Characteristics of decision-making process during prescribing in general practice

Background/Aim. The process of precribing decision-making by general practitioners requires numerous consultations in order to obtain maximal effects, minimal risks, and cost-effectiveness with the full appreciation of a patient's right to choose. The aim of our study was to describe the process of decision-making by general practitioners who decide on the treatment for an individual patient, and to relate the scope and nature of this process to the quality of the outcome of the decision. Methods. The study involved 53 general practitioners who worked in the Health Center, Kragujevac at the time of investigation (September-December 2002.). General practitioners made prescribing decisions, thinking aloud, for five patients with urinary tract infections (n = 2), or stomach complaints (n = 3). The resulting 265 transcripts were analyzed to determine the scope and nature of the decision-making processes. Differences in prescribing were related to the case or the practitioners′ working experience, and to their educational background. Results. Our results showed that the more years of practice the practitioners had the less treatments they prescribed, and the less additional aspects before prescribing they considered. The doctors with less experience, in most of the cases, considered the core aspects, while those with more experience more often considered the contextual and habitual aspects. Educational background of the general practitioners, and the type of a considered disease, had an influence on the decision-making process. The most optimal method for decision-making (marked as type F) was mostly used by the practitioners with the least experience, while the those with more experience mainly made their decisions in the ways considered the least acceptable. The optimal method for decision-making process does not necessarily provide the optimal therapy, so the least acceptable decision-making might not result in an inappropriate treatment. Conclusions. The observed prescribing decisions were mostly in disagreement with the Good Clinical Practice. Our study pointed out the need for the obligatory continuation of medical education of general practitioners in decision-making process during prescribing.


Introduction
Prescribing is a very important activity in general practice, considering that most consultations with a general practitoner result in the prescription of a drug to be issued 1,2 .According to Lexchin 3 , an appropriate prescribing means that the practitioners should try to maximize effectiveness, minimize the risks and costs, and also consider the patients' choices.Inappropriate prescribing could harm both an individual and the society.
It is obvious that decision-making in treatment options is a rather complex process influenced by many variables 4 .It requires enough time and a positive attitude toward the access toa critical appraise and getting of information for the benefit of both a patient and a doctor 5 .It is important to understand how physicians make decisions, in order to promote the appropriate delivery of effective services 6 .Although we know that attempts to change a physician's behavior in order to improve the quality of care are difficult to conduct and take many complex interventions 7 , we believe that it is a highly important endeavor for medicine − it reflects a current emphasis on the cost, evidence, and a patient's participation in his treatment 6 .The first fundamental step in an attempt to improve the quality of prescribing is to take an insight into a physician's drug prescribing patterns in order to identify prescribing problems 8 .
In the search for the optimal model of prescribing behaviour, which could enable the best possible care and treatment for our patients, we face the question: Does daily practice correlate with this theoretical model or (and to which extent) disagree with it?The aim of this investigation was to describe decision-making process of general practitioners who decide on the treatment for a patient, and to connect the scope and nature of that process with the quality of the final decision.

