Using explicit decision rules to manage issues of justice, risk, and ethics in decision analysis: when is it not rational to maximize expected utility?

Patients, doctors and risk attitudes

A lively topic of debate in decision theory over recent years concerns our understanding of the different risk attitudes exhibited by decision makers. There is ample evidence that risk-averse and risk-seeking behaviours are widespread, and a growing consensus that such behaviour is rationally permissible. In the context of clinical medicine, this matter is complicated by the fact that healthcare professionals must often make choices for the benefit of their patients, but the norms of rational choice are conventionally grounded in a decision maker's own desires, beliefs and actions. The presence of both doctor and patient raises the question of whose risk attitude matters for the choice at hand and what to do when these diverge. Must doctors make risky choices when treating risk-seeking patients? Ought they to be risk averse in general when choosing on behalf of others? In this paper, I will argue that healthcare professionals ought to adopt a deferential approach, whereby it is the risk attitude of the patient that matters in medical decision making. I will show how familiar arguments for widely held anti-paternalistic views about medicine can be straightforwardly extended to include not only patients' evaluations of possible health states, but also their attitudes to risk. However, I will also show that this deferential view needs further refinement: patients' higher-order attitudes towards their risk attitudes must be considered in order to avoid some counterexamples and to accommodate different views about what sort of attitudes risk attitudes actually are.

Nonconsequentialist decisions

According to a simple form of consequentialism, we should base decisions on our judgments about their consequences for achieving our goals. Our goals give us reason to endorse consequentialism as a standard of decision making. Alternative standards invariably lead to consequences that are less good in this sense. Yet some people knowingly follow decision rules that violate consequentialism. For example, they prefer harmful omissions to less harmful acts, they favor the status quo over alternatives they would otherwise judge to be belter, they provide third-party compensation on the basis of the cause of an injury rather than the benefit from the compensation, they ignore deterrent effects in decisions about punishment, and they resist coercive reforms they judge to be beneficial. I suggest that nonconsequentialist principles arise from overgeneralizing rules that are consistent with consequentialism in a limited set of cases. Commitment to such rules is detached from their original purposes. The existence of such nonconsequentialist decision biases has implications for philosophical and experimental methodology, the relation between psychology and public policy, and education.

Priority setting in surgery: improve the process and share the learning

Surgeons and surgical programs encounter priority-setting challenges every day, such as in regard to purchasing new technologies or managing waiting lists for elective surgery. The purpose of this paper was to explore priority setting in surgery. Traditionally in surgery, priority-setting decisions for new technologies have been based on evidence of effectiveness and cost-effectiveness; and decisions about managing waiting lists for elective surgery have been based on urgency rating scores. The fairness of priority-setting processes in surgical programs should be enhanced to permit all relevant information and values to be considered. The quality of these decisions can be improved by using an approach that captures and shares lessons from each priority-setting experience. The approach we propose in this paper- describe, evaluate, and improve using a leading conceptual framework for priority setting, called "accountability for reasonableness"-can be used by any surgical program to improve its priority setting, share lessons with others, and develop an evidence base for how these important health policy decisions should be made.

Cost-effectiveness analysis with risk aversion

This paper discusses why, in a medical context, the standard assumption of a risk-neutral social planner is inappropriate and develops a framework for conducting cost-effectiveness (CE) analysis when social planners are risk-averse. This framework demonstrates that if new medical interventions are variance increasing (decreasing), the risk-neutral approach will approve (reject) projects that should be rejected (accepted). This methodology is applied to two medical interventions that have been previously evaluated and considered cost-effective in the published literature. Since both conclusions assumed risk neutrality we determine the level of societal risk-aversion that would be necessary to reject these new interventions and compare these levels to previous estimates of risk-aversion in the economics literature. We find that for reasonable values of the risk-aversion parameter, only one of the two interventions should be approved. It is our recommendation that the cut-off risk aversion parameter (the level of risk-aversion above which a project would be rejected) should become a standard reported figure in future CE studies.

