Table 2.
Conditions and interventions predicting in-hospital mortality in patients with IPF patients
| Risk factors | Unadjusted | Adjusted Model* | ||||
|---|---|---|---|---|---|---|
| OR | p-value | 95%CI | OR | p-value** | 95%CI | |
| Age | 0.995 | < 0.001 | (0.993–0.997) | 1.007 | < 0.001 | (1.004–1.009) |
| Female | 0.72 | < 0.001 | (0.7–0.76) | 0.77 | < 0.001 | (0.73–0.81) |
| Race***: | ||||||
| Black | 0.83 | < 0.001 | (0.76–0.91) | 0.69 | < 0.001 | (0.62–0.76) |
| Hispanic | 0.9 | 0.008 | (0.83–0.97) | 0.67 | < 0.001 | (0.62–0.74) |
| Ever Smoker | 0.83 | < 0.001 | (0.8–0.86) | 0.79 | < 0.001 | (0.75–0.83) |
| Elective admission | 0.63 | < 0.001 | (0.58–0.68) | 1.28 | < 0.001 | (1.16–1.4) |
| Academic hospital | 1.21 | < 0.001 | (1.16–1.27) | 1.14 | < 0.001 | (1.09–1.2) |
| Any respiratory failure | 8.67 | < 0.001 | (8.13–9.25) | 4.77 | < 0.001 | (4.34–5.13) |
| Mechanical ventilation therapy | 11.36 | < 0.001 | (10.86–11.88) | 7.26 | < 0.001 | (6.9–7.64) |
| Bronchoscopy | 2.25 | < 0.001 | (2.11–2.39) | 1.23 | < 0.001 | (1.13–1.33) |
| Gastroesophageal reflux disorder | 0.82 | < 0.001 | (0.78–0.86) | 0.84 | < 0.001 | (0.8–0.89) |
| Obstructive sleep apnea | 0.78 | < 0.001 | (0.74–0.84) | 0.62 | < 0.001 | (0.58–0.67) |
| Diabetes mellitus | 0.94 | 0.008 | (0.9–0.99) | 0.91 | 0.001 | (0.87–0.96) |
| Frailty | 1.67 | 0.001 | (1.23–2.27) | 1.48 | 0.03 | (1.04–2.1) |
| Low body mass index | 1.65 | < 0.001 | (1.5–1.82) | 1.51 | < 0.001 | (1.34–1.69) |
| Obesity | 0.81 | < 0.001 | (0.76–0.87) | 0.77 | < 0.001 | (0.71–0.84) |
| Pneumonia | 2.31 | < 0.001 | (2.22–2.41) | 1.38 | < 0.001 | (1.31–1.45) |
| New pulmonary embolism | 2.18 | < 0.001 | (1.97–2.42) | 1.83 | < 0.001 | (1.62–2.08) |
| Dependence on long-term Oxygen | 1.23 | < 0.001 | (1.18–1.28) | 0.91 | < 0.001 | (0.87–0.96) |
*Adjusted for all variables mentioned in this table
** Statistically significant P-value cutoff after Bonferroni correction is (p < 0.003)
*** Compared to white
The logistic regression model was statistically significant, χ2 = 14,153.7, p < 0.001. The model explained 29.7% (R2) of the variance in mortality and correctly classified 89.4% of cases. Sensitivity was 16.1%, specificity was 98.3%, positive predictive value was 53% and negative predictive value was 90.7%. The area under the ROC curve was 0.835 (95% CI, 0.831 to 0.839), which is an excellent level of discrimination (Figs. E3)