Article Text

Download PDFPDF

Discourse ethics in practical medicine
  1. F Keller,
  2. G Allert,
  3. H Baitsch,
  4. G Sponholz,
  5. Ethics in Medicine Working Group at the University of Ulm
  1. Medical Faculty, University of Ulm, Germany
  1. Correspondence to:
 Dr F Keller
 Dvision of Nephrology, Medical Faculty, University Hospital, Robert Koch Strasse 8, D-89070 Ulm, Germany; frieder.keller{at}


Problems emerge in practical medicine because the binary ethics of the classic patient/doctor relationship has been replaced by multiagent interaction between those engaged in the process of diagnosis and treatment. New methods are required to deal with complex problems in every patient. Where and why the current practice can fail is illustrated with an example of an unspectacular routine case of cancer. The failure may result from basing the procedure on mechanistic methods or from the deficit and difficulty in communication. Whether rule based algorithms could have improved the treatment in the patient with cancer is discussed. How discourse ethics may fit better with the course of the case is described. Clinical Medicine follows a similar logic to that modelled by discursive ethics, ethics thinking should essentially contribute to the procedural logic of medical practice. Discourse ethics can be used as a procedural model that copes with the complexity and temporality of practical medicine. Applied discourse ethics can turn out to be both instrumental in mediating inherent conflicts and constitutive for value based problem solving in modern medical practice.

  • clinical medicine
  • rule based inference
  • discourse ethics
  • Juergen Habermas

Statistics from

Clinical medicine differs from medical science in so far as it is action principally regarding the patient. Today, the patient/doctor relationship ought to be a participatory and personal relationship, forming the basis of a “therapeutic alliance”.1 An egalitarian interaction, not primarily based on hierarchy, can lay the groundwork for mutual respect and autonomy. It is this interindividual relationship that has made clinical medicine an ethical discipline ever since Hippocrates and Aristotle.2

According to Donald A Schon (1930–97), the central problem in biomedicine is the failure of the technical and purely rational approach. He calls for the “reflection-in-action” approach, especially for medical professionals.3 Applying medicine to a patient is an act to which both knowledge and values contribute. Scientific knowledge must be harmonised with individual preferences. Ethics have some role in nearly every medical action. Practical medicine needs a new method to solve the inherent conflicts between fast growing medical knowledge and technologies on the one hand and diversifying individual and group values on the other. We would like to show how growing problems in practical medicine may result from both the inherent complexity of medical science and the simple fact that treating a patient is a relational encounter with the inscrutable “Other” (Emmanuel Levinas, 1906–95).4

We suggest that discourse ethics may have an essential role as a mediator and an integral component in practical medicine. For this purpose, we define our understanding of discourse ethics as a procedural model for communicatory action, where conflicting ethical principles apply simultaneously. Discourse ethics has been proposed by Juergen Habermas as a means of giving autonomy, initiative, and responsibility back to the patients, not to the institutions or to the systems represented by anonymous elites. As the discourse is based on equality and reciprocity, it facilitates both autonomy and responsibility. Shared responsibility and reciprocal autonomy are essential to the modern patient/doctor relationship.


Presenting a routine clinical case to exemplify the applicability of discourse ethics as a procedural method for case based diagnosis and treatment may be useful. The real-life case presented here shows how the clinical treatment failed. Technical rational approaches would not help because the failures in the process were mainly due to failures in communicative action.

First, the history ...

A 63 year old woman was admitted to the gynaecology department with impaired renal function. She had an inoperable carcinoma of her uterus three years previously, which was treated by palliative radiotherapy. Instead of consulting a nephrologist, an intravenous urogram was ordered. This caused a further increase in serum creatinine level and complete anuria. After this iatrogenic complication, the nurses refused to place a urinary catheter, as the patient had already experienced a misplacement during radiotherapy. The conflict escalated and the urologist was phoned, but he judged that this was the terminal stage of the patient’s disease and that the best solution for her would be to die of uraemia. The gynaecologist who had ordered the urographic investigation with contrast media, however, consulted and argued with the nephrologist. In view of the diagnosis, the nephrologist thought that it would be pointless to start renal replacement therapy. The gynaecologist and the nephrologist discussed whether they should talk again with the nurses or with the urologist. However, they finally agreed to talk with the patient, intuitively following a discursive approach.

