Application of Markov Modeling to Assess Outcomes in COPD

Rationale: To create rolling window frames to individually assess each hospitalization with the diagnosis of COPD in the MIMIC database and develop Markov models to assess outcomes.Data was extracted from a cohort of 697 patients using ICD.9 CODES for acute exacerbation of COPD where a total of 341 patients in the cohort experienced acute respiratory failure. A total of 181 (26%) received antibiotics of which 56 (31%) received antibiotics within 6 hours of admission. There was a 9.8% mortality rate in the group that did not receive antibiotics versus a 4.4% mortality rate in the group that received antibiotics. We evaluated time-to-antibiotics for the 4 groups and conducted Monte Carlo simulations for each group. There was a positive correlation of mortality rate with the length of time to initiation of antibiotics for within 24 hours and after 48 hours. The estimated a mortality rate for each of the 4 groups are (1) 2.9%, (2) 3.8%, (3) 1.5%, and (4) 5.5%.