Prediction of Physical Functioning and General Health Status Trajectories on Mortality Among Persons with Cognitive Impairment

doi:10.21203/rs.3.rs-1639214/v1

Download PDF

Research Article

Prediction of Physical Functioning and General Health Status Trajectories on Mortality Among Persons with Cognitive Impairment

https://doi.org/10.21203/rs.3.rs-1639214/v1

This work is licensed under a CC BY 4.0 License

You are reading this latest preprint version

Background

The concern posed by the confluence of aging and cognitive impairment is growing in importance as the U.S. population rapidly ages. As such, we sought to examine the predictive power of physical functioning (PF) and general health status (GHS) trajectories on mortality outcomes among persons with cognitive impairment (PCIs).

Methods

We used group-based trajectory models to identify latent group memberships for PF trajectories in 1,641 PCIs and GHS trajectories in 2,021 PCIs from the National Health and Aging Trends Survey (2011–2018) and applied logistic regressions to predict mortality using these memberships controlling for individual characteristics.

Results

We identified six trajectory groups for PF and four groups for GHS. Trajectory group memberships for both outcomes significantly predicted mortality. For PF, group memberships largely captured the average levels over time, and worse trajectories (i.e., lower baselines and faster declines) were associated with higher odds of death. The highest mortality risk was associated with the group experiencing a sharp decline early in its PF trajectory, although its average level across time was not the lowest. For GHS, we observed two groups with comparable average levels across time, but the one with a convex-shape trajectory had much higher mortality risks compared to the one with a concave-shape trajectory.

Conclusions

Our findings highlighted that health trajectories predicted mortality among PCIs, not only because of general levels but also because of the shapes of declines. Close monitoring health deterioration of PCIs is crucial to understand the health burden of this population and to make subsequent actions.

Cognitive Aging

Mortality

Physical Functioning

Health Trajectory

In the U.S. with a rapid increase of the aging population, cognitive impairment has become a major public health concern. Currently, more than 16 million people in the U.S. have various levels of cognitive impairment (1). Studies have established a strong link between cognitive impairment and mortality risks (2–4), as well as the associations between cognitive impairment and declines in physical functioning and general health status (5–7). In addition, prior literature shows that general health status and physical functioning are some of the strongest predictors of mortality in the general population of older adults (8–10). A handful of studies have also shown that self-rated health trajectories and physical activity trajectories significantly predict mortality among the general older adult population outside the U.S (11–13). However, few studies have examined whether and how physical functioning and general health status, particularly their trajectories, predict mortality among persons with cognitive impairment (PCIs) in the U.S.

It is important to study this topic especially among PCIs because health trajectories may be more complex in this population compared to the general aging population. For example, there is substantial heterogeneity in trajectories of general health status among PCIs and declines in general health status are more common in this population compared to the general aging population in the U.S. (14, 15). The substantially heterogeneous health trajectories may lead to substantially different mortality risks.

In addition, PCIs’ health trajectories can reveal a more complete picture of health status at different time points or at different stages of disease career that are masked by an average measure (11). For example, some PCIs may experience a greater variability in health status over time but have the same average health status as another group of PCIs. Differences in the shape of trajectories and the rates of change are therefore not captured by an average measure of health, and these differences could potentially have an independent effect on mortality later in life.

We fill the gap in literature by exploring the relationship between trajectories of general health status and physical functioning and later-life mortality among PCIs. Using a nationally-representative dataset of U.S. Medicare beneficiaries, we apply group-based trajectory models (GBTMs) to classify PCIs into distinct trajectory latent groups and estimate their trajectories (16). We then examine whether PCIs’ trajectory group memberships predict mortality, and whether this relationship is independent of a variety of sociodemographic and health characteristics.

Data and Measures

This study used data from the National Health and Aging Trends Study (NHATS) from 2011 to 2018. The NHATS is a nationally representative sample of Medicare beneficiaries aged 65 and above (17, 18). Respondents were selected using multistage sampling techniques and surveyed annually starting from 2011. Round 1 contains 8,245 home-interviewed respondents. If a sampled individual could not respond due to severe cognitive or physical impairment, a proxy respondent, either a family member or primary caregiver, was interviewed. In Round 5, the NHATS study added a refreshment sample of 4,182 respondents to maintain the representativeness. Throughout the study period, respondents were followed-up except for recorded death. On average, NHATS has response rate 89.5% for living adults and 95.5% for decreased adults (18).