Methods
Out of 70 general practitioners, who worked in the Health Center, Kragujevac at the time of the investigation, 53 accepted to participate in this study.Ten general practitioners did not accept to be involved in the study (5 specialists in general practice, and 5 medical doctors without specialization).Seven medical doctors (two of them were spe-cialists) were absent (for different reasons) for a long time, therefore they were not accessible at the time of the investigation.Out of the participants, 32 were doctors who had specialized in general practice, and 21 who had not.This specimen was stratificated according to the practical experience and the level of the doctors previous education (Table 1).
The investigation was conducted from September to December 2002 in the Health Center Kragujevac, in the working environment of the participating doctors.
To show the characteristics of decision-making process during prescribing, five written patient cases were used, and the doctors had to make their treatment decisions.The written cases were based on actual patient cases, analogous to those from the similar study 9 .The cases described the patients complaints, circumstances and diagnostic information.Two of them were dealing with uncomplicated urinary tract infections: two 36-year-old non-pregnant women -one with epilepsy treated with carbamazepine, and other one already treated for the similar urinary infection two months before.The other three cases were patients with stomach complaints: a 24-year-old female who asked for specific tablets for her non-specific stomach complaints, a 55-year-old accountant with minor gastritis, and a 43-year-old woman with the relapse episode of reflux oesophagitis, previously treated with H 2 antagonists.These written cases were considered valid for measuring decision processes, for which an actual interaction between a doctor and a patient was not strictly needed 10,11 .For the diagnoses described in these cases, there were no significant changes in the recommended first choice treatment over the recent years 9 .
To reveal the decision-making process, the "think aloud" method was used as the most valid and adequate one, because it gave the more accurate information about the decision processes than when doctors were asked to describe or explain their decisions 12−14 .The doctors were presented with the written patient cases and asked to make prescribing decisions for them.They should verbalize aloud all considerations that came into their minds while reading the case and while deciding upon the treatment.The verbalizations were tape recorded, wrote down in details and analyzed.
The aspects relevant to the choice of treatment were divided into core (effects, side effects, co-medication, comorbidity, other characteristics of the treatment itself and contraindications), contextual (previous experiences of the patient and patient's demands) and habitual (habits and standard treatments).By the "treatment mentioned" we implied every separate drug treatment option, which included one or more different medications.The number of treatments mentioned was classified in three categories: none, one or two, and more.The doctors' decision processes were categorized, depending on the number of treatments and aspects considered, as one of seven types, which were described by Simon in 1995 15 .When only one treatment is mentioned, it indicates a decision process in which no comparison is made between different alternatives: this could be a type A, in which no aspects are mentioned either, or a type B, or a C process, in which one, or two or more aspects are mentioned, respectively.If one tries to evaluate and compare several options in the search for the optimal choice, it could be the type D, E or F, depending on the number of aspects considered (none, one, or two, or more aspects, respectively).The transcripts that mention no treatment and no aspects at all are classified as type N. Type F processes are considered as closest to the optimizing strategy.
The mentioned aspects and treatments in the transcripts and types of the established decision process were compared according to the years of experience of the participating doctors, the level of their previous education, and the written cases used in this study.
The investigation involved the comparison of treatments prescribed by the participating doctors and the first choice or second choice treatments according to the Therapeutic Guidelines, the Royal Australian College of General Practition, State of Victoria, Australia, considering that there was no adequate local therapeutic guidelines at the time of investigation.There were also implied some additional local aspects and the influence of the cost of medications on decision process during prescribing, according to the local Essential Drugs List ("Official Gazette RS", number 60/2001), which comprised the names of medications that were covered by the basic health insurance.
For the description of general characteristics of the participating doctors and the results of measuring, the method of descriptive statistics was used: absolute numbers and frequencies.To determine whether the differences in the decision-making process were related to prescribing different treatments, the χ 2 test was used.The relation between the doctors' years of practical experience and the number of treatments considered was tested for each case separately with non-parametric correlations (Spearman's rho).The influence of the patient case on the number of treatments and the type of aspects considered during the decision process was tested with non-parametric tests for more than two related samples (Friedman's test for ordinal data and Cochran's Q test for dichotomous data).

