Self-Reported Dementia-Related Diagnosis Underestimates the Prevalence of Older Americans Living with Possible Dementia

Abstract

Background:

Dementia screening is an important step for appropriate dementia-related referrals to diagnosis and treat possible dementia.

Objective:

We sought to estimate the prevalence of no reported dementia-related diagnosis in a nationally representative sample of older Americans with a cognitive impairment consistent with dementia (CICD).

Methods:

The weighted analytical sample included 6,036,224 Americans aged at least 65 years old that were identified as having a CICD without history of stroke, cancers, neurological conditions, or brain damage who participated in at least one- wave of the 2010–2016 Health and Retirement Study. The adapted Telephone Interview of Cognitive Status assessed cognitive functioning. Those with scores ≤ 6 were considered as having a CICD. Healthcare provider dementia-related diagnosis was self-reported. Age, sex, educational achievement, and race and ethnicity were also self-reported.

Results:

The overall estimated prevalence of no reported dementia-related diagnosis for older Americans with a CICD was 91.4% (95% confidence interval (CI): 87.7%−94.1%). Persons with a CICD who identified as non-Hispanic black had a high prevalence of no reported dementia-related diagnosis (93.3%; CI: 89.8%−95.6%). The estimated prevalence of no reported dementia-related diagnosis was greater in males with a CICD (99.7%; CI: 99.6%−99.8%) than females (90.2%; CI: 85.6%−93.4%). Moreover, the estimated prevalence of no reported dementia-related diagnosis for non-high school graduates with a CICD was 93.5% (CI: 89.3%−96.1%), but 90.9% (CI: 84.7%−94.7%) for those with at least a high school education.

Conclusion:

Dementia screening should be encouraged during routine geriatric health assessments. Continued research that evaluates the utility of self-reported dementia-related measures is also warranted.

INTRODUCTION

Dementia and related brain disorders that decrease cognitive functioning are projected to impact approximately 14 million Americans by the year 2060 [1, 2]. These new cases will exacerbate the already staggering dementia-related monetary costs and threaten longevity for the rapidly growing older American population [3–5]. Developing effective preventive and treatment measures for dementia has become a public health priority [6]. As such, programs that emphasize surveillance and treatment for dementia through early diagnosis have been implemented to reduce the projected dementia-related economic and health burden in the United States [7].

Screening for dementia with questionnaires, interviews, and physical measures can be an initial step in helping to initiate and organize a comprehensive cognitive assessment [8–10]. Medicare annual wellness visits may include a cognitive functioning assessment [11], and Healthy People 2030 has targeted increasing the number of persons that have discussed cognitive functioning with their healthcare provider [12]. Consensus statements regarding screening for dementia in older adults, however, are mixed. For example, the United States Preventive Services Task Force does not recommend nor oppose dementia screening in routine health assessments for older adults [13]. Conversely, guidelines from the American Academy of Neurology suggest annual cognitive function screening for older adults to facilitate earlier detection of dementia [14]. Therefore, healthcare providers may choose to not include cognitive functioning as part of their routine health assessments, which in turn, may lead to a lack of referrals for possible dementia.

Access to healthcare is also important for screening and properly diagnosing dementia, particularly for those affected by healthcare disparities [15]. Demographic characteristics such as age, sex, race, and ethnicity are each linked to dementia prevalence and risk [16–18]. Thus, certain population sub-groups could be at greater odds for undiagnosed dementia due to poor healthcare access and quality of care [19]. Distinguishing those living with a cognitive impairment consistent with dementia (CICD), but no dementia-related diagnosis, may aid in identifying and treating severe cognitive impairment. We sought to estimate the prevalence of no reported dementia-related diagnosis in a nationally representative sample of older Americans with a CICD.

MATERIALS AND METHODS

Participants

The weighted analytic sample included 6,036,224 Americans aged ≥ 65-years that were identified as having a CICD according to an adapted version of the Telephone Interview of Cognitive Status (TICS without history of stroke, cancers, neurological conditions, or brain damage who participated in at least one-wave of the 2010–2016 Health and Retirement Study (HRS). The HRS is a longitudinal-panel study that monitors health factors during aging. Participants engage in core interviews biennially until death Interview response rates for the HRS have overall been >80% [20]. More details about the HRS an available elsewhere [21].

