. 2011 Apr 19;183(7):E391–E402. doi: 10.1503/cmaj.101860

Table 4:

Association between study-level factors and proportion of readmissions deemed avoidable in binomial regression models^*

Study-level factor	Weighted overall proportion of readmissions deemed avoidable
	Unadjusted			Adjusted
	In studies with factor	In studies without factor	p value	In studies with factor	In studies without factor	p value
Used administrative databases	59.0	11.7	< 0.001	–	–	–

Included patients on medical wards^†	59.0	20.0	< 0.001	–	–	–

Included surgical patients^†	9.3	18.0	< 0.001	–	–	–

Included geriatric patients^†	9.3	18.0	< 0.001	–	–	–

> 1 reviewer	24.6	9.3	< 0.001	–	–	–

Limited to specific diagnoses	34.2	10.0	< 0.001	74.0	23.1	< 0.001

Only readmissions because of physician factors considered avoidable	9.5	17.9	< 0.001	–	–	–

Publication year ≥ 2000	10.5	14.1	< 0.001	–	–	–

Follow-up period for readmissions of up to 1 yr after discharge^‡	9.0	20.9	< 0.001	36.8	59.4	< 0.001

> 2 sources of information used to determine avoidability of readmissions	24.6	9.6	< 0.001	–	–	–

Mostly teaching hospitals in study	8.7	53.4	< 0.001	20.8	76.4	< 0.001

Study from United States	25.5	9.9	< 0.001	–	–	–

Study from United Kingdom or Ireland	15.6	11.4	< 0.001	–	–	–

This table summarizes the results of univariable and multivariable binomial regression models that measured the association of study-level factors with the proportion of readmissions deemed avoidable. With the exception of the first factor (administrative database study), all analyses excluded the three studies that used administrative databases alone.¹⁸^,²⁷^,³³

^†

Compared with studies that excluded such patients or that did not specify patient type.

^‡

Compared with studies that had a follow-up period of up to 2.5 months after discharge.