Identification of Major Cognitive Disorders in Self-Reported versus Administrative Health Data: A Cohort Study in Quebec

Abstract

Background:

The first imperative in producing the relevant and needed knowledge about major neurocognitive disorder (MNCD) is to identify people presenting with the condition adequately. To document potential disparities between administrative health databases and population-based surveys could help identify specific challenges in this population and methodological shortfalls.

Objective:

To describe and compare the characteristics of community-dwelling older adults according to four groups: 1) No MNCD; 2) Self-reported MNCD only; 3) MNCD in administrative health data only; 4) MNCD in both self-reported and administrative health data.

Methods:

This retrospective cohort study used the Care Trajectories-Enriched Data (TorSaDE) cohort, a linkage between five waves of the Canadian Community Health Survey (CCHS) and health administrative health data. We included older adults living in the community who participated in at least one cycle of the CCHS. We reported on positive and negative MNCD in self-reported versus administrative health data. We then compared groups’ characteristics using chi-square tests and ANOVA.

Results:

The study cohort was composed of 25,125 older adults, of which 784 (3.1%) had MNCD. About 70% of people with an MNCD identified in administrative health data did not report it in the CCHS. The four groups present specific challenges related to the importance of perception, timely diagnosis, and the caregivers’ roles in reporting health information.

Conclusion:

To a certain degree, both data sources fail to consider subgroups experiencing issues related to MNCD; studies like ours provide insight to understand their characteristics and needs better.

Keywords

Alzheimer’s disease cross-sectional surveys data accuracy dementia linked data self-reported data

INTRODUCTION

Globally, major neurocognitive disorder (MNCD) affects as many as 50 million people, and this number is expected to double over the next 30 years [1]. The evolution of the disease is associated with disability and dependency, as it affects memory, cognitive abilities, and behavior [1]. Because of its high prevalence, associated multimorbidity, cost, and burden of care for families and the healthcare system, MNCD has risen as a priority for public health research [1].

The first imperative in producing the relevant and needed knowledge about MNCD is to identify people presenting with the condition adequately. Researchers may use various data sources, including administrative health databases, offering access to real-world data and a large population base. These sources, however, do not contain some relevant individual variables, such as financial status, education, and self-perceived variables [2]. MNCD case identification is also challenging, as the accuracy of MNCD codes in routinely collected administrative health data might vary. A systematic review by Wilkinson et al. [3] found wide variations in validation studies of MNCD coding using administrative health databases. Indeed, sensitivity ranged from 21% to 86%, indicating the inherent complexity of the condition and highlighting the absence of a solid gold standard for MNCD confirmation [3, 4].

Population-based surveys—another frequently used data source—is an accessible way to collect essential information about people living with MNCD, such as healthcare needs, health behavior, quality of life, and social support [5, 6]. Population-based surveys often rely on self-assessments. Thus, one challenge is the accuracy of self-reported health conditions since some conditions, like MNCD, are prone to biased responses [7]. Several factors can affect the accuracy of self-reported conditions, including age, type, disease severity, and memory capacity [8, 9]. Awareness and recognition of MNCD vary particularly between individuals, which can be accounted for by several factors. For example, about 40% of patients with Alzheimer’s disease exhibit some form of anosognosia, that is, an unawareness of the disease and its deficits [10 –12]. In addition, obtaining a diagnosis of MNCD in primary-care settings (the patient’s entry point in healthcare services) is often a complex process [12]. Indeed, significant MNCD cases go undetected for years after individuals start developing symptoms [13, 14]. Moreover, the practices and attitudes regarding MNCD diagnosis disclosure tend to vary significantly in terms of when and by whom the diagnosis is communicated, whom is told, and what terms are used [14].

Since survey and administrative health data present unique data quality challenges, disparities can be expected between diagnoses obtained from both sources.

Only a few studies have advanced our knowledge on the subject by considering awareness of the disease. However, the method used for MNCD identification varies as it is not based on a validated algorithm (for health administrative data), relies on screening tests, or does not consider unawareness of MNCD [15 –17].

In the absence of clinical cohort studies, which are costly to conduct on an extensive basis, the combination of population-surveys and administrative health data is of particular interest to examine these variations [2]. Thus, this combination could help identify strategies to overcome methodological shortfalls and develop an integrated approach to identifying people living with MNCD [18].

