Abstract
Neuropsychiatric symptoms (NPS) present in older adults with Alzheimer’s disease (AD) and other dementias are related to mortality. Research on the relationship between NPS and mortality in a non-dementia population is limited. This study examines NPS as a predictor of six-year mortality among community dwelling Mexican Americans aged 80 years and older. Data included 466 cognitively normal participants from Wave 7 of the Hispanic Established Population for the Epidemiological Study of Elderly. NPS were measured using the Neuropsychiatric inventory (NPI). Cox proportional hazard models were used to estimate the hazard ratio (HR) of mortality. The HR of death at 6 years was 1.02 (95% Confidence Interval—CI [1.00, 1.04]) as a function of NPI score and 1.09 (95% CI [1.02, 1.17]) for the number of NPI conditions, controlling for demographic and health characteristics. Apathy, irritability, and aberrant motor behavior were all independently predictors of mortality. NPS may be modifiable risk factors to increase survival time or may be indicative of underlying health problems. NPS may be related to underlying health conditions among older adults with normal cognitive functioning.
Introduction
Hispanic older adults represent the fastest growing segments of adults over 65 in the United States, with most older Hispanics being of Mexican origin (Ortman et al., 2014). The oldest old Hispanic Americans, those over age 85, are projected to number 3.4 million by 2060 (Administration for Community Living., 2020). Hispanics have a significantly longer life expectancy with cognitive impairment than non-Hispanic white and non-Hispanic black older adults (Garcia et al., 2019). While prior research shows a high burden of cognitive impairment in the older Hispanic population, there is a paucity of data on behavioral changes in late life and their impact on health outcomes—especially in cognitively robust Hispanics. Neuropsychiatric symptoms (NPS) refer to behavioral changes, such as apathy and anxiety, that are indicative of underlying neurological and medical disorders (Cummings et al., 1994). An emerging literature has highlighted the effects of caring for older adults who experience behavioral changes in late life, but the questions remain as to whether these behavioral changes themselves are predictive of future health declines in older adults who exhibit behavioral changes. Importantly, while changes in behaviors in late life are often indicative of cognitive decline, NPS are also present in cognitively normal older adults (Nunes et al., 2019).
Theoretical Framework
One critical marker of Alzheimer’s disease is a decline in cognitive functioning, including information processing, executive functioning, and memory domains (Aisen et al., 2017; Sperling et al., 2011). Figure 1 shows a continuum of decline in cognitive functioning, representing theoretical thresholds for diagnosis of Mild Cognitive Impairment (MCI) and dementia. Measures of cognitive functioning developed by neuropsychological and cognitive aging researchers do not fully capture the wide array of behavioral changes that may occur in the prodromal stage of dementia development, changes that have been shown to be independent predictors of future cognitive decline (Wise et al., 2019). Additionally, NPS are predictive of higher mortality for older adults with dementia (Bränsvik et al., 2020; Weiner et al., 2005). While research has established that cognitively impaired older adults have a higher risk of mortality when exhibiting NPS, little is known about older adults who have normal cognitive functioning but exhibit changes in behavior. NPS in cognitively normal older adults may be early symptoms of prodromal dementia, new psychiatric conditions, or other health issues, sometimes referred to as Mild Behavioral Impairment (MBI) (Creese et al., 2019). Previous research has assessed the independent association between behavioral changes and mortality in older adults with MCI or early dementia, but it is unclear whether this association applies to cognitively robust older adults.
Neurodegenerative continuum for Alzheimer’s disease and other dementias.
Neuropsychiatric Symptoms in Cognitively Robust Adults
Most research examining NPS draws upon samples of older adults with cognitive impairment and dementia, but less is known about the presentation of NPS in the cognitively normal older adult population. NPS are highly prevalent in individuals with dementia, and 97% of individuals with dementia experience at least one NPS within 5 years of the onset of dementia (Steinberg et al., 2008). Apathy symptoms are predictors of future conversion from MCI to dementia (Salem et al., 2023). Additionally, anxiety, apathy, and dysphoria predicted high odds of mortality for older adults with MCI (Huang et al., 2022).