Decision analysis in evidence-based decision making

Many real world decisions have to be made on a limited evidence base, and clinical decisions are at best problematic. We explored some of the reasons why decision making in health care is so complex, and examined how decision analytic techniques might contribute to problem structuring and to implementation of evidence-based practice. We argued that decision analysis could, to some extent, overcome complexity of decision making by a clear structuring of the problem and a formal analysis of the implications of different decisions. Decision-analytic techniques can guide the management of individual patients or can be used to address policy questions about the use of treatment for groups of patients. However, decision analysis is not without its criticisms, e.g. problems are narrowly defined, replacing judgement and dehumanizing care, neglect of process utility and lack of primary data to develop decision analytic models. The development of evidence-based guidelines is a key component of the UK Government's quality strategy led by the National Institute for Clinical Excellence (NICE). However, the guidelines approach may lead to conflict when assessments of the effectiveness of interventions for individuals (whether or not supported by a formal decision analysis) conflict with the recommendations made by NICE for cost and clinical effectiveness for aggregate groups of patients. Decision analysis may or may not help with this but if guidelines are derived from a decision analysis, then the implications of patient preferences should be made clearer. However, decision analysis-derived guidelines will make general recommendations that may not be appropriate for all individuals. Nonetheless, decision analysis does make such implications explicit and propose that the guidelines should be supported by some mechanism for determining individual patient preferences. It will now need to consider whether some of NICE resources should be directed beyond evidence-based guidelines into decision analysis-derived guidelines and into decision analytical techniques to provide support for clinical and cost effective decision making within the patient-clinician encounter.

The irrelevance of inference: a decision-making approach to the stochastic evaluation of health care technologies

The literature which considers the statistical properties of cost-effectiveness analysis has focused on estimating the sampling distribution of either an incremental cost-effectiveness ratio or incremental net benefit for classical inference. However, it is argued here that rules of inference are arbitrary and entirely irrelevant to the decisions which clinical and economic evaluations claim to inform. Decisions should be based only on the mean net benefits irrespective of whether differences are statistically significant or fall outside a Bayesian range of equivalence. Failure to make decisions in this way by accepting the arbitrary rules of inference will impose costs which can be measured in terms of resources or health benefits forgone. The distribution of net benefit is only relevant to deciding whether more information is required. A framework for decision making and establishing the value of additional information is presented which is consistent with the decision rules in CEA. This framework can distinguish the simultaneous but conceptually separate steps of deciding which alternatives should be chosen, given existing information, from the question of whether more information should be acquired. It also ensures that the type of information acquired is driven by the objectives of the health care system, is consistent with the budget constraint on service provision and that research is designed efficiently.

A method for evaluating health care providers' decision making: the Provider Decision Process Assessment Instrument

Valid and reliable assessment of the clinical decision-making process is essential for the evaluation of decision aiding methods and effective quality assurance programs. The Provider Decision Process Assessment Instrument is a 12-item questionnaire that measures a health care provider's degree of comfort with a medical decision. Its measurement properties were studied in two general internal medicine practices. Reliability, measured using Cronbach's alpha, was 0.90 (95% Cl = 0.87 to 0.92). Construct validity was also high, with expected negative correlations ranging from -0.53 to -0.67. The instrument also satisfied standard criteria for item homogeneity and was readily completed by clinicians. These results suggest that the Provider Decision Process Assessment Instrument will prove to be a valuable tool for assessing medical decision making in busy clinical settings.

The danger of applying group-level utilities in decision analyses of the treatment of localized prostate cancer in individual patients

The optimal management strategy for men who have localized prostate cancer remains controversial. This study examines the extent to which suggested treatment based on the perspective of a group or society agrees with that derived from individual patients' preferences. A previously published decision analysis for localized prostate cancer was used to suggest the treatment that maximized quality-adjusted life expectancy. Two treatment recommendations were obtained for each patient: the first (group-level) was derived using the mean utilities of the cohort; the second (individual-level) used his own set of utilities. Group-level utilities misrepresented 25-48% of individuals' preferences depending on the grade of tumor modeled. The best kappa measure achieved between group and individual preferences was 0.11. The average quality-adjusted life years lost due to misrepresentation of preference was as high as 1.7 quality-adjusted life years. Use of aggregated utilities in a group-level decision analysis can ignore the substantial variability at the individual level. Caution is needed when applying a group-level recommendation to the treatment of localized prostate cancer in an individual patient.

