Multiple imputation to quantify misclassification in observational studies of the cognitively impaired: an application for pain assessment in nursing home residents

Table 2 Distribution of pain indicators among newly admitted nursing home residents, by level of cognitive impairment (n = 1,930,192)^a

Pain Indicators	Level of Cognitive Impairment
	No/Mild, % (n = 1,160,053)		Moderate, % (n = 430,442)		Severe, % (n = 339,697)
	n	%	n	%	n	%
Staff-Assessed Pain	38,018		33,967		722,310
Yes		50.8		42.0		36.2
No		49.2		58.0		63.8
Resident-Reported Pain	1,117,134		397,072		271,192
Yes		66.0		49.4		34.0
No		34.0		50.6		66.0
Resident-reported pain details^b
Pain Frequency	1,115,210		394,128		265,555
No pain		34.1		51.0		67.4
Rarely		4.9		5.0		4.1
Occasionally		34.2		27.0		19.3
Frequently		21.0		13.3		7.4
Almost Constantly		5.8		3.7		1.8
Pain Numeric Rating	942,528		338,813		232,527
Median (25^TH,75^TH %)		4 (0–6)		0 (0–5)		0 (0–0)
Pain Verbal Descriptor Scale	564,584		258,391		212,453
No pain		67.3		77.7		84.2
Mild		11.9		9.1		7.1
Moderate		17.4		11.0		7.3
Severe		3.1		2.0		1.3
Very Severe		0.3		0.2		0.2
Pharmacologic Pain Management	1,159,468		430,198		339,472
Scheduled + PRN		28.0		19.3		13.6
Scheduled Only		8.0		10.0		11.7
PRN Only		40.5		34.5		28.5
None		23.6		36.3		46.3

^aResidents with at least one valid response to pain items were included in the overall study, but excluded from item-specific analyses if they lacked a valid response to the item of interest. As such, we present the sample size specific to each estimate
^bFor resident-reported details, values were set to “no pain” if explicitly documented as “no pain” or if missing and staff-assessed or resident-reported pain was “no pain”

ISSN: 1471-2288