It is estimated that more than 60% of community-dwelling Mexican Americans have one or more NPS, with the three most common being agitation/aggression (32%), depression (29%), and irritability (26%) (Salazar et al., 2015). Caregiver reports of NPS for Mexican Americans follow gendered patterns, with significantly higher reports of agitation in men and more depressive symptoms and anxiety in women (Milani et al., 2021). For caregivers of older Hispanics living with cognitive impairment and dementia, NPS in care recipients are associated with higher levels of depressive symptoms (Hinton et al., 2003). Likewise, caregivers of older Mexican American adults report more depressive symptoms when caring for older adults who have NPS (Flores et al., 2021). Furthermore, specific clusters of symptoms, specifically hyperactivity and psychosis, are related to negative relationship assessments among caregivers and care recipients (Cantu & Aranda, 2022). The role of NPS has also been explored in relationship to ethnic enclaves, showing that caregivers report fewer depressive symptoms related to NPS in census tracts with high proportions of Latinos (Rote et al., 2017). Research on NPS and Hispanics has focused on caregiver depressive symptoms with less emphasis on health outcomes in people experiencing NPS.
NPS occur more frequently in older adults with MCI and dementia but are present in some cognitively normal older adults (Nunes et al., 2019). The most common NPS for cognitively normal older adults included appetite changes (27.5%), depression (26%), and anxiety (23.9%) (Nunes et al., 2019). NPS behavioral disturbances are not exclusive to dementia, specifically depression, elation, irritability, disinhibition, and sleeping disorders are common co-occurrences with other health conditions (Squelard et al., 2012). Research in a cognitively normal population shows that NPS predict future cognitive impairment and dementia (Wise et al., 2019). Non-Hispanic white and Hispanic populations present NPS differently prior to MCI, with apathy being particularly important for whites and physical behaviors (nighttime behaviors and appetite changes) being more important for Hispanics (Thakur et al., 2021). While previous research shows how NPS are associated with cognitive outcomes, levels of cognition, and mortality in older adults with MCI and dementia, no research has examined how NPS are related to mortality in cognitively normal older adults.
Informant and family assessments may be especially important for Mexican Americans, who rely extensively on informal care to remain in the community in late life (Angel et al., 2014). This study examines the relationship between NPS and mortality in a community-dwelling Mexican American sample of cognitively normal older adults. We hypothesize that NPS will be predictive of mortality for older adults who do not show signs of cognitive impairment and that the presence of specific symptoms (e.g., depression, apathy, nighttime behaviors, and changes in appetite) will have differential impact on the prediction of mortality.
Methods
Data
Data for this analysis came from the Hispanic Established Population for the Epidemiological Study of the Elderly (H-EPESE), an ongoing longitudinal study of older Mexican American adults living in Texas, California, New Mexico, Arizona, and Colorado (Markides et al., 1997). A total of 3,050 participants were interviewed during the first observation wave in 1993 to 1994, and 9 observation waves have been made public as of 2024 (Cantu & Markides, 2019). The H-EPESE was replenished at Wave 5 (2004–2005) with a representative sample of 902 adults aged 75 and older. At Wave 7 (2010–2011), study participants, then aged 80 and over were asked to name the person closest to them who helped the most, and they were asked to complete an informant questionnaire about the participant’s health (n = 925). Data for this study is from both the respondents’ direct assessments of health at wave 7 and the informant questionnaires.
Sample Selection
There were 1,078 participants at Wave 7 (hereafter referred to as baseline), and of those, 925 also had an informant caregiver who completed a qualified informant survey. Our final sample included older adults with normal cognition whose caregivers completed the qualified informant survey (n = 477). We excluded 11 participants who were missing other covariates for a final sample of 466. All participants included in the study included an informant interview who reported on the health of the target participants.
