I am trying to create a statistical model for my research. I am dealing with a surgery that has the chance of causing damage to the patient. I have found 5 signs, which I believe are indicators that the surgery will fail. Both the signs and the result of the surgery are binary. I am trying to create a model that will tell me the probability of surgery failing given that 1-5 of the signs are present. I was thinking of using multinomial logistic regression to create my model. Is there better distribution to use to create my model? Any other thoughts? Thank you for your help!
So it appears that the result of surgery has been over-simplified down to a 0/1 outcome, meaning that a much larger sample size will be required in order to build a reliable model, e.g., at least a few hundred surgical failures. The choice of a model for binary Y is the binary logistic regression model. Multinomial logistic regression would be used if Y were categorical with more than 2 unordered possible values. You can use the 5 predictors in the binary logistic model without assuming they are equally important, i.e., without just counting the number present.
To do a credible job will require a large amount of background reading. I suggest you start by reading Ewout Steyerberg's Clinical Prediction Models book as well as introductory papers or books on logistic regression.