Meditation and Cognitive Outcomes: A Longitudinal Analysis Using Data From the Health and Retirement Study 2000–2016

Abstract

Objective

We aimed to assess the association between meditation practice and cognitive function over time among middle-aged and older adults.

Method

We included Health and Retirement Study (HRS) participants assessed for meditation practice in the year 2000 as part of the HRS alternative medicine module (n = 1,160) and were followed up for outcomes over 2000–2016 period. We examined the association between meditation ≥ twice a week vs none/less frequent practice and changes in the outcomes of recall, global cognitive function, and quantitative reasoning using generalized linear regression models. Stratified analyses among persons with/without self-reported baseline depressive symptoms were conducted to assess the link between meditation and cognitive outcomes.

Results

Among our full study sample, meditation ≥ twice a week was not significantly associated with total recall [$\beta$; 95% CI: −0.97, 0.57; p = 0.61], global cognitive function [$\beta$; 95% CI: −1.01, 1.12; p = 0.92], and quantitative reasoning [$\beta$; 95% CI: −31.27, 8.32; p = 0.26]. However, among those who did not have self-reported depressive symptoms at baseline, meditation ≥ twice a week was associated with improvement in cognitive outcomes such as total recall [$\beta$; 95% CI: 0.03, 0.18; p = 0.01] and global cognitive function [$\beta$; 95% CI: 0.05, 0.40; p = 0.01] over time.

Conclusions

Frequent meditation practice might have a protective effect on cognitive outcomes over time, but this protection could be limited to those without self-reported baseline depressive symptoms. Future studies could incorporate more precise meditation practice assessment, investigate the effect of meditation on cognitive outcomes over time, and include more rigorous study designs with randomized group assignment.

Pre-registration

This study is not preregistered.

Cognitive decline impacts health-related quality of life (Roehr et al., 2017), affects the ability to function independently (Njegovan et al., 2001), increases the risk of hospitalization (Chodosh et al., 2004), and mortality (Lavery et al., 2009). Lifestyle interventions such as meditation have been employed to promote cognitive health (Chen et al., 2019; Laditka et al., 2012). Meditation practice has been defined by Cardoso et al. (2004) as a set of essential characteristics: being self-induced, involving specific well-defined techniques (e.g., regulating attention and posture), cultivating self-focus, using muscle relaxation during some part of the process, and logic relaxation (relaxing intentions to analyze, judge, or hold on to any expectations during the process of meditating). The pathways through which meditation practice may influence cognitive function include biological aging biomarkers, brain structure and function, and coping with negative emotions (Berk et al., 2017).

Most studies investigating the association of meditation with cognitive outcomes focus on adults aged 18 and above (Cásedas et al., 2020) making it unclear whether the cognitive health benefits may extend to the older population aged ≥ 65 years (Sperduti et al., 2017). Preliminary evidence exists suggesting that meditation practice among older adults aged 65 years and over may decelerate age-related cognitive decline (Gard et al., 2014a). Also, meditation may enhance some domains of cognitive function among older adults (Moynihan et al., 2013). The previous studies focusing on older adults have several limitations: Most studies recruit persons with cognitive impairment (Berk et al., 2018; Fam et al., 2020; Innes et al., 2018, 2021; Lenze et al., 2014), or residents of assisted living facilities (Pandya, 2020). Secondly, many studies are cross-sectional analyses or interventional studies that investigated the change in cognitive outcomes in the short-term periods (outcomes measured immediately post-intervention or after a few months of follow-up time) and the results have been inconsistent (Malinowski & Shalamanova, 2017). Therefore, more evidence is needed to understand the long-term effect of meditation on cognitive function (Farhang et al., 2019; Klimecki et al., 2019), especially among healthy older adults.

Depression is associated with cognitive decline (Dickinson et al., 2011; Koenig et al., 2014; Saenz et al., 2020; Sullivan et al., 2013). Previous studies among older adults have suggested an association between having depression at baseline and cognitive decline as compared to persons without depression at baseline (Dickinson et al., 2011; Saenz et al., 2020). Even within disease subgroups such as diabetes, experiencing depression has been associated with greater cognitive declines across all domains of cognition, treatment groups, and within all participant subgroups (Sullivan et al., 2013). Studies have suggested that a history of having depression and changes in depression status over time may modify the level of health-related benefits obtained through the practice of lifestyle interventions (Pischke et al., 2010), including meditation interventions (Guglani et al., 2012). Therefore, it is important to explore whether meditation practice improves cognitive outcomes separately among persons with and without depression.

In the current study, we aimed to explore the effect of meditation practice on cognitive health outcomes using a sub-sample from the longitudinal Health and Retirement Study (HRS) to understand whether meditation may be associated with changes in cognitive outcomes among healthy middle-aged and older adults in the long term. Many meditation interventions incorporate a frequency of practice of at least twice per week to experience the benefits of the practice (Kim et al., 2013; Prakhinkit et al., 2014; Robert-McComb et al., 2015; Teut et al., 2013). Therefore, we investigated the effect of practicing meditation at a frequency of 2–3 times per week on the cognitive outcomes. Our secondary aim was to investigate whether depressive symptoms at baseline may modify the effect of meditation on the cognitive outcomes.

