Table 4.

Multivariable analyses of adverse health outcomes by frailty (F), dysphagia (D), and their interaction.

Model	Length of stay‡ odds ratio (95% CI)	Hospital cost odds ratio (95% CI)
Model 1: ACG binary *
D vs. No D for F	1.51 (1.45, 1.58)	1.27 (1.23, 1.31)
D vs. No D for non-F	1.67 (1.63, 1.70)	1.30 (1.28, 1.32)
Model 2: FRS index categories †
D vs. No D for F	1.30 (1.11, 1.53)	1.11 (0.98, 1.24)
D vs. No D for pre-F	1.66 (1.60, 1.72)	1.37 (1.33, 1.41)
D vs. No D for non-F	1.66 (1.62, 1.70)	1.29 (1.27, 1.31)

Model	Odds ratio of non-routine discharges (95% CI)	Odds ratio of medical complication (95% CI)	Odds ratio of surgical complication (95% CI)
Model 1: ACG binary *
D vs. No D for F	2.36 (1.95, 2.85)	1.85 (1.58, 2.17)	1.24 (1.06, 1.45)
D vs. No D for non-F	2.69 (2.49, 2.90)	1.98 (1.85, 2.12)	1.76 (1.57, 1.96)
Model 2: FRS index categories †
D vs. No D for F	3.08 (1.20, 7.89)	2.48 (1.20, 5.13)	1.33 (0.67, 2.66)
D vs. No D for pre-F	3.06 (2.64, 3.55)	1.99 (1.76, 2.26)	1.33 (1.15, 1.55)
D vs. No D for non-F	2.61 (2.41, 2.82)	1.99 (1.86, 2.13)	1.86 (1.66, 2.09)

^*The ACG binary is the Johns Hopkins Adjusted Clinical Groups (ACG) 10 frailty-defining diagnosis indicators and is a binary frail yes/no variable.

^†A FRS index was created from the FRS measure by number of positive categories divided by 19. The non-frail, pre-frail, frail cut-points were ≤ 0.08, > 0.08 but < 0.25, and ≥ 0.25, respectively.

^‡N for LOS models = 2,979,475, after excluding outliers and observations with missing covariates. N for total hospital encounter cost models = 2,910,880, after excluding outliers and observations with missing covariates. N for discharge disposition models = 3,023,335, after excluding missing values for discharge disposition. N for medical complications models = 3,024,415, after excluding observations with missing covariates. Due to low prevalence, AIDS, Chronic blood loss anemia, Congestive heart failure, Drug abuse, Lymphoma, Paralysis, Pulmonary circulation disorders, Solid tumors without metastasis, Peptic ulcer disease excluding bleeding, and Psychoses were not considered as covariates in the models of medical complications. N for the surgical complications models = 1,332,180, after excluding observations with missing covariates.

In the surgical complications subset, due to low prevalence, AIDS, Lymphoma, Peptic ulcer disease excluding bleeding, and Psychoses were not considered as covariates in these models.

Otherwise, all models adjusted for: age, sex, race, primary payer, hospital bed-size, hospital geographic region, median household income, admission type, surgical procedure, and the Elixhauser Comorbidities.