Table 5. Dementia odds (Log_e transformed) and all-cause mortality: four-way decomposition models by poor sleep quality, overall, by sex and by race: HRS 2006–2020^a,b.

	TE		CDE		INTREF		INTMED		PIE
Y: All-cause mortality; M: Poor sleep quality score	β ± SE	P	β ± SE	P	β ± SE	P	β ± SE	P	β ± SE	P
Overall
Reduced Model 1A–1C (N = 6,991)
X: Hurd	+0.39525 ± 0.01906	<0.001	+0.39470 ± 0.01884	<0.001	−0.00014 ± 0.00049	0.78	+0.00004 ± 0.00028	0.88	+0.00064 ± 0.00059	0.28
X: Expert	0.62033 ± 0.02985	<0.001	+0.62222 ± 0.02984	<0.001	−0.00023 ± 0.00121	0.85	−0.00339 ± 0.00154	0.028	+0.00172 ± 0.00092	0.061
X: LASSO	0.58242 ± 0.02680	<0.001	+0.58335 ± 0.02676	<0.001	−0.00070 ± 0.00089	0.43	−0.00115 ± 0.00086	0.18	+0.00092 ± 0.00065	0.16
Full model 2A–2C (N = 6,510)
X: Hurd	+0.50781 ± 0.04782	<0.001	+0.46837 ± 0.04583	<0.001	+0.01866 ± 0.00609	0.002	+0.01585 ± 0.00517	0.002	+0.00493 ± 0.00187	0.008
X: Expert	+0.38284 ± 0.03203	<0.001	+0.35900 ± 0.03112	<0.001	+0.01024 ± 0.00345	0.003	+0.00947 ± 0.00278	0.001	+0.00414 ± 0.00153	0.007
X: LASSO	+0.44570 ± 0.03455	<0.001	+0.41699 ± 0.03368	<0.001	+0.01151 ± 0.00389	0.003	+0.01224 ± 0.00335	<0.001	+0.00496 ± 0.00174	0.004
Male
Reduced Model 1A–1C (N = 2,931)
X: Hurd	+0.56784 ± 0.04557	<0.001	+0.55737 ± 0.04524	<0.001	+0.01378 ± 0.00707	0.051	−0.00581 ± 0.00328	0.076	+0.00251 ± 0.00170	0.14
X: Expert	+0.60686 ± 0.05046	<0.001	+0.60202 ± 0.05059	<0.001	+0.00639 ± 0.00618	0.30	−0.00446 ± 0.00385	0.24	+0.00291 ± 0.00221	0.19
X: LASSO	+0.57377 ± 0.04415	<0.001	+0.56788 ± 0.04412	<0.001	+0.00730 ± 0.00583	0.21	−0.00373 ± 0.00274	0.17	+0.00233 ± 0.00167	0.16
Full model 2A–2C (N = 2,698)
X: Hurd	+0.41642 ± 0.07246	<0.001	+0.36637 ± 0.06879	<0.001	+0.03971 ± 0.01513	0.009	+0.00806 ± 0.00666	0.23	+0.00228 ± 0.00210	0.28
X: Expert	+0.26959 ± 0.04759	<0.001	+0.24091 ± 0.04630	<0.001	+0.02081 ± 0.00885	0.019	+0.00534 ± 0.00328	0.10	+0.00254 ± 0.00189	0.18
X: LASSO	.3520528 ± .0514405	<0.001	.3215936 .0500869 ±	<0.001	.0218936 ± .0099902	0.028	.0057126 ± .003571	0.11	.0028529 ± .0020199	0.16
Female
Reduced Model 1A–1C (N = 4,060)
X: Hurd	+0.37651 ± 0.02293	<0.001	+0.37352 ± 0.02191	<0.001	+0.00283 ± 0.00294	0.34	+0.00007 ± 0.00039	0.86	+0.00009 ± 0.00054	0.86
X: Expert	+0.63482 ± 0.03735	<0.001	+0.64365 ± 0.03736	<0.001	−0.00729 ± 0.00256	0.004	−0.00278 ± 0.00172	0.11	+0.00123 ± 0.00096	0.20
X: LASSO	+0.60279 ± 0.03461	<0.001	+0.60793 ± 0.03429	<0.001	−0.00511 ± 0.00281	0.068	−0.00024 ± 0.00080	0.76	+0.00022 ± 0.00071	0.76
Full model 2A–2C (N = 3,812)
X: Hurd	+0.56909 ± 0.06359	<0.001	+0.54232 ± 0.06188	<0.001	−0.00075 ± 0.00514	0.88	+0.02080 ± 0.00754	0.006	+0.00673 ± 0.00282	0.017
X: Expert	+0.46129 ± 0.04325	<0.001	±0.44589 ± 0.04242	<0.001	−0.00320 ± 0.00340	0.35	+0.01338 ± 0.00440	0.002	+0.00522 ± 0.00223	0.019
