Skip to Content
Interactive Textbook on Clinical Symptom Research Logo


Home Button

Tools for Decision Making Sections
Author Bio
Introduction
Probability Theory
Case Study 1: Patient History
Bayes' Theorem
Currently selected section: Methods for Estimating Pre-test Probability
Estimating Likelihood Ratios
Sensitivity and Specificity
Interpreting Test Results
Calculating Post-test Probabilities
Post-test Probabilities in Clinical Practice
Conclusions: Case Study 1
Part II
Part III
References


Chapter 14: Tools for Decision Making: Methods for Estimating Pre-test Probability
        

Clinical Prediction Rule Estimates

Such rules are based on analysis of a standardized set of data, including:

  • Clinical findings, and
  • The final diagnosis, for each of many patients with a diagnostic problem.

One type of clinical prediction rule uses regression analysis to identify the best clinical predictors and their diagnostic weights. The sum of the diagnostic weights corresponding to a patient's findings is a score, and the probability of disease for each patient is equivalent to the prevalence of disease among patients with a similar score.

To create a clinical prediction rule, follow these steps:

  • Cohort assembly - Because you must have a standard data set on all patients, it necessary to assemble the cohort prospectively. You must decide what symptom qualifies a patient for inclusion. You need a way to identify patients with the symptom and notify the research assistant to obtain informed consent.

  • Data collection - A standard data set is essential. Record the data on a form that specifies the data to be obtained and the standard way to obtain the data (standard phrasing of questions and definitions of abnormal physical findings).

  • Implement protocol - Next, implement a protocol for gathering the data necessary to establish the diagnosis (the gold standard test). The data may include a definitive test, careful long-term follow-up, or a combination of the two.

  • Establish the patient's diagnosis - By a means that does not use the findings that you will use to estimate the probability of disease (predictor variables).

  • Data entry - Finally, enter the data in a file. Divide the population into a training set (on which to define the key predictors of disease and their relationship to one another) and a test set (on which to test the clinical prediction rule). Use a multivariate method (logistic regression analysis, neural net analysis, or recursive partitioning).
  • An excellent reference for investigators interested in pursuing this work is:

    Wasson JH, Sox HC Jr, Goldman L, Neff RK. Clinical prediction rules: applications and methodologic standards. N Engl J Med. 1985;313:739-99.


     

    Page 10 of 43
          Previous Page