Measures
Dependent Variable: Mortality
The mortality status of the H-EPESE participants was determined through linkage with the National Death Index (NDI) and reports from relatives in subsequent waves of interviews through December 31, 2016. It has been established that NDI mortality linkages for Hispanic populations are “reasonably good,” with 5% higher mortality in survey data compared to NDI estimates (Arias et al., 2010). In total, 192 death dates were obtained. Survival time is calculated as days from interview in Wave 7 to date of death or censoring (December 31, 2016) for 274 individuals who had neither NDI matches nor family reports of death prior to 2016 in subsequent waves of the H-EPESE (up to Wave 9 in 2018).
Neuropsychiatric Symptoms
NPS were assessed using the Neuropsychological Inventory (NPI) (Cummings et al., 1994). The Neuropsychiatric Inventory (NPI) is a well-established measure of psychopathologies and behavioral changes that accompany neuropathology (Cummings et al., 1994). The NPI covers 12 symptoms, including agitation/aggression, dysphoria/depression, irritability/lability, apathy/indifference, anxiety, disinhibition, aberrant motor behavior, delusions, hallucinations, euphoria/elation, nighttime behavioral disturbances, and appetite/eating disturbances. Previous research has validated the NPI for Hispanics when comparing them to non-Hispanic Whites (Sayegh & Knight, 2014). In the H-EPESE, the NPI is asked of informant caregivers, who answer questions about the presence, severity, and their level of stress for the 12 domains in the past month. The total NPI score is the sum of the severity, mild, moderate, severe (1–3), and informant distress, none, minimal, mild, moderate, severe, and extremely severe (0–5) for each item for a total possible score of 0 to 106. We examined NPS status four ways: (1) using total NPI score (severity and distress summed together [0–106], participants with no NPI symptoms were coded 0), (2) the number of NPI symptoms (total count of informant reports of symptoms. 0–12), (3) any NPI symptoms compared to no NPI symptoms (binary, 0–1), and (4) each individual NPI symptoms separately (binary, 0–1 for each symptom).
Cognitive Functioning
Cognitive function was assessed using the Mini Mental State Exam (MMSE; Folstein et al., 1975). The MMSE has a maximum score of 30 points, with a higher score indicating better cognitive functioning. We classified participants as having normal cognitive functioning if their MMSE was at least 21. Previous research has defined an MMSE cut-off of less than 21 as representing impaired or poor cognition for populations with low educational attainment in general and specifically in older Mexican Americans (Downer et al., 2018; Raji et al., 2010).
Covariates
Covariates included in our analysis were age (continuous years), nativity (categories, US born and foreign born), gender (male and female), education (continuous years), self-reported health status( good and excellent vs. fair and poor), disability in activities of daily living (ADL)(do you need help from another person or special equipment to: walk across a small room, bathing, grooming, dressing, eating, getting out of bed, using the toilet, coded no help, help on at least one 0–1), instrumental activities of daily living (IADL) (can you: use the telephone, drive a car or take a bus, go shopping, prepare meals, do light housework, take medication, handle money, walk up stairs, walk a mile, coded can do all or cannot do all 0–1), and self-reported medical conditions (arthritis, diabetes, hypertension, heart attack, and stroke). All covariates were measured at baseline. Participants who did not report having been told by a doctor that they had high blood pressure but had a measured systolic blood pressure of ≥140 mmHg or diastolic blood pressure of ≥90 mmHg were also classified as having hypertension.
Statistical Analysis
Baseline differences according to the presence of NPI symptoms for participants with normal cognition were assessed using ANOVA and chi-square test for continuous and categorical variables, respectively. Additionally, we examined the prevalence of each NPI symptom among those with any NPI symptoms. The risk of mortality according to NPI symptoms at baseline was estimated using Cox proportional hazard models for (1) total NPI score, (2) the number of NPI symptoms, (3) any NPI symptom, and (4) individual NPI symptoms using time to mortality (date of death) or censoring (December 31, 2016). All models include individuals with and without NPI symptoms, with the interpretation of hazards being (1) the increased risk of mortality for each point in the NPI score, (2) the increased risk of mortality for each additional NPI symptom, (3) the increased risk of mortality for having any NPI symptom compared to having none, and (4) the increased risk of mortality for each individual symptom compared to not having the symptom. All Cox proportional hazard models control for age, education, gender, nativity, ADL and IADL disability, self-reported health status, and self-reported chronic conditions. All variables included in the analysis met the proportional hazards assumption as determined by visual inspection of plots for Schoenfeld residuals and tests of non-zero slopes for the Schoenfeld residuals (all p values > .05). All analyses were conducted using STATA (StataCorp LLC, College Station, TX).
