Predicting hospital patient readmission

For most hospitals, a relatively small group of patients with chronic conditions make up a disproportionate amount of medical costs. One study attempted to determine if an algorithm (in this case, logistic multivariate regression) based on routine inpatient data could be used to identify the at-risk patients who are more likely to be readmitted and form this chronic patient group (Howell, Coory, Martin and Duckett 2009). The model only ‘performed moderately’; plagued by false negatives, the system over aggressively denied benefits from many patients who should receive them (ibid). Researchers concluded that a statistical algorithm should never serve as a unilateral means of identifying at-risk patients. However, the model might be useful for identifying patients who could use closer clinical inspection.