... then the patient

When the nephrologist saw the patient, she was in unexpectedly good physical and mental condition. At the bedside, the gynaecologist (G) and the nephrologist (N) discussed the options with the patient (P):

 G: Because of your kidney problems we called the nephrologist.
 N: We have the option to dialyse, or to wait.
 P: Does waiting mean that I could die?
 N: To be honest, this could happen.
 P: Would dialysis be needed for the rest of my life?
 N: Probably not.
 G: But dialysis would not solve your tumour problems.
 P: I am inclined to undergo dialysis.

All the three consented to start haemodialysis immediately. The next day, the urologist was willing to place a urinary catheter and, as often happens, the patient’s urine output resumed. One week later, however, a diagnosis of peritoneal carcinoma was confirmed. The standard course of chemotherapy was ordered, with high dose cyclophosphamide and epirubicin. The patient was discharged after two weeks. At home, her condition deteriorated rapidly and she died one month later.

Relativity and temporality of medical values

Looking at the results objectively, this patient’s course can be characterised by terms such as incompetence and malpractice (intravenous urography), negligence and mismanagement (no catheter placement), overtreatment (haemodialysis), and futility (chemotherapy). Shortcomings emerge from the urologist’s rule based view of the disease in general terms instead of visiting the patient, talking to her, and taking a look at her as an individual.5 Haemodialysis prepared the way for ordering chemotherapy without discussing other options. The point at which it would have been best to reduce treatment and start palliative care was missed. One event—renal impairment—led to a cascade of subsequent procedures: urography, haemodialysis, and chemotherapy. Such a dynamic emerged from the logic of “If you say <A>, then you must also say <B>”, or from the law of “all-or-none”.

On the other hand, when looking at each discrete action, the patient’s welfare was primarily in mind. Ordering urography, for example, was equivalent to taking over responsibility. The decision to carry out dialysis was made easier as contrast nephrotoxicity, presumably, would resolve. After high tech dialysis, the urologist agreed to place a simple catheter. Chemotherapy was ordered to give her the chance of any benefit. What was done, however, should have been done in a more communicative way; but would this have amounted to a more scientific approach—to more than common sense?


In recent years, various scientific models have been proposed and different rule based algorithms have been developed to deal with complex clinical situations such as the case presented here. We have considered whether and how any of the algorithm-like methods may have improved the treatment of this patient with cancer.

1. In the patient’s treatment course, a simple intervention (urography) led to an unpredictable cascade of actions. We may be inclined to model the unpredictable decision making process by using chaos theory. New approaches to chaos theory, however, are deterministic in nature—a simple rule governing only the apparently chaotic processes.6 Chaos theory does not describe unpredictability.

2. A tree for decision analysis can be constructed to find the presumably most beneficial solution by using mathematical algorithms.7 This attributes a probability value to the benefit and complication risk (for example, from dialysis). Such probabilistic data to quantify the different options, however, were lacking from this case. The prospective likelihood values would also depend very much on the subspecialties, such as gynaecology, urology, nephrology, or oncology. The nearer the subspecialty is to the patient’s problem and the closer the examination, the wider and better the prospect appears in medicine.

3. Regarding quality-of-life assessment, the mortality from uraemia contrasts with the discomfort caused by haemodialysis. Different qualities such as mortality or discomfort are subjective judgements, depending on non-uniform cultural mores, traditions, and standards that are difficult to harmonise and cannot be evaluated on the same scale.8 Quality adjusted life years can be used to estimate and compare cost effectiveness, but not to assess the preferences of patients.9 This is a matter of asking, not of measuring.