Results
Sample Characteristics
Table 1 shows the overall descriptive characteristics of the sample and by any NPI symptoms. The mean age was 85.3 years, the majority were female, the mean years of education was 5.7, and 54% had at least one NPI symptom. Participants with NPI symptoms (NPI ≥ 1) had a mean NPI score of 8.10 and, on average, 2.8 symptoms. The number of deaths was higher among those with NPI symptoms (n = 118, 47%) than those without NPI symptoms (n = 74, 34.4%). Those with NPI symptoms were significantly more likely to report more ADL limitations and had significantly lower MMSE scores than those without NPI symptoms. Participants with NPI symptoms did not differ from those without NPI symptoms for age, gender, nativity, level of education, self-rated health, IADL disability, arthritis, diabetes, hypertension, stroke, and heart attack.
Baseline Sociodemographic Characteristics for Older Mexican Americans from the Hispanic Established Population for the Epidemiologic Study of the Elderly (n = 466; 2010–2011).
Note. NPI = neuropsychiatric symptoms; SD = standard deviation; MMSE = mini-mental state examination; ADL = activities of daily living; IADL = instrumental activities of daily living.
Table 2 shows the prevalence of individual NPI symptoms for participants with normal cognition from most to least prevalent. The most common symptoms were agitation (27%), depression (22.1%), and irritability (20%); the least common symptoms were euphoria (5.8%), delusions (4.9%), and hallucinations (2.4%).
NPI Symptoms for Older Mexican Americans with Normal Cognition (n = 466).
Note. NPI = neuropsychiatric symptoms.
Neuropsychiatric Inventory and Mortality
The findings from the multivariate Cox hazard model for mortality are presented in Table 3. Model 1 shows that the total NPI score was predictive of mortality, indicating that for each additional point of the total NPI score, there was an increase in the hazard of mortality (Hazard Ratio [HR] 1.02, 95% Confidence Interval—CI [1.00, 1.04]). Model 2 shows that the number of NPI symptoms was also a significant predictor of mortality, meaning that each additional NPI symptom was associated with higher hazards of mortality (HR 1.09, 95% CI [1.02, 1.17]). However, there was no significant increase in hazard of mortality for having a single NPI symptom compared to having no NPI symptoms (Model 3) and mortality over the follow-up period when controlling for demographic characteristics and health covariates, although the relationship was in the expected direction of an increased hazard. Figure 2 shows the pattern of earlier mortality for adults with NPI symptoms compared to those without. Other variables that were significant predictors of mortality were age, ADL disability, and diabetes.
Mortality According to Baseline NPI Among Older Mexican Americans (n = 466).
Note. NPI = neuropsychiatric symptoms; HR = hazard ratio; CI = confidence interval; ADL = activities of daily living; IADL = instrumental activities of daily living.
p < .05. **p < .01. ****p < .001.
Cox proportional hazard of mortality by informant reports of at least 1 NPI symptom.
Not all NPI symptoms were predictive of mortality. Table 4 shows the results from separate Cox HR models for each of the NPI symptoms. While all individual symptoms had an HR over 1, only apathy (HR 1.82, 95% CI [1.26, 2.64]), irritability (HR 1.54, 95% CI [1.10, 2.17]), and aberrant motor behavior (HR1.74, 95% CI [1.09, 2.80]) were significant predictors of mortality (p < .05).
Mortality According to Individual NPI Symptoms (n = 466).
