in 1996, when e-prescribing was still emerging and many operational
At a Glance Synopsis: Pharmacy personnel (pharmacists, technicians, and interns; n = 1,094 surveys returned) practicing in 276 chain community pharmacies reported being generally satisfied with but also perceiving key weaknesses in the current status of electronic prescribing (e-prescribing). Pharmacists, compared with pharmacy technicians, considered e-prescribing technology to be significantly more positive in terms of safety, efficacy, and efficiency. Of 2,235 written comments from survey respondents, 57% mentioned negative (prescribing errors cited most frequently cited) and 43% positive (improved clarity and/or legibility of prescriptions most frequently cited) features of e-prescribing. Respondent comments were used to develop 11 best-practice recommendations for improving e-prescribing in the chain community pharmacy setting. Analysis: Several technology-related measures can improve safe, effective, and efficient transmitting of e-prescriptions, including prescriber-side error-checking applications, electronic “bundling” of e-prescriptions by physicians to indicate to the receiving pharmacy the number of prescriptions being transmitted for a patient, and electronic alerts at the pharmacy that provide an obvious indication on the main prescription-processing screen that an e-prescription has been received and needs to be processed. Other recommendations include affording pharmacists the ability to electronically request supplemental or clarifying information from the prescriber and adopting means of reducing the need for physician callback and/or editing of e-prescriptions by pharmacy personnel. Eliminating the routine printing of e-prescriptions, which then must be re-entered by pharmacy staff, is one important goal of e-prescription processing; however, until this change occurs, e-prescriptions should be received by pharmacies in a standardized format that closely matches the manner in which information is organized on paper prescriptions.
Journal of the American Pharmacists Association
Research
and technical specifics had yet to be defined. Results indicated that pharmacists believed e-prescribing would save time and decrease medication errors, especially those caused by misinterpretation of handwritten prescriptions. Despite these positive results, only 49% of the pharmacists indicated that they would welcome e-prescribing. The authors conjectured that this was partly due to fears of undue influence or interference by pharmacy benefit managers or insurance companies. Indeed, 97% of respondents believed that the integrity of a prescription could be compromised if a third party were to intercept the prescription. Despite these concerns, most (54%) pharmacists agreed that e-prescribing was probably inevitable. Another study by Murray et al.2 evaluated the impact of e-prescribing on pharmacist work patterns in the outpatient pharmacy of a hospital. Using multidimensional work sampling, pharmacists were signaled randomly throughout the day and instructed to record their activity, the functional purpose of the activity, and the contact with whom they were engaged when signaled. Data collection occurred immediately before and after an e-prescribing system was implemented in the hospital. A total of 4,687 observations were recorded before and 4,735 after the implementation. The results demonstrated some important changes in work-related activities and functions after e-prescribing began: Pharmacists spent 12.9% more time correcting prescription problems, had 3.9% less idle time, and spent 2.2% less time in discussions with others. Pharmacists also spent 34.0% less time filling prescriptions, 45.8% more time in problem-solving activities involving prescriptions, and 3.4% less time providing advice. Other analyses in community pharmacy have tended to confirm the effect of e-prescribing on pharmacy workflow and personnel productivity.3,4
Objectives The analysis reported here was part of a federally funded national pilot to evaluate the status of e-prescribing in the community practice setting. The objectives of this analysis were to measure the attitudes and beliefs of community-based pharmacy personnel toward e-prescribing and the processing of e-prescriptions and generate best-practice recommendations to improve e-prescribing in the community setting.