With this in mind, we aimed to describe and compare the characteristics of community-dwelling older adults according to four groups: 1) No MNCD; 2) Self-reported MNCD only; 3) MNCD in administrative health data only; 4) MNCD in both self-reported and administrative health data.

METHODS

Design and data source

This is a retrospective cohort study in the province of Quebec, Canada. The Commission d’accès à l’information of the Institut de la statistique du Québec, and McGill’s Research Ethics Board approved the study. This study used data from the Care Trajectories -Enriched Data (TorSaDE) cohort (n = 81,093 distinct participants), a linkage between two complementary sources: Canadian Community Health Survey (CCHS) and administrative health data from the provincial health-insurance board (Régie de l’assurance maladie du Québec: RAMQ) [2].

The TorSaDE cohort includes all Quebec (Canada) respondents who participated in at least one of the five cycles (2007–08, 2009–10, 2011–12, 2013–14, 2015-16) of the CCHS and agreed to their data to be shared (sharing consent rate of about 96% for all CCHS cycles). Briefly, the CCHS is an annual Canadian cross-sectional survey of self-reported data, and its sampling base represents about 97% of the Canadian population aged 12 years or more. It gives access to health status, healthcare utilization, and health determinants. In addition to the health component, the survey includes questions about respondent characteristics. The CCHS questionnaires responses were linked with the participants’ health administrative data over 21 years (1996–2016) [2]. The RAMQ provides universal health-insurance coverage to all residents of the Province of Quebec, thus a nearly exhaustive sample. Coverage includes services provided in emergency departments, hospitals, and medical clinics, including primary health care clinics.

The RAMQ administrative health register gives access to a large range of variables, including: 1) patient demographic information (date of birth, date of death if applicable, postal code); 2) medical services register (data on medical services provided by Quebec fee-for-service physicians, including diagnoses coded according to the International Classification of Diseases 9 (ICD-9)); 3) pharmaceutical services (data on pharmacy-claimed drugs, for patients covered by the public drug insurance plan –about 85% of people 65 years old and over); 4) MED-ECHO registry (information on hospitalization, including diagnosis coded in ICD-10).

Study population

The study population included individuals from Quebec (Canada) who: 1) participated in at least one cycle of the CCHS, 2) were living in the community, and were at least 65 years old at the time of the CCHS completion.

Variables

Identification of people living with MNCD

Using the RAMQ data, we identified MNCD cases before the date of CCHS completion using a validated algorithm. The latter was developed and validated in Ontario by the Institute for Clinical Evaluative Science (ICES) and adopted by the Public Health Agency of Canada. The definition of MNCD used to develop the algorithm includes Alzheimer’s disease, vascular dementia, dementia in other diseases classified elsewhere (frontotemporal dementia, idiopathic normal pressure hydrocephalus), and unspecified dementia (senile dementia, presenile dementia). The ICD-9 and 10 codes used are: ICD-9 (46.1, 290.0, 290.1, 290.2, 290.3, 290.4, 294.x, 331.0, 331.1, 331.5, 331.82); ICD-10 (F00.x, F01.x, F02.x, F03.x, G30.x). The algorithm showed good psychometric properties, with a sensitivity of 79.3% and a specificity of 99.1% [19]. An individual was considered as having a MNCD if they met a least one of the following three criteria: 1) the presence of a primary or secondary diagnosis code for MNCD during hospitalization (in MED-ECHO); 2) the presence of three MNCD diagnosis codes register at least 30 days apart in a two-year period (in the medical services register); 3) a prescription claimed for an Alzheimer’s disease-related drug in the pharmaceutical services register (cholinesterase inhibitor subclass, including donepezil, galantamine, rivastigmine, tacrine, and memantine). The date of MNCD identification was when the first of the three criteria became positive.

Using the surveys data, we identified self-reported diagnoses of MNCD using the answer to a specific question: Do you have Alzheimer’s Disease or any other dementia? (Yes/No/ Don’t know/ Don’t want to answer).

Socio-demographic characteristics

In the surveys, we considered various variables at CCHS completion: age, sex, living alone, education level, residential area, and CCHS completed by a third party (allowed only if the initial respondents’ physical or mental condition prevented them from answering).

We also considered individual income with a threshold of 20,000$ (CAD), corresponding approximatively to the mean low-income cut-offs for the years of our study [20].

We also considered self-perceived health and mental health status, and it is to note that the question relative to perceived mental health status was passed over when a third-party completed the CCHS.