Method

Participants

This study included a sub-sample of persons from the HRS interview sample in 2000 (Health & Retirement Study, 2004). The HRS is sponsored by the National Institute on Aging (grant number NIA U01AG009740) and is conducted by the University of Michigan. The HRS sample in the year 2000 (n = 42,052) is representative of the US population in 2000, born in year 1947 or earlier. The HRS interview is conducted biennially and includes: (1) Core interview questions that are consistent across the years; (2) Additional Modules on other health-related topics not covered under the core interview questions. The topics for the additional modules vary across years. All participants from the HRS sample receive the core interview and any one module included for a particular year (selection for any module is a random subsample of the total HRS sample) (Health & Retirement Study, 2004). In 2000, the HRS interview included an additional “Alternative Medicine” module. A total of 1,462 persons were assigned to the “Alternative Medicine” module, of which, 1,160 participants responded. These 1,160 respondents were included as our study sample.

Procedure

This was a longitudinal study involving secondary data analysis using the HRS data from years 2000 through 2016 for our study sample. This study used the HRS data products (Health & Retirement Study, 2004, 2020). The RAND HRS longitudinal data is based on the core interview data and includes cognitive health outcomes. For the purpose of this study, the “Alternative Medicine” module of the HRS administered in 2000 was merged with the RAND HRS longitudinal data for the years 2000 through 2016.

Measures

This study analyzed three cognitive outcomes, viz., total recall, global cognitive function, and quantitative reasoning which are available in the HRS longitudinal data. For each of these cognitive outcomes, higher scores indicate better functioning with respect to the outcome.

Total recall (Lekhak et al., 2020; Ofstedal et al., 2005) was calculated using a summary score for the immediate word recall and delayed word recall tasks. These tasks required participants to recall a list of ten nouns. The range of possible scores for total recall is 0–20. Total recall was measured biennially over nine time points between years from 2000 through 2016 in the HRS. The average number of repeated measures for total recall per participant was 6.07 (SD = 2.88, min = 1, max = 9).

Global cognitive function (G. Kim et al., 2019; Ofstedal et al., 2005) was calculated using a summary score of the total recall outcome and mental status index (which includes counting, naming, and vocabulary tasks). The range of possible scores for global cognitive function is 0–35. In the HRS, global cognitive function was measured biennially over nine time points between years from 2000 through 2016 among the ≥ 65 years age group participants. The average number of repeated measures for global cognitive function per participant in the ≥ 65 years age group was 5.12 (SD = 2.83, min = 1, max = 9).

Quantitative reasoning is a specific facet of fluid intelligence (Paul et al., 2016) which relates to the ability to reason using concepts based on mathematical relationships. In the HRS, quantitative reasoning is measured using a number series instrument and the calculated scores are scaled to be comparable to the W scores in the Woodcock-Johnson III (WJIII) test battery. The quantitative reasoning measure was introduced in the HRS interview in addition to crystallized intelligence measures only in the years 2010, 2012, and 2016 within our study timeframe. The average number of repeated measures for quantitative reasoning per participant was 0.87 (SD = 1.15, min = 0, max = 3). Note that n = 348 of the study sample was lost due to mortality by 2010.

We also included secondary analyses of some of the individual components of cognitive function as separate outcomes: immediate recall (score ranges 0–10), delayed recall (score ranges 0–10), and mental status (scores range 0–15).

The primary predictor of interest was self-reported meditation practice, operationalized as a binary variable in our analyses: (1) ≥ twice a week and (0) none/less frequent practice (Kim et al., 2013; Prakhinkit et al., 2014; Robert-McComb et al., 2015; Teut et al., 2013). We created this variable using data from two questions participants were asked in the survey in 2000 about whether they had ever engaged in meditation practice (responses measured in a “yes”/ “no” format), and what was the frequency of their meditation practice (measured using five response categories: Daily, ≥ twice a week, ≥ once a week, ≥ once a month, < once a month). Our secondary predictors were: (1) self-reported breathing exercise, and (2) self-reported personal prayer practice outside of religious services attendance. Breathing exercises include a wide range of techniques such as pacing the breath, actively controlling the breath, or allowing the breath to flow naturally (Ospina et al., 2007). Breathing exercises are considered part of meditation practice (Ospina et al., 2007). Prayer, as defined by Jors et al. (2015), involves a conscious effort to open oneself to and establish a connection with a higher being. Each of the secondary predictors were measured and operationalized in the same format as our primary predictor.

Our analyses adjusted for the following self-reported measures: (1) participants’ age at the midpoint of the baseline interview period, (2) gender, (3) race/ethnicity [Non-Hispanic (NH) White, NH Black, NH other, Hispanic], (4) relationship status [with partner, no partner], (5) education level [< High School, Highschool, Some College, ≥ Bachelor’s degree], (6) annual household income [< $20,000; $20,000 to < $35,000; $35,000 to < $55,000; $55,000 to < $100,000; ≥ $100,000]) any attendance at religious services (yes, no), (8) comorbidity index, and (9) depressive symptoms score. The HRS interview collects information on high blood pressure, diabetes, cancer, lung disease, heart disease, stroke, psychiatric problems, and arthritis. The comorbidity index was a count of the number of comorbidities ever reported, including information reported in the baseline year 2000 as well as waves of interviews prior to year 2000. The HRS interview measures depressive symptoms using items from the Center for epidemiological Studies Depression Scale (8-item CES-D) (Karim et al., 2015). The score range is 0 to 8, indicating the number of self-reported depressive symptoms. Reference groups were males, non-Hispanic White, having no partner, annual household income < $20,000, less than high school education, and no religious service attendance. For each of the predictors, the none or less frequent practice group was the reference.

