The Influence of Genotype Information on Psychiatrists’ Treatment Recommendations: More Experienced Clinicians Know Better What to Ignore

Physician-Specific Maximum Acceptable Risk in Personalized Medicine: Implications for Medical Decision Making

“…(The remaining attributes were analyzed in our previous articles. 10,11 ) This approach is often presented in the literature because it is simple to describe and directly computed from the outputs of the model. However, results from the RPL model are much richer, as they consider the entire sequence of choices and the heterogeneity across clinicians.…”

“…9 Additionally, in estimating preference weights for different treatment characteristics, DCEs allow researchers to investigate how patients or clinicians trade off the risks and benefits of treatment. In 2 previous articles, 10,11 we investigated what thresholds psychiatrists use when determining whether a patient is responding to treatment and whether psychiatrists were influenced by the patient's genotype when recommending a treatment. However, DCEs also enable the researcher to derive a marginal rate of substitutions between positively perceived attributes (such as benefits from a treatment) and negatively perceived attributes (such as costs or risks of adverse events linked to a treatment).…”

“…[13][14][15] This study presents 2 methods widely used in preference analysis applied to other fields, such as transportation and environmental economics to incorporate preference heterogeneity when computing marginal rates of substitution (e.g., MAR). We demonstrate this using data from a previous study with a sample of psychiatrists 10,11 in which we aim to explore the extent of MAR heterogeneity across our sample of psychiatrist clinicians when prescribing treatments to their patients. The idea is to exploit the outcome from the RPL model, which allows us firstly to derive the unconditional distributions for each parameter-and therefore the MAR-and second, to retrieve individual posterior estimates conditional to the sequence of choices observed for each psychiatrist in the sample.…”

Scientific challenges for precision public health

“…This makes it challenging to define a ‘responder’29 in the context of personalised interventions. For example, we ourselves have shown how health professionals struggle to distinguish signal from noise when trying to identify ‘a responder’ to treatment in even simple scenarios 34. Some tend to overestimate the effects of a patients’ genotype on ‘responder’ status 34.…”

“…For example, we ourselves have shown how health professionals struggle to distinguish signal from noise when trying to identify ‘a responder’ to treatment in even simple scenarios 34. Some tend to overestimate the effects of a patients’ genotype on ‘responder’ status 34. But even if we properly define an individual-level MCID, the calculation of a final sample size (in this case how many epochs are necessary to observe and characterise the time trace of behaviour) will also depend on an individual’s engagement and willingness to be followed up and record data over prolonged periods.…”