Participants included in the current study were those with cognitive impairment whose cognitive scores were below the threshold for identifying Mild Cognitive Impairment (MCI) or dementia for at least two waves. In order to be able to model their health trajectories, individuals with fewer than three rounds of follow-up after the first observation of cognitive impairment were excluded. Detailed sample construction can be found in Appendix Figure S1. The final sample included 1,641 individuals for modeling physical functioning trajectories and 2,361 individuals for general health status trajectories (not every respondent had assessment-based physical functioning).

Cognitive Impairment

Cognitive impairment was defined as having either MCI or dementia. At each wave, participants were asked to perform three tasks: 10-item immediately and delayed recall to assess memory, clock drawing to assess executive function, and questions about time and place to assess orientation. An individual was considered to have MCI if he or she scored less or equal to 1.5 standard deviation below the mean in one domain. Dementia was defined as scoring below or equal to 1.5 standard deviation below the mean in at least two domains. Dementia status was additionally determined from direct reports from respondents or based on a proxy-reported score of at least 2 on the AD8 Dementia Screening Interview (19).

Physical Functioning

Physical functioning was measured using the NHATS Expanded Short Physical Performance Battery (SPPB) (20). Designed to capture a wide range of physical functioning, the NHATS Expanded SPPB included three components: 1) nested balance test, 2) a 3-meter usual walking speed to measure locomotion, and 3) rapid chair stands to measure lower body muscle function. Each of the 3 components scores from minimum, 1, to maximum, 4, for non-missing responses. This scoring system reflects the distribution quartiles of the weighted NHATS sample (17). A score of 0 was given if the participant was unable to stand or walk. The physical functioning measure used in this study was generated by summing up the scores of the three components, ranging from 0–12. A higher score indicates better physical functioning.

General Health Status

General health status was assessed by asking respondents to self-report their general health status from 1 (excellent) to 5 (poor). 26.73% of 10,701 person-year observations for general health status were reported by a proxy respondent (the baseline proportion for proxy-report was 14.1%, as shown in Table 1). Robustness checks without proxy reports show highly consistent patterns of trajectories with the results including them. We used Diehr’s recoding strategy to transform general health status to a 15–95 scale (excellent = 95, very good = 90, good = 80, fair = 30, and poor = 15) to improve the interpretability of the general health status measure (21).

Table 1

Baseline characteristics for the general health status (GHS) and physical functioning (PF) samples
	N		Mean (SD)/ %		Missing rate
	GHS	PF	GHS	PF	GHS	PF
Female	2021	1641	59.77%	59.72%	0.00%	0.00%
Age	2021	1641	81.05	80.76	0.00%	0.00%
			(7.59)	(7.54)
Race/Ethnicity	1986	1616			1.73%	1.52%
White			55.99%	55.88%
Black			30.06%	29.95%
Other			3.98%	4.02%
Hispanic			9.97%	10.15%
Education attainment	1977	1608			2.18%		2.01%
< High school			43.10%	41.67%
High school			24.73%	25.12%
> High school			32.17%	33.21%
# siblings	2002	1629			0.94%	0.73%
0			26.32%	26.21%
1–3			51.80%	51.75%
4+			21.88%	22.04%
# children	2021	1641			0.00%	0.00%
0			9.60%	8.78%
1–3			53.59%	53.69%
4+			36.81%	37.54%
Medicare drug coverage	1858	1520	69.32%	68.82%	8.07%	7.37%
Medicaid	1922	1568	27.63%	28.19%	4.90%	4.45%
Tricare	1961	1595	4.49%	4.95%	2.97%	2.80%
Comorbidity	1982	1611			1.93%	1.83%
0			8.17%	8.19%
1–3			63.62%	63.63%
4+			28.20%	28.18%
Marital status	2017	1638			0.20%		0.18%
Never married			5.26%	4.88%
Married/live with a partner			37.88%	38.89%
Separated, divorced, widowed, never married			56.87%	56.23%
Smoking regularly	1535	1272	45.60%	45.05%	24.05%	22.49%
Self-report	2021	1641	85.90%	--	0.00%	0.00%
MCI	2021	1641	38.69%	40.28%	0.00%	0.00%
Note: Mean (SD) for continuous variables and % for categorical variables.

Covariates

We selected sociodemographic and health covariates based on prior studies on mortality risks (22–24). They included sex, age, race/ethnicity (non-Hispanic White (hereafter “White”), non-Hispanic Black (hereafter “Black”), non-Hispanic other (hereafter “Other”), and Hispanic), educational attainment (less than high school, high school graduate, or beyond high school), number of siblings (0, 1–3, 4+), number of children (0, 1–3, 4+), enrollment in Medicare, enrollment in Medicare Part D, and enrollment in Tricare, comorbidity (0, 1–3, 4+), marital status (never married, married or living with a partner, separated/divorced/widowed), smoking regularly (i.e. smoking at least 1 cigarette per day), and presence of MCI rather than dementia. Number of siblings, number of children, and comorbidity were measured as continuous variables. Using these, we constructed categorical variables to capture potential nonlinear relationships between them the mortality, based on their distributions in our sample and previous studies (25).

