The four principles: Can they be measured and do they predict ethical decision making?

Medical ethical principles

Research on the importance of medical ethical principles is generally one of three types. The first type, where most of the research has been focused, concerns the moral development and orientation of medical students throughout their medical training[6–9]. However, this body of literature is mainly focused on the developmental nature of morality rather than the specific principles valued, or their role in personal ethical decision making.

The second type of research has examined differences between groups of students with respect to the ethical principles they value[10, 11] and compares whether the same principles are important for different professions. These findings have shown that females value autonomy more than males and that medical students are generally more beneficent than lawyers.

The third kind of research investigates medical students’ evaluations of certain ethical principles[8, 12]. The study by Herbert[12] had students read several ethical scenarios and identify as many ethical issues as they could in each scenario. The responses were evaluated post hoc against a “gold standard” marking scheme, where issues were classified as reflecting one of the three principles of autonomy, justice and beneficence. Results indicate that students were able to identify ethical issues in the cases consistent with the principles, however, the responses were often limited in range and tended to be of the same type. Price, Price, Williams and Hoffenberg[8] examined changes in medical students’ attitudes as they progressed through their medical course, where attitudes were defined by the assessment of the ethical principles.

Westin and Nilstun[13] discuss the principles and whether a certain type of ethical conflict (patient with ischemic heart disease) could be resolved by using “normative” reasoning with the ethical principles. This study was useful in identifying the conflicting nature of the principles for different stakeholders and the difficulties encountered with applying them in practice to a specific case, but it did not take an empirical approach to identifying the importance of the principles and their application. Moreover, it has been criticised for the manner in which the principles were applied or interpreted (principles examined in isolation) in the given case[14].

Landau and Osmo[15] investigated the hierarchies of ethical principles among Israeli social workers. The ethical principle of “Protection of life” was the most important principle guiding the social workers’ decision making with 45% of respondents rating this principle as the most important. However, there was little consensus among the rankings of the ethical principles by the social workers and there was no consistent pattern of application across cases. Professional and personal hierarchies were not significantly different from each other. Social workers responded mainly to the case information and changed the importance of their ethical principles based on the situational information available in each case.

Twelve ethical principles were used in the Landau and Osmo[15] study including “Autonomy & freedom”, “Truthfulness & full disclosure”, “Least harm” and “Public good”. The principles overlap considerably with the four principles but had a more specific social work focus. The 12 principles are also not conceptually distinct. The present study focuses only on six principles with an effort to prevent any conceptual overlap between principles. From a methodological point of view, the Landau and Osmo[15] study was based only on the ranking of principles and simple Likert scales. The present study proposes an additional way to assess the importance of principles by computing the relative weight of principles for each individual. This has the advantage of being able to assess the importance of the principles when two very important principles conflict, which was not the case in the Landau and Osmo[15] study.

Because of the novelty of this approach and the introduction of a new measure, no specific predictions can very clearly be extrapolated from past findings or empirical research. However, there is considerable theoretical discourse concerning the importance of the principles in medical ethics. Several arguments and pieces of evidence exist that may point to the possible empirical results.

First, the debate on medical principles can indicate which medical principles are the most important. In the Hippocratic Oath the principle of primary importance is Primum non nocere, above all do no harm. This principle has a long history of being very important in health care contexts. The Hippocratic injunction to do no harm has been an axiom central to the education of medical and graduate students[16].

In more recent times, the primary importance of this principle has been contested. In a reflection and discussion of the principles and their merits, Gillon[17] argues for the principle of autonomy to ‘trump the rest’. He provides compelling reasons why respect for autonomy should be of primary importance in medical ethics and applied ethics in general. Different people argue for different hierarchies and yet others argue for no hierarchy at all[18], just the application of all the relevant principles to a case on a more relativistic basis. Prediction from the theoretical literature is therefore not straightforward.

In summary, this study examines the medical ethical principles and uses the AHP as a methodological tool to derive individual weightings for ethical principles. There are two aims of this study. First, to develop and evaluate a measure of the four principles from medical ethics, and second, to determine whether individuals’ rankings of these principles are used in decision-making in ethical scenarios where these principles conflict.

Analytic hierarchy process: an overview

The measure of the medical ethical principles developed here uses pairwise comparisons to elicit the weightings for the principles. This methodology is part of the AHP. The AHP is a multi-criteria decision making tool originally developed by Saaty[5] that has been widely applied to many areas in the field of decision making[19] including resource allocation[20], business performance evaluation[21], project selection[22], and auditing[23].

In the AHP, a judgement or a comparison is the numerical representation of a relationship between two elements that share a common parent. In this study there is only one parent (ethical principles) and a judgement consists of a rating of the relative importance of one principle over another. Through trade-offs the technique enables the explication of the advantages and disadvantages of options under circumstances of risk and uncertainty.

