Parkkinen, Veli-Pekka (2025) Choice of effect measure, extrapolation, and decision-making in patient care and in public health. [Preprint]

Text efmes.pdf Download (174kB)

Abstract

Decision-theoretic arguments show that only absolute effect measures, and not relative ones, suffice for utility-maximizing clinical or public health decisions (Jäntgen 2023, Sprenger & Stegenga 2017}. This paper argues that despite the validity and importance of these results, no general conclusions about the policy-relevance of different measures follow from such arguments. This is because in a typical decision situation, the decision-relevant risks or summary effects are not directly available for the target patient or population, but must be inferred via extrapolation. Since absolute effects always depend on baseline risk, reliably extrapolating them requires controlling for all causes of the outcome that vary between the source and the target, not just for effect modifiers that mechanistically interact with the exposure. Relative effects depend on baseline risk in some circumstances but not in others, depending on the nature of the exposure (e.g. Huitfeldt et al. 2018, Sheps 1958). An appropriately chosen relative measure may thus be extrapolatable with far less auxiliary evidence or fewer assumptions, and given the target baseline risk, provides the risk information needed for rational decisions (Jäntgen 2023). Hence, estimating baseline risk in the target directly and extrapolating a relative effect may in many cases be the most reliable and efficient way to obtain decision-relevant evidence. Absolute measures are thus not generally superior in real-life decision contexts where extrapolation error must be dealt with. Instead, appropriate choice of effect measure to be used in inferring the decision-relevant risks depends on the properties of the exposure of interest. Lastly, the paper outlines some implications for philosophical treatments of the problem of extrapolation itself.

---

Export/Citation:

EndNote | BibTeX | Dublin Core | ASCII/Text Citation (Chicago) | HTML Citation | OpenURL

Social Networking:

Share |

---

Item Type:	Preprint
Creators:	Creators Email ORCID Parkkinen, Veli-Pekka
Keywords:	causality; effect measures; extrapolation
Subjects:	General Issues > Causation Specific Sciences > Medicine > Epidemiology
Depositing User:	Veli-Pekka Parkkinen
Date Deposited:	21 Feb 2025 15:12
Last Modified:	21 Feb 2025 15:12
Item ID:	24785
Subjects:	General Issues > Causation Specific Sciences > Medicine > Epidemiology
Date:	2025
URI:	https://philsci-archive.pitt.edu/id/eprint/24785

Monthly Views for the past 3 years

Monthly Downloads for the past 3 years

Plum Analytics

Actions (login required)

View Item