Methods A self-administered survey was constructed to collect data from pharmacists, technicians, and student interns who practice in chain community pharmacies that routinely receive, process, and dispense e-prescriptions (see Appendix 1 in the electronic version of this article, available online at www.japha.org). Of particular interest were respondents’ perceptions regarding the effect of e-prescribing on seven specific outcome criteria: Safety of patient care Effectiveness of patient care Efficiency of patient care Communications with the patient w w w.p h a r m a c i s t . c o m
M ay / Ju n e 2 0 0 8 • 4 8 : 3 •
JAPhA • 365
Research
Evaluation of e-prescribing in chain community pharmacy
Communications with the physician/prescriber Effect on relations with the patient Effect on relations with the physician The terms used to represent the seven outcome criteria were left broad intentionally to allow respondents to apply their own interpretation. Respondents were asked to compare e-prescriptions to conventional prescriptions according to each criterion by selecting the appropriate response from a 5-point Likert-type rating scale (1, much worse, to 5, much better). Data were analyzed using multiple analyses of variance (ANOVA) to determine whether the three personnel classes included in the study (pharmacists, technicians, and interns) perceived the effect of e-prescribing on selected outcomes differently than their coworkers. Overall satisfaction with e-prescribing and the processing of e-prescriptions was measured using a 6-point Likerttype scale (1, very dissatisfied, to 6, very satisfied). Stepwise multiple linear regression modeling was performed to identify the best model for predicting pharmacy staff satisfaction with e-prescribing from a linear combination of the seven outcome variables according to the following formula: Y = β0 + β1X1 + . . . + βkXk. Comments were also solicited from respondents regarding the major positive and negative features of the e-prescribing implemented in their organization. Comments received from respondents were classified by the investigators according to their content and major theme. The interpretation of respondent comments was aided by consultation with key operations and management contacts at participating chain pharmacy organizations. Based on comments received and follow-up interviews by the investigators with key chain contacts, a series of bestpractice recommendations were distilled from respondents’ comments for improving the implementation of e-prescribing in the community practice setting. Pharmacies in the sampling frame were identified from a comprehensive list of those operated by seven participating chain pharmacy organizations in six states where e-prescribing is most active: Florida, Massachusetts, New Jersey, Nevada, Rhode Island, and Tennessee. Eligibility for participation in the study required that the pharmacy dispense an average of five or more e-prescriptions each day. While somewhat arbitrary, this value was selected after discussions with chain management and selected pharmacy staff and was intended to ensure that pharmacy personnel were sufficiently familiar with e-prescription processing to provide meaningful responses to survey questions. Across the seven participating chain pharmacy organizations, 553 units met the minimum daily e-prescription volume for inclusion in the sample. Requests were subsequently made to each of the seven chain pharmacy organizations to distribute the survey to prescription-dispensing personnel in each pharmacy. The survey was made available in both paper and online 366 • JAPhA • 4 8 : 3 • M a y / J u n e 2 0 0 8
w w w.j a p h a . o r g
electronic format. In most chain pharmacy organizations, pharmacy personnel do not have access to the Internet. For the few chains in which personnel have Internet access, restrictions and firewalls generally prevent them from accessing unauthorized noncorporate sites. As a result, the distribution and collection of surveys was conducted almost entirely via fax or mail using the paper version. One chain pharmacy organization with 95 eligible units ultimately agreed to distribute the survey to only 25 because of concerns related to ongoing changes in computer systems resulting from the recent acquisition of a competing chain. Another chain with 111 eligible units subsequently decided to limit distribution of the survey to 50 pharmacies as a result of concerns about disruptions in workflow. Thus, from an initial sampling frame of 553, surveys were ultimately made available to personnel in 422 pharmacies. Survey data were entered into a database by a research assistant beginning in April 2006 and continuing through July 2006. Entered data were examined by the investigators on a daily basis for out-of-range values, and approximately 20% of surveys were randomly selected for duplicate input to check for data entry errors. Data were analyzed using Microsoft Excel 2002 (10.4302.4219) SP-2 and SAS 9.1.3.
Results A total of 1,094 surveys were returned from 276 of 422 sampled pharmacies (65.4% response rate). On average, four responses were received per participating pharmacy. Technicians represented the largest respondent group (55.3%), followed by pharmacists (40.8%), student interns (3.2%), and eight personnel classified as “other” (0.7%). The majority of pharmacist respondents were baccalaureate trained (72.2%) and women (56.9%). A substantially smaller proportion (22.4%) were doctoral trained (i.e., PharmD), and several also had advanced graduate degrees (e.g., MS, PhD). Pharmacists responding to the survey had been practicing pharmacy for a mean (±SD) of 13.9 ± 3.4 years at their current location. Technician respondents were predominantly women (86.4%) with a high school diploma, general equivalency diploma, or some college (68%). An additional 26.7% had completed either a baccalaureate or associate degree. Technicians responding to the survey had been working as pharmacy technicians for an average of 5.7 years, of which 3.2 years were spent at their current location. ANOVA was used to determine whether pharmacists, technicians, and interns differed in their perceptions of how e-prescribing affects key patient-relevant outcomes. Results of this analysis appear in Table 1. Pharmacists and technicians differed in important ways in their views of the effects of e-prescribing. Pharmacists viewed the technology significantly more positively than their technician counterparts on the key aspects of safety, effectiveness, Journal of the American Pharmacists Association
Evaluation of e-prescribing in chain community pharmacy
Research
Table 1. Pharmacy personnel attitudes toward e-prescribing/e-prescriptions Outcome n Safety of patient care Effectiveness of patient care Efficiency of patient care Communications with patient Communications with prescriber Overall relations with patient Overall relations with prescriber
Pharmacists, mean ± SD 446 3.92 ± 1.00 a 3.85 ± 0.90 b 3.91 ± 0.97c 3.33 ± 0.89 3.37 ± 1.14 3.43 ± 0.88 3.34 ± 1.00
Technicians, mean ± SD 605 3.71 ± 1.01a 3.67 ± 0.99 b 3.66 ± 1.09c 3.28 ± 1.06 3.36 ± 1.14 3.36 ± 0.99 3.39 ± 1.00
Interns, mean ± SD 35 3.94 ± 0.84 3.89 ± 0.83 4.06 ± 0.95 3.17 ± 1.07 3.49 ± 1.07 3.54 ± 0.85 3.57 ± 0.88
All, mean ± SD 1,086 3.81 ± 1.00 3.75 ± 0.95 3.77 ± 1.05 3.30 ± 0.99 3.37 ± 1.13 3.39 ± 0.94 3.38 ± 1.00
Scale: 1, much worse; 2, somewhat worse; 3, no change; 4, somewhat better; 5, much better. Analysis of variance (Pr > F): aP = 0.0038, bP = 0.0073, cP = 0.0002.