Data Analyses

Descriptive analysis was conducted for the analytic sample. Means and SD for continuous variables and frequencies and percentages for categorical variables were included. Relationship status and annual household income were treated as time varying variables. Age at baseline, gender, race, years of education, baseline comorbidity index, and depression score at baseline were treated as time-invariant variables. We used generalized linear latent and mixed-effects models (GLLAMM) with random intercept (Rabe-Hesketh et al., 2002, 2005) to investigate the association of the predictors with the outcomes. The linear effect of time (defined as duration from 2000 through 2016) on each of the outcomes was explored. Each model included the primary predictor, secondary predictors, and were adjusted for all covariates. The interaction terms included in the models were time*meditation ≥ twice a week, time*breathing exercise ≥ twice a week, time*personal prayer ≥ twice a week, and baseline depression score*meditation ≥ twice a week.

In the GLLAMM models, the measurement occasions were the level 1 units, and the individuals were the level 2 units. The following is an illustration of the model specification for the total sample analyses, wherein, is the outcome and $e_{\mathrm{i},\mathrm{j}}$ is the error term at the i^th year for the j^th individual, ${\beta }_0$ through ${\beta }_{23}$ are the level 1 regression coefficients (the fixed-effects), ${\beta }_{0,\mathrm{j}}$ is the random intercept for the j^th individual:

$$\begin{gathered} {\mathrm{Y}}_{\mathrm{i},\mathrm{j}}=({\beta }_0+b_{0,j})+{\beta }_1{\text{Meditation}}_{\mathrm{i},\mathrm{j}}+{\beta }_2{\text{Breathing}}_{\mathrm{i},\mathrm{j}}+{\beta }_3{\text{Prayer}}_{\mathrm{i},\mathrm{j}} \\ +{\beta }_4{\text{Time}}_{i,j}+{\beta }_5{\text{Religious Service Attendance}}_{\mathrm{i},\mathrm{j}}+{\beta }_6{\text{Relationship Status}}_{\mathrm{i},\mathrm{j}} \\ +{\beta }_7{\text{Gender}}_{i,j}+{\beta }_8\text{Race Category}{\mathrm{A}}_{\mathrm{i},\mathrm{j}}+{\beta }_9\text{Race Category}{\mathrm{B}}_{\mathrm{i},\mathrm{j}} \\ +{\beta }_{10}\text{Race Category}{\mathrm{C}}_{i,j}+{\beta }_{11}\text{Education Level}{\mathrm{A}}_{\mathrm{i},\mathrm{j}}+{\beta }_{12}\text{Education Level}{\mathrm{B}}_{\mathrm{i},\mathrm{j}} \\ +{\beta }_{13}\text{Education Level}{\mathrm{C}}_{i,j}+{\beta }_{14}\text{Income Level}{\mathrm{A}}_{\mathrm{i},\mathrm{j}}+{\beta }_{15}\text{Income Level}{\mathrm{B}}_{\mathrm{i},\mathrm{j}} \\ +{\beta }_{16}\text{Income Level}{\mathrm{C}}_{i,j}+{\beta }_{17}\text{Income Level}{\mathrm{D}}_{\mathrm{i},\mathrm{j}}+{\beta }_{18}{\text{Age}}_{\mathrm{i},\mathrm{j}}+{\beta }_{19}{\text{Number of Comorbidities at baseline}}_{i,j} \\ +{\beta }_{20}{\text{Meditation}}_{i,j}\ast {\text{Time}}_{i,j}+{\beta }_{21}{\text{Breathing}}_{i,j}\ast {\text{Time}}_{\mathrm{i},\mathrm{j}}+{\beta }_{22}{\text{Prayer}}_{\mathrm{i},\mathrm{j}}\ast {\text{Time}}_{i,j} \\ +{\beta }_{23}{\text{Meditation}}_{i,j}\ast {\text{Baseline Depression Score}}_{i,j}+e_{i,j} \end{gathered}$$

Since depression is known to be associated with cognitive function, we conducted two subgroup analyses: (1) Non-depressed, i.e. participants with depression score = 0 at baseline; (2) Depressed, i.e., participants with depression score ≥ 1 at baseline (≥ 1 baseline depressive symptoms was used as a cut-off point to capture persons at risk of mild depression or more (Doshi et al., 2008)). We conducted weighted analyses using inverse probability weighting to adjust for selection biases including bias due to treatment selection, and non-response due to mortality or attrition for other reasons. The weight used in our analyses was the product of the inverse probabilities for (1) selection for meditation practice ≥ twice a week, (2) selection for breathing practice ≥ twice a week, (3) selection for personal prayer practice ≥ 2 twice a week, (4) mortality at each wave if alive in the previous wave, and (5) attrition due to other reasons at each wave if alive in the previous wave. We have reported two decimal places for figures in all results unless the value was less than 0.005. For p-values, we also reported more than two decimal places if the value was between 0.045 and < 0.05 since a p-value of less than 0.05 was defined as statistically significant. Analyses were performed using STATA 13 (StataCorp LLC, College Station, TX).