Health characteristics

Using RAMQ data, we calculated 1) the Combined Charlson and Elixhauser Comorbidity index (CI) as proposed by Simard et al. [21] in the year preceding the date of completion of the CCHS (a lower score represents lower comorbidity); 2) the proportion of MNCD identified during an episode of hospitalization; 3) MNCD identified in the two years following the index date.

Variables on the use of services

Using the RAMQ data, the number of visits to a family physician, the number of hospitalization days and the mean of drugs taken over a month, including dementia-related drugs, were computed one year before the CCHS completion.

Analysis

Statistical analyses were performed using SAS Enterprise Guide 8.3 (SAS Institute, Cary, NC). Data were validated and checked for outliers and missing values. The study sample was divided into four groups: 1) No MNCD; 2) Self-reported MNCD only; 3) MNCD in administrative health data only; 4) MNCD in both self-reported and administrative health data. We compared MNCD through self-reported and administrative health data and compiled the characteristics of our four groups. The difference between subgroups was tested using the chi-square test for categorical variables and ANOVA for continuous variables.

As a complement, we accounted for the effect of age and sex using 1) logistic regression analysis for categorical variables, and 2) general linear models for continuous variables. Finally, where significant differences were identified, we conducted post hoc analysis using multiple pairwise comparisons (for chi-square test) and Tukey’s test (for ANOVA) [22]. We chose a significance level of 0.001 for all analysis.

RESULTS

The study cohort was composed of 25,125 older adults, of which 784 (3.1%) self-reported MNCD in the surveys and/or had the condition identified in the administrative health data. A low proportion of missing values was reported in the data (less than 10%) for all variables, so no further actions were taken.

Table 1 shows the differences between MNCD cases reported in the CCHS survey and administrative health data. There is discordance for 551 individuals (70.3%). Only about 34% of individuals with an MNCD identified in the administrative health data self-reported the condition. By comparison, 70.4% of individuals who self-reported MNCD also presented the condition in the administrative health data.

Table 1

Major cognitive disorders in self-reported versus administrative health data

		Self-reported assessment in CCHS survey		Total
		Negative MNCD	Positive MNCD
Administrative health data	Negative MNCD	24,341*	98*	24,439 (97.3)
		(99.6)	(0.4)
		(98.2)	(29.6)
	Positive MNCD	453*	233*	686 (2.7)
		(66.0)	(34.0)
		(1.8)	(70.4)
Total		24,794 (98.7)	331 (1.3)	25,125 (100)

*The information represent respectively: Frequency / Row percentage/ Column percentage. CCHS, Canadian Community Health Survey; MNCD, major neurocognitive disorder.

Characteristics of individuals

Table 2 shows the characteristics of the cohort, distributed according to four distinct groups: 1) No MNCD (n = 24,341; 96.9%); 2) Self-reported MNCD only (n = 98; 0.4%); 3) MNCD in administrative health data only (n = 453; 1.8%); and 4) MNCD in both self-reported and administrative health data (n = 233; 0.9%).

Table 2

Characteristics of the overall sample and individuals assigned to the four groups

