Objectives: Statistical models predicting outcome after intraparenchymal hemorrhage include patients irrespective of do-not-attempt-resuscitation orders. We built a model to explore how the inclusion of patients with do-not-attempt-resuscitation orders affects intraparenchymal hemorrhage prognostic models.
Design: Retrospective, observational cohort study from May 2001 until September 2003.
Setting: University-affiliated tertiary referral hospital in Seattle, WA.
Patients: Four hundred twenty-four consecutive patients with spontaneous intraparenchymal hemorrhage.
Measurements and main results: We retrospectively abstracted information from medical records of intraparenchymal hemorrhage patients admitted to a single hospital. Using multivariate logistic regression of presenting clinical characteristics, but not do-not-attempt-resuscitation status, we generated a prognostic score for favorable outcome (defined as moderate disability or better at discharge). We compared observed probability of favorable outcome with that predicted, stratified by do-not-attempt-resuscitation status. We then generated a modified prognostic score using only non-do-not-attempt-resuscitation patients. Records of 424 patients were reviewed: 44% had favorable outcome, 43% had a do-not-attempt-resuscitation order, and 38% died in hospital. The observed and predicted probability of favorable outcome agreed well with all patients taken together. The observed probability of favorable outcome was significantly higher than predicted in non-do-not-attempt-resuscitation patients and significantly lower in do-not-attempt-resuscitation patients. Results were similar when applying a previously published and validated prognostic score. Our modified prognostic score was no longer pessimistic in non-do-not-attempt-resuscitation patients but remained overly optimistic in do-not-attempt-resuscitation patients.
Conclusions: Although our prognostic model was well-calibrated when assessing all intraparenchymal hemorrhage patients, predictions were significantly pessimistic in patients without and optimistic in those with do-not-attempt-resuscitation orders. Such pessimism may drive decisions not to attempt resuscitation in patients in whom a favorable outcome may have been possible, thereby creating a self-fulfilling prophecy. To be most useful in clinical decision making, intraparenchymal hemorrhage prognostic models should be calibrated to large intraparenchymal hemorrhage cohorts in whom do-not-attempt-resuscitation orders were not used.