4. By using high capacity computers and multivariate statistics, predictive score models, such as the Acute Physiology And Chronic Health Evaluation (APACHE) Score, can be derived from accumulated clinical observations for making predictions of patient survival in the intensive care unit.10 We could not make use of this type of model as this, or a comparable system (for example, acute tubular necrosis ATN-ISS Score), is not calibrated for our hospital, as would be needed for a decision to withhold dialysis.11 The problem remains: discontinuing treatment in 100 patients with a 5% survival probability would mean that all 100 patients would die, but if 20 cases were treated, one patient could survive.12 Doctors, however, tend to see the actual case as the one that survived but not the other 19.

5. Priorities can be chosen by fuzzy-set analysis based on cost effectiveness and cost utility in healthcare.13 Counting values on an ordinal scale by modern computer methods may be valuable in deciding whether a hospital should, in the long run, be equipped with dialysis facilities, but such an institutional solution does not solve the actual conflict in the patient with cancer.

6. One stage further from generalisation to individualisation may be achieved by applying Bayes’s theorem.14 Patients with advanced cancer do not usually benefit from dialysis. But in the case described earlier, the conditional probability that renal function would recover was higher than the primary probability without the history of radiocontrast investigation. The bayesian approach, however, requires an assumption of conditional independence and a priori knowledge for advanced statistical calculations. In fact, we could not be sure about such conditions and data in the patient with cancer.

7. Completely assumption-free artificial neural networks predict from different previous cases what will happen to a prospective patient.15 Artificial neural networks are a kind of case based “learning by doing” by using multivariate statistics. Even the patient’s individual preferences can be taken as an input parameter to calculate the dialysis proposal as an output. Neural networks, however, are restricted to making proposals, whereas all people participating must take responsibility for the decision and the patient must consent to the proposal.

8. All the models mentioned above cannot take into account the time dependent sequence of momentary actions. For the patient, the open course of discrete moments was different from the fatal final event, and “final event” means death, in this and many other cases. What momentarily seemed to be right to all those concerned—such as haemodialysis—was debatable after chemotherapy started. The temporality of clinical processes needs a special kind of reasoning.16 Such temporal logic would require, for example, the application of “Petri nets” instead of linear first order logic.17 We had no such grids at hand.

9. In contrast with the above computer assisted methods, the theory of the four bioethical principles does not depend on numerical or mathematical models,18 but the clinical application of the four basic moral principles must also be considered a rule based method. Although these principles do not try to give ultimate reasons but are rather middle-level principles, they are not yet applicable by themselves. Autonomy is in conflict with justice (if you treat one patient with cancer, why not all?), beneficence is in conflict with maleficence (haemodialysis justifies chemotherapy).

Medical practice is influenced by the four bioethical principles as well as by several medical virtues, such as prudence, courage, responsibility, self-control, discretion, and veracity.19 All the knowledge and preferences, principles, and virtues must be contributed by more people than one doctor to the patient. This is no longer a binary but a multiactor scenario that needs to communicate on how to act.


The obvious misfits between the nine algorithms and the case of the patient with cancer may allow some general conclusions to be drawn. In scientific medicine, knowledge is acquired by statistically synthesising the observations made in a group of patients. In practical medicine, a knowledge based answer must be given on pending options, which is influenced simultaneously by value based judgements (fig 1). Treating a patient is action, and this action must not be objectified, but must be personalised like good journalism or narrative story writing.

Figure 1

 Statistics and ethics. Statistical inference from a group of patients A, B, C, ... can be used to generate rule based knowledge and to validate hypothesis driven medical science (X). All available knowledge G, H, I, ... must be synthesised and individualised to a medical treatment for a single patient (Y) by the knowledge and value based discourse.

