gms | German Medical Science

Learning is a lifelong process which today does not end with the achievement of a certain qualification. It requires innovative forms of learning which independent of time and location (E-Learning [1]) and teaching materials where the content is tailored to the needs of the target audience. We developed educational materials for the training of healthcare workers in anatomy for an e-learning platform on Moodle [2]. Therefore we were interested in the question of which type of learning ideally promotes growth of knowledge. According to Winteler [3], learning should be interactive and constructive and ideally take place in learning communities. In terms of e-learning, Ellaway and Masters [4] see a need for not only storing documents online but also to enable didactically meaningful access and for encouraging users to interact, collaborate and communicate.

Stark et al. [5] compared to instruction-oriented and problem-based learning in teacher education. Here, the instruction-oriented approach proved particularly effective for participants with little prior knowledge. This superiority was, however, applied to tasks which require reproducible knowledge. Harter et al. [6] compared a larger cohort in their study (421 students of the 4^th semester at their own university) across 11 biochemistry seminars regarding the impact and acceptance of talk-and-chalk versus interactive group work. They found that students both judge talk-and-chalk more highly and performed significantly better in the reproduction of knowledge compared to those who had gained their knowledge in groups. But in a second test 4-6 weeks after the knowledge test, the initially found superiority of talk-and-chalk teaching is no longer detectable. Lohrmann and Herbert [7] analyzed 10 curricula for the training of healthcare workers. Those who used a large repertoire of active learning strategies and spent enough time on adequate instruction performed best, with discussion in small groups of 2-6 students playing an important role. Hopkins et al. [8] found no significant effect in terms of knowledge gain in three groups of students who acquired knowledge on the masticatory system via prosection in a dissection course, through a virtual model or such using both methods. Their learning was also observed. It was found that groups which used with the computer model split into subgroups which worked independently, regardless of the original group size (3-6 students). Beyond the learning success, it also seems to improve social interaction between participants. Hopkins et al. [8] highlight the differences between team and (small) group learning.

The focus of our work was not the analysis of group dynamics but the design of material which will allow newly graduated healthcare workers (novices) and even physiotherapists with long-term experience as part of CPD (experts), the highest possible knowledge gain. We therefore compared a learning offer containing the solutions (reproductive) was compared with material in which these solutions had to be worked out (interactive).

The primary research question was whether interactive learning leads to better success than reproductive learning of comparable content. A second research question aims to establish whether both groups outperform the control group.

According to these research questions, the following null hypotheses were tested:

1.

The results of the post-test do not differ between the interactive and reproductive intervention group.

2.

The post-test results differ in the intervention groups on the one hand but not in relation to the control group.

Furthermore we examined what other factors influenced the results of the post-test.

Graduates of medical vocational colleges (In training) in the fields of healthcare with comparable knowledge in anatomy participated in the study a few months before their final exam.

The second group consisted of graduates in physiotherapy with long-term experience (Training completed) who were participating in osteopathy CPD.

The written consent of the Ethics Commission is available.

The study had a two-stage quasi-experimental study design (see Figure 1 [Fig. 1]). The prior knowledge of the participants was determined through a test before the intervention. Since the results were very balanced amongst the various vocational classes and the participants of the osteopathy CPD course for organisational reasons (schedules, travel to Leipzig etc.) a decidion was made not to mix them. The vocational college students remained as a group, the participants of the postgraduate CPD course were divided into three numerically comparable groups. The first worked interactively, i.e. they solved tasks in small groups of 4-6 with the help of various sources (textbooks and anatomy atlases, see Figure 2d [Fig. 2]) in 45 minutes. Interactions between participants was not required. In general, they sought the answers to each question individually in various books and formulated their answers. As far as they were observed, questions were dealt with in their predefined order. They were not observed regrouping them (e.g. by topic). The second group read the material which contained the solutions for the same period of time. Learning success was measured through a test about one week after the intervention, relating directly to the content of the working materials.

The control group was recruited from vocational college graduates or participants in postgraduate training who had had the same training as those in the experimental groups. To determine this, we obtained reliable data from the vocational college's administration and the head of osteopathy training. The control group took part in both tests but not in the intervention.