Results
The analysis of the total number of transcripts showed that the nature of cases had a significant influence on the treatments mentioned during decision-making (Friedman's test; p = 0.0001).The effect of practical experience on the mentioned treatments was demonstrated only during the consideration of the fifth written case by the doctors with specialization in general practice (χ 2 test; p = 0.01).The statistical relevance between the practical experience and the number of mentioned treatments was shown (Spearman's ρ = -0.195;p = 0.001), so that the doctors with more years of practice considered less number of treatments more often.Between the doctors specialized in general practice, and those without specialization, no significant difference (χ 2 test; p = 0.362) considering the number of the mentioned treatments, was demonstrated.In 265 transcripts, considerations of two or more medical treatments were made only in 24.2%, no treatment was mentioned in 37.4%, and only one treatment in 38.7% of all the transcripts.
The influence of practice experience on the mentioned aspects was shown only during the consideration of the fourth written case by the doctors with specialization in general practice (Spearman's p = 0.302; p = 0.024).The coefficient of correlation showed that the doctors with less experience more often considered core aspects, while the doctors with more practice more often considered contextual and habitual aspects.The considerations of aspects were shown only in 205 of the transcripts.The significant difference among the frequencies of the appearance of core, contextual and habitual aspects during decision-making about the treatments for each case separately was also shown (Cochran's Q test; p = 0.0001; Table 2), as the total frequency of all the aspects in 265 transcripts (χ 2 test; p = 0.0001).The frequencies of appearance of different kinds of core aspect, as their total Cohran's Q test; p = 0.0001 frequency, during decision-making process for each case separately, also demonstrated a statistical relevance (χ 2 test; p = 0.0001; Table 3).
The significant difference in appearances of different types of decision-making process (Table 4) was shown for each case separately (Friedman's test; p = 0.0001), as in total number of transcripts (χ 2 test; p = 0.0001).The practical experience showed no significant influence on the type of decision-making process of the doctors without specialization in general practice (χ 2 test; p = 0.589), but of the specialists it did (χ 2 test; p = 0.0001).Between these two groups of the doctors no statistically significant difference was shown in the appearance of various types of decision-making processes (χ 2 test; p = 0.475).
The frequencies of appearance of the first and second choice of treatments or inadequate therapy as final decision among cases, as the total number of transcripts, were different (χ 2 test; p = 0.0001; Table 5).Between the final decisionmaking and the type of decision process no such difference was demonstrated, neither in the doctors without specialization (χ 2 test; p = 0.504), nor in specialists (χ 2 test; p = 0.453).
Among the different kinds of the mentioned additional aspect in different written cases during decision-making by the doctors with specialization in general practice, a statis-tically significant difference (χ 2 test; p = 0.001; Table 6) was noticed.In the same group of the participants, a certain influence of practical experience on the additional aspects was shown (Spearman's p = -0.142;p = 0.014): the coefficient of correlation showed that the doctors with more practice often considered less the additional aspects.The frequencies of appearance of different additional aspects showed the difference in the whole specimen too (χ 2 test; p = 0.0002).
The relation between prescribing medications out of the essential drugs list and the other medications showed no significant difference among the cases (χ 2 test; p = 0.004).The frequencies of the mentioned drugs, including their combinations, showed statistical relevance according to the final decision (χ 2 test; p = 0.0001), also in the total number of transcripts (χ 2 test; p = 0.0001).

