The results of this study, which involved 217 students in seven different medical schools in developed and developing countries, show that a short training course presenting a normative model for pharmacotherapy can significantly improve the ability of students to solve written patient problems.

The results of the same open question A at T1, T2 and T3 showed that the study group remembered what was learned (retention effect), even after six months; in the same period the scores of the study group remained virtually the same. The better score was obtained in each of the six steps of the problem solving routine. No such effect was seen in the control group.

The results of the open questions A at T1, B at T2 and C at T3 are proof of a transfer effect: students in the study group were able to apply this knowledge to new patient problems. It is especially important to note that this effect was also maintained for at least six months. Again, no such effect was seen in the control group.

A last indication of the success of the method is the outcome of the questionnaire which indicated that the students felt confident in prescribing skills, although less in choosing a dosage for a drug and the duration of treatment As there is no control group in this subjective measurement the outcome needs to be interpreted with caution.

How do we know what has been transferred; knowledge about the pharmacotherapy of pain, or the skill in solving problems of patients with pain? The highly structured questions X, Y and Z were specifically designed to measure the knowledge on the treatment of pain, as compared to problem solving ability. The difference between study and control groups was largest in the open questions A, B and C, and much less in the structured questions. This suggests that the transfer effect in the study group is mainly due to better problem solving skills, rather than increased knowledge.

The significant and persistent positive impact of such a short training course may meet with some scepticism. Objections against the results could be: (1) students are self-selected and not representative; (2) teachers were biased in their scoring; (3) “cross-contamination” occurred between the study and control groups, especially in the period between T2 and T3; and (4) control students could learn from the test, and score better in subsequent tests.

All participating students were self-selected and all were probably eager to learn; in general they may have been better than average students. However, assignment to one of the two groups was random and their mean scores for questions A and X at T1 were identical. There is therefore no evidence that students in the study group were better than those in the control group.

Possible teacher scoring bias was separately tested in two universities (New Delhi and Yogyakarta), through an additional blind scoring of all test results by independent experts. In both cases the independent scores gave exactly the same results as the teachers’ scores, there were no systematic differences between the teachers’ and independent scores, and there was a significant correlation between the teachers’ and independent scores. We therefore conclude that there was no teacher scoring bias.

Cross-contamination between the two groups was prevented as much as possible by careful instruction of the students. Often the control students were very disappointed to be excluded from the training and were eager to learn as well. However, they were promised their own copy of the guide (after T3) and in two universities the same training programme was given to them as well, after T3. Cross-contamination can therefore not be excluded completely, but it should be stressed that, if it occurred, it would have decreased the difference between the two groups and would have hidden rather than flattered the positive impact of the training.

A last source of bias is the fact that control students can learn from the tests and score better in subsequent tests. This could only have been prevented by enrolling a second control group at T2 and a third at T3, which was practically impossible. However, the nearly identical scores of control students for problem A in all three tests indicate that this effect is probably limited and restricted to prescription writing only. Moreover, as with cross-contamination, it would have hidden rather than increased the positive outcome.

It has been mentioned already that the results in this study are only valid for patient problems concerning pain. Nevertheless, the results are very similar to those of a previous study with mixed problems.²⁷ This would suggest that the normative problem solving method is generally applicable in developed and developing countries, although a separate study with mixed patient problems would be needed to give definitive proof. It should also be stressed that our study assessed student competence to solve a written patient problem, which does not necessarily predict their actual performance in subsequent clinical practice. In view of the promising results, it seems worthwhile to study the impact of this training method on subsequent performance.

It may seem surprising that a short training course of only eight hours of teaching can induce the persistent transfer effect demonstrated by this study. In view of the impossibility of teaching students all basic knowledge on the thousands of drugs available, this approach seems to constitute an efficient way of improving the rationality of prescribing. However, this method can only be successful when it is accompanied by a fundamental change in the teaching methods of the trainers, away from the habit of transferring knowledge towards real problem-based teaching of pharmacotherapy. The principles of rational prescribing should also be reinforced in the clinical phases of training.