Results

Our descriptive results for characteristics of our study sample (Table 1) indicated that a majority of the study participants were female (58.02%), non-Hispanic White (79.22%), having a partner (67.59%), with high school level of education (34.40%), and household income < $20,000 (30.60%). The population had an average age of 66.99 (SD = 10.50) years, with an average comorbidity index of 1.71 (SD = 1.36), and mean baseline depression score of 1.51 (SD = 1.85). In our study population, 23.45% of the participants reported meditation practice ≥ twice a week, 13.36% reported breathing exercise practice ≥ twice a week, and 72.93% reported personal prayer practice ≥ twice a week. As compared to the none/less frequent meditation practice group, the group with meditation practice ≥ twice a week had a higher proportion of females (62.13% vs 56.76%), persons with household income < $20,000 (37.87% vs 28.38%), and a lower proportion of NH-White population (68.38% vs 82.55%), persons with partner (61.76% vs 69.37%), and high school level education (31.62% vs 35.25%).

Table 1.

Baseline characteristics of population who attempted the HRS 2000 alternative medicine module

		Total			None/ < 2 times per week meditation practice		≥ 2–3 times per week meditation practice
			n	%	n	%	n	%	p
			1160	100	888	76.55	272	23.45
Gender	Male		487	41.98	384	43.24	103	37.87	0.12
	Female		673	58.02	504	56.76	169	62.13
Race	NH White		919	79.22	733	82.55	186	68.38	< 0.001
	NH Black		143	12.33	81	9.12	62	22.79
	NH Other		14	1.21	11	1.24	3	1.1
	Hispanic		84	7.24	63	7.09	21	7.72
Relationship status	No partner		376	32.41	272	30.63	104	38.24	0.02
	With partner		784	67.59	616	69.37	168	61.76
Education level	< High School		314	27.07	233	26.24	81	29.78	0.60
	Highschool		399	34.4	313	35.25	86	31.62
	Some College		210	18.1	162	18.24	48	17.65
	≥ Bachelor's deg.		237	20.43	180	20.27	57	20.96
Household income	< $20k		355	30.6	252	28.38	103	37.87	0.05
	$20k to < $35k		237	20.43	191	21.51	46	16.91
	$35k to < $55k		229	19.74	181	20.38	48	17.65
	$55k to < $100k		207	17.84	160	18.02	47	17.28
	≥ $100k		132	11.38	104	11.71	28	10.29
Breathing exercises	< twice a week		1005	86.64	804	90.54	201	73.9	< 0.001
	≥ twice a week		155	13.36	84	9.46	71	26.1
Personal prayer	< twice a week		314	27.07	291	32.77	23	8.46	< 0.001
	≥ twice a week		846	72.93	597	67.23	249	91.54
Attend religious services	No		307	26.49	260	29.31	47	17.28	< 0.001
	Yes		852	73.51	627	76.53	225	82.72
Baseline depression symptoms	None		465	40.12	364	40.99	101	37.27	0.27
	≥ 1		694	59.88	524	59.01	170	62.73
		n	M	SD	M	SD	M	SD
Age		1160	66.99	10.50	66.73	10.44	67.87	10.69	0.12
Education years		1160	12.22	3.27	12.23	3.25	12.18	3.34	0.83
Baseline depression symptoms scorea		1159	1.51	1.85	1.45	1.79	1.71	2.01	0.06
Total recall indexb		1160	9.88	3.76	9.91	3.74	9.78	3.84	0.61
Comorbidity indexa		1160	1.71	1.36	1.66	1.30	1.88	1.55	0.04
Total cognition scorec		632	21.50	5.32	21.64	5.24	21.08	5.54	0.24
Mental status indexd		632	12.68	2.46	12.85	2.42	12.16	2.50	0.002
Immediate recall indexe		1160	5.48	1.81	5.49	1.80	5.41	1.84	0.51
Delayed recall indexe		1160	4.41	2.17	4.42	2.15	4.37	2.21	0.73
Quantitative reasoning		628	491.04	43.02	493.19	42.48	483.80	44.16	0.02

Higher score to be interpreted as better for all the cognitive outcomes. Range of possible scores: ^a0-8; ^b0-20; ^c0-35; ^d0-15; ^e0-10. All variables in this table include values for year 2000, except for the quantitative reasoning outcome which includes year 2010 values, which was the first occasion when it was measured. Non-Hispanic (NH)

Our study had a 16-year timeframe and the average number of non-responding participants who were alive in each follow-up wave was 60.88 (SD = 10.82). The average number of participants who died in each follow-up wave was 67.88 (SD = 11.92). By 2016, 48.97% (n = 568) of our study population was alive, and 4.22% (n = 49) had dropped out of the sample.

The results of our longitudinal analyses are presented in Table 2.

Table 2.

Longitudinal analyses results for association of complementary health practices with cognitive outcomes