Discussion
This study showed how the general practitioners practice experience and the level of their education, as well as the nature of cases they considered, influenced their decisionmaking during prescribing.The practical experience had an influence both on the number of medical treatments and the different considered aspects.The most optimal way of decision-making, named as the type F 15 , was the most frequent in the subgroup of the doctors with the least practical experience, while the doctors with the most experience most frequently decided in the ways considered inappropriate.The study also revealed that the medical doctors with less experience more often considered the core aspects, those with more practice more often considered the contextual and habitual aspects; an additional aspect was less frequently considered if a medical doctor had more practice.
Therefore, practical experience showed to have a significant influence on the medical doctors prescribing behaviour, confirming the fact already verified by a great number of investigations as, for example, Ely et al. 26 who established that the doctors with more years of practice more often considered less number of aspect during prescribing, Denig et al. 9 showed that there was an effect of practice experience on the considering aspects, Veninga et al. 16 explained 11% of all variations in prescribing by the years of practice.
The level of physicians' previous education had some influence on the decision-making process during prescribing, since it was shown that some additional aspects had a different significance in decision-making by the doctors with different levels of previous education.On the other hand, the results did not verify an effect of doctor's previous education on the aspects or treatment considered.These results complied with Denig's study 9 , but the fact that the level of previous education had no influence on the number of treatments considered dissagreed with the conclusion made by Robinson 17 , who believed that it was one of the most significant factors that could explain differences in prescribing behaviour.The importance of a doctor's previous education for prescribing, particularly for prescribing antibiotics, was also shown by Steinke et al. 18 .The disagreement of this study with the others could be explained by the characteristics of an education system and working enviroment in Serbia, that do not provide enough motivation to the more educated doctors.
The nature of the considered cases had a significant influence on decision-making process during prescribing, affecting both the number of treatments and the considered as-pects.Our data showed that the general practitioners often choose the aspects that would consider both the case itself and the number of information they received from the patients, although the optimal decision-making demanded consideration of all the core aspects mentioned.Moreover, this study showed that most of the general practitioners for the urinary tract infection cases and for the reflux esophagitis case prescribed drugs recommended as the first choice; patients with non-specific stomach complaints were most often treated with the second choice medications.
The influence of a case itself on the type of decisionmaking during prescribing had been previously recognized in a similar study, published by Denig et al. 9 , which showed that the urinary tract infection cases predominantly triggered more simplier decision processes (types A and B), while the stomach complaint cases were followed by the more complex way of decision making.This study also confirmed that the nature of cases did effect the treatments and aspects considered and a doctor's choice of drugs, which was in compliance with our results.
This investigation also showed that in more than 2/3 of all 265 transcripts the aspects were considered: core aspects were mentioned in about 1/3 of the cases, contextual a bit more than that, and habitual were considered in 1/10 of all the analyzed transcripts.Investigating only the transcripts with the considered aspects, we noticed that, although the core aspects were the most significant for optimal prescribing behaviour 27,28 , the contextual aspect were the most frequent.This confirmed the fact that the patients demands and other circumstances linked to them significantly affected the doctors prescribing behaviour.
Other authors also examined the infuence of the factors on decision-making, and many of them showed that a patient's preferences, which belonged to the contextual aspects, were among the most powerful factors during prescribing 19−24 , even when they were incorrectly interpreted 25 .The results of Denig's study 9 showed the similar appearances in the domain of contextual and habitual aspects, but the core aspects were noticed in more than 1/2 of all the transcripts.Also, their study showed that no aspects at all were considered in 1/4 of the cases (our investigation showed 1/3).
In the total number of transcripts, it was shown that 1/3 of the prescribed medications belonged to the essential drugs list alone, which indicated that the doctors did not always consider the financial aspect of a medical therapy.Surprisingly, the frequency of prescribing the essential drugs observed in our study was much lower than in other countries 26 , even lower than results shown by other studies in our country 27 .Within the whole specimen, over a half of all the doctors recommended the first choice drugs.Altogether with the second choice treatments prescribed, it made 70% of the decisions which resulted in the treatments recommended by the guidelines.Other 30% were inadequate medical treatments.Many investigations, that analyzed inappropriate prescribing, were reviewed by Buetow et al. 28 .The authors concluded that, although inappropriate prescribing did exist, the proposition that it was broadly disseminated was groundless.However, 1/3 of inadequate medical treatments prescribed in our study could be qualified as a deviation from the optimal prescribing.Yet, these deviations did not always result in suboptimal prescribing behaviour, and the optimal way of decision-making did not always provide the optimal therapy.Although other authors came to a similar conclusion 9 , it should not be an excuse for an inappropriate prescribing.
The most important result of this study was a fact that, in more than 60% of the transcripts analyzed, the most common way of decision-making during prescribing was detected, while the optimal prescribing behaviour accomplished only 1/10 of the general practitioners.Moreover, 1/3 of all the patients left surgery without any therapy and consideration of any aspect at all.
On the basis of the results obtained by this study, we can conclude that in the whole specimen of the participating physicians, the observed behavior mostly did not comply with the theoretical decision-standards.Therefore, our study suggests the need for the obligatory continuation of medical education in general medical practice, as well as in the usage of clinical guidelineses.The main idea was given in the WHO Guide to Good Prescribing 29 : "First you need to define carefully the patient's problem (the diagnosis).After that, you have to specify the therapeutic objective, and to choose a treatment of proven efficacy and safety, from different alternatives.You then start the treatment, for example by writing an accurate prescription and providing the patient with clear information and instructions.After some time you monitor the results of the treatment; only then will you know if it has been successful.If the problem has been solved, the treatment can be stopped.If not, you will need to re-examine all the steps".The optimalization of decision making process during prescribing will undoubtedly improve the quality of medical care.

Conclusion
The observed prescribing decisions were mostly in disagreement with the Good Clinical Practice.Our study pointed out the need for the obligatory continuation of medical education of general practitioners in decisionmaking process during prescribing

Aknowledgment
The authors are grateful to the general practitioners who participated in this study.

Table 4 Types of decision processes observed for each case
All the groups were tested, Friedman's test; p = 0.0001