0
$\begingroup$

I'm trying to create a logistic regression model for predicting future admissions based on historic clinical/utilization/demographic information. Although I have three years history available, for some of my potential predictors the data is not recorded for more than 50% of the patients. One such example would be blood pressure record. I think they will add value to my predictions. What strategies should I consider in treating such variables? I'm considering the following as of now but would like to hear from experts about pros, cons and other options.

  1. consider these as missing values and impute

  2. consider these patients to have normal blood pressure and use normal values

  3. consider these patients to have normal blood pressure and treat it as a dichotomous variable (in/out of normal range)

Any help is appreciated. Thank you.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.