	Total sample				Non-depressed Subgroup				Depressed Subgroup
	Unweighted		Weighted		Unweighted		Weighted		Unweighted		Weighted
	$\beta$	p	$\beta$	p	$\beta$	p	$\beta$	p	$\beta$	p	$\beta$	p
	Outcome: Total Recall Index
	n = 1158		n = 1150		n = 464		n = 459		n = 694		n = 691
Meditation	−0.40	0.09	−0.20	0.61	−0.49	0.14	−0.73	0.09	−0.17	0.62	0.06	0.92
Meditation*time	0.03	0.06	0.01	0.65	0.07	0.004 **	0.11	0.01	−0.001	0.97	−0.04	0.31
Breathing	0.40	0.08	0.60	0.047 *	0.74	0.07	0.69	0.10	0.30	0.35	0.63	0.16
Breathing*time	−0.01	0.68	−0.01	0.69	−0.05	0.15	−0.05	0.39	0.02	0.36	0.02	0.66
Personal prayer	0.36	0.049 *	0.14	0.54	0.49	0.07	0.20	0.66	0.32	0.18	0.20	0.51
Personal prayer*time	−0.04	0.003 *	−0.001	0.97	−0.01	0.78	−0.03	0.46	−0.07	< .001 ***	−0.01	0.70
time	−0.13	< .001 ***	−0.16	< .001 ***	−0.15	< .001	−0.15	< .001 ***	−0.10	< .001 ***	−0.15	< .001 ***
Baseline depression score	−0.21	< .001 ***	−0.22	< .001 ***					−0.19	0.001 **	−0.20	0.02 *
Meditation*baseline depressions score	0.12	0.11	0.08	0.52					0.13	0.18	0.05	0.80
	Outcome: Total Cognition Score (age > = 65)
	n = 634		n = 628		n = 251		n = 247		n = 383		n = 382
Meditation	−0.20	0.57	0.05	0.92	−1.02	0.17	−1.41	0.04 *	0.76	0.15	−0.27	0.73
Meditation*time	0.07	0.02 *	0.07	0.25	0.14	0.003 **	0.22	0.01 *	0.02	0.56	0.02	0.75
Breathing	0.27	0.52	0.51	0.38	0.04	0.93	0.50	0.48	0.26	0.60	−1.26	0.24
Breathing*time	−0.06	0.12	−0.11	0.19	−0.14	0.01 *	−0.13	0.20	0.01	0.86	0.09	0.23
Personal Prayer	0.82	0.02 *	−0.08	0.88	1.16	0.02 *	0.38	0.61	0.39	0.36	0.50	0.66
Personal prayer*time	−0.09	0.004 **	−0.005	0.94	−0.05	0.26	−0.12	0.20	−0.11	0.01 *	−0.03	0.69
Time	−0.31	< .001 ***	−0.41	< .001 ***	−0.29	< .001 ***	−0.35	< .001 ***	−0.32	< .001 ***	−0.45	< .001 ***
Baseline depression score	−0.25	0.01 *	−0.30	0.01 *					−0.21	0.04 **	−0.32	0.03 **
Meditation*baseline depression score	0.33	0.01 *	0.39	0.01 *					0.20	0.21	0.76	0.01 *
	Outcome: Quantitative Reasoning (years 2010, 2012, and 2016)
	n = 668		n = 661		n = 303		n = 299		n = 365		n = 362
Meditation	−13.62	0.14	−11.48	0.26	−3.64	0.77	3.21	0.84	−24.63	0.08	−23.74	0.10
Meditation*time	0.94	0.19	0.80	0.24	0.10	0.91	−0.34	0.74	1.59	0.13	1.56	0.11
Breathing	−18.43	0.10	−23.71	0.06	1.63	0.92	5.94	0.73	−25.47	0.11	−34.89	0.07
Breathing*time	1.07	0.23	1.54	0.10	0.47	0.71	−0.03	0.98	1.03	0.41	1.78	0.24
Personal prayer	−6.92	0.40	−14.35	0.17	−5.76	0.60	−22.88	0.19	−9.40	0.43	−21.30	0.18
Personal prayer*time	0.25	0.70	0.62	0.38	0.43	0.62	1.13	0.33	0.15	0.87	0.83	0.46
Time	2.82	< .001 ***	2.52	< .001 ***	2.62	< .001 ***	2.63	0.02 *	3.02	< .001 ***	2.22	0.03 *
Baseline depression score	−2.76	0.001 **	−2.50	0.03 *					−2.58	0.02 *	−1.49	0.36
Meditation*baseline depression score	1.89	0.18	0.74	0.66					3.00	0.13	2.16	0.33
	Outcome: Immediate Recall Index
	n = 1158		n = 1150		n = 464		n = 459		n = 694		n = 691
Meditation	−0.12	0.28	−0.09	0.54	−0.27	0.046 *	−0.42	0.02 *	0.01	0.97	0.08	0.74
Meditation*time	0.01	0.05*	0.01	0.50	0.03	0.02 *	0.05	0.01 *	0.002	0.85	−0.02	0.39
Breathing	0.19	0.07	0.24	0.07	0.43	0.01 *	0.33	0.08	0.09	0.49	0.22	0.21
Breathing*time	−0.003	0.75	−0.01	0.64	−0.02	0.24	−0.01	0.69	0.01	0.46	0.01	0.52
Personal prayer	0.10	0.23	−0.01	0.96	0.18	0.17	0.03	0.85	0.09	0.43	0.01	0.95
Personal prayer*time	−0.01	0.04 *	0.01	0.68	0.001	0.91	−0.01	0.68	−0.03	0.004 **	0.003	0.86
time	−0.06	< .001 ***	−0.08	< .001 ***	−0.07	< .001 ***	−0.08	< .001 ***	−0.06	< .001 ***	−0.08	< .001 ***
Baseline depression score	−0.10	< .001 ***	−0.11	< .001 ***					−0.09	< .001 ***	−0.10	0.004 **
Meditation*baseline depression score	0.03	0.32	0.03	0.57					0.02	0.63	−0.001	0.99
	Outcome: Delayed Recall Index
	n = 1158		n = 1150		n = 464		n = 459		n = 694		n = 691
Meditation	−0.07	0.58	0.002	0.99	−0.23	0.20	−0.38	0.06	0.01	0.97	0.38	0.59
Meditation*time	0.02	0.06	0.005	0.73	0.04	0.01 *	0.06	0.01 *	−0.001	0.93	−0.02	0.27
Breathing	0.24	0.09	0.32	0.06	0.48	0.04 *	0.38	0.13	0.08	0.63	0.38	0.15
Breathing*time	−0.005	0.66	−0.01	0.69	−0.02	0.20	−0.04	0.34	0.01	0.59	0.004	0.87
Personal prayer	0.23	0.03 *	0.18	0.21	0.26	0.09	0.15	0.52	0.18	0.23	0.22	0.39
Personal prayer*time	−0.03	0.002 **	−0.01	0.67	−0.01	0.55	−0.03	0.34	−0.04	< .001 ***	−0.01	0.46
time	−0.06	< .001 ***	−0.08	< .001 ***	−0.08	< .001 ***	−0.07	0.002 **	−0.05	< .001 ***	−0.07	< .001 ***
Baseline depression score	−0.10	< .001 ***	−0.10	0.003 **					−0.10	0.01 *	−0.09	0.04 *
Meditation*baseline depression score	0.02	0.57	0.01	0.89					0.04	0.54	−0.08	0.67
	Outcome: Mental Status Index (age > = 65)
	n = 634		n = 628		n = 251		n = 247		n = 383		n = 382
Meditation	−0.52	0.02 *	−0.42	0.29	−0.57	0.03 *	−0.21	0.41	−1.09	0.002 **	−1.13	0.04 *
Meditation*time	0.04	0.01 *	0.03	0.46	0.03	0.17	0.05	0.16	0.04	0.09	0.02	0.74
Breathing	−0.04	0.85	−0.40	0.29	0.48	0.10	−0.53	0.18	−0.10	0.71	−0.41	0.28
Breathing*time	−0.02	0.25	−0.07	0.17	−0.03	0.26	−0.06	0.29	−0.01	0.70	0.06	0.08
Personal prayer	0.19	0.26	−0.34	0.29	0.25	0.33	0.34	0.31	0.23	0.31	−0.23	0.54
Personal prayer*time	−0.03	0.10	−0.01	0.77	−0.02	0.33	−0.04	0.37	−0.03	0.20	−0.05	0.11
Time	−0.12	< .001 ***	−0.15	< .001 ***	−0.09	< .001 ***	−0.12	0.01 *	−0.14	< .001 ***	−0.16	< .001 ***
Baseline depression score	−0.12	0.01 *	−0.12	0.07					−0.15	0.01 *	−0.11	0.04 *
Meditation*baseline depression score	0.22	0.01 *	0.33	0.01 *					0.34	0.001 **	0.44	< .001 ***

