Table 5.

Multivariable Analysis of Risk Factors for Corneal Abrasion in 2009–2011

Reference	Covariate	Odds Ratio	95% confidence interval	Logistic Regression P-value	Hosmer and Lemeshow P-value	Area under ROC curve
Prostatectomy*	Robotic prostatectomy	1.508	0.987–2.302	0.057	0.9435	0.6003
	Race	0.062	0.009–0.446	0.006
Hysterectomy‡	Modified hysterectomy	R/Hyst (1) vs O/Hyst (0)			0.6558	0.7850
		6.505	4.323–9.788	0.0001
		L/Hyst (2) vs O/Hyst (0)
		3.821	2.594–5.630	0.0001
	Age	1.020	1.007–1.034	0.003
	Malignant	1.031	0.657–1.618	0.895
	# of chronic conditions	1.139	1.065–1.219	0.0001
	Race	0.275	0.128–0.589	0.001
	Year	1.119	0.916–1.368	0.271

^*Open prostatectomy is the reference group;

^‡Open hysterectomy is the reference group

The multivariable model, which considers all of the covariates (see Materials and Methods, Statistical Analysis subsection, paragraph 4, lines 1–5), used terms from the univariate model table 4 with P < 0.2. Results using either forward or backward elimination were the same. Among patients with prostatectomies, the risk of corneal abrasion was significantly less in African Americans compared to Whites. Among hysterectomy patients, robotic or laparoscopic procedure along with increasing age and number of chronic conditions was associated with higher risk for corneal abrasion; additionally, corneal abrasion was significantly less likely in African Americans compared to Whites. The Hosmer- Lemeshow and the area under the ROC curve values suggested good fit of the model.

ROC = receiver operating characteristic.