and efficiency. Comparing e-prescriptions with conventional prescriptions, pharmacists rated its effect on safety of patient care at 3.92, with technicians rating it at 3.71 (P < 0.05). Likewise, pharmacists assessed e-prescriptions significantly higher than technicians with regard to its impact on the effectiveness of patient care (3.85 versus 3.67, P < 0.05) and efficiency of patient care (3.91 versus 3.66, P < 0.05). Student pharmacist interns evaluated all outcome criteria higher than either pharmacists or technicians, but the small sample size prevented meaningful statistical comparison with the other two personnel classes. Although pharmacy personnel rated e-prescriptions more favorably than conventional prescriptions on all seven outcome criteria, they evaluated the impact of e-prescribing on communications with the patient least favorably (3.30), followed by communications with the physician (3.37), overall relations with the physician (3.38), and overall relations with the patient (3.39). Among these four outcomes, no significant differences between pharmacists and technicians were observed. Figures 1 through 3 illustrate the distribution of responses by pharmacists, technicians, and interns to the seven evaluative outcomes assessed in the survey. No differences were found among pharmacists, technicians, and interns in terms of overall satisfaction of pharmacy person-
nel with e-prescribing and the processing of e-prescriptions, as all three groups were somewhat to moderately satisfied with e-prescribing and the processing of e-prescriptions (means 4.43, 4.42, and 4.66, respectively, on a 6-point Likert scale) (Figure 4). Regression modeling
In our multiple linear regression analysis, model selection was conducted using the RSQUARE option in PROC REG; initially, all possible models were retained. Models were subsequently ranked using adjusted R2 values, and the CMALLOWS option was applied to determine model bias value Cp ≤ p, where p equals the number of parameters in the model, including the intercept. For prediction, Mallows recommended choosing the model with the fewest independent variables, where the value of Cp approaches p. Among models that met the above selection criteria, collinearity was assessed using the variance inflation factor (VIF) and condition index (COLLIN) options. In general, a VIF value >10 and/or COLLIN value >30 suggest the presence of moderate to strong collinearity in the model. A model that predicted pharmacy staff satisfaction with e-prescribing from a linear combination of five outcome variables was ultimately selected on the basis of good predictive
Physician Relations
Physician Relations
Patient Relations
Patient Relations
Physician Communication
Physician Communication
Much W orse
Much W orse
Somewhat W orse
Somewhat W orse No Change
Patient Communication
No Change
Patient Communication
Somewhat Better
Somewhat Better
Much Better
Much Better Efficiency
Efficiency
Effectiveness
Effectiveness
Safety
Safety
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Figure 1. Pharmacists’ attitudes on impact of e-prescribing/e-prescriptions (n = 446) Journal of the American Pharmacists Association
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Figure 2. Technicians’ attitudes on impact of e-prescribing/e-prescriptions (n = 605) w w w.p h a r m a c i s t . c o m
M ay / Ju n e 2 0 0 8 • 4 8 : 3 •
JAPhA • 367
Research
Evaluation of e-prescribing in chain community pharmacy
Physician Relations
Intern (n=35) Patient Relations
Physician Communication Very Dissatisfied
Much W orse
Moderately Dissatisfied
Somewhat W orse No Change
Patient Communication
Somewhat Dissatisfied
Technician (n=605)
Somewhat Satisfied
Somewhat Better
Moderately Satisfied
Much Better
Very Satisfied
Efficiency
Effectiveness Pharmacist (n=446)
Safety
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Figure 3. Student pharmacists’ attitudes on impact of e-prescribing/e-prescriptions (n = 35) power, low collinearity, and adequate substantive explanatory power and model parsimony (adjusted R2 = 0.5961, F = 313.55, P < 0.0001 for the overall model and all variables, VIF