Although the HRS implements a panel design, we analyzed the last wave in which each person participated to best estimate the prevalence of no reported dementia-related diagnosis for older Americans with a CICD. Participants provided written informed con sent before entering the HRS and the University’: Behavioral Sciences Committee Institutional Review Board approved study protocols.

Measures

Dementia-related diagnosis

Starting in the 2010 wave, the HRS began including additional self-report measures of diagnosed health conditions. Respondents told interviewers if doctor (i.e., healthcare provider) had ever diagnosed them with dementia, senility, or any other serious memory impairment. Those indicating an affirmative healthcare provider diagnosis were considered as having dementia. When participants are re-interviewed in the HRS, they can dispute their previously recorded self-reported dementia response. This single-item self-report measure of dementia-related diagnosis from the HRS has been used in other similar studies [22–24] and has shown a moderate level of agreement with Hurd et al. [4] models for dementia [22].

Cognitive impairment consistent with dementia

The adapted TICS, which is a well-validated screening tool that was designed for population-based studies such as the HRS [25], was used to assess cognitive functioning. A 27-point composite scale that included immediate and delayed word recall from list of 10 noun-free words to measure memory, serial sevens subtraction test beginning with the number 100 to measure working memory, and counting back ward at maximal speed for 10 consecutive number: starting from 20 to measure mental processing speed Those with scores ≤ 6 were considered as having a CICD [26]. This CICD cut-point correctly classifies approximately 80% of respondents as having dementia or not [26, 27]. More information about how cognitive functioning was assessed in the HRS is accessible elsewhere [28].

Demographic characteristics

Participants reported their age, sex, educational achievement (did not complete high school; at least completed high school or passed an equivalency exam), race and ethnicity (Hispanic, non-Hispanic black, non-Hispanic white, non-Hispanic other (includes American Indian, Alaskan Native, Asian, Native Hawaiian, and Pacific Islander)).

Statistical analysis

All analyses were conducted with SAS 9.4 (SAS Institute; Cary, NC). The analyses accounted for the complex sampling design to produce population-based estimates that were nationally representative for older Americans according to HRS analytic guidelines [29]. Logistic regression was used to model the crude proportions of no reported dementia-related diagnosis for those living with a CICD. Multiple logistic regression was used to model age-, sex-, education-, and race- and ethnicity-adjusted proportions of no reported dementia-related diagnosis for those with a CICD. Adjusted estimates were reported for making comparisons across demographic characteristics. Unadjusted and adjusted logistic regression models were also performed for each individual sex, educational achievement, race and ethnicity subgroup. The results from the adjusted models were considered our principal results. Given that HRS waves were analyzed cross-sectionally in our study, we examined the consistency in responses for affirmative self-reported dementia-related diagnosis over the 2010–2016 HRS study period.

As a sensitivity analysis, proxy respondents with information for self-reported dementia-related diagnosis were included (weighted n =894,950). An 11-point scale that evaluated the proxy’s assessment of the respondent’s memory ranging from poor-to-excellent, the proxy’s assessment of whether the respondent had limitations in five instrumental activities of daily living, and the survey interviewer’s examination of if the respondent had difficulty completing the interview because of a cognitive impairment was utilized [26, 27]. Scores ≥ 6 were considered as having a CICD [26, 27]. With proxy respondents included, multiple logistic regression was used to model the adjusted proportions of no reported dementia-related diagnosis for those with a CICD. In cases where proxy respondents had missing information for self-reported dementia-related diagnosis (an additional weighted n = 684,097), we imputed such missing cases as either having or not having a self-reported dementia related diagnosis. Separate multiple logistic regression models were again conducted to quantify the adjusted proportions of no reported dementia-related diagnosis for persons with a CICD. A present or absent self-reported dementia diagnosis from the previously missing response was used to demonstrate a best- and worst-case scenario for these respondents.

As an additional sensitivity analysis, persons that were identified as having a CICD but missing self-reported dementia-related diagnosis measurements (weighted n = 208,328) were examined. For these individuals, we first imputed that they had a self-reported dementia-related diagnosis, and then we imputed as no self-reported dementia-related diagnosis. The logistic models were likewise conducted and the range in the prevalence of no self-reported dementia-related diagnosis when accounting for these missing responses was presented as supplementary. An alpha level of 0.05 was used for all analyses.