Variables	Total population	Group 1 No MNCD	Group 2 Self-report only	Group 3 AHD only	Group 4 Self-report and AHD	p ^†	p adjusted for age and sex	Pairwise comparisons^‡
Total	25,125 (100)	24,341 (96.9)	98 (0.4)	453 (1.8)	233 (0.9)	NA	NA	NA
Age	73.8±6.9	73.6±6.8	79.5±8.3	81.0±6.8	80.6±6.4	<0.0001	NA	1¬=2,3,4
Female	14,455 (57.5)	13,971 (57.4)	51 (52.0)	307 (67.8)	126 (54.1)	<0.0001	NA	1¬=3 3 ¬= 4
Living alone	11,878 (47.3)	11,533 (47.4)	36 (36.7)	259 (57.2)	50 (21.6)	<0.0001	<0.0001	1¬=3,4 2 ¬= 3 3 ¬= 4
Yearly income
20,000 or less	5,841 (23.2)	5,602 (23.0)	31 (31.6)	167 (36.9)	41 (17.6)	<0.0001	<0.0001	1¬=3
More than 20000	19,302 (76.8)	18,757 (77.0)	67 (68.4)	286 (63.1)	192 (82.4)	3 ¬= 4
Education level
High school not completed	10, 956 (44.5)	10,503 (44.0)	56 (62.8)	268 (61.1)	126 (57.8)	<0.0001	0.0002	1¬=2, 3,4
High school at least completed	13,676 (55.5)	13,378 (56.0)	35 (37.2)	171 (39.0)	92 (42.2)
Residential area
Metropolitan	12, 899 (51.3)	12,498 (51.3)	48 (49.0)	232 (51.2)	118 (50.6)	0.2571	0.4102	NA
Small town	4,356 (17.3)	4,196 (17.2)	18 (18.4)	90 (19.9)	52 (22.3)
Rural	7,892 (31.4)	7,665 (31.5)	32 (32.7)	131 (28.9)	63 (27.0)
Survey completed by a third party	1059 (4.2)	797 (3.3)	45 (45.9)	55 (12.1)	162 (69.5)	<0.0001	<0.0001	1¬=2,3,4 1 ¬= 2,3,4 3 ¬= 4
Comorbidity index	0.72±1.8	0.67±1.7	1.49±2.4	2.37±3.0	2.45±2.9	<0.0001	<0.0001	1¬=2,3,4 2 ¬= 3,4
MNCD identified in the 2 years following the index date	NA	342 (1.4)	15 (15.3)	NA	NA	<0.0001	<0.0001	NA
Years since Dx RAMQ	NA	NA	NA	3.18±3.2	2.94±2.5	0.3253	0.5202	NA
MNCD identified during an episode of hospitalization	NA	NA	NA	140 (30.9)	31 (13.3)	<0.0001	<0.0001	NA
Perceived health status
Poor/ Fair	5,191 (20.7)	4,846 (19.9)	54 (55.1)	157 (34.7)	134 (57.8)	<0.0001	<0.0001	1¬=2,3 2 ¬= 3 3 ¬= 4
Good/Very good/ Excellent	19,919 (79.4)	19,481 (80.1)	44 (44.9)	296 (65.3)	98 (42.2)
Perceived mental health status
Poor/Fair	904 (3.8)	826 (3.5)	52 (9.4)		26 (36.6)	<0.0001	<0.0001	1¬=2,3,4 3 ¬= 4
Good/ Very good/ Excellent	23,143 (96.2)	22,700 (96.5)	398 (90.6)		45 (63.4)
Past family physician visits	3.14±3.12	3.11±3.10	3.43±3.11	4.24±3.60	3.68±3.22	<0.0001	0.0002	1¬=3
Past hospitalization days	1.48±7.45	1.31±6.5	2.7±8.86	6.9±17.66	7.7±28.28	<0.0001	<0.0001	1¬=3,4 2 ¬= 3,4
Average number of monthly medications	8.51±5.95	8.39±5.89	10.5±7.42	12.51±6.71	11.17±6.36	<0.0001	<0.0001	1¬=2,3,4
MNCD-related medication (Yes)	NA	NA	NA	227 (50.1)	174 (74.7)	<0.0001	<0.0001	NA

^†The χ² test for dichotomous variables and ANOVA analysis for continuous variables. ^‡ Pairwise comparisons: equal except stated otherwise. ADD, administrative health data. MNCD, major neurocognitive disorder. NA, non-applicable. Percentages in parenthesis are relative to the column total. Statistical significance (p≤0.001).

The results show statistically significant differences between the groups for all variables, except residential area and years since MNCD identification. Also, for each variable, controlling for age and gender did not significantly modify our results. The results of the post hoc analysis are presented in Supplementary Tables 1 and 2

Group 1 had higher educational levels (about 56% had completed at least high school) and mental-health status (n = 22,700; 96.5%), as well as the lowest number of medications ( $\bar{x}$ = 8.4±5.9) and comorbidity index ( $\bar{x}$ = 0.7±1.7). Individuals from this group also showed the highest perceived health status (n = 19,481; 80.1%). Regarding individuals in group 2, 15.3% had a MNCD identified in the RAMQ data in the two-year following the index date. More than half of them reported a Poor/Fair perceived health status (n = 54; 55.1%), and they showed the lowest comorbidity index ( $\bar{x}$ = 1.5±2.4) among individuals with identified or self-reported MNCD.

Group 3 contained the highest number of people living alone (n = 259; 57.2%) and the highest proportion of people diagnosed during hospitalization compared to other groups. It should also be noted that people in groups 2 and 3 reported a better perceived mental health status than individuals in group 4 (n = 398; 90.6%). Compared to other groups, individuals in group 4 were the least prone to live alone (n = 50; 21.6%), and most of their CCHS surveys were completed by a third party (n = 162; 69.5%). They also presented the highest comorbidity index ( $\bar{x}$ = 2.5±2.9) compared to groups 1 and 2. In addition, most reported poor/fair perceived health (n = 134; 57.8%) and mental-health status (n = 26; 36.6%).

