Heuristics structure and pervade formal risk assessment

The role of INTERCheckWEB digital innovation in supporting polytherapy management

The study aims at defining the factors affecting the clinicians' decision of changing or confirming the treatment options for frail patients in polytherapy, supporting prescribing patterns, thus also figuring out if the inclination of the clinicians towards digital solutions (INTERCheckWEB) and specific guidelines, could play a role in their decision. A literature review was performed, revealing the main individual, organizational and decisional factors, impacting on the clinicians' propensity to change the current patients' therapy: the clinician perceptions of support in case of clinical guidelines use or INTERCheckWEB use were studied. A qualitative approach was implemented, and thirty-five clinicians completed a questionnaire, aimed at evaluating fifteen different clinical cases, defining if they would change the patient's current therapy depending on the level of information received. Three methodological approaches were implemented. (1) Bivariate correlations to test the relationships between variables. (2) Hierarchical sequential linear regression model to define the predictors of the clinician propensity to change therapy. (3) Fuzzy Qualitative Comparative Analysis-fsQCA, to figure out the combination of variables leading to the outcome. Patient's age and autonomy (p value = 0.000), as well as clinician's perception regarding IT ease of use (p value = 0.043) and seniority (p value = 0.009), number of drugs assumed by the patients (p value = 0.000) and number of concomitant diseases (p value = 0.000) are factors influencing a potential change in the current therapy. The fsQCA-crisp confirms that the clinical conditions of the patients are the driving factors that prompt the clinicians towards a therapy change.

Inductive Risk, Science, and Values: A Reply to MacGillivray

The argument from inductive risk (AIR) is perhaps the most common argument against the value-free ideal of science. Brian MacGillivray rejects the AIR (at least as it would apply to risk assessment) and embraces the value-free ideal. We clarify the issues at stake and argue that MacGillivray's criticisms, although effective against some formulations of the AIR, fail to overcome the essential concerns that motivate the AIR. There are inevitable trade-offs in scientific enquiry that cannot be resolved with any formal methods or general rules. Choices must be made, and values will be involved. It is best to recognize this explicitly. Even so, there is more work to be done developing methods and institutional support for these choices.

Null Hypothesis Testing ≠ Scientific Inference: A Critique of the Shaky Premise at the Heart of the Science and Values Debate, and a Defense of Value-Neutral Risk Assessment

Many philosophers and statisticians argue that risk assessors are morally obligated to evaluate the probabilities and consequences of methodological error, and to base their decisions of whether to adopt a given parameter value, model, or hypothesis on those considerations. This argument is couched within the rubric of null hypothesis testing, which I suggest is a poor descriptive and normative model for risk assessment. Risk regulation is not primarily concerned with evaluating the probability of data conditional upon the null hypothesis, but rather with measuring risks, estimating the consequences of available courses of action and inaction, formally characterizing uncertainty, and deciding what to do based upon explicit values and decision criteria. In turn, I defend an ideal of value-neutrality, whereby the core inferential tasks of risk assessment-such as weighing evidence, estimating parameters, and model selection-should be guided by the aim of correspondence to reality. This is not to say that value judgments be damned, but rather that they should be accounted for within a structured approach to decision analysis, rather than embedded within risk assessment in an informal manner.

Characterising bias in regulatory risk and decision analysis: An analysis of heuristics applied in health technology appraisal, chemicals regulation, and climate change governance

In many environmental and public health domains, heuristic methods of risk and decision analysis must be relied upon, either because problem structures are ambiguous, reliable data is lacking, or decisions are urgent. This introduces an additional source of uncertainty beyond model and measurement error - uncertainty stemming from relying on inexact inference rules. Here we identify and analyse heuristics used to prioritise risk objects, to discriminate between signal and noise, to weight evidence, to construct models, to extrapolate beyond datasets, and to make policy. Some of these heuristics are based on causal generalisations, yet can misfire when these relationships are presumed rather than tested (e.g. surrogates in clinical trials). Others are conventions designed to confer stability to decision analysis, yet which may introduce serious error when applied ritualistically (e.g. significance testing). Some heuristics can be traced back to formal justifications, but only subject to strong assumptions that are often violated in practical applications. Heuristic decision rules (e.g. feasibility rules) in principle act as surrogates for utility maximisation or distributional concerns, yet in practice may neglect costs and benefits, be based on arbitrary thresholds, and be prone to gaming. We highlight the problem of rule-entrenchment, where analytical choices that are in principle contestable are arbitrarily fixed in practice, masking uncertainty and potentially introducing bias. Strategies for making risk and decision analysis more rigorous include: formalising the assumptions and scope conditions under which heuristics should be applied; testing rather than presuming their underlying empirical or theoretical justifications; using sensitivity analysis, simulations, multiple bias analysis, and deductive systems of inference (e.g. directed acyclic graphs) to characterise rule uncertainty and refine heuristics; adopting "recovery schemes" to correct for known biases; and basing decision rules on clearly articulated values and evidence, rather than convention.