RESULTS

The weighted characteristics for older Americans living with a CICD are shown in Table 1. Participants were aged 80.4 (95% confidence interval (CI): 79.8–81.0) years. Table 2 presents the weighted characteristics for persons with a CICD by race and ethnicity. Interestingly, those identifying as non-Hispanic white were older (81.6 years; CI: 80.8–82.4) than persons that identified as Hispanic (78.2 years; CI: 76.8–79.6), non-Hispanic black (78.3 years; CI: 77.3–79.4), and non-Hispanic other (76.5 years; CI: 73.3–79.8). Further, there was a lower proportion of reported educational achievement at high school graduate or above for persons identifying as Hispanic (23.2%; CI: 16.3%−30.1%) compared to those identifying as non-Hispanic black (45.6%; CI: 39.7%−51.6%), non-Hispanic white (73.8%; 70.6%−77.0%), and non-Hispanic other (54.5%; CI: 35.4%−73.6%). Table 3 lists the weighted characteristics for those living with a CICD by reported dementia-related diagnosis status. Those with no reported dementia-related diagnosis were aged 80.0 (CI: 79.4–80.6) years, while those with a reported dementia-related diagnosis were aged 83.6 (CI: 82.0–85.2) years.

Table 1.

Weighted Characteristics for Those with a Cognitive Impairment Consistent with Dementia

	Weighted Frequency	Weighted Percentage	95% Confidence Interval
Ethnicity and Race
Hispanic	895,868	14.8	12.7–17.0
Non-Hispanic Black	1,112,518	18.4	16.3–20.5
Non-Hispanic White	3,847,805	63.8	60.9–66.5
Non-Hispanic Other‡	180,033	3.0	1.9–4.1
Sex
Male	2,383,366	39.5	36.5–42.5
Female	3,652,858	60.5	57.5–63.5
Educational Achievement
Not a High School Graduate	2,379,881	39.4	36.6–42.3
High School Graduate or Above	3,656,343	60.6	57.7–63.4

^‡includes American Indian, Alaskan Native, Asian, Native Hawaiian, and Pacific Islander.

Table 2.

Weighted Characteristics for Persons with a Cognitive Impairment Consistent with Dementia by Ethnicity and Race

	Age (y)	95% Confidence Interval	Female (%)	95% Confidence Interval	High School Graduate or Above(%)	95% Confidence Interval
Ethnicity and Race
Hispanic	78.2	76.8–79.6	65.5	58.1–73.0	23.2	16.3–30.1
Non-Hispanic Black	78.3	77.3–79.4	60.8	55.0–66.5	45.6	39.7–51.6
Non-Hispanic White	81.6	80.8–82.4	59.4	55.5–63.3	73.8	70.6–77.0
Non-Hispanic Other‡	76.5	73.3–79.8	56.6	37.8–75.5	54.5	35.4–73.6

Table 3.

Weighted Characteristics for Those with a Cognitive Impairment Consistent with Dementia by Reported Dementia-Related Diagnosis Status

	Living with a Cognitive Impairment Consistent with Dementia
	Reported Dementia-Related Diagnosis (n = 683,571)			No Reported Dementia-Related Diagnosis (n = 5,352,653)
	Weighted Frequency	Weighted Percentage	95% CI	Weighted Frequency	Weighted Percentage	95% CI
Ethnicity and Race
Hispanic	107,030	15.7	9.7–21.6	788,838	14.7	12.5–17.0
Non-Hispanic Black	80,225	11.7	6.9–16.6	1,032,293	19.3	17.0–21.5
Non-Hispanic White	489,181	71.6	64.2–78.9	3,358,624	62.7	59.7–65.8
Non-Hispanic Other†	7,135	1.0	0.1–2.5	172,898	3.3	2.0–4.4
Sex
Male	287,712	42.1	33.5–50.6	2,095,654	39.2	36.0–42.3
Female	395,859	57.9	49.3–66.5	3,256,999	60.8	57.7–64.0
Educational Achievement
Not a High School Graduate	227,122	33.2	25.4–41.1	2,152,759	40.2	37.1–43.3
High School Graduate or Above	456,449	66.8	58.9–74.6	3,199,894	59.8	56.7–62.9

^†includes American Indian, Alaskan Native, Asian, Native Hawaiian, and Pacific Islander. CI, confidence interval.

