Table 4. Univariable and multivariable logistic regression by characteristics among 99 persons with intellectual disabilities who died of COVID-19, the Netherlands*.

Characteristics	Univariable			Multivariable†
	Odds ratio (95% CI)	p value		Odds ratio (95% CI)	p value
Sex
M	1.0 (0.7–1.6)	0.82		ND
F	Referent
Age	1.1 (1.1–1.1)	<0.001		1.09 (1.07–1.12)	<0.001
Disability level‡
Borderline to mild	Referent
Moderate	0.5 (0.3–0.9)	0.03		ND	NA
Severe to profound	0.6 (0.4–1.1)	0.13		ND	NA
Disability etiology‡
Down syndrome	2.86 (1.6–4.9)	0.001		5.6 (2.9–10.6)	<0.001
Concurrent conditions
Diabetes	3.4 (1.9–6.0)	<0.001		ND	NA
Hypertension	2.6 (1.5–4.6)	0.001		ND	NA
Heart disease	3.9 (2.1–7.1)	<0.001		2.3 (1.2–4.5)	0.01
Lung disease	4.0 (1.9–8.3)	<0.001		4.6 (2.0–10.7)	<0.001
Epilepsy	1.5 (0.8–2.7)	0.17		ND	NA
Overweight, BMI >25 kg/m2	0.9 (0.6–1.5)	0.82		ND	NA

*Because of nonresponses for some patient data among 2,586 persons included in the study, these data reflect missing values for sex, n = 60; age n = 67; and disability level, n = 114. BMI, body mass index; NA, not applicable; ND, not done. †Variables with p<0.1 in univariable analyses were included in multivariate logistic regression analysis. Because we used stepwise backward selection, we removed nonsignificant variables from the multivariable model and we could not provide estimates. The area under the curve was 0.844 (95% CI 0.808–0.880; p<0.001). We used Hosmer-Lemeshow goodness-of-fit test to assess the model fit for logistic regression and considered p>0.05 nonsignificant. Variance inflation factor (VIF) diagnostics indicated no evidence of collinearity (all VIF<1.2) among variables in final model. ‡Both the variable disability level and the variable etiology concern the level of intellectual disability. To avoid interdependency, we only included etiology in the multivariable model, because this variable shows a stronger univariable relationship and had no missing values.