A learning curve of LMA^® ProSeal™ insertion: a prospective analysis of cumulative sum method

Abstract

Purpose

LMA^® ProSeal™ (pLMA) has been used as an alternative to tracheal tubes. It is unclear how many cases are required to achieve proficiency in performing pLMA insertion among novice residents. Therefore, we analyzed the learning curve of pLMA insertion using a cumulative sum (CUSUM) chart and assessed the effects of learning.

Methods

In this single-center, prospective, observational study, we included 15 novice residents. Staff anesthesiologists recorded success or failure; insertion time; and incidences of bleeding or reflex including cough, hiccups, and limb movement. A successful pLMA insertion was defined as effective ventilation within two attempts with an insertion time of ≤ 120 s. Regarding CUSUM, we set acceptable and unacceptable failure rates as 20% and 40%, respectively. Further, α and β errors were designated as 0.1. We stratified the number of cases encountered by each resident into four groups of 10 cases each (1–10, 11–20, 21–30, and ≥ 31 cases) and evaluated the effects of learning.

Results

Each resident encountered 44 ± 5 (mean ± SD) cases of pLMA insertion, and 14/15 achieved proficiency in performing pLMA insertion after 20 ± 8 cases. Success rate (76%, 86%, 91%, and 93%; P < 0.001) and insertion time (45 s, 35 s, 31 s, and 26 s; P < 0.001) significantly improved with increased experience; however, incidences of bleeding (16%, 10%, 8%, and 10%; P = 0.124) and reflex (5%, 3%, 3%, and 3%; P = 0.54) remained unchanged.

Conclusion

Experience with 20 ± 8 cases is needed to achieve proficiency in performing pLMA insertion for novice residents in a tertiary teaching hospital.

Appendix

The cumulative sum (CUSUM) method is a statistical method that focuses on the outcome rather than the process of performing procedural skills, and it has been performed to assess an individual’s procedural performance [12,13,14]. For CUSUM, acceptable (p0) and unacceptable (p1) failure rates and type I and II errors (α and β) were predefined [12].

Upper and lower decision limits (h1 and h0) were calculated as follows:

$$h1 = a/\left( {P + Q} \right){\text{ and }}h0 = - b/\left( {P + Q} \right),$$

where a = ln [(1 − β)/α], b = ln [(1 − α)/β] and

$$P = \ln \left( {p1/p0} \right){\text{ and }}Q = \ln \left[ {\left( {1 - p0} \right)/\left( {1 - p1} \right)} \right].$$

CUSUM charts were plotted with case number on the x-axis and CUSUM on the y-axis. CUSUM values started at 0, and the quantity S [Q/(P + Q)] was subtracted from the previous value when a successful attempt was recorded, resulting in a downward trend. When a failed attempt was recorded, the quantity 1 – S was added to the previous value, resulting in an upward trend. If the line crossed the upper decision limit (h1) from below, the true failure rate was regarded as significantly greater than the unacceptable failure rate. If the line crossed the lower decision limit (h0) from above, the true failure rate was considered not significantly different from the acceptable failure rate. If CUSUM remained between two boundary lines, statistical inference could not be determined.