The AHP is used in this study as a pragmatic tool to assess the relative preferences that individuals have for the principles. The technique of weight computation for the principles can be considered an alternative way to assess the importance of the principles in the individual decision making process. Prior research has tended to only measure the importance of principles either in scenarios, in isolation (one principle at a time), or with post-hoc matching of responses to set criteria. The AHP methodology is a novel approach in this area.

It should be noted that no behavioural hypothesis about the way people cognitively use the principles is made in order to use the AHP. The numerical results must therefore be seen as an approximation and, to some extent, still qualitative, in spite of their quantitative nature.

Participants

Participants were 94 first year psychology students from the University of Queensland, Australia. Participation in this study formed part of their course requirements. There were 65 females (69%) and 29 males. The age of the participants ranged from 17 to 58 years with an average age of 21.5 (SD = 7.01) years. Eighty-two percent of participants were Australian and the other 18% were from predominantly Asian backgrounds. This study was cleared in accordance with the ethical review processes of the University of Queensland and within the guidelines of the National Statement on Ethical Conduct in Human Research, clearance number 03-PSYCH-P-02.

Measure of the medical ethical principles

The measure of medical ethical principles that uses the AHP methodology was designed to measure the importance of the medical ethical principles in a general and global sense, that is, in a context without specific situational information or cues. In the past, given the complicated nature of the principles and the importance of all of the principles, it has generally been the case that they have been discussed and debated in situations where the principles come into conflict within a specific case scenario. The scenario is then assessed from a principles perspective as the extent to which the principles bear on the case. This approach is understandable and often informative but in this study the interest is in knowing whether people hold more general preferences for the principles that supersede specific case information. That is, do people weight some of the principles as more important than others irrespective of the situation at hand?

The four standard principles proposed by Beauchamp and Childress[2] were used in the new measure, as well as two other principles; confidentiality and truth-telling, which are within the Beauchamp and Childress[2] framework embedded within the principle of autonomy. The concepts are often discussed separately in medical ethical discourse, and they exist as separate entities in the medical code of ethics from the American Medical Association. Moreover, previous empirical research has also separated these principles[10, 15]. Therefore, the new measure was developed to assess the importance of six medical ethical principles; non-maleficence, beneficence, autonomy, justice, confidentiality and truth-telling.

For each principle statements were developed that defined the principle in a general sense without reference to characteristics of a specific situation. This provided participants with precise knowledge of the principles necessary for making an informed and accurate judgement. There were two statements for each principle, each expressing the concept in a slightly different way with essentially the same meaning - for example: Beneficence - An obligation to convey benefits and to help others to further their legitimate interests (i.e. “One has a moral obligation to help other people”). For each principle there was also two examples of the principle to demonstrate the meaning in a broad context, one for medicine and the other for business. The measure is a pair-wise ranking task that forces participants to choose between the ethical principles (in the form of generalised statements) when they conflict. All possible pairs are formed from the six principles, making a total of 15 pairs of statements. In this way, the measure allows the assessment of the relative importance of the principles. The measure is analysed using the AHP[5] which yields relative weights for the six principles. In addition to this new measure, participants were also asked to rank order the principles from most to least important.

The scenarios

There were four scenarios used in this study all containing ethical issues framed in a medical context and involving medical ethical principles. The first was an IVF scenario dealing with issues of ownership, autonomy, and privacy. Two scenarios (hereafter referred to as Confidentiality and End of Life) were from a questionnaire on medical ethics[8]. The Confidentiality scenario primarily concerns issues of privacy and trust (Dr Heron has a right to his confidentiality) weighed against the principle of non-maleficence (the possibility of future harm to potential patients). In contrast, the End of Life scenario concerns patient autonomy and the right of a patient to choose to end their own life. The ethical conflict in this case arises because of the conflict between autonomy and professional duty and non-maleficence. The fourth scenario is a commonly cited and discussed case in the field of medical ethics and involves the process of a blood transfusion for a child of Jehovah’s Witnesses[24]. This case involves the principles of beneficence (helping the child’s interests) versus patient autonomy or the parents’ right to decide for their child. Together these four scenarios were thought to provide a good basis for, and be representative of, the salient issues in medical ethics. All four scenarios can be seen in Additional file1. At the end of each scenario participants were asked two questions, the first about the ethicality of the action (1) How ethical is this action? (rated on a seven point Likert scale from very unethical to very ethical), and the second concerning their intentions to act in that way if they were in the same situation, (2) I would act in the same way (rated on a seven point Likert scale from strongly disagree to strongly agree).

Demographic questionnaire

The demographic questionnaire consisted of several items which asked questions about age, gender, religious commitment, and ethical training. Responses were given in either a generated written format or by circling or ticking an appropriate scale response option. These items were included in order to control for the effects of some possible explanatory variables (age and gender) and/or to investigate gender differences where appropriate.