Statistical Analyses

We applied GBTMs to identify distinct trajectory groups among PCIs and classify PCIs into these groups, assuming that there exist multiple latent subgroups with distinct trajectories within the PCI population (26). Consistent with prior studies, we used the first observed cognitive impairment diagnosis in our data as the baseline, with years since first observed diagnosis as the time variable (14). Based on the distributions of the general health status and physical functioning variables in our data (see Appendix Figure S2), we modeled general health status using a censored normal distribution and physical functioning using a zero-inflated Poisson distribution. We first fitted models with varying numbers of latent groups and included linear, quadratic, and cubic terms to select the best functional forms of the trajectories. We selected the best fitted model based on the Bayesian information criterion (BIC) and diagnosis statistics including the average posterior probability (AvePP > 0.7) and the odds of correction classification (OCC > 5) (16). Appendix Table S1 shows the BICs and Appendix Table S2 shows the diagnosis statistics. Each participant was assigned to the latent group with the highest posterior probability. All analyses were performed using the ‘traj’ package in Stata 15 (27).

After identifying latent trajectory groups and classifying PCIs into these groups, we applied multinomial logistic regression models to predict mortality using latent trajectory group memberships with and without controlling for covariates. To impute a small number of missing values in the covariates, we performed multiple imputations by chained equations using the Stata package ‘ice’(28). The regression coefficients from 10 imputed datasets were then combined based on the Rubin’s rule (29). Results without multiple imputations, as shown in Appendix Table S3 and S4, were highly consistent. We conducted two sets of analyses. In the baseline model, we predicted mortality using group memberships without any covariates. In the second model, we additionally controlled for sociodemographic and health covariates.

Physical Functioning

We identified six latent trajectory groups, as shown in Fig. 1. Estimates of model parameters are shown in Appendix Table S5, and Appendix Table S6 shows the individual characteristics of each latent group. Approximately 8.45% PCIs were in Group 1 (“high start, moderate decrease”), 22.4% were in Group 2 (“high-medium start, moderate decrease”), 27.56% were in Group 3 (“medium start, moderate decrease”), 6.07% were in Group 4 (“medium-low start, early fast decrease”), 22.02% were in in Group 5 (“medium-low start, moderate decrease”), and 13.66% were in Group 6 (“low start, stable”). Groups 1 (“high start, moderate decrease”), 2 (“high-medium start, moderate decrease”), 3 (“medium start, moderate decrease”) and 5 (“medium-low start, moderate decrease”) started with baseline values 8.86, 6.32, 3.98, and 2.03, respectively, and declined at moderate rates (the slopes of the declines were − 0.15, -0.34, -0.33, and − 0.24, respectively). Group 4 (“medium-low start, early fast decrease”) started with a medium-low level of baseline functioning (2.95) but experienced an acute decline to an SPPB score of 0 in the first 3 years after the first observation of cognitive impairment (the slope was − 1.43). Group 6 (“low start, stable”) stayed at a level of almost no physical functioning over the 8 years.

Table 2 shows the results on predicting mortality using trajectory group memberships. We used Group 1 (“high start, moderate decrease”) as the reference group because this group had the best physical functioning over time. Model 1 estimated the association without covariates. There were significant associations between group memberships and the odds of dying. The odds of dying for PCIs in Group 2 (“high-medium start, moderate decrease”) were 5.04 times as large as the odds for Group 1 (“high start, moderate decrease”). For Groups 3 (“medium start, moderate decrease”), 4 (“medium-low start, early fast decrease”), 5 (“medium-low start, moderate decrease”), 6 (“low start, stable”), the corresponding odds ratios were 8.73, 28.88, 14.84, and 21.48, respectively. This pattern persisted despite attenuated magnitudes after the inclusion of a variety of sociodemographic and health characteristics (Model 2), suggesting that the predictive power of trajectory group memberships was independent of these characteristics. In general, except for Group 4 (“medium-low start, early fast decrease”), the odds of dying increased as the group membership number increased, indicating that belonging to a trajectory group with ‘worse trajectories’ (i.e., lower baselines and/or faster declines) was associated with higher odds of dying. Group 4 (“medium-low start, early fast decrease”) was associated with the highest mortality risks, although its average level across time was higher compared to Group 6 (“low start, stable”). Among all the covariates, conditional on physical functioning trajectory group memberships, being female, Black, or Hispanic, and having MCI rather than dementia were significantly predictive of lower mortality whereas smoking regularly was significantly predictive of higher mortality.

