Time to consensus: the effect of the stomach on consensus decision-making at large conferences

Methods

The authors sat on opposite sides of the room to provide a different perspective on the meeting process. The time taken to make a consensus decision was defined as the amount of time that passed from when the chair moved the meeting to address the specific recommendation to the time when the final vote was counted and the motion passed. Timing was done using synchronized Timex watches. Given the margin of error for defining the start of the discussion, it was felt that the use of a stopwatch was not necessary. Both authors did the timing and the average measure of time was used.

The time of discussion and decision-making was correlated with the morning and afternoon coffee breaks, lunchtime, and with the end of the day. The day was broken into 15-minute increments, numbered sequentially starting after the introduction at the beginning of each session. The quality of the morning snack versus the afternoon snack was assessed using the Frank–MacKnight Adequacy of Food at Breaks Rating Scale1 and correlated with the association between time to consensus and snack on the second day.

Given that different topics could be more contentious than others, the authors developed a PITA (pain in the ass) score based on the self-rated report of frustration by each reviewer. This is a score out of 10 with 1 being blissfully content and 10 being “really ticked off.” The PITA scoring system was informally pre-tested at a Medical Staff meeting 1 month before the consensus conference.

Data were analyzed according to standard methods, including inputting (i.e., fabricating) all missing data.2 Our work was approved by the Institutional Review Board of the host hotel.

Results

The meeting began at 0830 each day with breaks at 1030 and 1430 for 15 minutes. On the morning of the first day, 8 recommendations were voted on. The mean amount of time taken on discussion and voting during this period was 9 minutes and 12 seconds (standard deviation 7 minutes, 4 seconds). The amount of time taken varied over the course of the morning, with a mean of 14 minutes and 14 seconds in the first six 15-minute time intervals compared with a mean of 8 minutes and 27 seconds in intervals 7 and 8 immediately before lunch (p < 0.05).

Photo by: Alan King

A similar pattern was seen in the periods between morning break and lunch, and was repeated in the afternoon before the break and at the end of the day (Table 1). The most rapid consensus was achieved immediately before the end of the day.

The association between the rating of food quality and speed of consensus is seen in Table 2. Overall, there was a trend (p = 0.59) toward an increased speed of consensus when food quality was rated high, especially as it related to lunch versus snack. The impact of uncertainty on the part of participants about the quality of the snack was factored in using similar scores before the morning and afternoon snack breaks.

PITA scores measuring association between frustration with the recommendation being discussed and the time to consensus are seen in Table 3. There was a significant association between decreased time and a PITA score > 8. Overall, PITA scores tended to be higher late in the day, but this was not statistically significant.

Interpretation

This is the first paper to study subtle variables in the time-honoured tradition of consensus conferences. We found that there was an association between speed of consensus and the proximity to meal/food times. The impact of this factor on the quality of recommendations was not investigated, but this study raises concerns about the validity of recommendations made immediately before lunch and breaks. Likewise, there was an association between the self-rated frustration of participants and the TTC. This factor has been identified by participants in such conferences for years, but this is the first time that such an association has been rigorously studied.

It is unclear what strategies can be used by conference organizers to mitigate the effect of these findings. Providing hospital food as the snack, or preferably lunch meal, would decrease participants' interest in hastening a discussion before break. Eliminating food altogether may end up being counterproductive and may increase participants' “crankiness,” leading to greater impact of the PITA factor. Identification of the association between the PITA and the TTC may allow organizers to intervene when the tension is rising in the room and to use strategies such as group hugs, “7th inning stretches” or primal scream therapy to decrease frustration.

This study has several limitations. Observations were only possible on day 1 of the conference, and there may be different findings for longer conferences. Participants may tire and become apathetic, possibly adding to the effect of snack and lunch breaks on TTC. The results may also have been confounded by drowsiness, and a major fault in our paper is that we paid no attention to the ground-breaking drowsiness work of Rockwood and colleagues.3 Alternatively, the PITA factor may be greater the longer the conference. The differing culinary tastes of the researchers may have affected their observations. It is worth noting that both authors come from families with Scottish backgrounds; researchers with family from other countries, such as France or Italy, may have higher standards for the food provided.4

Future research in this field should include studies of the effect of these factors on the quality of recommendations being produced. The current study does not assess this and, indeed, the recommendations arrived at most quickly may be the most valid.

Conclusions

The time taken to agree on consensus recommendations is strongly associated with time to food breaks, especially lunch. The quality of the food and the level of participants' frustration with the question at hand also influenced TTC. Organizers should take these findings into account when planning consensus conferences.