"≥ twice a week" of meditation practice compared to the "none or < twice a week" meditation practice (reference group). " ≥ twice a week" of breathing practice compared to the "none or < twice a week" breathing practice (reference group). " ≥ twice a week" of personal prayer practice compared to the "none or < twice a week" personal prayer practice (reference group). The non-depressed and depressed subgroups based on baseline depression score = 0 and baseline depression score ≥ 1, respectively

^*0.01 < = p < 0.05

^**0.001 < = p < 0.01

^***p < 0.001

Total Sample Analyses

Meditation ≥ twice a week was not significantly associated with any of the outcomes: total recall [$\beta$; 95% Confidence Interval (CI): −0.97, 0.57; p = 0.61], global cognitive function [$\beta$; 95% CI: −1.01, 1.12; p = 0.92], and quantitative reasoning [$\beta$; 95% CI: −31.27, 8.32; p = 0.26]. Also, meditation ≥ twice a week was not associated with changes in the outcomes over time: total recall [$\beta$; 95% CI: −0.04, 0.06; p = 0.65], global cognitive function [$\beta$; 95% CI: −0.05, 0.19; p = 0.25], and quantitative reasoning [$\beta$; 95% CI: −0.54, 2.15; p = 0.24]. Among persons with none/less frequent meditation practice, higher level of baseline depression was adversely associated with total recall [$\beta$; 95% CI: −0.34, −0.11; p < 0.001], global cognitive function [$\beta$; 95% CI: −0.51, −0.09; p = 0.01], quantitative reasoning [$\beta$; 95% CI: −4.80, −0.21; p = 0.03], immediate recall [$\beta$; 95% CI: −0.17, −0.05; p < 0.001] and delayed recall [$\beta$; 95% CI: −0.16, −0.03; p = 0.003]. However, meditation ≥ twice a week at higher levels of baseline depression had a protective effect on global cognitive function [$\beta$; 95% CI: 0.12, 0.65; p = 0.01] and mental status [$\beta$; 95% CI: 0.07, 0.58; p = 0.01]. With respect to the secondary predictors, breathing exercise ≥ twice a week was associated with higher total recall [$\beta$; 95% CI: 0.01, 1.20; p = 0.047]. But, both breathing exercise ≥ twice a week and personal prayer ≥ twice a week were not associated with any changes in the outcomes over time.

