We conducted this study in a 26-bed medical-surgical ICU in a community hospital in Mississauga, ON, Canada. The ICU operated using an intensivist-led interdisciplinary team model. The ICU, containing only individual patient rooms and 2 nursing stations, is physically divided into two geographic areas separated by approximately 150 m: one 20-bed unit and one 6-bed satellite unit. The area of the main ICU is approximately 480 m2. The study population consisted of 4 intensivists and 121 interdisciplinary staff members (nurses, respiratory therapists, pharmacists, social workers, unit clerks, and ICU nurse-leaders) who worked daytime shifts. Night staff did not participate in the trial. With the exception of respiratory therapists, who worked in all hospital departments, staff members worked almost exclusively in the ICU.
Baseline Communication Methods
Prior to study implementation, ICU staff members used various communication methods, including receive-only numeric pagers, overhead paging, physical searches for staff, handwritten messages left in or near patient charts, and e-mail, which required the use of the hospital e-mail system (Meditech, Westwood, MA) and could only be accessed using desktop computers. Intensivists acquired personal wireless e-mail devices (BlackBerry, Research in Motion [RIM], Waterloo, ON) at their own expense over a six-month period before the start of the trial.
All baseline communication methods remained available during the study period. ICU staff members were free to choose the method of communication depending on the clinical situation and could use several different methods if needed. For example, the initial notification of a problem could be via wireless e-mail with a preliminary response via wireless e-mail followed by subsequent direct in-person communication. We did not record usage data of baseline communication methods before, during or after the study period, but subjective experience suggests that all methods were used throughout the study depending on the clinical context.
During the study, staff accessed wireless e-mail using BlackBerry 7,290 handheld devices using cellular phone frequencies (Global System for Mobile Communications [GSM] 850/1900 MHz) with maximum power output of 2 W at 850 MHz and 1 W at 1,900 MHz. This wireless device uses push technology to send e-mails without delay to the recipient's device, eliminating the need for users to log into an account to check for new messages. In this study, device vibration alerted users to the receipt of a new message. Messages could be sent to one or multiple recipients. A server encrypted all messages to ensure secure message transmission and compliance with provincial and national privacy regulations. All devices were password-protected, with a 1-hour time-out period, and stored data could be remotely erased from any device by an administrator in the event of loss. Because of these security measures, there were no restrictions on the inclusion of personal health information in messages. We did not use other device functionality such as voice or calendar capabilities.
We used wireless e-mail as our method for real-time push text communication for the following reasons: (1) security (as discussed above), (2) form factor: the selected devices had a full character keyboard that facilitated text entry, and (3) integration with hospital e-mail system: e-mails created with the wireless device were also stored on the hospital e-mail server. In contrast, other methods of text communication, such as text messaging from mobile phones (short message service, or SMS), are limited by lack of encryption, slow message creation using numeric key pads, and lack of integration with the hospital e-mail system.
Devices were assigned to specific roles or to individual ICU leadership staff. The former category comprised 28 devices, distributed to bedside nurses (up to 19/daytime shift), the charge nurse (n = 1), respiratory therapists (n = 2), social worker (n = 1), pharmacists (n = 2), and unit clerks (n = 3). These users received a specific device at the beginning of the shift and returned it at the end of the shift. Six devices were assigned to leadership staff members, who retained the same device for the entire study period. This category included clinical educators (n = 2), clinical leaders (n = 2), ICU nursing director (n = 1), and ICU administrative assistant (n = 1). Intensivists (n = 4) supplied their own personal devices and cellular network access during the study period. They could send orders to specific nurses by wireless e-mail, which were transcribed into the hospital's paper chart.
Prior to the study, we created an implementation group with 15 members representing all user types. The group met 4 times over 3 months before device deployment and developed policies regarding device use, message content, device security and implementation.
We introduced wireless e-mail into the ICU on Aug 1, 2005 for a 6-month study period. Implementation group members provided device training to ICU staff on an as-needed basis. Devices were used during the day 7:30 am to 7:30 pm shift only and charged at night.
We recorded the number of e-mail messages transmitted and received during approximately 5 months of the 6-month study period by reviewing the secure e-mail account associated with each role-specific device. We also recorded e-mail volume during a subsequent 5-month nonstudy period (Jan 6, 2007–Jun 6, 2007). Specific times that messages were received and opened by the recipient were not available from the wireless service provider or the local wireless e-mail server.
We developed the questionnaire (see Appendix, available as an online data supplement at http://www.jamia.org
) using standard survey methodology. 42,43
We first generated survey items by searching MEDLINE (1966 to second week October 2005) for papers on wireless communication in the hospital (search terms included e-mail, wireless e-mail, BlackBerry, communication) and by conducting semi-structured interviews with 12 ICU staff (nursing, ICU leadership, pharmacy and respiratory therapy) and 4 ICU physicians. We developed content areas of interest (domains) and specific questions (items) within domains. Questions contained stems with ordinal response frames using 7-point Likert scales, with 7 indicating the highest level of agreement or most favorable response, 1 indicating the least favorable response, and 4 indicating a neutral response. Three intensivists with methodology training (NA, JF, DS) reviewed the instrument for redundant items, face and content validity, comprehensiveness, and clarity. A focus group of 17 ICU staff led by the hospital patient safety director provided a similar review. We then pretested the penultimate version using 11 members of the implementation group and 18 attendees at an interdisciplinary staff meeting. Finally, we assessed clinical sensibility of the final draft by administering it to 5 ICU staff and obtaining feedback using a 6-question instrument. The final questionnaire comprised 49 questions.
A unit clerk administered the questionnaire to ICU staff, tracked respondents, and collected completed forms over a 5-week period (Dec 22, 2005–Jan 27, 2006). Completion of the questionnaire was voluntary, and all responses were confidential. We provided an incentive (Can $2 gift certificate for coffee) for questionnaire completion. The hospital research ethics board approved the study.
We present e-mail usage data as the mean (standard deviation, SD) number of messages transmitted and received daily and survey responses as median (interquartile range, IQR) scores on 7-point ordinal Likert scales, unless otherwise stated. We used t tests to examine for differences in the number of e-mail messages sent and received during and after the study period. For each survey question, we used the sign test to test the null hypothesis that the median response equals 4, the neutral position on the Likert scale. All p values are two-sided. We also collapsed items on the Likert scales into categories representing favorable (Likert = 5,6,7) versus unfavorable (including neutral) responses (Likert ≤ 4). Graphical representations of our results were constructed. The small number of nonnursing respondents precluded reliable comparison of results among healthcare disciplines. Calculations were carried out using SAS 8.0 (Cary, NC).
Role of the Funding Sources
Research in Motion (RIM, Waterloo, ON) provided devices for the study on an unrestricted basis. Rogers Wireless (Toronto, ON) provided unlimited data communication over a local GSM cellular network on an unrestricted basis. After the study period, the hospital provided operational funds for ongoing wireless e-mail usage, but did not have to pay for the devices. The funding sources had no role in the conception, design, or conduct of the study; collection, management, analysis, or interpretation of the data; or preparation, review, or approval of the manuscript.