Note. NPI = neuropsychiatric symptoms; HR = hazard ratio; CI = confidence interval.
p < .05. **p < .01.
Discussion
This study found that cognitively normal older Mexican Americans with NPI symptoms of apathy, irritability, and aberrant motor behaviors were more likely to die over the follow-up period compared to those without these symptoms. We found that the total NPI score and the number of NPI symptoms were significant predictors of mortality. However, a binary indicator of any NPI compared to no NPI was not significantly associated with increased mortality hazards. These cognitively robust older adults who exhibit changes in behavior in late life represent a population experiencing MBI. These changes in behavior that are not accompanied by significant cognitive decline were associated with increased mortality, underscoring a need for early recognition of these symptoms and the development of timely intervention for both the healthcare system and caregivers.
Our results are consistent with previous studies that observed an increase in risk of mortality among older adults with dementia who have NPI symptoms (Bränsvik et al., 2020; Weiner et al., 2005). Using Swedish health registry data, Bränsvik and colleagues observed that older adults with dementia had a 1% increase in mortality per point of NPI score, and we found a similar magnitude of increased risk for older adults without dementia. Additionally, there was a stepwise increase in mortality with each category of NPI severity. The categories used in their study were no NPI, mild NPI symptoms (0–4), moderate NPI symptoms (5–8), and severe NPI symptoms (9–12). In sensitivity analysis, we examined mortality using the same typology and found that only participants with moderate NPI symptoms had a significantly higher HR of mortality; however, in our study, only seven participants had severe NPI. The lack of participants with more than 9 NPI symptoms in our sample led us to conclude that there are likely different meaningful NPI symptom cutoffs for cognitively normal adults compared to older adults with dementia. Additionally, our null finding that a binary indicator for any NPI compared to no NPI was not related to mortality suggests that the relationship between NPS and mortality is significant only for individuals with several or specific symptoms. We did not find a significant relationship using a binary indicator of no symptoms compared to any symptoms, but the relationship was in the expected direction.
While an extensive body of research has examined the role of cognitive impairment in mortality in late life, relatively less attention has been paid to the role of neuropsychiatric behaviors and their role in mortality in late life, and to our knowledge, no research has been published on the relationship between mortality and NPS for cognitively normal older adults. Our motivation for studying this population is to ascertain if NPS are predictive of mortality for older adults who were robust to cognitive decline. The question remains: are NPS causal mechanisms for mortality in late life, as such that their presence indicates a higher risk of mortality in late life, or are they symptomatic of underlying etiologies for which interventions can lessen the risk of mortality? Evidence suggests that both are possible. One explanation for increased mortality with NPS is the use of antipsychotic medication for NPS. Older adults who have symptoms of psychosis treated by antipsychotics in late life have higher all-cause mortality over follow-up (Gill et al., 2007). Additionally, specific antipsychotics have also been shown to increase mortality in older adults with dementia (Kales et al., 2012). However, similar research with cognitively normal adults has not been conducted.
We also find evidence for the alternative hypothesis: that the symptoms themselves are responsible for higher mortality (Bränsvik et al., 2020; Lopez et al., 2013), in the finding that apathy, irritability, and aberrant motor behaviors were the only specific symptoms that predicted mortality for cognitively normal older adults. Previous literature found that apathy was less prevalent for older Hispanic populations compared to non-Hispanic whites (Thakur et al., 2021), but we found that apathy was a particularly important predictor of mortality for older Mexican Americans. Apathy predicted a 1.82 higher hazard of mortality and is not likely to be treated with antipsychotics. Apathy is likely related to depression, which is a predictor of mortality for older Mexican Americans (Mutambudzi et al., 2016), but the more general question on depression was not. It is possible that the NPI measure of apathy is a more accurate predictor of depression than the “general low spirits question” that is asked. Likewise, the only previously published study on MBI and mortality found that apathy symptoms were significant predictors of mortality for older adults with MCI (Huang et al., 2022). However, there is little research on the association between specific NPS behaviors and mortality (Bränsvik et al., 2020), and there are no clear explanations for why irritability or aberrant motor behavior predict mortality for cognitively normal older adults. It is possible that the NPS symptoms indirectly impact mortality by reducing adherence to guideline-recommended medications and lifestyle interventions for common co-occurring cardiovascular (e.g., coronary artery disease), metabolic (e.g., diabetes) and other age-associated comorbidities. Future longitudinal research is needed to understand how these specific symptoms lead to mortality for cognitively normal older adults and interventions that might improve outcomes.