Table 2

Predicting death using physical functioning trajectory group memberships, Odds Ratios and 95% Confidence Intervals
	Model 1	Model 2
Group membership (Ref. Group 1)
Group 2	5.04**	4.40**
	(1.98, 12.82)	(1.69, 11.44)
Group 3	8.73***	6.59***
	(3.49, 21.87)	(2.56, 16.98)
Group 4	28.88***	19.62***
	(10.90, 76.50)	(7.14, 53.94)
Group 5	14.84***	10.81***
	(5.91, 37.25)	(4.16, 28.09)
Group 6	21.48***	16.03***
	(8.47, 54.50)	(6.05, 42.46)
Female		0.70*
		(0.53, 0.93)
Age		1.23
		(0.91, 1.65)
Age squared		1.00
		(1.00, 1.00)
Race/Ethnicity (Ref. White)
Black		0.64**
		(0.47, 0.87)
Other		0.71
		(0.37, 1.37)
Hispanic		0.52**
		(0.32, 0.84)
Educational attainment (Ref. < High school)
High school		1.02
		(0.74, 1.40)
> High school		1.21
		(0.89, 1.64)
# of siblings (Ref. 0)
1–3		0.99
		(0.76, 1.31)
4+		0.88
		(0.61, 1.27)
# of children (Ref. 0)
1–3		1.02
		(0.64, 1.63)
4+		0.87
		(0.54, 1.40)
Medicare		0.95
		(0.73, 1.23)
Medicaid		0.97
		(0.70, 1.33)
Tricare		1.35
		(0.77, 2.35)
Comorbidity (Ref. 0)
1–3		1.09
		(0.68, 1.74)
4+		1.34
		(0.81, 2.20)
Marital status (Ref. Never married)
Married/live with a partner		1.28
		(0.64, 2.56)
Separated, divorced, widowed		1.14
		(0.59, 2.21)
Smoke regularly		1.33*
		(1.01, 1.75)
MCI		0.64**
		(0.50, 0.83)
Note: * p < 0.05; p < 0.01; * p < 0.001. N = 1,641.

General Health Status

We identified four trajectory group memberships for general health status (Fig. 2). Estimates of model parameters are shown in Appendix Table S7, and Appendix Table S8 shows the individual characteristics of each latent group. Approximately 44.79% PCIs were in Group 1 (“high start, slight decrease”), 18.82% were in Group 2 (“high start, convex”), 14.05% were in Group 3 (“low start, concave”), 22.33% were in Group 4 (“low start, slight increase”). Group 1 (“high start, slight decrease”) started with good health, experienced only slight decreases over time, and still maintained good health at the end of our study period. Group 2 (“high start, convex”) started with close to good health, experienced sharp decreases first, reaching close to fair health, and then recovered a little after the 5th year since the first observation of cognitive impairment. Starting with a baseline value just above fair, Group 3 (“low start, concave”) experienced sharp increases first but then experienced declines starting from the 5th year since the first observation of cognitive impairment. At the end of the study period, this group dropped to its starting value. Group 4 (“low start, slight increase”) started with just below fair health and experienced increases slightly over the study period, maintaining the level of fair health.

Table 3 shows the results predicting mortality using general health trajectory group memberships. In Model 1 without any covariates, two of the three group memberships were significantly associated with increased odds of dying as compared to Group 1. Unlike the case for physical functioning, we did not find a general gradient in the magnitudes of the associations. Group 2 (“high start, convex”) had the greatest odds of dying, 1.79 times as large as the odds for Group 1 (“high start, slight decrease”). Group 4 (“low start, slight increase”) had a comparable magnitude, which was 1.74 times as large as the odds for Group 1 (“high start, slight decrease”). After including covariates, as shown in Model 2, the magnitudes generally became bigger for all groups, suggesting the existence of negative confounders (30). Compared to Group 1 (“high start, slight decrease”), the odds of dying were 1.94 times, 1.46 times, and 2.32 times as large for Groups 2 (“high start, convex”), 3 (“low start, concave”) and 4 (“low start, slight increase”), respectively. The average level of general health status for Groups 2 (“high start, convex”) and 3 (“low start, concave”) were comparable with values of 53.5 and 54.3. Conditional on general health status trajectory group memberships, being Blacks or Hispanics, having at least 4 children, self-report, and having MCI rather than dementia were significantly predictive of lower mortality, whereas smoking regularly was significantly predictive of higher mortality. These results suggest that the relationship between these covariates and mortality was not completely mediated by general health status trajectories.