Evidence-informed patient choice. Practical issues of involving patients in decisions about health care technologies

Evidence-informed patient choice involves providing people with research-based information about the effectiveness of health care options and promoting their involvement in decisions about their treatment. Although the concept seems desirable, the processes and outcomes of evidence-informed patient choice are poorly understood, and it should be carefully evaluated.

[Economic analysis of secondary prevention of coronary heart disease with simvastatin (Zocor) in Germany]

BACKGROUND

Coronary artery disease (CAD) has enormous financial, medical, as well as economical consequences in Germany. Important risk factors include raised total cholesterol and LDL cholesterol blood levels. The 4S Study has demonstrated the clinical effectiveness in the secondary prevention of CAD.

METHOD

Based on the health economics data of this study, we undertook a cost-effectiveness analysis of the secondary prevention of CAD with simvastatin in Germany.

RESULTS

On average, the costs per life year gained were DM 18,500.-(sensitivity analysis: DM 9,340.- to DM 29,374.-). The consequences of this result are discussed.

CONCLUSION

It is necessary from a health economists' point of view to assess the efficiency of a clinically effective measure in a standardised manner. This permits a comparison of efficiency with other, competing forms of health care, which is necessary in areas like CAD where there are different approaches to combat the disease. Simvastatin is highly efficacious in the secondary prevention of CAD in a defined patient population and, in comparison to other interventions in this area, it also proves to be cost-efficient.

Primer on medical decision analysis: Part 4--Analyzing the model and interpreting the results

This paper is the fourth of a five-part series that describes the principles of construction and evaluation of valid decision models. In this review, the authors describe the key principles of detecting and eliminating structural and programming errors in decision trees (debugging). In addition, they offer guidelines to facilitate the interpretation of analytic results of decision models.

Safe working practices and HIV infection: knowledge, attitudes, perception of risk, and policy in hospital

OBJECTIVES--To assess the knowledge, attitudes, and perceptions of risk of occupational HIV transmission in hospital in relation to existing guidelines. DESIGN--Cross sectional anonymous questionnaire survey of all occupational groups. SETTING--One large inner city teaching hospital. SUBJECTS--All 1530 staff working in the hospital in October 1991 and 22 managers. MAIN MEASURES--Knowledge of safe working practices and hospital guidelines; attitudes towards patients with AIDS; perception of risk of occupational transmission of HIV; availability of guidelines. RESULTS--The response rate in the questionnaire survey was 63% (958/1530). Although staff across all occupational groups knew of the potential risk of infection from needlestick injury (98%, 904/922), significantly more non-clinical staff (ambulance, catering, and domestic staff) than clinical staff (doctors, nurses, and paramedics) thought HIV could be transmitted by giving blood (38%, 153/404 v 12%, 40/346; chi 2 = 66.1 p < 0.001); one in ten clinical staff believed this. Except for midwives, half of staff in most occupational groups and 19% (17/91) of doctors and 22% (28/125) of nurses thought gloves should be worn in all contacts with people with AIDS. Most staff (62%, 593/958), including 38% (36/94) of doctors and 52% (67/128) of nurses thought patients should be routinely tested on admission, 17% of doctors and 19% of nurses thought they should be isolated in hospital. One in three staff perceived themselves at risk of HIV. Midwives, nurses, and theatre technicians were most aware of guidelines for safe working compared with only half of doctors, ambulance, and paramedical staff and no incinerator staff. CONCLUSIONS--Policy guidelines for safe working practices for patients with HIV infection and AIDS need to be disseminated across all occupational groups to reduce negative staff attitudes, improve knowledge of occupational transmission, establish an appropriate perception of risk, and create a supportive and caring hospital environment for people with HIV. IMPLICATIONS--Managers need to disseminate policy guidelines and information to all staff on an ongoing basis.

