Table 4.

OR for use of each category of psychotropic drugs at one time-point given use one or two time-points earlier, respectively, lag 1 and lag 2, covariates were measured at baseline

Variable	Number of users		Unadjusted		Adjustedb
	Baseline	A2	OR (95% CI)	p-value	OR (95% CI)	p-valueo
Lag 1 (N=565a)
Antipsychoticsc	26	27	76.8 (28.5; 207.2)	< 0.001	80.5 (28.9; 224.0)	< 0.001
Antidepressantsd	97	118	53.3 (28.6; 90.1)	< 0.001	47.3 (25.1; 89.1)	< 0.001p
Anxiolyticse	44	67	25.9 (12.0; 55.5)	< 0.001	23.6 (10.7; 52.1)	< 0.001
Sedativesf	109	118	17.3 (10.5; 28.6)	< 0.001	18.2 (10.9; 30.4)	< 0.001
Lag 2 (N=432a)
	Baseline	A3
Antipsychoticsi	17	29	24.2 (7.1; 83.1)	< 0.001	20.8 (5.8; 75.0)	< 0.001
Antidepressantsj	66	97	29.8 (15.0; 59.2)	< 0.001	28.1 (13.8; 57.4)	< 0.001p
Anxiolyticsk	32	50	10.4 (4.5; 23.9)	< 0.001	11.4 (4.5; 28.7)	< 0.001
Sedativesl	74	98	6.6 (3.8; 11.3)	< 0.001	6.6 (3.8; 11.4)	< 0.001

Lag 1: Two consecutive assessment time points (baseline and A2)

Lag 2: One time point between selected time points (baseline and A3)

^a Cases with at least one missing on covariates were excluded

^bFollowing adjustment variables considered: Age, gender, marital status, CDR-SoB, GMHR, PSMS, NPI Affective sub-syndrome, NPI Psychosis sub-syndrome, NPI Agitation sub-syndrome all measured at A1 (baseline); Nursing home admission was not included as adjustment variable (no baseline values available); both models reduced by AIC

^cAdjusted for Gender

^dAdjusted for CDR-SoB and NPI Affective sub-syndrome

^eAdjusted for PSMS and NPI Affective sub-syndrome

^fAdjusted for NPI Affective sub-syndrome and NPI Agitation sub-syndrome

^gAdjusted for PSMS

^hAdjusted for CDR-SoB and NPI Psychosis sub-syndrome

ⁱAdjusted for Gender and PSMS

^jAdjusted for CDR-SoB, NPI Affective sub-syndrome and NPI Psychosis sub-syndrome

^kAdjusted for Marital status, GMHR and NPI Affective sub-syndrome

^lAdjusted for Age and CDR-SoB

^mAdjusted for Marital status and CDR-SoB

ⁿAdjusted for Gender, Age, Marital status and CDR-SoB

^op-values were calculated by generalized linear mixed model adjusting for municipality level if present

^p no cluster effect at municipality level present