Day 2: Miguel Hernan - How to Make People Immortal & Why it's Not a Good Idea | ADIA Lab Symposium

Day 2: Miguel Hernan - How to Make People Immortal & Why it's Not a Good Idea | ADIA Lab Symposium

by Miguel Hernán

Life & Natural Systems Medicine & Health MBNPublic health and preventive medicineMBNSEpidemiology and Medical statistics
🎙 Miguel Hernán 👥 824 📅 November 5, 2025 ⏱ 42 min 👁 627 📄 expert opinion 🧭 2026-08-16
Available in: English (current) Français

Keywords

immortal timetarget trialcausal inferencesurvival analysisbias

Summary

In this talk, Miguel Hernán explains the concept of immortal time bias, a common error in the analysis of healthcare databases that can lead to incorrect conclusions about treatment effects. He illustrates the problem with historical examples, such as the observation that generals live longer than soldiers, and more recent examples like heart transplant studies and the association between Oscar winners and longevity. Hernán emphasizes that immortal time arises when the analysis design creates periods during which individuals cannot experience the outcome, often due to selection based on future events or misclassification of treatment groups. He proposes a unified framework based on the explicit specification and emulation of a target trial, which aligns the start of follow-up with eligibility and treatment assignment, thereby preventing this bias. The talk underscores the importance of proper study design in causal inference, especially in the era of AI and big data, and highlights that even sophisticated analytical methods cannot correct for fundamental design flaws.

160 words

Critical Evaluation

Value of the Information & Strength of the Argument

The talk provides a clear and compelling explanation of a subtle but critical bias in medical research. Hernán uses intuitive examples and a step-by-step logical progression to demonstrate how immortal time bias arises and why it leads to erroneous conclusions. The argumentation is solid, grounded in established epidemiological principles, and the speaker effectively communicates the practical implications for data analysis. The value lies in its educational clarity and the emphasis on the target trial framework as a solution, which is a powerful and actionable concept for researchers.

96 words

Title / Content Match

The title is catchy and accurately reflects the core topic of immortal time bias, though it may initially seem sensational.

Quality & Reliability

9/10

The speaker is a leading expert in causal inference and epidemiology, and the talk is based on well-established methodological principles. The content is rigorous and clearly explained, with references to historical and contemporary examples.

Key Moments

Markers derived by PSI from the transcript: the creator did not define chapters.

Cited Sources

Concurring Sources

  • Target trial emulation — The framework discussed in the talk is a well-established concept in epidemiology.

Contribution & Novelties

The talk provides a clear and accessible explanation of immortal time bias, a common but often overlooked issue in observational data analysis. It emphasizes the importance of explicitly specifying a target trial to avoid this bias, which is a key principle for reliable causal inference. The talk is particularly relevant in the context of AI and big data, where such biases can be amplified.

Pour aller plus loin :

  • Target trial emulation — A framework for designing observational studies to mimic randomized trials.
  • Immortal time bias — A detailed overview of the bias and its implications.
  • Causal inference — The broader field of methods for determining cause-and-effect relationships.

108 words

Radar Profile

The radar profile shows high scores in quality, reliability, and technical level, with a slightly lower score in quantity of information, reflecting the focused nature of the talk. The overall profile indicates a highly informative and trustworthy presentation.

Reliability 9/10

💬 No comments were provided for analysis.