The weighted crude and adjusted prevalence of no reported dementia-related diagnosis for those with a CICD are in Table 4. The overall adjusted prevalence of no reported dementia-related diagnosis for those with a CICD was 91.4% (CI: 87.7%−94.1%). Persons with a CICD who identified as non-Hispanic black had a high prevalence of no reported dementia-related diagnosis (93.3%; CI: 89.8%−95.6%). The estimated prevalence of no reported dementia-related diagnosis was higher in males with a CICD (99.7%; CI: 99.6%−99.8%) than females (90.2%; CI: 85.6%−93.4%). Moreover, the estimated prevalence of no reported dementia-related diagnosis for non-high school graduates with a CICD was 93.5% (CI: 89.3%−96.1%), but 90.9% (CI: 84.7%−94.7%) for those with at least a high school education. The evaluation of the consistency in responses for our self-reported dementia-related diagnosis measure revealed that there were no disputes (i.e., changes) in affirmative self-reported dementia-related diagnosis, such that no persons who reported a dementia-related diagnosis later changed their response to not having a dementia-related diagnosis.

Table 4.

Weighted Crude and Adjusted Prevalence of No Reported Dementia-Related Diagnosis for Those with a Cognitive Impairment Consistent with Dementia

	Crude Prevalence		Adjusted Prevalence
	%	95% Confidence Interval	%	95% Confidence Interval
Overalla	88.7	86.7–90.4	91.4	87.7–94.1
Ethnicity and Race b
Hispanic	88.1	82.6–91.9	92.0	86.5–95.3
Non-Hispanic Black	92.8	89.2–95.2	93.3	89.8–95.6
Non-Hispanic White	87.3	84.5–89.6	88.3	85.3–90.7
Sex c
Male	87.9	84.6–90.6	99.7	99.6–99.8
Female	89.2	86.5–91.3	90.2	85.6–93.4
Education d
Not a High School Graduate	90.5	87.7–92.7	93.5	89.3–96.1
High School Graduate or Above	87.5	84.7–89.9	90.9	84.7–94.7

^aadjusted for age, sex, education, ethnicity and race

^badjusted for age, sex and education

^cadjusted for age, education, ethnicity and race

^dadjusted for age, sex, ethnicity and race.

Supplementary Table 1 shows the adjusted prevalence of no reported dementia-related diagnosis for those with a CICD when including proxy respondents. The overall adjusted prevalence of no reported dementia-related diagnosis for those with a CICD when including proxy respondents was 83.2% (CI: 79.6%−86.3%). Further, the overall adjusted prevalence of no reported dementia-related diagnosis for persons with a CICD when including proxy respondents imputed as not having a self-reported dementia related diagnosis was 84.3% (CI: 80.9%−87.2%), but the prevalence estimate decreased to 74.9% (CI: 70.8%−78.6%) when proxy respondents were imputed as having a self-reported dementia-related diagnosis. Supplementary Table 2 presents the adjusted prevalence of no reported dementia-related diagnosis for those with a CICD when evaluating individuals with missing self-reported dementia-related diagnosis. The overall adjusted prevalence of no reported dementia-related diagnosis for persons with a CIDC when imputing previously missing responses as not having a self-reported dementia-related diagnosis was 91.7% (CI: 88.1%−94.3%), but the prevalence estimates similarly decreased to 84.8% (CI: 80.6%−88.3%) when previously missing responses were imputed as having a self-reported dementia-related diagnosis.

DISCUSSION

The principal results of this investigation revealed that the estimated prevalence of no reported dementia-related diagnosis among older Americans living with a CICD was substantial. Specifically, the adjusted overall estimated prevalence for no reported dementia-related diagnosis in those with a CICD was 91.4%. The sub-group analyses also revealed similar prevalence estimates for no reported dementia-related diagnosis. These findings suggest that self-reported dementia-related diagnoses underestimates the number of older adults living with a CICD. We recommend that healthcare providers screen for low cognitive functioning during routine health assessments when possible, and not rely entirely on self-reported dementia diagnoses. Improving access to care for disadvantaged sociodemographic sub-groups may likewise support the occurrence of screening for severe cognitive impairment.

