Table 2. Effect estimates of WHELD intervention in comparison to TAU on primary outcome and key secondary outcome measures (multiple imputation analysis).

Outcome measure	Adjusted effect (SE)*	p-Value	Mean difference (SEM)	95% CI of mean difference	Effect size (Cohen’s D)	Number needed to treat∆
DEMQOL-Proxy (n = 553)	R = 0.12 Z = 2.82	0.0042	2.54+ (0.88)	0.81, 4.28	0.24	9
CMAI (n = 553)	R = 0.11; Z = 2.68	0.0076	4.27+ (1.59)	−7.39, −1.15	0.23	6
NPI-NH (n = 547)	R = −1.5; Z = 3.52	<0.001	4.55+ (1.28)	−7.07, −2.02	0.30	9

*Adjusted effect takes into account baseline value, age, sex, Clinical Dementia Rating, site, and clustering within care homes.

^∆Based on binary outcome: better than mean overall outcome versus mean outcome or worse than overall mean outcome for DEMQOL and CMAI.

⁺DEMQOL: improvement in WHELD group from baseline to 9 months 4.78, improvement in TAU group 2.24, mean difference 2.54. CMAI: improvement in WHELD group from baseline to 9 months −4.13, worsening in TAU group 0.14, mean difference 4.27. NPI-NH: improvement in WHELD group from baseline to 9 months −2.64, worsening in TAU group 1.91, mean difference 4.55.

CMAI, Cohen-Mansfield Agitation Inventory; NPI-NH, Neuropsychiatric Inventory–Nursing Home Version; TAU, treatment as usual.