Table 4.
Risk-adjusted OR1/predicted mean difference (95% CI)B | |||||||||
---|---|---|---|---|---|---|---|---|---|
Outcome | Any complication | Mechanical wound | Infection | Urinary | Pulmonary | Gastrointestinal | Cardiovascular | Systemic | Surgical |
Non-routine | 2.65 (2.58-2.72) | 3.03 (2.81-3.27) | 5.18 (4.91-5.46) | 1.51 (1.40-1.63) | 2.44 (2.31-2.57) | 1.51 (1.46-1.56) | 1.87 (1.77-1.99) | 1.87 (1.77-1.99) | 1.52 (1.44-1.59) |
discharge | |||||||||
In-hospital | 6.08 (5.79-6.39) | 0.95 (0.84-1.07) | 8.60 (8.13-9.11) | 1.36 (1.17-1.58) | 2.93 (2.73-3.15) | 1.01 (0.95-1.08) | 2.55 (2.34-2.78) | 2.45 (2.14-2.82) | 2.08 (1.90-2.27) |
mortality | |||||||||
LOSA (days) | 5.54 (5.45-5.64) | 4.80 (4.61-4.99) | 6.47 (6.35-6.61) | 1.31 (1.10-1.52) | 3.41 (3.27-3.54) | 3.64 (3.55-3.74) | 2.01 (1.87-2.15) | 1.25 (1.00-1.51) | 1.83 (1.69-1.98) |
Abbreviation: LOS - length of stay
Modeling used NIS-provided population weights generalized with STATA’s “svy” command to account for patient clustering within hospital-level variables and to extrapolate the sample to a nationally representative version of the US population.
Interpretation: Risk-adjusted odds of non-routine discharge (in-hospital mortality) are x times higher among patients with a given type of complication relative to patients without that type of complication; Risk-adjusted predicted mean length of stay is x amount higher among patients with a given type of complication relative to patients without that type of complication.
To account for non-nonnally distributed data, modified Park tests were used to determine selection of a gamma distribution (lambda not significantly different from 2). Predicted mean differences were calculated using post-estimation average marginal effects following GLM with link log, family gamma.
Risk-adjusted models accounted for potential confounding due to other complications, patient characteristics (age, sex, race/ethnicity, insurance type, median income quartile, indication, primary diagnosis, CCI, and year), and hospital-level factors (hospital volume, teaching-location, and geographical region).