Lyndon Garrett and Dr. Curtis LeBaron, Organizational Leadership and Strategy

CO-AUTHORS: Michael D. Cohen, PhD; Roy Ilan, MD, MSc; Lyndon Garrett; Curtis D. LeBaron, PhD; Marlys K. Christianson, MD, PhD

Hospital handoffs have been recognized by both regulators and researchers as a locus of potential communication failure with risks to patient safety and quality of care. The initial objectives of my orca grant research were to explore the roles of both the incoming and outgoing physicians in the successful accomplishment of a shift-change patient handoff. As we got into our study, we realized an important oversight in the literature. While most empirical studies on shift-change have been done in settings where multiple patients are transferred, theoretical discussions in the literature have focused on how best to hand off a single patient. As a result, research has overlooked what has been called the portfolio problem, how best to allocate available discussion time across multiple patients.

We have used video recordings of experienced attending physicians in the Intensive Care Unit (ICU) of a tertiary medical center to conduct the first study of this issue. We found that our participants allocated disproportionate time to cases that happened to come early in the order of discussion. This occurred despite the order of cases being unrelated to severity of illness.

METHODS

Between December 2008 and July 2009, 23 handoff sessions were recorded at week’s end, just prior to the transfer of responsibility for the 21-bed ICU from two outgoing to two oncoming physicians. Characteristics of our participants are shown in Table 1. Our video data were collected before we formulated the hypothesis that discussion order is related to duration.

The procedure followed in the ICU was to discuss patients in bed list order. With unpredictable patient arrivals and all rooms equally equipped, the order of the cases was effectively randomized. Our main measures of interest were constructed from the videos: the number of patient-discussions in each of the 23 sessions; the ordinal position of each discussion in its session; and the duration of each patient-discussion in seconds. The durations were measured by three independent coders, including myself, then subsequently analyzed by myself and Michael Cohen while I was in Michigan for a month working on this project. Any discrepancies between the coders of more than 3 seconds were reviewed by Michael Cohen.
We ran into several problems with the analysis, for example, regarding the patient order, to be the 3rd patient discussed in a group of 6 is different than to be 3rd of 12. We also had issues with patients being discussed multiple times in a handoff or patients who were skipped. We met several times with a statistical consultant at U of M who guided us in how to do the analysis.

RESULTS

We observed a total of 262 patient-discussions. Mean duration was 142.73 seconds (SD 98.20). A median session had 11 discussions (range 6 to 23).

Kendall taus were less than zero for 19 sessions. Their overall mean was -0.186 (median -0.282). A 100,000 replication random replacement simulation using our dataset’s exact structure of sessions and of patients-per-session, determined that the observed pattern of Kendall tau coefficients would occur by chance with p < .0001. To determine the magnitude of the portfolio effect, we used three approaches that produced highly similar estimates. We compared the mean fraction of a session used by patient-discussions that occurred first (0.116) and last (0.075) (The Wilcoxon signed rank test gives p< 0.0373.)

Because our predictor variables are ordinal, the strictly appropriate estimation using all data is monotonic regression applied separately for each session size. The weighted averages of first and last estimated duration proportions using this method are 0.126 and 0.075. Linear regression is often used to approximate relationships with ordinal predictor variables, although they may not strictly meet its assumption. The estimated first and last duration proportions for session size 11 are 0.118 and 0.071.
The three approaches all estimate that in a median session covering 11 randomly ordered patients, the first discussion used at least 50 percent more time than the last, while the average time received declined steadily with increasing ordinal position.

CONCLUSION

This study of discussion order and duration is the first of its kind. If replication confirms our findings, it suggests that shift-change handoffs, and programs that train clinicians in handing off, should include methods for explicitly controlling the allocation of scarce time across the portfolio of patients being handed off. Examples of such methods include discussing the “sickest” patients first, or the newest first, or deliberately concluding with a block of time reserved for returning to cases requiring further discussion.

COMMENT

This paper has been submitted to the Journal of the American Medical Association (JAMA), one of the highest impact journals in academia, and is currently under review. My involvement included coding and analyzing the data, learning various ways to compute statistical analyses, participating in conference calls with the research team, spending a month in Michigan working through various components of the project, and preparing conference presentations of our project. I not only learned valuable research skills, but I feel like this is a significant contribution to the medical field, and also offers potential insights to the routines literature in the field of Organizational Behavior, which is my field. The contributions to Organizations research may include a better understanding of how knowledge transfers through interaction are structured, and how knowledge resources may be constrained by the structure of the interaction. We are currently working on a second paper from the same dataset on how routines are negotiatively maintained and evolved through interaction. This second paper will be sent to an Organizations journal for submission in the next few months.