In the case of the patient with cancer, knowledge and principles must not be created but applied. The problem with modern clinical medicine may result from the need to individualise the available knowledge to make it applicable to the individual patient. The patient as an autonomous subject must participate in this process.20 The autonomy of the patient is based on informed consent, and thus the patient participates in a communicatory act with the gynaecologist, nephrologist, and oncologist.

In contrast with classical physics (“where one body is, no other can be”), and in line with modern relativistic thinking, diverging possibilities such as death and dialysis apply simultaneously in clinical medicine. The etymology of “medi-cina” indicates that the golden mean holds true more than the principle of “either right or wrong, there is no third way”. The chronology of medical decision making requires a chronological or temporal logic for the case based process.21 Temporal logic seems to be constitutive to the theory of action.22 Every following decision is dependent on what has been decided before. Premises and decisions are referential, as are prognosis and treatment: treatment can change the prognosis, but prognosis also influences the treatment. If a decision is made (urography), the assumptions—true or false—result in actions (haemodialysis) and create new conditions (pushing chemotherapy). Every decision selected out of several possibilities is usually irreversible. This chronology of events brings the temporality of the logic to factual perception.

With the patient, a scientific “true or false” testing could not be applied to the haemodialysis treatment.* It was rather a “more or less right” answer and a decision of “now, later, or never”. Temporal logic obviously applies to clinical medicine as one main reason for starting dialysis was to gain time. Clinical problem solving is a process with multivalent logic as time and values play a part. The logical structure of medical practice is analogous to the logic of ethical reasoning, and ethical deliberations are constitutive to the logic of practical medicine. Clinical medicine and applied ethics are both analogous and complementary; both are looking for a solution that is individualised from general knowledge and normative presumptions. The application of medical science must be based on ethical reasoning to fit the temporality and individuality in the patient/doctor relationship. A universal model is needed for the individualised proceeding and communicatory action.

Discourse model

The case of the patient with cancer may illustrate the reifying and temporal mode of medical logic. In this case, we find that a discourse process would have come closer to the temporal logic than any of the other nine algorithms discussed above. The discursive approach may be not only the procedural but also the ethical approach, most fitting to the type of action that was needed in this case. According to Christine K Cassel,23

 “Communicative ethics says ... that striving for the best among imperfect results is admirable, especially when strict adherence to principles does not allow for negotiation and compromise essential to the sharing of power and the framing of decision.”

Discourse ethics emerged as a universal approach to overcome the loss of a metaphysical foundation in philosophical ethics and in postmodern societies. It works as a procedural model for finding a consensus, where a plurality of traditional, religious, moral, cultural, philosophical, political and juridical values, norms, and principles are in conflict.24 In postmodern ethical theory, a universal and simultaneously fundamental foundation of ethics is regarded as no longer possible or desirable.25 The purely rational foundation of ethics must even be doubted after the “trolley experiments” disclosed the strong emotional engagement influencing moral judgement.26 According to newer concepts of ethics, all people have some ethics—otherwise it cannot be evoked—but some personal ethics may differ. For postmodern cultures, moral authority primarily derives from the actual consent of those concerned.27 Also, the modern theory of science is more influenced by the discursive consent model than by the traditional congruence concepts of truth (Charles S Peirce, 1839–1914).28 The discourse model stands for both rational reasoning in, and the practical application of, ethics.

Going back to the case, early communication and consultation with the nephrologist could have avoided contrast nephrotoxicity. This would be time consuming only for the moment, but not in the long run. Fundamental but conflicting values (beneficence v non-maleficence) were in competition for the urologist and the nurses, when they are made responsible for placing a urinary catheter. Moral obligations, such as to “save life and dialyse”, are becoming relative through the dialogue with the patient and ethical reasoning on preferences.29 The discourse model allows the expert (urologist) to change his mind without losing authenticity. The discourse model is a case based dialogue implicitly or explicitly applying ethics, which does not affect the validity of personal ethics or medical expertise. The discourse corresponds to the temporality and individuality of medical practice.

