Page 28 - Journal of Special Operations Medicine - Winter 2016

one user trained the other before the data collection, In 2013, we reported results of an experiment in which,
the two users may have removed similar but unequal if the variable chosen for the effectiveness outcome was
amounts of slack before turning the windlass. binary (yes-no for hemorrhage control) as opposed to
continuous (blood volume loss), then the learning curve
When assessing the number of replications necessary for was flat instead of the expected classic power curve. The
22
the inexperienced user to match the proficiency of the flat learning curve started at or near the desired outcome
expert user, the large number of replications required (a yes answer to hemorrhage control) from the first or
(n = 101) was surprising because we anticipated that 50 second iteration of use and remained there. On the other
might suffice (Table 2). If the goal of instruction is to hand, and at the same time for the same user, performance
produce expert users, then the number of training uses improved when measured by continuous data (blood
may need to exceed 100, approximately the threshold loss volume and time to effectiveness). 21,22 A flat learn-
where the less experienced user attained expert levels ing curve appears good superficially because it implies
of performance in this study (Table 2). The two orders that, in this case, the user was competent and remained
of magnitude difference in thresholds between time (or so. However, the flat curve also implies an absence of
blood loss, as continuous parameters) versus effective- learning. On the contrary, more sensitive variables (e.g.,
ness, a binary yes-no parameter, would entail about two blood volume loss and time to effectiveness) simultane-
orders of magnitude difference in requirement for both ously showed classic learning and improvement with in-
instruction time and user learning time. If these results creasing experience. These findings from the prior study
22
are extrapolated to a broader area, then a lower goal were replicated and also developed in the present study;
expectation—becoming less than expert—may be at- altogether, such findings indicated to us that the choice
tained approximately where the trend line crosses the of metric to be assessed may affect how performance is
chosen threshold, as shown in Figure 5. For example, interpreted. Performance assessment may be faster and
if the instructor chooses that the difference is simply to simpler if yes-no metrics are chosen, whereas assessment
remain under the threshold of 20 seconds for each of the may be slower, yet more comprehensive, with continuous
first five or 10 tests, then that may suffice. as it did in variables. For example, instructors may take more time
this case. However, such a method would require expert in assessing blood loss associated with tourniquet use for
values for each test, which may burden the instructor. a time involving more than 100 uses to better assess ex-
pertise. Furthermore, instructors may desire trainees to
Study of learning may be done with data to quantify become effective, fast, and safe, so assessing effectiveness,
learning, but selection of metrics, validation of values time, and pressure may be necessary.
(e.g., normal range limits, competence thresholds, expert
levels), and best practices of learning assessment all need Multiple hypotheses were generated from this study’s
future study. This study is hypothesis-generating and results (Table 3):
may stimulate further developments in first-aid learning
such as in hemorrhage control interventions or in airway 1. Learning curves of tourniquet users may vary (e.g.,
procedures. Such exploration into tourniquet use may flat, curved) by the metric (effectiveness, time) that
aid best practice development in training or caregiving. has been selected by the assessor to evaluate learning;
effectiveness may require very few uses, whereas time
Minor findings of this study deal with four learning top- to effectiveness might require many more repetitions.
ics that may aid in the design of future studies. First, 2. Selection of more than one metric may assess learn-
measures of learning that can be made readily available ing more comprehensively, but such selection may
to tourniquet instructors include effectiveness, time to take more instruction time. Selection of both effec-
effectiveness, and blood loss volume; such measures tiveness and time (uses must be both effective and
may also be candidates for useful feedback to trainees. within the time limit) may require more assessment
Second, not every metric studied in this study had an time by the instructor.
obvious value in assessing learning. For example, us- 3. There may be interinstructor variability in their
ers showed little learning associated with the amount awareness of the utility of different metrics. These
pressure needed when applying the tourniquets. On the results are new; therefore, few instructors may be
other hand, for one user—as an example of a sizeable aware that metrics may have specific value in assess-
learning association—not only was more user speed as- ing learning.
sociated with more experience but most (72%) of the 4. Instruction goals (e.g., expert performance or a spe-
improvement was associated with that change in experi- cific threshold for competence) may affect the number
ence. Finally, differences (i.e., interuser, intra-user, and of uses required. If expert use is sought, many it-
intermetric differences) were occasionally large (e.g., the erations may be required. If a specific threshold of
difference in user learning in mean blood loss for one time or blood loss is used, then fewer uses may be
tourniquet model was 284mL). required. However, if the goal is complex (e.g., to be

12 Journal of Special Operations Medicine Volume 16, Edition 4/Winter 2016