It should also be noted that Table 1 merges the information for groups 2 and 3 related to perceived mental health status because of the small cell size and confidentiality concerns. However, post hoc analysis showed no significant differences between groups 2 and 3 for this variable.

DISCUSSION

This study used distinctive groups to compare the identification of MNCD in self-reported and administrative health data in community-dwelling older adults. The results show discordance and significant differences in several aspects, as 70% of people with an MNCD identified in administrative health data did not report it in the CCHS. We also found differences between our four groups: 1) No MNCD; 2) Self-reported MNCD only; 3) MNCD in administrative health data only; 4) MNCD in both self-reported and administrative health data.

Such inconsistencies are reported in geriatric populations for some health conditions. A study by Singh [23] aimed to examine how the self-reporting of chronic disease agreed with database diagnoses in a Veterans Affairs healthcare setting. Among individuals presenting with a clinical diagnosis, 91% with diabetes self-reported the condition, while this proportion was 86%, 83%, and 73% for heart disease, depression, and pulmonary disease, respectively. A higher level of concordance is expected in conditions representing a higher care burden, such as diabetes, implying daily management [7, 21]. Moreover, a recent longitudinal study by Amjad et al. [15] aimed to determine whether undiagnosed dementia or unawareness affects the risk of hospitalization or ED visits, using a large national study on aging. In parallel to our results, a majority of individuals were not aware of their condition (≈45% versus ≈57% in our study), only about 26% were aware (versus about 30% in our study), and about 28% were undiagnosed. This last information is harder to compare with our results, as they used screening questionnaires and cognitive testing to identify undiagnosed individuals. In contrast, our self-reported data may reflect more on the individual perception of the cognition and indicate underdiagnosis.

The literature also raises interesting hypotheses regarding our results. Group 2 was composed of people only self-reporting MNCD. This could reflect the presence of conditions whose evolution might lead to MNCD—such as subjective cognitive decline (SCD), which has a prevalence of around 10% in older adults—or a mild cognitive impairment (MCI), which affects up to 18% of people aged 60 years or older [24, 25]. Parallel to this, individuals in group 2 mainly reported poor/fair perceived health status, and a poor/fair perceived mental-health status that was three times higher than that of the general population. Presenting with SCD or MCI is related to adverse effects, such as lower health-related quality of life, which is a concept closely related to perceived health and mental-health status [24, 26].

As the condition is frequently under-recognized, challenges in obtaining an MNCD diagnosis must also be considered. Indeed, Lang et al. [12] performed a meta-analysis showing that about 63% of MNCD cases went undetected in community settings and that the detection rate was even lower with milder conditions. Moreover, only 38% of older adults presenting with confusion or memory loss reported discussing their concerns with a healthcare professional. The mean duration between the onset of MNCD symptoms and assessment or diagnosis can be up to 2 years [25, 27].

In the early stage of MNCD, once the situation has been discussed with healthcare professionals, it can take more than a year to get a diagnosis, as that requires multiple assessments to eliminate other conditions with dementia-like symptoms [28, 29]. With respect to our results, this might have resulted in the diagnosis going unrecognized, as some individuals might have obtained a diagnosis in the subsequent months or years. Thus, even without a positive MNCD condition in administrative health data, we must still consider that people reporting the condition experienced related challenges. Thus, in community settings, even the early stages of MNCD have been associated with unmet needs [30]. Individuals with milder MNCD have even reported significantly higher unmet needs, as those can be less apparent to care providers [31].