Table 3

Predicting death using general health status trajectory group memberships, Odds Ratios and 95% Confidence Intervals
	Model 1	Model 2
Group membership (Ref. Group 1)
Group 2	1.79***	1.94***
	(1.40, 2.30)	(1.47, 2.55)
Group 3	1.19	1.46*
	(0.88, 1.61)	(1.04, 2.02)
Group 4	1.74***	2.32***
	(1.37, 2.20)	(1.74, 3.09)
Female		0.93
		(0.73, 1.18)
Age		1.27
		(0.98, 1.65)
Age squared		1.00
		(1.00, 1.00)
Race/Ethnicity (Ref. White)
Black		0.68**
		(0.54, 0.91)
Other		0.63
		(0.35, 1.11)
Hispanic		0.50**
		(0.33, 0.76)
Educational attainment (Ref. < High school)
High school		1.05
		(0.79, 1.39)
> High school		1.26
		(0.97, 1.65)
# of siblings (Ref. 0)
1–3		0.95
		(0.75, 1.22)
4+		0.84
		(0.61, 1.16)
# of children (Ref. 0)
1–3		0.74
		(0.51, 1.07)
4+		0.67*
		(0.46, 1.00)
Medicare		0.96
		(0.76, 1.21)
Medicaid		0.93
		(0.70, 1.22)
Tricare		1.09
		(0.65, 1.83)
Comorbidity (Ref. 0)
1–3		1.24
		(0.83, 1.85)
4+		1.39
		(0.90, 2.14)
Marital status (Ref. Never married)
Married/live with a partner		1.10
		(0.63, 1.90)
Separated, divorced, widowed		0.99
		(0.58, 1.68)
Smoke regularly		1.52**
		(1.17, 1.97)
Self-report		0.61**
		(0.46, 0.82)
MCI		0.59***
		(0.47, 0.73)
Note: * p < 0.05; p < 0.01; * p < 0.001. N = 2,021.

This study examined whether and how trajectory group memberships for PCIs’ physical functioning and general health status predict mortality risks. Using group-based trajectory modeling, we identified six distinct trajectory groups for physical functioning and four groups for general health status among PCIs. We found that trajectory group memberships for both physical functioning and general health status significantly predicted mortality. For physical functioning, except for Group 4 (“medium-low start, early fast decrease”), group memberships were largely reflective of the average levels across time among PCIs, and worse trajectories (i.e., lower baselines and faster declines) were associated with higher odds of death. However, although Group 4 (“medium-low start, early fast decrease”) had a higher average level across time than Group 6 (“low start, stable”), the former had much higher mortality. For general health status, we observed two groups with comparable average levels across time, but the one with convex-shape trajectories had higher mortality risks compared to the one with concave-shape trajectories. For both health outcomes, the patterns largely persisted after the inclusion of various sociodemographic and health characteristics.

Our findings on physical functioning highlight the importance of considering trajectories, capturing information on both the general level of physical functioning and the rate of change over time. The results provide examples in which the shape of health deterioration is more predictive of mortality than are the average levels across time. For example, Group 4 (“medium low start, early fast decrease”) had higher mortality risks than Group 6 (“low start, stable”), which had lower average levels across time. Group 4 also had substantially higher mortality risks than Group 5 (“medium low start, moderate decrease”). There are several explanations to consider. One is that the declining function of Group 4 (“medium low start, early fast decrease”) is a consequence of worsening chronic conditions, which is responsible for increased mortality risks. Alternatively, it is possible that there is a negative feedback loop for some individuals. In this case, declining functioning decreases resilience, causing worsening of chronic conditions, which in turn further decreasing functioning. In any case, these findings suggest that rapidly declining functioning is robust indicator of increased mortality risks. The clinical challenge requiring additional research is understanding when interventions can change this trajectory and when these trajectories are irreversible and should trigger consideration of a more palliative approach to care.

Results for general health status similarly highlight the importance of examining trajectories. For example, despite comparable average levels across time, Group 2 (“high start, convex”) had higher mortality risks than Group 3 (“low start, concave”). It is possible that improving general health status (mostly self-rated, so their actual health may even have declined but was perceived to have improved) earlier after developing cognitive impairment, despite a relatively lower baseline value, leads to positive psychological consequences, which reduce mortality risks. In general, the relationship between respondent- or proxy-reported general health status and mortality is more complex than the one for assessment-based physical functioning. Future research is needed to determine the mechanism by which different trajectories are associated with differential risks.