Our results also speak to important gender differences in behavioral changes in late life and specific risks for men and women. We found that apathy was a particularly important for mortality over follow-up, which is a significantly more common symptom for older Mexican American women (Milani et al., 2021). Likewise, irritability was also found to be a predictor of mortality, which is a more common symptom for older men (Milani et al., 2021). These specific symptoms may be important behaviors for caregivers to pay attention to in late life as these symptoms that are often normalized as part of aging are also related to the mortality process—underscoring the need for clinicians to investigate for potential underlying medical and neuropsychiatric causes of these NPI symptoms
Furthermore, our sample was restricted to the very old Mexican Americans over the age 80, which might shape our results. Health of the oldest old tends to reflect a robust survivorship bias, and increases in the proportion of adults living to advanced age has been accompanied by increases in overall health in late life (Thinggaard et al., 2020). However, examinations of the health of oldest old Mexican Americans have revealed that this population had significantly more diabetes, cognitive impairment, and heart attack than non-Hispanic whites. However, cognitive impairment was only a predictor for mortality for non-Hispanic white oldest-old adults (Samper-Ternent et al., 2012). Our results suggest that within cognitively robust oldest-old adults, changes in behavior are important precursors to mortality. While it remains unclear if these changes in behavior are linked to underlying etiologies that cause death, our findings of NPS-mortality association offer an important avenue to addressing health in late life. Behavior changes that might seem normal to caregivers should be topics of conversation between caregivers and health professionals during clinic visits—a key step to mitigating future decline in health and function among the oldest old.
Limitations
This study has several limitations to the generalizability of our findings. The MMSE was not available for all participants, and we determined that participants had cognitive impairment if they were missing MMSE information because of cognitive impairment reported elsewhere in the survey. Individuals missing MMSE were likely drawn from the most impaired and would not have been included in the cognitively normal sample. A second concern is the use of an MMSE cut-off of 21 for cognitively normal. We conducted sensitivity analysis using a cut-off of 24, and our results were similar but lacked statistical power since few older adults in the sample had an MMSE over 24. Additionally, NPI symptoms are reported by caregivers who had a range of relationships with the respondents. It is possible that NPI reports from children or paid caregivers, for example, could be qualitatively different from those of spouses. In modeling mortality, we attempted to include cancer as a covariate; however, it did not meet the proportional hazards assumption.
Conclusions
In summary, our examination reveals that the NPI total score, the number of NPI symptoms, and the specific NPI symptoms of apathy, irritability, and aberrant motor behaviors are all significant predictors of mortality in cognitively normal older Mexican Americans. These results offer potential areas of intervention that could reduce mortality among cognitively normal older adults. While previous research has shown the importance of these behavioral symptoms in caregivers’ well-being and future cognitive decline, we have found that there is an independent pathway to mortality for these behavioral symptoms that could be incorporated into models of care for older adults. Specifically, current data collection efforts are not sufficient to understand what types of behavioral changes older adults are experiencing in late life and how care partners understand these changes. Efforts to educate care partners that changes in how behavioral changes may reflect previously unrecognized pathologies that are outside of normal aging are critical for improving outcomes for older adults and their caregivers. Greater awareness of MBI is invaluable to understanding changes in behavior in late life. A tendency to see changes in behavior as part of aging, especially for Mexican American families, could result in delayed diagnosis of more serious but potentially modifiable health problems in late life.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The study was supported by NIA grants K01AG086589, R01AG010939-25, and P30AG059301.