Another relevant part of our results points to the importance of an informant or caregiver in reporting health information. The input of a close informant or caregiver is widely favored in population research on MNCD, as it ensures reliable data [32]. The significant absence of an informant in responding to the survey could help explain the discordance in group 3; individuals only presenting MNCD in administrative health data. This could indicate a subgroup of individuals failing to report their cognitive conditions, notably because of anosognosia [33]. Anosognosia was notably associated with increased neuropsychiatric symptoms and deficits, higher caregiver burden, and also with a better-perceived quality of life, and less depression [34, 35]. Parallel to this, individuals in group 3 reported the highest perceived health and mental health compared to groups 2 and 4. The variations in this group could also point to nondisclosure practices with the MNCD diagnosis. Often concerned about the negative impact on the patients, only 34% of general practitioners and 48% of specialists reported to usually inform the patient, while 89% and 97%, respectively, inform the family [27]. On the other hand, individuals from group 3 are the most prone to be diagnosed with MNCD during an episode of hospitalization. This could highlight the fact that MNCD is less severe for this group. Indeed, hospitalization is a stressful episode that could lead patients to show MNCD symptoms for the first time [36]. The fact that these individuals also live alone in a high majority could also highlight the critical role of caregivers in the early detection of dementia, as their concerns are known to drive MNCD recognition in primary care [37]. Indeed, professionals in primary care may not recognize earlier symptoms of MNCD through routine examination, and screening practices are still variables in these settings [38].

In group 4, MNCD was concordant in self-reported and administrative health data. This group is characterized by the fact that third parties responded to the survey for the most part. Parallel to this, the characteristics of group 4 could point to higher MNCD severity, as group members presented with related aspects such as the highest comorbidity index, number of medications, and MNCD-related drug prescriptions [39].

From a methodological standpoint, the limitations and strengths of both self-reported and administrative health data must be considered. Administrative health databases are regularly used for research objectives [40]. However, the fact that they were not created for this purpose can lead to issues with the quality of diagnostic data [41]. To illustrate, the physician enters the diagnostic codes in the RAMQ’s medical services register, so a particular diagnosis might not have been made according to clinical practice standards. Moreover, only one diagnosis can be reported at a time, likely giving priority to unstable or more severe conditions [42]. However, data from an administrative health database are considered the benchmark to assess the prevalence of medical conditions in the geriatric populations [43].

For their part, population-based surveys provide the opportunity to monitor health trends, including determinants of health and inequalities, but are especially prone to bias [9]. One significant limitation regarding CCHS is the absence of cognitive assessment to detect MNCD. Indeed, self-reported information is not privileged since awareness of MNCD varies between individuals [18]. The use of self-assessment questionnaires has proven effective in detecting MNCD and MCI, even without an informant [44]. The CCHS was not explicitly designed for a cognitively impaired population, which is another barrier. While the CCHS requires around 45 minutes to administer and includes more than 100 questions, surveys should be kept short and simple for individuals with MNCD [45].

Implications

According to our results, health administrative data appears to provide broader detection of MNCD. However, to a certain degree, both data sources fail to consider subgroups experiencing issues related to MNCD. They also offer a unique and different perspective: studies like ours provide the insight to understand their characteristics better. Our results could also support calls to action to improve screening and early diagnosis of MNCD.

Our work also opens avenues for further research, such as 1) deepening our understanding of the reported discordance from the standpoint of health results, inequalities and informant participation; 2) determining to what extent individuals exclusively self-reporting MNCD might receive a clinical diagnosis in the future; and 3) highlighting the effect of survey best practices for the older population, including those with MNCD.

Strengths and limitations

One of the study’s strengths involved the use of an innovative, enriched cohort, which yielded significant complementarity [2]. We also used a validated algorithm to detect MNCD conditions in the administrative health data. As for limitations, little information was available regarding MNCD severity. Moreover, we did not consider other types of cognitive impairment, such as MCI.

CONCLUSION

People living with MNCD represent a complex population whose characteristics need to be understood and thoroughly managed to enhance care quality and effectiveness. Further studies should help bridge the knowledge gaps raised by our results, especially related to databases combining self-evaluated and professionally evaluated variables.

Footnotes

ACKNOWLEDGMENTS

The authors thank the Alzheimer Society of Canada for their contribution to the financing of this publication. The authors are grateful to Dre. Josiane Courteau for her assistance in data management and to Matthew A. Garriss for his editorial assistance.

ID received a postdoctoral scholarship from the Alzheimer Society Research Program (ASRP) of Canada to conduct this study. The TORSADE Cohort was funded by the Quebec SUPPORT Unit (Support for People and Patient-Oriented Research and Trials), an initiative funded by the Canadian Institutes of Health Research (CIHR), the Ministère de la santé et des services sociaux du Québec, and the Fonds de recherche du Québec –Santé.

Authors’ disclosures available online ).

The supplementary material is available in the electronic version of this article: .

References

World Health Organization [WHO] (2017) Global action plan on the public health response to dementia 2017–2025. https://apps.who.int/iris/rest/bitstreams/1092215/retrieve, Last updated December 7, 2017, Accessed on October 6, 2021.