Non-Depressed Subgroup

Meditation ≥ twice a week was associated with lower global cognitive function [$\beta$; 95% CI: −2.77, −0.06; p = 0.04] and immediate recall [$\beta$; 95% CI: −0.76, −0.08; p = 0.02]. However, there was a significant protective effect of meditation ≥ twice a week on changes over time in total recall [$\beta$; 95% CI: 0.03, 0.18; p = 0.01], global cognitive function [$\beta$; 95% CI: 0.05, 0.40; p = 0.01], immediate recall [$\beta$; 95% CI: 0.01, 0.08; p = 0.01], and delayed recall [$\beta$; 95% CI: 0.02, 0.10; p = 0.01]. The secondary predictors, breathing exercise ≥ twice a week and personal prayer ≥ twice a week were not associated with any outcomes or changes in the outcomes over time.

Depressed Subgroup

Meditation ≥ twice a week was associated with lower mental status index [$\beta$; 95% CI: −2.22, −0.05; p = 0.04] on average as compared to the none/less frequent meditation practice group. Among persons with none/less frequent meditation practice, higher level of baseline depression was adversely associated with total recall [$\beta$; 95% CI: −0.36, −0.04; p = 0.02], global cognitive function [$\beta$; 95% CI: −0.60, −0.03; p = 0.03], immediate recall index [$\beta$; 95% CI: −0.17, −0.03; p = 0.004], delayed recall index [$\beta$; 95% CI: −0.17, −0.01; p = 0.04], and mental status [$\beta$; 95% CI: −0.21, −0.004; p = 0.04]. However, meditation ≥ twice a week at higher levels of baseline depression had a protective effect on global cognitive function [$\beta$; 95% CI: 0.21, 1.32; p = 0.01], and mental status index [$\beta$; 95% CI: 0.22, 0.65; p < 0.001]. The secondary predictors, breathing exercise ≥ twice a week and personal prayer ≥ twice a week, were not associated with any outcomes or changes in the outcomes over time. See Supplementary Tables S1 and S2 for additional analyses using a 2-year timeframe and cross-sectional analyses.

Discussion

This study is among the few studies that investigate the influence of complementary health practices such as meditation on cognitive outcomes after a considerably long follow-up period in a US sample population with a majority of older adults. In addition, this study adds to the literature about how having reported depressive symptoms may modify the effectiveness of complementary health practices for cognitive health.

The protective effect of meditation ≥ twice a week over time on some cognitive outcomes such as total recall (including immediate and delayed recall) and global cognitive function was observed in the non-depressed subgroup but not in the depressed subgroup. However, some preliminary evidence suggests that mindfulness-based interventions such as Mindfulness-based Cognitive Therapy (MBCT) may be associated with improved overall cognitive function, and cognitive flexibility, among persons with mild to severe depressive symptoms (Shapero et al., 2018). A prior study found some benefits of meditation for immediate word recall among older adults with anxiety and depressive symptoms (Wetherell et al., 2017).

Prior studies have focused on specific domains of cognition such as executive function or memory or on overall performance including all domains of cognition. Among adults, meditation has been associated with a positive influence on executive control and working memory (Cásedas et al., 2020). Even among adults 60–85 years with mild cognitive impairment, meditation has been linked to positive cognitive outcomes (Yu et al., 2021). Among middle-aged and older adults with memory complaints but no confirmed diagnosis of cognitive disorder, meditation was associated with improvements in self-reported memory function (Berk et al., 2018). Studies among older adults with some levels of cognitive impairment have found meditation was associated with positive changes in memory and cognitive performance (Innes et al., 2018), subjective memory (Innes et al., 2021), and immediate and delayed word recall (Lenze et al., 2014), and overall cognitive outcomes (Fam et al., 2020). Among assisted living facility residents, meditation was linked with a positive effect on cognitive outcomes (Pandya, 2020). Among older adults with anxiety and depressive disorders, meditation was associated with improvements on immediate word recall, but no other cognitive domains (Wetherell et al., 2017). It is possible that meditation may not influence all domains of cognition uniformly as reflected in the results of our study. A study among healthy older adults with no cognitive complaints found a small but significant benefit of meditation on executive function (Moynihan et al., 2013), and another study showed no benefit on executive function (Mallya & Fiocco, 2016). Psychological stressors such as anxiety and depression may moderate the effectiveness of meditation interventions on cognitive outcomes, and this may provide an explanation for inconsistencies in reports of meditation efficacy.

The results of our analyses in our total sample and within the depressed subgroup did not reflect these benefits of meditation practice on cognitive outcomes over time as suggested in prior studies. This inconsistency between our study and prior studies is possibly because the length of follow-up time included in our analyses extends to 16 years. The consistency of meditation practice over time is important to sustain the benefits. In our study, meditation practice may not have been maintained in the long term. Also, we did not restrict our population to persons with cognitive health issues. Our study findings for ≥ twice a week of meditation practice compared to no/less-frequent practice are consistent with a prior study using the same data (exploring outcomes changes over age, not over waves of interview) and a lower cut-off for defining meditation practice wherein no difference was found in the episodic memory (total recall) between those who reported ever engaging in meditation compared with those who did not (Lekhak et al., 2020).

Our results suggested that breathing exercise ≥ twice a week practice compared to none/less-frequent practice was associated with higher total recall on average. The patterns of change over time, however, were not significantly different between the two comparison groups. Few studies have explored the link between breathing exercises and cognitive outcomes. Prior studies have explored the effect of breathing exercises on spatial and verbal memory among school children (Manjunath & Telles, 2004; Naveen et al., 1997), and among adult males (Joshi & Telles, 2008) and found that breathing exercises were associated with spatial memory but not verbal memory (Joshi & Telles, 2008; Manjunath & Telles, 2004; Naveen et al., 1997). Our memory measures were verbal, and therefore, our findings for association of breathing exercises with memory (recall) in the long term are consistent with prior studies in populations different from ours. On the other hand, a study among medical students found that a 6-week breathing exercise intervention was associated with improved memory (Chandla et al., 2013). The reasons for our results being different from this study (Chandla et al., 2013) could be the differences in the type of population, intensity and type of breathing exercises practiced.