X=LASSO	+0.51349 ± 0.04663	<0.001	+0.49309 ± 0.04586	<0.001	−0.00386 ± 0.00357	0.28	+0.01806 ± 0.00535	0.001	+0.00619 ± 0.00260	0.017
White
Reduced Model 1A–1C (N = 5,666)
X: Hurd	+0.40191 ± 0.02179	<0.001	+0.40169 ± 0.02137	<0.001	+0.00032 ± 0.00128	0.81	−0.00002 ± 0.00019	0.91	−0.00008 ± 0.00083	0.92
X: Expert	+0.65884 ± 0.03420	<0.001	+0.66085 ± 0.03412	<0.001	−0.00201 ± 0.00104	0.053	−0.00132 ± 0.00109	0.23	+0.00132 ± 0.00091	0.15
X: LASSO	+0.59003 ± 0.02964	<0.001	+0.59204 ± 0.02956	<0.001	−0.00203 ± 0.00100	0.042	−0.00009 ± 0.00065	0.89	+0.00012 ± 0.00083	0.89
Full model 2A–2C (N = 5,324)
X: Hurd	+0.52716 ± 0.05367	<0.001	+0.48794 ± 0.05143	<0.001	+0.01634 ± 0.00657	0.013	+0.01836 ± 0.00614	0.003	+0.00452 ± 0.00200	0.043
X: Expert	+0.43335 ± 0.03742	<0.001	+0.40639 ± 0.03639	<0.001	+0.00910 ± 0.00388	0.019	±0.01379 ± 0.00380	<0.001	+0.00406 ± 0.00188	0.030
X: LASSO	+0.47401 ± 0.03910	<0.001	+0.44245 ± 0.03806	<0.001	+0.01079 ± 0.00438	0.014	+0.01629 ± 0.00431	<0.001	+0.00448 ± 0.00201	0.026
Black/Hispanic/Others
Reduced Model 1A–1C (N = 1,325)
X: Hurd	+0.36174 ± 0.03989	<0.001	+0.36399 ± 0.04016	<0.001	+0.00134 ± 0.00387	0.73	−0.00105 ± 0.00274	0.70	−0.00254 ± 0.00336	0.45
X: Expert	+0.47987 ± 0.06066	<0.001	+0.48313 ± 0.06060	<0.001	+0.01265 ± 0.00903	0.16	−0.01126 ± 0.00714	0.12	−0.00465 ± 0.00564	0.41
X: LASSO	+0.53338 ± 0.06233	<0.001	+0.53659 ± 0.06229	<0.001	+0.00674 ± 0.00757	0.37	−0.00578 ± 0.00582	0.32	−0.00417 ± 0.00461	0.37
Full model 2A–2C (N = 1,186)
X: Hurd	+0.41028 ± 0.10579	<0.001	+0.37851 ± 0.10205	<0.001	+0.01963 ± 0.01694	0.25	+0.00605 ± 0.00738	0.41	+0.00608 ± 0.00652	0.35
X: Expert	±0.22589 ± 0.06172	<0.001	+0.21525 ± 0.05966	<0.001	+0.00927 ± 0.00966	0.34	+0.00042 ± 0.00205	0.84	+0.00095 ± 0.00453	0.83
X: LASSO	+0.33939 ± 0.07511	<0.001	+0.32810 ± 0.07360	<0.001	+0.00591 ± 0.01086	0.59	+0.00133 ± 0.00252	0.60	+0.00405 ± 0.00560	0.47

Abbreviations: CDE: Controlled Direct Effect; HRS: Health and Retirement Study; INTMED: Mediated Interaction; INTREF: Interaction Referent; PIE: Pure Indirect Effect; SE: Standard Error. ^aCox PH regression models with mortality as the main outcome and poor sleep quality as the potential mediator/moderator allowed to interact with the main exposure. Log_e (odds (dementia probability)) using Hurd, expert and LASSO algorithms were the main alternative exposures of interest, sample size N = 6,991, four-way decomposition analysis. 1 SD of the “poor sleep quality score” corresponded to 2.18 point higher score. 1 SD of Hurd, expert and LASSO algorithm Log_e (dementia probability) corresponded to 3.93, 3.13 and 2.40, respectively. ^bExogenous variables are the ones included in Table 2, Models 1A–1D and 2A–2D, as covariates for the reduced and full models, respectively. See Covariates section for detail.