Vanasse

, Chiu

, Courteau

, Dorais

, Bartlett

, Zawaly

, Benigeri

(2021) Cohort Profile: The Care Trajectories-Enriched Data (TorSaDE) Cohort.. Int J Epidemiol 50, 1066–1066h.

Wilkinson

, Ly

, Schnier

, Rannikmäe

, Bush

, Brayne

, Quinn

, Sudlow

CLM

(2018) Identifying dementia cases with routinely collected health data: A systematic review. Alzheimers Dement 14, 1038–1051.

Hashimoto

, Brodt

, Skelly

, Dettori

(2014) Administrative database studies: Goldmine or goose chase? Evid Based Spine Care J 5, 74–76.

Canuel

, Gosselin

, Duhoux

, Brunet

, et Lesage

(2019) Post-Disaster Mental Health Impacts Surveillance Toolkit, https://www.inspq.qc.ca/en/publications/2676. Last updated January 12, 2012, Accessed on October 6, 2021.

Bhandari

, Wagner

(2006) Self-reported utilization of health care services: Improving measurement and accuracy. Med Care Res Rev 63, 217–235.

Fortin

, Haggerty

, Sanche

, Almirall

(2017) Self-reported versus health administrative data: Implications for assessing chronic illness burden in populations. A cross-sectional study.-E. CMAJ Open 5, E729–E733.

Short

, Goetzel

, Pei

, Tabrizi

, Ozminkowski

, Gibson

, Dejoy

, Wilson

(2009) How accurate are self-reports? Analysis of self-reported health care utilization and absence when compared with administrative data. J Occup Environ Med 51, 786–796.

Knäuper

, Carrière

, Chamandy

, Xu

, Schwarz

, Rosen

(2016) How aging affects self-reports. Eur J Ageing 13, 185–193.

10.

Turró-Garriga

, Garre-Olmo

, Calvó-Perxas

, Reñé-Ramírez

, Gascón-Bayarri

, Conde-Sala

(2016) Course and determinants of anosognosia in Alzheimer’s disease: A 12-month follow-up. J Alzheimers Dis 51, 357–366.

11.

Wilson

, Sytsma

, Barnes

, Boyle

(2016) Anosognosia in dementia. Curr Neurol Neurosci Rep 16, 77.

12.

Lang

, Clifford

, Wei

, Zhang

, Leung

, Augustine

, Danat

, Zhou

, Copeland

, Anstey

, Chen

(2017) Prevalence and determinants of undetected dementia in the community: A systematic literature review and a meta-analysis. BMJ Open 7, e011146.

13.

Liss

, Seleri Assunção

, Cummings

, Atri

, Geldmacher

, Candela

, Devanand

, Fillit

, Susman

, Mintzer

, Bittner

, Brunton

, Kerwin

, Jackson

, Small

, Grossberg

, Clevenger

, Cotter

, Stefanacci

, Wise-Brown

, Sabbagh

(2021) Practical recommendations for timely, accurate diagnosis of symptomatic Alzheimer’s disease (MCI and dementia) in primary care: A review and synthesis. J Intern Med 290, 310–334.

14.

Phillips

, Pond

, Paterson

, Howell

, Shell

, Stocks

, Goode

, Marley

(2012) Difficulties in disclosing the diagnosis of dementia: A qualitative study in general practice. Br J Gen Pract 62, e546–e553.

15.

Amjad

, Samus

, Huang

, Gundavarpu

, Bynum

JPW

, Wolff

, Roth

(2022) Acute care utilization risk among older adults living undiagnosed or unaware of dementia. J Am Geriatr Soc 70, 470–480.

16.

Savva

, Arthur

(2015) Who has undiagnosed dementia? A cross-sectional analysis of participants of the Aging, Demographics and Memory Study. Age Ageing 44, 642–647.

17.

Michalowsky

, Eichler

, Thyrian

, Hertel

, Wucherer

, Hoffmann

, Flessa

(2016) Healthcare resource utilization and cost in dementia: Are there differences between patients screened positive for dementia with and those without a formal diagnosis of dementia in primary care in Germany? Int Psychogeriatr 28, 359–369.

18.

Jeong

, Yim

, Lee

S-Y

, Lee

, Potenza

, Kwon

J-H

, Koo

, Kweon

Y-S

, Bhang

S-Y

, Choi

J-S

(2018) Discordance between self-report and clinical diagnosis of Internet gaming disorder in adolescents. Sci Rep 8, 10084.