Procedure

Questionnaire booklets were compiled containing the measures outlined above and these were administered to participants in a classroom environment by an experimenter. The ordering of the materials remained consistent across participants with the demographic measure first, followed by the scenarios, and finally a measure of medical ethical principles was presented last. Participants were given a maximum of one hour for the completion of the task.

The pairwise matrices

From the list of 15 pairs of statements, a reciprocal matrix was constructed for each person from the preferences that participants indicated (see Table1). For example, if the participant rated principle A (“Autonomy”) as preferable to principle B (“Beneficence”) by a strength of “5” then the upper section of the matrix would be a 5 and the corresponding lower cell would be 1/5. This would be repeated for all cells until a full reciprocal matrix was obtained. Each column and each row represent one of the six medical principles so the resulting matrix is a 6×6 reciprocal asymmetric matrix where each number represents the participants preference for one principle over another and by a given magnitude on a scale from 1-9.

An SPSS syntax file was generated for the purpose of calculating the matrices, the corresponding preference weights for each person and their consistency index. These calculations were derived from the formulas presented by Saaty[5]. For example, take a person A, whose answers correspond to the matrix presented in Table1. The weights shown in the last column are computed from the 6 × 6 matrix using the AHP. In this example, A prefers autonomy most, making it his or her strongest principle, with truth telling their least important principle.

Overall weightings of the ethical principles

Table2 shows the weights for the six ethical principles for the entire sample. The weights seem to indicate that five of the principles are equally important because of the roughly equal weightings. Moreover, there is clearly a preference for non-maleficence over the other five principles. This implies that when in conflict there is a strong preference for causing no harm as an a priori principle.

Participants were also asked to provide their own rankings of the six principles on a scale from least important to most important. The percentages of people who ranked the principles in one of the first two positions of importance and in one of the last two positions are also shown in Table2.

These self report rankings are relatively consistent with the rankings from the AHP. Approximately 57% of participants rank the principle of non-maleficence as one of the two most important principles and this dominance is reflected by the high weighting in the AHP task. The principle of justice is also rated as very important by 50% of the sample however, this principle, as measured by the AHP method, is less important than the self report measure would indicate. Both truth-telling and confidentiality are the least important principles using the self rankings. Again, this is partially consistent with the AHP rankings.

A series of paired sample t-tests with bonferroni corrections were conducted to test the differences between the weightings. The results of this analysis can be seen in Table3. There are significant differences between non-maleficence and all of the other principles. There are also significant differences between autonomy and truth-telling, justice and truth-telling and confidentiality and truth-telling. Therefore, non-maleficence is the most important principle and truth-telling the least important principle.

Gender differences and individual heterogeneity

A one-way ANOVA was conducted to test if males and females differed in their AHP weightings of the ethical principles. Results revealed that the only difference between the sexes was for the truth-telling principle such that females ($\bar{X}=.14$) weighted the principle as significantly more important than males ($\bar{X}=.09$), t(90) = −1.93,p < .05. There was also a significant correlation between years of university education and the ranking of the autonomy principle. People who have spent more years at university rate the principles of autonomy as more important, r = .21,p < .05.

While overall the most important principle is non-maleficence, there is considerable heterogeneity between people. The majority of the sample has high weightings for non-maleficence over the other principles but there are sub-groups of people whose pattern of weightings are quite different. Table4 shows the weightings for three people. Their profiles are quite different. Person B represents the more typical member of the sample with a strong preference for non-maleficence over the other principles. Person A, by contrast, has a more balanced profile with the strongest preference for autonomy but also a strong preference for justice. Person C holds an altogether different uncommon profile with a very strong preference for confidentiality and a second less strong preference for truth-telling.

The AHP methodology also yields a consistency ratio (CR), which is an indication of how consistent people are in their judgements (0 being perfectly consistent and 1 being totally random). Saaty[5] recommends a threshold of .10 for this index. In this task people were much more inconsistent with an average CR of around .30 ($\bar{X}=.32,\mathrm{SD}=.26$). However, when the most inconsistent people were removed from the sample and the average weights recalculated, the order and magnitude of the weights remained unchanged.

Moreover, because people are inconsistent does not imply that they are not doing the task properly. It may be the case that there is more than one dimension on which they are basing their judgements of the principles. The AHP methodology is not sensitive to this distinction because it uses only the first eigenvector and eignevalue. In general, these results show that considerable heterogeneity can exist between people in their preferences for the principles.

Predicting ethical judgments and intentions

The correlations between judgements for the four medical ethical scenarios, and the medical ethical principles are shown in Table5. There are no significant correlations between the weightings of the principles and the participants’ ethical judgements. These relationships were also explored for participants’ ethical intentions and the same results were obtained.

The absence of any relationship between preferences for medical ethical principles and scenario judgements precludes the possibility of the principles predicting judgements and intentions. This lack of predictiveness for the medical ethical principles may be due in part to the limited variance in the weightings. Also, the calculation of the weights using the method described by Saaty[5] results in small dependent weights with limited variance. This can be problematic in regression equations.