Resumen de STATISTICALLY Speaking. Predicted Probabilities' Relationship to Inclusion Probabilities.
Di Fang, Jenny Chong, Jeffrey R Wilson
The authors determine the probability of hospital readmission within 30 days of being discharged using binary data from Medicare. They mention that deriving predicted probabilities using retrospective data could result in biased estimates. They present the relationship between anticipated probabilities based on data treated retrospectively using logit, probit, and complementary log-log binary models. They suggest improvements in decreased rehospitalization could be a statistical artifact.
