SSI_model_functions.R

# model from Sheyn et al. https://pubmed.ncbi.nlm.nih.gov/35830590/
# predictions recalibrated by Rhodea et al. [unpublished, will add link when published]
# set recalibrate = FALSE for original model
cms_model = function(Age, PVD=0, Diabetes_complicated=0, COPD=0, 
                     Mesh=0, Abdominal_sacrocolpopexy=0, Sling=0, 
                     Transfusion=0, Bowel_injury=0, Hemiplegia_or_paraplegia=0, 
                     return_lin_pred=FALSE, recalibrate = TRUE, intercept = -2.18, slope = 0.66){
  # Age = in years
  # PVD = Current peripheral vascular disease
  # Diabetes_complicated = Current diabetic complications
  # COPD = Current chronic obstructive pulmonary disease (COPD)
  # Mesh = Mesh prosthesis implanted during prolapse surgery
  # Abdominal_sacrocolpopexy = Abdominal (laparotomy) sacrocolpopexy performed
  # Sling = Sling performed
  # Transfusion = Perioperative blood transfusion
  # Bowel_injury = Intraoperative bowel injury
  # Hemiplegia_or_paraplegia = Current hemiplegia or paraplegia
  # return_lin_pred = return linear predictor (TRUE) or risk (FALSE)
  # recalibrate = return recelibrated model predictions (TRUE) or original (FALSE)
  # intercept, slope = logistic recalibration parameters (ignored if recalibrate = FALSE)
  
  LP = -1.2905403 - 0.035923587 * Age + 0.00014583438 * pmax(Age - 57, 0)**3 - 
    0.001212714 * pmax(Age - 68, 0)**3 + 0.0014428642 * pmax(Age - 71, 0)**3 - 
    0.00041141731 * pmax(Age - 76, 0)**3 + 3.5432762e-05 * pmax(Age - 84, 0)**3 + 
    0.27159901 * PVD + 0.30138923 * Diabetes_complicated + 0.20282055 * COPD + 
    0.25400704 * Mesh + 0.82215628 * Abdominal_sacrocolpopexy + 
    0.46605468 * Sling + 1.0624051 * Transfusion + 1.036331 * Bowel_injury + 
    0.80865929 * Hemiplegia_or_paraplegia
  
  if (recalibrate){
    LP <- intercept + slope*LP
   }
  
  p = plogis(LP)
  
  if (return_lin_pred){
    return(LP)
  } else{
    return(p)
  }
}