Discourse ethics is a universal model for non-dominated communicatory action; it is a “sharing of power”.23 The gynaecologist could not order, but could ask for dialysis; the nephrologist could not refuse dialysis, but he could argue. The patient, gynaecologist, nurses, urologist, nephrologist, and oncologist can communicate; everybody has ethics; no expert, mediator, or ethics committee was actually needed. Discourse is a method and also a target. For practical medicine, the discourse is both a way to proceed to a consent and a mode to import ethical principles such as respectful reciprocity.

The discourse, admittedly, depends on a set of conventions such as equality and reciprocity, and on rules to be followed, such as argumentation and listening.30 The discourse is both a rule based procedure and a case based ethics. Discourse is also learning by doing, how to instil general principles and values into practice. The discursive dialogue must be learnt and taught; the discourse will change the mind and educate the participants. Equal rights are the specific prerequisite of the discourse process. Domination-free communication between the people specifically forming part of the discourse may increase the chance that a creative and adequate solution can be found. It defines the aim of ethics, as the procedural rules promote mutual respect: the communicative rules establish reciprocal equity. Respect and reciprocity may be seen as both the procedural means and the ethical ends of the discourse process.


Treating patients is solving problems. An obvious asymmetry exists in the patient/doctor relationship: the patient has the problem and the doctor holds the solution to heal or to help. But the principle of reciprocity also applies to this asymmetry, where the problem is becoming that of the doctor and the solution must fit the patient. The objection is made that discourse ethics resemble only an arbitrary, ephemeral and decisionistic counting of votes.31

What happens if people are unable or unwilling to participate in the discursive process of arguing, if it is not the better argument but the higher status or the more eloquent rhetoric that prevails? Possibly, people could consent to something very stupid and irrational. Such a failure of the discourse process, however, applies only to the momentary action. When no objection is raised, experts too can make foolish decisions (intravenous urography in renal failure). The discourse provides a good chance of finding out which decision needs the expert with medical knowledge (type of chemotherapy) and which needs communicatory seeking for the best solution in a special situation (any chemotherapy?). As compared with deterministic or authoritarian decisions, the discursive arguing process may also be more scientific in terms of theoretical fallibility and practical testability.

Major objections have been made against discourse ethics and some of these objections need to be discussed in the light of the logic of medical action. In discourse ethics, the universal ability to generalise the reasoning must be demanded for the theoretical discourse, whereas adequacy must be demanded for the practical discourse process applied in an individual case.32 Answers to the objections are given below to exemplify, by using the case of the patient with cancer, what we think that adequacy could be.

Objection I: Medical decisions must be made on a basis of equity. This cannot happen if each case is settled on its own merits. Answer I: Equity comes first, but individuality in addition makes for better decisions. Formal justice must be made factual by a concrete action (to dialyse or not). The theory of justice as fairness is derived from an individualised approach (John Rawls, 1921–2002).33

Objection II: Medical decisions must be made impersonally and professionally. This cannot happen if decision making is seen as a unique personal encounter between doctor and patient. Answer II: Medical professionalism is not impersonal (Emmanuel Levinas).3 Besides cognitive factors, emotional factors, which enhance and reaffirm the patients’ expectations, will make the patient/doctor relationship successful even in therapeutic terms.34 As others are involved, the discourse model permits personal “distance” (for example, the urologist); as the doctor is responsible for, and face to face with, the patient (for example, the gynaecologist), the discourse also calls for personal empathy (Thomas Percival, 1740–1804).

Objection III: Medical decisions must be made transparently and in a way that is publicly defensible. Answer III: Discretion stands before publicity. Transparency of medical decision making can be established better by all those concerned than by one paternalistic and solipsistic decision maker. The discourse will make a treatment easier to defend publicly (for example, dialysis in the case of a patient with advanced cancer).

Objection IV: The method advocated by discourse ethics is subjective and arbitrary. Answer IV: Strictly speaking, there is no objective decision. All action must be taken by subjects. After discursive seeking of consent, such a decision must have a reason and is no longer arbitrary with regard to the individual case.

