Electronic pharmacist-physician communications in a highly computerised hospital: timeliness of action in practice

Background: Errors in communication have been found to be one of the leading causes of adverse events in healthcare. In the context of electronic health records, communication can be asynchronous, where participants do not need to be present at the same time. This modality of communication means fewer interruptions are required to transfer information. However, there can be losses of interaction, sense of immediacy, or feedback. PICS is a locally developed ePrescribing system in use at the University Hospitals Birmingham NHS Foundation Trust (UHBFT) throughout all inpatient beds. Pharmacists can communicate medication-related information to the physician with the use of a free-text review message assigned to a single prescription on a patient’s profile. We aimed to determine the effects of temporal, message and prescription factors on the time to action of medication-related communications.

Methods: Data were captured from the PICS audit database between 1st January 2012 and 31st December 2012 on: 1) Review messages assigned by pharmacists; 2) Details of the prescription on which the messages were assigned; and 3) Details of any changes made to the prescription following a review message being assigned. Data were coded for temporal, message and prescription factors. Messages associated with the prescription and that requested a measurable action that could be determined from the data were analysed to investigate two outcomes: 1) Action; and 2) Time taken to action. We conducted a multivariable analysis using Generalised Estimating Equations (GEE) to account for the effects of multiple factors simultaneously, and to adjust for any potential correlation between outcomes for repeated review messages on the same prescription. All analyses were performed using SPSS 22 (IBM SPSS Inc., Chicago, IL, USA), with p<0.05 considered significant.

Results: Over the one year period there were 71 994 inpatient episodes, for which 1 291 773 prescriptions were generated within PICS. Pharmacists assigned a total of 36 245 review messages to prescriptions; 34 506 were coded for analysis after excluding for blank or duplicate messages. A total of 9340 messages requested a measurable action that could be further analysed in the GEE model, for which selected results are presented here. Just under a quarter of messages generated an intended action within 24 hours (22.5%, n=2039). Messages were significantly less likely to be actioned within 24 hours if they were assigned on a Friday (p<0.001, OR 0.663, 95% CI 0.530–0.828) or at the weekend (p=0.001, OR 0.276, 95% CI 0.130–0.585), with a median time to action of 22.8 hours (range 1.9–94.1) and 37.3 (10.1–54.1) hours respectively, relative to Monday with a median time of 20.2 hours (2.2–48.2). The time taken also changed depending on drug type (p=0.001), with messages assigned to anti-infective drugs actioned nearly four times quicker, median time 6.0 hours (1.2–26.4), compared to the reference category cardiovascular medicines with a median time of 23.0 hours (3.4–68.8) (p<0.001, OR 2.055, 95% CI 1.546–2.732).

Conclusion: The proportion of messages actioned within 24 hours decreased significantly for those assigned on a Friday and at weekends, suggesting that such medication-related tasks may not be prioritised by physicians during this time. The lack of ward pharmacist presence at the weekend may also contribute to the time delay as interventions are not followed up with physicians. It was however reassuring, with the current focus on antimicrobial stewardship, to find that messages assigned to drugs for infection were actioned in a timelier manner compared to other medicines. This study is independent research arising from a Doctoral Fellowship supported by the National Institute for Health Research.