The pre-test (10 min) consisted of 20 questions in multiple choice format on different areas of the anatomy (locomotor system, gastrointestinal tract, nervous system, etc.) with three possible answers, one of which was true. A total of 20 points could be achieved.

The working materials for the intervention on various topics from the field of macroscopic anatomy was identical for the reproductive and the interactive group. 22 questions dealt with the gastrointestinal tract, 19 dealt with the spinal-cord and 20 with the cardiovascular system. The reproductive group was given the learning content while the interactive group had to work it out themselves. The following primary task types were given: Labelling of arrows, completing gaps in texts, checking correct answers and labelling of images.

Figure 2 [Fig. 2] shows the same task, the left designed for reproductive learning (see Figure 2a [Fig. 2]), the right for interactive learning (see Figure 2b-d [Fig. 2]).

The materials were accompanied by a short evaluation form to gather data on the views of the participants regarding topic selection, processing time, memorability, etc. using a 5-point Likert scale. On this scale, A=does not agree at all, B=agree somewhat, C=broadly agree, D=agree, E=agree fully and F=no response.

The post-test consisted of 27 open questions (example in Figure 3 [Fig. 3]) in which they were required to give answers to the questions or label illustrations. The test content was based on the themes of the work materials. A total of 12 questions related to the gastrointestinal tract, 10 to the spinal-cord and 5 to the cardiovascular system. Processing time was 30min. A total of 36 points could be achieved.

Cronbach's α to assess the internal consistency for the pre-test is 0.476. The difficulty of the items should be between 0.4 and 0.8. There were no very difficult tasks (x<0.2) but six very easy tasks (x>0.8). The selectivity was over 0.2 for only four items.

Regarding the item analysis of the post-test results Cronbach's α=0780, the values for the item difficulty and discriminatory power for almost all items are >0.2.

The data was analysed using Microsoft Office Excel 2003 and SPSS 18 for Windows. The composition of the groups was described by absolute and relative frequencies and with respect training type and test results in terms of group affiliation. Mean, median, standard deviation, standard error, minimum and maximum were taken for the (quasi-)continuous parameter distributions. As a screening test the Kolmogorov-Smirnov test (KS) was used. With p>0.1 it was assumed that no deviation from normal distribution was present. When deviating from the normal distribution, the data was transformed in a suitable manner. For tests regarding distribution differences, the transformed data was used. For descriptive statistics and graphic representation the original values were retained; the median was then given as the expected value of the distribution.

The main instrument was variance analysis (ANOVA). It was used both for the comparison of multiple groups and to determine the influencing factors (intervention group, training group, pre-test result or a combination thereof). First, the models were calculated with all interactions and then gradually the non-significant factors were removed. The error probability was evaluated as p=0.05, consequently, p<0.05 was treated as significant. In addition, 0.05<p<0.1 was judged as a tendency.

For the graphic representation of the post-test results as a function of the intervention group, training and pre-test result, Box-Whisker Plots were used. The blocks indicate the 25%-75% percentile (interquartile range). The average mark indicates the median. The beams are limited by minimum and maximum. The outlying circles and stars indicate outliers and extreme cases.

Of the 301 participants, results for pre-and post-test were available for 216 participants (71.8%) to which the analysis refers (see Table 1 [Tab. 1]).

Age did not play a central role but accounted for, as the Training Group factor, in the analysis via the different levels of training of the participants. The average age was 20.8 years for graduates from the vocational college (min. 18, max. 31), the participants of the postgraduate course 35.1 years (min. 23, max. 58).

Gender was also not considered explicitly as only 40 of the 216 participants were male (30 graduates from the vocational college and 10 future osteopaths).

The following table shows the descriptive statistics of the distributions for the pre-test (see Table 2 [Tab. 2]):

The participants were split into two groups based on the results of pre-tests (median border </=15 points, >15 points).

The following table shows the descriptive statistics of the distributions for the post-test (all questions) (see Table 3 [Tab. 3]):

This analysis looks at subjects which belonged to the interactive or reproductive intervention groups. Table 4 [Tab. 4] shows the results of the univariate trifactorial variance analysis with which the impact of the intervention (interactive vs. reproductive), the level of education (vocational college graduates or postgraduates) and the pre-test results (</=or>15 points) with all interactions with the result of the post-test.