We also found that being female, Black, or Hispanic, and having MCI rather than dementia were significantly predictive of lower mortality whereas smoking regularly was significantly predictive of higher mortality, conditional on physical functioning trajectories. Conditional on general health status trajectories, being Blacks or Hispanics, having at least 4 children, self-report, having MCI rather than dementia, and not smoking regularly were significantly predictive of lower mortality. Other covariates were not significant likely because trajectories largely mediated the relationship between them and mortality. One explanation for the lower mortality risks among Blacks and Hispanics is mortality selection: Blacks and Hispanics disproportionately die early before entering our sample, and therefore the ones left in our sample are positively selected in terms of health. The finding that persons with MCI had lower mortality risks than those with dementia, conditional on trajectory group memberships, adds to the literature demonstrating associations between declining cognition and increased mortality (31, 32).

Our findings may add to the literature in several ways. First, we found that physical functioning trajectory group memberships were highly predictive of mortality, especially when we compared the magnitudes of the coefficients to those for general health status. Indeed, the predictive power was so strong that physical functioning trajectories might have overpowered the many other known risk factors in the models. The comparison between Groups 4 (“medium-low start, early fast decrease”) and 6 (“low start, stable”) further confirms the notion that the shape of trajectories matters to mortality risks on top of the average levels across time. Therefore, clinicians should pay particular attention to the mortality risks of those who experienced sharp declines in physical functioning in a relatively short period of time.

Second, prior literature has shown that general health status is associated with mortality in the general population (33, 34) but such association was not found among PCIs (35). Nielsen, Siersma (35) raised an explanation that general health status is not a valid health indicator among PCIs due to “loss of insight”. However, alternatively, the lack of predictive power of general health status to mortality may be driven by the lack of information on changes of general health status over time, and no previous studies have explored whether and how its trajectories predict mortality among PCIs. Our results suggest that trajectories of general health status are a robust predictor of mortality among PCIs because they capture both the average health level across time and the shape of health deterioration. Therefore, our findings underscore the need to collect general health status trajectory data, rather than cross-sectional ones, for policymakers to better estimate mortality risks for the PCI population.

Third, our results on general health status without adjusting for covariates also suggest that those who perceived their health to be declining over time were at about the same risk for mortality as those with persistently poor general health status, while those who perceived their health to be improving were not statistically different from those who had persistently high general health status. Therefore, previous studies that only measured general health status at the baseline when predicting mortality can lead to misleading results. For example, the baseline value for Group 2 (“high start, convex”) was much higher than that for Group 3 (“high start, concave”), but the former had higher mortality risks than the latter due to differential shapes in the trajectories. When comparing these two groups (see Appendix Table S8), the former tended to be White and have relatively higher socioeconomic status and fewer comorbidities compared to the latter. Future studies are needed to examine what determines the differential shapes of trajectories between these two groups. Perhaps mental health is a determinant here, as Black Americans tended to be happier than Whites (36).

Our study has several limitations. First, although we controlled for a variety of individual sociodemographic and health characteristics when predicting mortality, we may still have missed some important characteristics, such as living arrangements. Second, our sample restriction criteria may lead to unhealthy individuals less likely being included in our sample. Our investigation shows that individuals excluded because of having less than two rounds of cognitive impairment were a mix of high- and low-socioeconomic subgroups (see Appendix Table S9), but individuals excluded because of having fewer than three rounds of data were more likely to have low socioeconomic statuses (see Appendix Table S10).

Despite these limitations, this study suggests several avenues for future research and policy making. Our findings highlighted that health trajectories predicted mortality among PCIs, because of both the general levels and the shapes of declines. Close monitoring health deterioration of PCIs is crucial to understand the health burden of this population and to make subsequent actions.

Ethics Approval and Consent to Participate: Not applicable

Consent for Publication: Not applicable

Availability of Data and Materials: Part of the datasets generated and analyzed during the current study are not publicly available because they are restricted access (see NHATS website: https://nhats.org/researcher/nsoc), but are available from the corresponding author on reasonable request and with NHATS’ permission.

Competing Interests: The authors declare that they have no competing interests

Funding: The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the National Institute on Aging grant numbers 1R21AG074238-01, 1R56AG067619-01, and P30AG021342.

Authors’ Contributions: EZ, BW, and TF designed and conceptualized the study. EZ and YS prepared the data. YS conducted the data analyses. EZ and XW drafted the manuscript. All authors contributed to finalizing the manuscript.

Acknowledgements: We thank Nancy Lu and Patrick Disman for their excellent research assistance.