There was no significant difference in any of the outcomes or change in the outcomes for personal prayer ≥ twice a week vs none/less frequent practice. While a lot of interest has focused on the neural processes that are activated or areas of the brain that are recruited during the act of praying (Schjoedt et al., 2009; van Elk & Aleman, 2017), few studies have investigated the effect of engaging in praying on performance at specific cognitive domain tasks. The prior study by Lekhak et al. (2020) found that personal prayer had significant protective effects for Total Recall index (episodic memory) at increased age. Another study using later waves of HRS data (2006–2012), wherein personal prayer frequency was measured in 2006, found that there was an initial protective effect of frequency of personal prayer on episodic memory (total recall), however, there was no significant association of personal prayer frequency with changes in episodic memory over time (Kraal et al., 2019). In our analysis, the average effect of personal prayer practice on Total Recall Index in the total sample analyses leaned in the positive direction, but was not significant, and the effect of personal prayer practice on total recall index was not significant over time.

Prior studies have suggested that meditation-based practices such as mindfulness and yoga are positively correlated with fluid intelligence among middle-aged persons (Gard et al., 2014b). Meditation practices (which include breathing exercises) may correlate with fluid intelligence through the aspect of mindfulness, since mindfulness and fluid intelligence both relate to cognitive flexibility (Gard et al., 2014b). Prayer has been associated with cognitive outcomes in previous studies (Jeynes, 2020). Novel problem-solving capacity is the essence of fluid intelligence (Kent, 2017). Religious behavior including engagement in personal prayer is often used as a coping mechanism and plays an important role in the problem-solving processes (Pargament et al., 1988). Our results suggested no significant associations between quantitative reasoning and more frequent meditation-based practices over time. It is possible that the effect of baseline meditation practice did not sustain in the long term (after a minimum of 10 years, since the first measure of quantitative reasoning was available in 2010).

We observed a difference in the effect of meditation ≥ twice a week on cognitive outcomes by the level of baseline depressive symptoms. Among persons reporting none/less frequent meditation practice, increased levels of depressive symptoms at baseline had an adverse impact on cognitive outcomes. However, at increased levels of baseline depressive symptoms, meditation ≥ twice a week had a protective effect on the outcomes of global cognitive function and mental status. This protective effect of meditation ≥ twice a week at higher levels of baseline depressive symptoms was also observed within the depressed subgroup.

Prior literature provides evidence associating higher education level with better cognitive abilities across the lifespan (Lövdén et al., 2020). Also, education level may predict use of meditation practice (Cramer et al., 2016). Therefore, we recognized educational level as a potential confounder in the association of meditation and cognitive outcomes and controlled for it in all our analyses by including it as a covariate. Our results for education level are consistent with prior studies in suggesting that it predicts better cognitive abilities. In our results, as compared to those with high school education, global cognition, total recall, and quantitative reasoning performance was poorer for those with less than high school education, but better for those with a bachelor’s degree or higher. Those with some college had better global cognition, and quantitative reasoning as compared to those with high school level education. Although total recall did not significantly differ between some college and high school levels of education, delayed recall was significantly higher in the group with some college as the highest completed level of education as compared to the high school level group.

Limitations and Future Directions

There are several limitations to this study: First, this is an observational study, and therefore, our results may be affected by selection bias due to attrition over time and bias due to differential mortality between groups. Besides attrition and mortality, sample size was also limited due to some of the outcomes being measured only within a restricted age group. However, we tried to address this concern through the inverse probability weighting approach. Second, since the analytic sample used in this study is a random sample of the baseline HRS participants in 2000, the generalizability of the study results to the US population is limited. Third, all our key independent variables were measured only at baseline. Therefore, we were not able to control for the time-varying influence of the predictors in our analyses. It is possible that the effect of complementary health practice only at baseline may not have lasted for the length of time analyzed in our study. Fourth, the dose of the practice is also important to consider in order to derive the benefit of meditation practice on cognitive outcomes (Jha et al., 2007). But we were unable to account the dose of meditation practice with accuracy due to the limited availability of information. Fifth, different types of mediation practices may influence different domains of cognition (Gard et al., 2014a), but we did not have information about meditation type in the data. Similarly, we did not have information on the kind of breathing exercise or personal prayer practice. Sixth, there is also a possibility of type-1 error due to the number of analyses conducted.

Future research could address some of the limitations in our study through more rigorous study designs, larger sample sizes, more reliable and frequent measurement of the key predictors, and specificity with respect to the type of exposure—whether it is meditation, breathing exercise, or personal prayer. Also, future studies could use different thresholds for baseline depression symptoms score for subgroup analyses. It was not within the scope of our study to investigate the mechanisms whereby meditation may influence cognitive outcomes. More research is needed to understand the mechanisms through which the various complementary health practices may influence the cognitive outcomes, especially through mental health factors such as depression and stress (Dickinson et al., 2011) and personal well-being.

Data Availability

This study used publicly available data from the Health and Retirement Study. Accessed from: https://hrs.isr.umich.edu/data-products.