The training group (p=0.005) and the pre-test result (p<0.0005) have a significant influence on the results of the post-test. For the intervention group, a trend was found (p=0.089) and an interaction with the results of pre-test. When summarizing the results of the pre-test, a significant difference (p=0.003) between the intervention groups (22.8 (interactive) vs 20 points (reproductive)) is found for one pre-test result </=15 points. For one pre-test result >15 points on the other hand, a small, non-significant difference between the intervention groups is detected: 24.6 (interactive) vs 25.6 points (reproductive). The null hypothesis of no differences between the interactive and reproductive intervention can thus be rejected for subjects with bad pre-test results.

When comparing the vocational college graduates with the participants of the postgraduate course, test subjects who have completed vocational training and usually with several years of professional experience fare better in both intervention groups (interactive: 25.7; reproductive: 25.4) compared to those still in training (interactive: 22.9; reproductive: 21.7). Figure 4 [Fig. 4] and 5 [Fig. 5] illustrates the results: Using Box-Whisker Plots, the distribution of the post-test results in relation to the intervention group, the pre-test results and the training group is shown.

This analysis compares subjects from the intervention group with those in the control group. There was a significant influence by the intervention (yes/no) (p<0.0005). On average, participants in the intervention group attained 23.3, the control group 19.8 points in the post-test, i.e. the null hypothesis of a lack of difference between intervention and control group can be rejected.

Figure 6 [Fig. 6] shows how many of the 71 participants solved the problems correctly (red), partially correct (blue) or incorrectly (black) or not at all (grey) using bar charts.

The first section (Gastro) was tackled by most participants, for the most part correctly. The number of unprocessed tasks begins to sharply increase in the second section (Spinal) already and the tasks from the third section (Cardio) were tackled by only a few participants.

While there was sufficient time to read the questions (reproductive group), the interactive group needed more time.

The difference between the means (interactive: 1.27 and reproductive: 4.14) were statistically significant in the T-test for independent samples (p<0.0005, T=-25.359 and df=155.823) (see Table 5 [Tab. 5]).

The participants’ views in answer to the questionnaire item asking if 45min was sufficient processing time reflects this (interactive: 17 participants; reproductive: 3 students: “too little time”).

If the univariate trifactorial variance analysis is applied only to questions on the gastrointestinal tract, the effects shown for the entire dataset are enhanced.

A significant effect for the intervention group (p=0.006) and the training group (p=0.042) can be seen. The intervention groups differ significantly, as already shown for the entire dataset, from the control group (p<0.0005).

The null hypothesis of a lack of differences between interactive and reproductive group can thus be rejected with reference to the partial results (Gastro) for the entire sample, not only for subjects with weak prior knowledge (interactive vocational college graduates: median=15, reproductive vocational college graduates: median=14; participants of postgraduate course: median=16, reproductive postgraduate course participants: median=15).

The results show that interactive learning leads to better results but initially takes more time. This also applies to a “specialist rote learning subjects” such as anatomy. When carrying out the intervention, it was noted that 45 minutes was a very tight schedule for interactive work, but sufficient time for reading the reproductive material. For reasons of comparability, however, the same processing time was retained for both groups. The option of splitting the topics was not used by the participants. There was, however, no explicit mention of this option. The topics were processed in their given order and not by degree of difficulty. Given the limited amount of time available, this is viewed as a failing in retrospect. It is recommended hat the option of group work (a number of people in the small groups focussing on different areas and combining the results, efficient use of resources, etc.) be discussed in detail with the participants before the intervention.

It should also be noted that the post-test tended to test skills which the students of the interactive group had practiced. Thus, the interactive learners had an advantage over the reproductive students. Had the post-test consisted of multiple-choice questions (as the pre-test), the results might have been different. Follow-up studies should analyse the learning behaviour in groups, in analogy to Hopkins et al. [8].