Centers for Disease Control and Prevention. Cogntive impairement: Call for action, now! 2011 [Available from: https://www.cdc.gov/aging/pdf/cognitive_impairment/cogimp_poilicy_final.pdf.
van de Vorst IE, Koek HL, Stein CE, Bots ML, Vaartjes I. Socioeconomic disparities and mortality after a diagnosis of dementia: Results from a nationwide registry linkage study. Am J Epidemiol. 2016;184(3):219–26. doi:10.1093/aje/kwv319.
Zilkens RR, Davis WA, Spilsbury K, Semmens JB, Bruce DG. Earlier age of dementia onset and shorter survival times in dementia patients with diabetes. Am J Epidemiol. 2013;177(11):1246–54. doi:10.1093/aje/kws387.
Helmer C, Joly P, Letenneur L, Commenges D, Dartigues J-F. Mortality with dementia: Results from a french prospective community-based cohort. Am J Epidemiol. 2001;154(7):642–8. doi:10.1093/aje/154.7.642.
Clouston SAP, Brewster P, Kuh D, Richards M, Cooper R, Hardy R, et al. The dynamic relationship between physical function and cognition in longitudinal aging cohorts. Epidemiol Rev. 2013;35(1):33–50. doi:10.1093/epirev/mxs004.
Auyeung TW, Kwok T, Lee J, Leung PC, Leung J, Woo J. Functional decline in cognitive impairment— the relationship between physical and cognitive function. Neuroepidemiology. 2008;31(3):167–73. doi:10.1159/000154929.
Stephan Y, Sutin AR, Luchetti M, Aschwanden D, Terracciano A. Self-rated health and incident dementia over two decades: Replication across two cohorts. J Psychiatr Res. 2021;143:462–6. doi:10.1016/j.jpsychires.2021.06.036.
Wu C, Geldhof GJ, Xue Q-L, Kim DH, Newman AB, Odden MC. Development, construct validity, and predictive validity of a continuous frailty scale: Results from 2 large us cohorts. Am J Epidemiol. 2018;187(8):1752–62. doi:10.1093/aje/kwy041.
Lima-Costa MF, Cesar CC, Chor D, Proietti FA. Self-rated health compared with objectively measured health status as a tool for mortality risk screening in older adults: 10-year follow-up of the bambuí cohort study of aging. Am J Epidemiol. 2011;175(3):228–35. doi:10.1093/aje/kwr290.
Smits CHM, Deeg DJH, Kriegsman DMW, Schmand B. Cognitive functioning and health as determinants of mortality in an older population. Am J Epidemiol. 1999;150(9):978–86. doi:10.1093/oxfordjournals.aje.a010107.
Zimmer Z, Martin LG, Jones BL, Nagin DS. Examining late-life functional limitation trajectories and their associations with underlying onset, recovery, and mortality. J Gerontol B Psychol Sci Soc Sci. 2014;69(2):275–86. doi:10.1093/geronb/gbt099.
Mok A, Khaw K-T, Luben R, Wareham N, Brage S. Physical activity trajectories and mortality: Population based cohort study. BMJ. 2019:l2323. doi:10.1136/bmj.l2323.
Falconer J, Quesnel-Vallée A. Pathway from poor self-rated health to mortality: Explanatory power of disease diagnosis. Soc Sci Med. 2017;190:227–36. doi:10.1016/j.socscimed.2017.08.008.
Zang E, Guo A, Pao C, Lu N, Wu B, Fried TR. Trajectories of general health status and depressive symptoms among persons with cognitive impairment in the united states. J Aging Health. 2022:08982643211060948. doi:10.1177/08982643211060948.
Feenstra M, van Munster BC, Vroomen JLM, de Rooij SE, Smidt N. Trajectories of self-rated health in an older general population and their determinants: The lifelines cohort study. BMJ Open. 2020;10(2):e035012. doi:10.1136/bmjopen-2019-035012.
Nagin DS. Group-based modeling of development. Cambridge, Mass: Harvard University Press; 2005.
Kasper JD, Freedman VA. National health and aging trends study user guide: Rounds 1–9 beta release 2020 [Available from: https://nhats.org/sites/default/files/2021-04/NHATS_User_Guide_R9_Final_Release_0.pdf.
Freedman VA, Kasper JD. Cohort profile: The national health and aging trends study (nhats). Int J Epidemiol. 2019;48(4):1044-5g. doi:10.1093/ije/dyz109.
National Academies of Sciences E, Medicine. Families caring for an aging america: National Academies Press; 2016.
Guralnik JM, Simonsick EM, Ferrucci L, Glynn RJ, Berkman LF, Blazer DG, et al. A short physical performance battery assessing lower extremity function: Association with self-reported disability and prediction of mortality and nursing home admission. J Gerontol. 1994;49(2):M85-M94. doi:10.1093/geronj/49.2.m85.