A variety of factors may explain why we found that the estimated prevalence of no reported dementia-related diagnosis in older Americans with a CICD was 91.4%. First, dementia-related conditions may be undiagnosed. Caregivers, including adult children (or similar) who may provide home healthcare, should not ignore symptoms of declining cognitive functioning because this may delay dementia diagnosis and relevant treatment [30]. Jacobson et al. [31] also revealed that Medicare annual wellness visits (Part B) requires detection of cognitive impairment, but approximately half of beneficiaries reported having an annual wellness visit, and less than a third reported having a structed cognitive assessment. These findings are alarming because older adults may not be visiting their healthcare provider routinely, and required assessments of cognitive functioning during these visits are not being conducted. Acknowledging risk factors, maintaining communication with a healthcare provider, and conducting cognitive function screenings during routine health visits may help to lower the prevalence of no dementia-related diagnoses for Americans living with a CICD.

Undiagnosed dementia-related conditions could similarly be explained by other risk factors [19]. The higher estimated prevalence of no dementia-related diagnosis for non-Hispanic black adults living with a CICD further highlights diagnosis issues and health disparities in risk factors [32]. Studies that examine dementia care and include community-based participatory research may help to uncover strategies that lower undiagnosed dementia prevalence [32]. Given that there could be sex-specific differences in dementia risk, screening for dementia in both men and women will help to provide the necessary treatment and care for slowing dementia [17]. Continuing to study race- and sex-specific risk factors may provide new opportunities for targeted dementia care interventions.

Alternatively, it is possible that persons with a healthcare provider dementia diagnosis did not self-report their diagnosis. Indeed, this lack of self-reporting their diagnosis may reflect an inability to recall having been told they have such a diagnosis. However, this has important implications. When older adults visit clinical settings, their self-report should be verified using cognitive screening and examination of prior medical records to ensure that dementia is not being missed. This study utilized a single-item measure for self-reported dementia-related diagnosis from the HRS. Similar self-reported measures such as subjective cognitive decline in population-based studies (National Health and Nutrition Examination Survey [33]) have been shown to be independently associated with objectively-measured low cognitive performance [34]. Our study used the adapted TICS and identified a high prevalence of persons with CICD, but no self-reported dementia-related diagnosis. Another study found that survey-based cognitive tests yielded equivalent dementia prevalence estimates using claims-based dementia diagnoses [35]. Yet, others have shown that self-reported dementia using survey data (e.g., HRS) is a critical source for case ascertainment that may identify more dementia cases than diagnosis codes in medical claims [22, 36]. Thus, it is possible that our study revealed additional dementia-related cases than what a claims-based diagnosis would have revealed. Nevertheless, continued research into the utility of self-reported measures of dementia related diagnoses, especially in population-based studies, is warranted.

Some study limitations should be acknowledged. The diagnostic criteria used for determining dementia in clinical and epidemiological assessments may vary and have implications on how dementia is operationalized [37]. We used the last wave that each person participated in the HRS to estimate the prevalence of no self-reported dementia-related diagnosis for older Americans with a CIDC, so examining changes in self-reported dementia-related diagnosis and adapted TICS scores was not possible. While the adapted TICS is well-validated for assessing cognitive functioning and the scoring criteria we used for determining CICD is robust [25–27], the CICD designation does not indicate that an individual has dementia, but does suggest that cognitive functions are sufficiently impaired and a comprehensive cognitive assessment should be urgently pursued [38]. Proxy respondents were excluded from the principal results because they did not have information for the adapted TICS. Although healthcare provider dementia diagnosis was self-reported, HRS self-report information is generally reliable and valid [39]. Further, persons with severe cognitive impairment such as those in our study are more likely to report dementia relative to individuals with milder cognitive impairment [22, 40]. Adjusting for additional variables in our analyses may have taken interpretability away from our estimates.

CONCLUSIONS

This investigation revealed that the overall estimated prevalence of no reported dementia-related diagnosis in older Americans living with a CICD was 91.4%. Similar estimates were found when examining specific race and ethnicity, education, and sex sub-groups. Our findings indicate that self-reported dementia diagnosis underestimates the number of Americans living with a CICD. We recommend more awareness and efforts be given to diagnosing dementia in clinical settings, including screening for dementia during routine health assessments. Such awareness and efforts may lead to appropriate referrals for possible dementia, and treatment after diagnosis.