Objection V: Discourse ethics suggest no clear method of decision making at all. Answer V: The discourse process does not replace the will to act, but it helps in finding the way to decide. What should be done if the discourse results in neverending deliberations? If the patient needs more time to decide, the staff must accept this. The more the course of the disease needs an immediate treatment (as was seen for chemotherapy), the more is it still up to the treating doctors to make the decisions to the best of their knowledge, virtue, and responsibility. But such pressure often also needs to be communicated retrospectively.

Discourse ethics may be the most generic way of applying ethics as all those concerned can participate in the discourse process. Other ethics (deontological, utilitarian, principled, or communitarian) may be required and must be considered for more complex and more fundamental problems on a societal level with general validity, such as decisions on embryonic stem cell research or euthanasia. Such a process, however, may also be best set in motion by general discourse within the entire community. The discourse is a procedural technique that is applicable to reasoning and conflict mediation (at the bedside or at a scientific meeting), to parliamentary politics and even to the teaching of ethics.35 This universal applicability of the discourse model is a surprising experience that may result from the “Universalpragmatik”, as presumed and claimed for the theoretical discourse by Juergen Habermas.36 The resultant consent, or dissent, found by the practical application of discourse ethics must only hold temporarily for the singular and individual case. The result of the practical discourse must only hold temporarily for the singular and individual case.


In medicine, a new method for case based proceeding is needed. Such a method should use everyday speech more than complicated rule based algorithms. Communicatory action in clinical medicine is learning by doing, guided by a structured proceeding and ethical reasoning. Clinical ethics resembles a time dependent process of three sequential stages (fig 2).

Figure 2

 Discursive ethics as a procedural model for clinical medicine. The charter of clinical ethics and medical professionalism have been condensed into 10 commitments and three fundamental guiding principles as in panel (2).37 The relative weight of commitments and principles, however, must be balanced for the special clinical circumstances by the case based discourse.

1. Clinical problems pose diverse moral dilemmas where ethical principles are in conflict (beneficence v autonomy). The resulting dilemma can be external or internal to the patient, related to medical knowledge as well as to professional or individual values.

2. Medical expertise, ethical virtues, professional commitments, and normative principles are involved.37 For the process of ethical reasoning, every person participating has expertise, is guided by private values, and has to accept professional and personal responsibility.

3. Conflicting values and intentions should result in concrete action (“wait and see” could turn out to be the worst option in the anuric case). The discourse model stands for both, as it is instrumental in finding a solution between irreconcilable medical experts and it integrates ethics as a constituent in the individual treatment.

Clinical ethics is “a pragmatic seeking for the best thing to do now”.38 Discourse ethics applies especially well to procedural logic in clinical medicine. Given the generalisation achieved by medical statistics in the last 50 years, an additional and complementary way to accomplish the individualisation process is needed for the future. Discursive ethics may support this addition to clinical medicine (fig 1).


We thank Anthony S Avery, Medical Faculty of the University of Nottingham, who supported us with the concrete verbalisation of our rather abstract thinking. This work is dedicated to Heidrun House.



  • * Clinical medicine could be handled as the mapping of the whole scale of true premises (+x) or false premises (−x) into a decision (y) with veracity values between zero (y = ±0) and one (y = +1). Possible conclusions are all more or less “right”, more or less likely, but decisively so (0<y<1). For such mapping, any sigmoid or logistic function could be proposed, such as the following Fermi function: Decision (y) = 1/[1+exp(−Σ±Premise (xi))]. With often ambiguous premises (±Premise xi = 0), as for example in clinical medicine, an equivocal truth value is calculated (Decision y = 0.5). With any changing premise (xi), the veracity value of a decision can change (y). Thus, a multigranular or a temporal logic is less a problem of too complex mathematics than a problem of parameterising the real situation.

  • Competing interests: none declared.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.