In addition, this was only a (time-limited) intervention, which arose mostly from the test conditions (e.g. participants mostly not from Leipzig). A retention test was planned for as part of this investigation. Some participants stated that they would have expected a higher knowledge gain had they had the opportunity to study the material in preparation for the post-test. However, considering the fact that the interactive participants for the most part only managed to complete the first third (gastrointestinal tract) in the available 45 minutes, it remains an open question what the knowledge gain would have been if enough time had been available or how the results would have looked in the post-test had they been instructed to process the three sub-areas as a team effort and to subsequently collate the answers.

The post-test differences between the two intervention groups might have been even more pronounced if the option of interactive work, as possible in e-learning, had been better used. The participants filled out cloze texts, labelled drawings and answered open questions. Tools and group work were permitted. In terms of computer-assisted learning, Schworm and Renkl [9] and Atkinson and Renkl [10] went much further by working with case studies, explanations and additional structured support. Here we would expect differences between the work patterns of vocational graduates who have grown up with computers and the participants of the postgraduate course [8].

The change in question format meant that the pre- and post-test were not directly comparable. The original idea was for a screening to give a quick overview and for splitting participants into similar groups beforehand with the post-test deliberately working a lot with visual material. Unfortunately, this is not easily compatible with the MC. Further investigations should therefore use open questions in the pre-test.

Several studies question whether the customary evaluation of learning success using exams, grades or oral examinations is suitable at all to adequately reflect the quality of knowledge [11]. Dochy [12] reported that in 91.5% of the reviewed studies prior knowledge had a positive effect on learning performance. The differences are particularly evident when comparing novices and experts. We assume that in the minds of novices, such as those attending the vocational college in the present study, learning processes are still very much conscious, while among experts, in this case the participants of the postgraduate course, patterns have emerged in the long-term memory which largely automated. As a result, the cognitive load might have reduced and learning performance improved [13]. We have reason to believe this as studies of the learning strategies of the participants using the established LIST (Learning at University) questionnaire [14]. The analysis explicitly includes items which allow classification into surface and deep learners [11]. It turned out that the vocational college students tended to be surface learners versus the participants of the postgraduate course who tended to be deep learners. They organise their knowledge better, elaborate and critically review. Vocational students learn more through repetition.

Beder and Darkenwald [15] described adults as more motivated, more pragmatic, self-driven and task-oriented to a greater degree in comparison with “pre-adults”. They worked harder and took more responsibility for the learning process. The general conditions certainly also played a role (CPD, the need to reconcile work and family life, etc.). The evaluation sheet accompanying the work material questioned the motivation for dealing with the topics presented using items such as “I enjoyed working through the material” or “I would like to have more training materials on topics such as ...”.

It was uniformly agreed that the topics were well chosen, clearly presented and that there was a desire for additional content. A dedicated investigation as per Artelt [16] was not carried out.

Especially in the context of postgraduate training, participants must actively form the learning process (self-control, Friedrich and Mandl [17]). Dreer [18] ascribes great potential to e-learning for encouraging self-directed learning and relies in particular on the ability to set individual priorities or omit themes with which one is already familiar. This enables learning which is independent of time and place. Regarding the use of materials on the e-learning platform (http://anatomie-sammlung.uni-leipzig.de/index.php?seite=virtueller-rundgang) information on whether the material was more interactive (time-consuming but more efficient) or reproductive (faster reading but probably also less retention) was extremely important.

Further insights are expected from the implementation of the teaching material on the e-learning platform. The interactive format for this resource has been chose, i.e. users can label photos, complete cloze texts, etc. and thus must work out solutions themselves. In case a participant gets stuck, however, there will be the option to quickly look up the solutions to minimise delays. In this way the individual pace of learning and the differences in background knowledge are duly reflected. Short self-tests should enhance learning.

By collecting data on access time and frequency additional findings are expected through the use of the content.

We would like to thank Dr Ekkehard Geipel (DGMM/ÄMM Doctors’ Seminar Berlin e.V.) for his dedicated collaboration and Sebastian Löffler (media technology) and Björn Weiler (technical editing) for the creation of the photographs for the work material as part of the project led by Adelgunde Graefe (Institute of Forensic Medicine, University of Leipzig), funded by the European Social Fund.

The authors declare that they have no competing interests.