¿Cómo se calculan y ponderan los árboles de decisión en medicina?

Respuesta generada por Itaca · 15 fuentes · 14 de octubre de 2026

Pregunta original

Dame articulos del area de medicna donde muestren arboles de toma de decisiones, como lo calculan y los pesos

Summary

There are two different things often called a “clinical decision tree.” Decision-analysis trees compare clinical actions by combining event probabilities with the value of health outcomes; machine-learning decision trees split patient data using predictor thresholds to make predictions. The word “weights” can mean outcome utilities in the first type, or feature importance in the second—they are not interchangeable. The articles below cover both, with practical explanations of how the calculations are made. [1,2]

Articles on clinical decision-analysis trees

  1. Aleem IS, Schemitsch EH, Hanson BP. “What is a clinical decision analysis study?” Indian Journal of Orthopaedics, 2008.
    Full text (PMC) [1] [2]
    A useful starting point: it explains decision nodes, chance nodes, probabilities, utilities, and sensitivity analysis, and describes a clinical example comparing treatment options for femoral neck fracture. It explains “rolling back” a tree: multiply each outcome’s value by its probability, then add the results along each option’s pathway.

  2. “Making Sense of Decision Analysis Using a Decision Tree.” Journal of Bone and Joint Surgery, 2009.
    Full text (PMC) [3] [1]
    A clinician-oriented tutorial on representing choices, uncertain events, and outcomes in a tree, and comparing options using expected outcomes. It is useful for understanding the structure and interpretation of a decision tree.

  3. “Interpretation of diagnostic data: 6. How to do it with more complex maths.” BMJ, 2003.
    Full text (PMC) [4] [5]
    A practical overview of constructing a tree, assigning probabilities and utilities, combining them, and testing whether the preferred decision changes when assumptions vary.

How the probabilities and “weights” are calculated

For a decision-analysis tree, a simple expected-value calculation is:

Expected value of an option = sum of (probability of each outcome × value assigned to that outcome).

For example, if an option leads to two possible outcomes, the expected value is calculated by multiplying the probability of each outcome by its value and adding those products. If outcomes occur in sequence, multiply probabilities along each pathway to obtain the probability of reaching its terminal outcome.

  • Probabilities estimate how likely events or outcomes are; they should come from relevant trials, systematic reviews, or other appropriate clinical data.
  • Utilities or outcome values represent how desirable the outcomes are. They may be expressed as quality-adjusted life-years, costs, or preference scores. They are not automatically the same as “importance weights”; they need a defined scale and a stated source, such as published estimates or elicited patient preferences.
  • Sensitivity analysis checks whether the preferred option changes when uncertain probabilities or outcome values are varied. [1–3]

If you mean machine-learning trees

  1. Luu J, et al. “Practical guide to building machine learning-based clinical prediction models using imbalanced datasets.” Trauma Surgery & Acute Care Open, 2024.
    Full text (PMC) [6] [7]
    This article shows how a clinical prediction tree splits data using predictor-specific thresholds and discusses SHAP, a method for describing how predictors contribute to model predictions. Those contributions are not probabilities of clinical outcomes or patient-valued utilities. A feature-importance score also does not, by itself, establish that a factor causes the outcome or should guide treatment.

Practical distinction: If you want to compare treatment strategies and account for benefits, harms, or patient preferences, look for decision analysis and expected utility. If you want a model that predicts an outcome from patient variables, look for classification/regression trees, random forests, or gradient-boosted trees.

Fuentes

  1. https://pmc.ncbi.nlm.nih.gov/articles/PMC2759613/
  2. https://www.sciencedirect.com/science/article/abs/pii/S0025619621001282
  3. https://pmc.ncbi.nlm.nih.gov/articles/PMC2669856/
  4. https://pmc.ncbi.nlm.nih.gov/articles/PMC1875498/
  5. https://academic.oup.com/jamia/article/33/8/1510/8699511
  6. https://pmc.ncbi.nlm.nih.gov/articles/PMC11177772/
  7. https://pmc.ncbi.nlm.nih.gov/articles/PMC12916885/
  8. https://pmc.ncbi.nlm.nih.gov/articles/PMC2386204/
  9. https://pmc.ncbi.nlm.nih.gov/articles/PMC13174778/
  10. https://pmc.ncbi.nlm.nih.gov/articles/PMC11959216/
  11. https://onlinelibrary.wiley.com/doi/full/10.1111/jep.13953
  12. https://pmc.ncbi.nlm.nih.gov/articles/PMC9738643/
  13. https://pmc.ncbi.nlm.nih.gov/articles/PMC2094783/
  14. https://onlinelibrary.wiley.com/doi/abs/10.1111/jep.13915
  15. https://pmc.ncbi.nlm.nih.gov/articles/PMC7390703/

Respuesta generada por inteligencia artificial a partir de las fuentes citadas. Verifica las fuentes antes de aplicarla.

¿Tienes una pregunta sobre tu propio caso? Itaca responde con fuentes en segundos.

Hacer una pregunta

Más preguntas sobre Medicina interna