Diehr P, Patrick DL, Spertus J, Kiefe CI, McDonell M, Fihn SD. Transforming self-rated health and the sf-36 scales to include death and improve interpretability. Med Care. 2001:670–80.
Stolz E, Hoogendijk EO, Mayerl H, Freidl W. Frailty changes predict mortality in 4 longitudinal studies of aging. J Gerontol A. 2020;76(9):1619–26. doi:10.1093/gerona/glaa266.
Bassuk SS, Berkman LF, Amick BC, III. Socioeconomic status and mortality among the elderly: Findings from four us communities. Am J Epidemiol. 2002;155(6):520–33. doi:10.1093/aje/155.6.520.
Fried LP, Kronmal RA, Newman AB, Bild DE, Mittelmark MB, Polak JF, et al. Risk factors for 5-year mortality in older adults: The cardiovascular health study. JAMA. 1998;279(8):585–92. doi:10.1001/jama.279.8.585.
Zang E, Guo A, Pao C, Lu N, Wu B, Fried TR. Trajectories of general health status and depressive symptoms among persons with cognitive impairment in the united states. J Aging Health. 2022. doi:10.1177/08982643211060948.
Nagin DS, Odgers CL. Group-based trajectory modeling in clinical research. Annu Rev Clin Psychol. 2010;6(1):109–38. doi:10.1146/annurev.clinpsy.121208.131413.
Jones BL, Nagin DS. A note on a stata plugin for estimating group-based trajectory models. Sociol Methods Res. 2013;42(4):608–13. doi:10.1177/0049124113503141.
Royston P, White I. Multiple imputation by chained equations (mice): Implementation in stata. J Stat Soft. 2011;45(4). doi:10.18637/jss.v045.i04.
Rubin DB. Multiple imputation for nonresponse in surveys. New York: John Wiley & Sons; 1987.
Mehio-Sibai A, Feinleib M, Sibai TA, Armenian HK. A positive or a negative confounding variable? A simple teaching aid for clinicians and students. Ann Epidemiol. 2005;15(6):421–3. doi:10.1016/j.annepidem.2004.10.004.
Lv X, Li W, Ma Y, Chen H, Zeng Y, Yu X, et al. Cognitive decline and mortality among community-dwelling chinese older people. BMC Med. 2019;17(1):1–10. doi:10.1186/s12916-019-1295-8.
Hui J, Wilson R, Bennett D, Bienias J, Gilley D, Evans D. Rate of cognitive decline and mortality in alzheimer’s disease. Neurology. 2003;61(10):1356–61. doi:10.1212/01.wnl.0000094327.68399.59.
DeSalvo KB, Bloser N, Reynolds K, He J, Muntner P. Mortality prediction with a single general self-rated health question. J Gen Intern Med. 2006;21(3):267–75. doi:10.1111/j.1525-1497.2005.00291.x.
Idler EL, Benyamini Y. Self-rated health and mortality: A review of twenty-seven community studies. J Health Soc Behav. 1997;38(1):21–37. doi:10.2307/2955359.
Nielsen ABS, Siersma V, Waldemar G, Waldorff FB. Poor self-rated health did not increase risk of permanent nursing placement or mortality in people with mild alzheimer’s disease. BMC Geriatr. 2016;16(1):1–8. doi:10.1186/s12877-016-0262-x.
Tang F, Jang H, Rauktis MB, Musa D, Beach S. The race paradox in subjective wellbeing among older americans. Ageing Soc. 2019;39(3):568–89. doi:10.1017/S0144686X17001064.

No competing interests reported.

Appendix20220305.docx

Download PDF

Editorial decision: Major revision
15 Jul, 2022
Reviews received at journal
05 Jul, 2022
Reviewers agreed at journal
05 Jul, 2022
Reviews received at journal
03 Jun, 2022
Reviewers agreed at journal
23 May, 2022
Reviewers invited by journal
23 May, 2022
Editor assigned by journal
23 May, 2022
Editor invited by journal
10 May, 2022
Submission checks completed at journal
10 May, 2022
First submitted to journal
09 May, 2022

You are reading this latest preprint version

Prediction of Physical Functioning and General Health Status Trajectories on Mortality Among Persons with Cognitive Impairment

Status:

Version 1

Abstract

Background

Methods

Results

Conclusions

Figures

Introduction

Methods

Data and Measures

Cognitive Impairment

Physical Functioning

General Health Status

Covariates

Statistical Analyses

Results

Physical Functioning

General Health Status

Discussion

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1