Towards understanding treatment preferences of hospital physicians

Seventy-two physicians working in a university hospital in The Netherlands were interviewed to clarify their decision-making process when choosing drugs of preference. Each physician was questioned about the treatment choices for either one or two general case descriptions. The physicians considered only a limited set of different treatment options, on an average 1.7-5.0. Further, the physicians expressed their expectancies as regards various treatment alternatives, and the value or weight they attached to the principle aspects of a treatment. An analytical decision model was used as a reference to gain insight into the extent to which the physicians make decisions based on their own subjective expectancies and values. This model assumes that physicians follow a maximizing strategy by choosing the treatment they personally assess as optimal. It was found that a model including only biomedical expectancies and values predicted the preferred treatment correctly in no more than 53% of the cases. Sometimes, biomedical aspects were disregarded that should have been relevant according to the physicians themselves. Adding aspects of the social environment and experiences improved the prediction of the model substantially; 3 out of 4 treatment preferences could be understood by following an analytical maximizing strategy including biomedical aspects and social aspects and experiences. In the remaining cases, the physicians were not able to describe their decision in terms of this maximizing strategy, which points at the use of alternative decision strategies. One alternative decision strategy mentioned by the physicians was a 'follow-the-routine' decision rule.

Using the analytic hierarchy process (AHP) to develop and disseminate guidelines

To be effective, practice guidelines must accommodate the unique circumstances of individual patients. This article describes how the analytic hierarchy process (AHP), a decision-making technique, could be used to create flexible guidelines by linking guideline developers and clinical decision makers as coworkers in a common decision-making process. The advantages of using this approach for guideline dissemination are discussed and compared with other methods for disseminating and implementing guidelines. The clinical feasibility of the AHP approach is also reviewed.

Therapeutic decision making of physicians

In this review the therapeutic decision-making process of physicians is described. This process is divided into two steps: the generation of a limited set of possible options (the 'evoked set') and the selection from this evoked set of a treatment for a specific patient. Factors that are important in both steps are reviewed. Behavioural and decision-making theories in general and decision-making analysis of physicians in particular are discussed to identify possible shortcomings in their decision-making process. Based on this information a model of the drug choice process is presented. With reference to this model possible ways of influencing drug choices of physicians are discussed.

Fear of contagion and AIDS: nurses' perception of risk

Nurses' fear of contagion when caring for persons with AIDS remains high despite increased levels of knowledge. This paper examines the multiple factors that contribute to nurses' perception of risk within the workplace. The authors suggests that constructs from theories such as decision making, psychoanalysis and cognitive psychology can provide insight into the assessment of risk. Findings from a recent survey of nurses are used to illustrate the complex nature of fear of contagion. Understanding this complexity may be an essential first step in order to provide opportunities for resolution of fears and modification of behaviors.

The role of public values in setting health care priorities

Public values must play a substantial role in any attempt to deal with the health care resource allocation problem. This article examines how preferences for the health outcomes of care (e.g. improved or worsened physical suffering) can provide a coherent basis upon which set explicit health care priorities. Preferences for health outcomes could be mapped onto information concerning the outcomes expected from the specific health services when used for particular clinical conditions. These 'preference-weighted' outcomes would determine the relative priority given to health services for each specific condition. Generic outcome measures would be used in order to permit comparison of benefits and harms across different services. It is argued herein that allocation rules cannot be based on individual patients' preferences. Instead, average population preferences should be used to evaluate the relative importance of services--as occurs in other insurance contexts--despite theoretical concerns about the aggregation of preferences. Patients' preferences might also be estimated by reference to relevant demographic factors, but only if population subgroups are identified with relatively homogeneous preference patterns and if the use of such subgroups is deemed socially acceptable. Concerns about stereotyping and discrimination might limit the tractability of this approach.