19.

Jaakkimainen

, Bronskill

, Tierney

, Herrmann

, Green

, Young

, Ivers

, Butt

, Widdifield

, Tu

(2016) Identification of physician-diagnosed Alzheimer’s disease and related dementias in population-based administrative data: A validation study using family physicians’ electronic medical records. J Alzheimers Dis 54, 337–349.

20.

Institut de la statistique du Québec (2022) Seuils du faible revenu selon la Mesure de faible revenu (MFR), avant impôt, selon la taille du ménage, Québec, 1996-2019. 2022. https://statistique.quebec.ca/fr/produit/tableau/seuilsdu-faible-revenu-mfr-seuils-avant-impot-selon-lataille-du-menage-quebec. Accessed on June 1, 2022.

21.

Simard

, Sirois

, Candas

(2018) Validation of the Combined Comorbidity Index of Charlson and Elixhauser to predict 30-day mortality across ICD-9 and ICD-10. Med Care 56, 441–447.

22.

Lee

, Lee

(2018) What is the proper way to apply the multiple comparison test? Korean J Anesthesiol 71, 353–360.

23.

Singh

(2009) Accuracy of Veterans Affairs databases for diagnoses of chronic diseases. Prev Chronic Dis 6, A126.

24.

Roehr

, Luck

, Pabst

, Bickel

, König

H-H

, Lühmann

, Fuchs

, Wolfsgruber

, Wiese

, Weyerer

, Mösch

, Brettschneider

, Mallon

, Pentzek

, Wagner

, Mamone

, Werle

, Scherer

, Maier

, Jessen

, Riedel-Heller

(2017) Subjective cognitive decline is longitudinally associated with lower health-related quality of life. Int Psychogeriatr 29, 1939–1950.

25.

Centers for Disease Control and Prevention. (2018) Subjective Cognitive Decline: A Public Health Issue. https://www.cdc.gov/aging/data/subjective-cognitivedecline-brief.html. Last updated February 27, 2019, Accessed Octerber 6, 2021.

26.

Wion

, Hill

, Reed Bell

, Mogle

, Yates

, Bhang

(2022) The role of cognitive self-report measure type in predicting cognitive decline among older adults: A systematic review. J Geriatr Psychiatry Neurol 35, 487–511.

27.

Low

L-F

, McGrath

, Swaffer

, Brodaty

(2019) Communicating a diagnosis of dementia: A systematic mixed studies review of attitudes and practices of health practitioners. Dementia (London) 18, 2856–2905.

28.

Rizzuto

, Feldman

, Karlsson

, Dahl Aslan

, Gatz

, Pedersen

(2018) Detection of dementia cases in two Swedish health registers: A validation study. J Alzheimers Dis 61, 1301–1310.

29.

Rasmussen

, Langerman

(2019) Alzheimer’s disease - why we need early diagnosis. Degener Neurol Neuromuscul Dis 9, 123–130.

30.

Stein

, Pabst

, Luck

, Lühmann

, Heser

, Jessen

, Bickel

, Mösch

, Pentzek

, Fuchs

, Wiese

, Mamone

, König

H-H

, Brettschneider

, Werle

, Scherer

, Maier

, Weyerer

, Riedel-Heller

(2017) Unmet care needs in the oldest old primary care patients with cognitive disorders: Results of the AgeCoDe and AgeQualiDe Study. Dement Geriatr Cogn Disord 44, 71–83.

31.

Black

, Johnston

, Leoutsakos

, Reuland

, Kelly

, Amjad

, Davis

, Willink

, Sloan

, Lyketsos

, Samus

(2019) Unmet needs in community-living persons with dementia are common, often non-medical and related to patient and caregiver characteristics. Int Psychogeriatr 31, 1643–1654.

32.

Galvin

(2018) Using informant and performance screening methods to detect mild cognitive impairment and dementia. Curr Geriatr Rep 7, 19–25.

33.

Kelleher

, Tolea

, Galvin

(2016) Anosognosia increases caregiver burden in mild cognitive impairment. Int J Geriatr Psychiatry 31, 799–808.

34.

Conde-Sala

, Reñé-Ramírez

, Turró-Garriga

, Gascón-Bayarri

, Campdelacreu-Fumadó

, Juncadella-Puig

, Rico-Pons

, Garre-Olmo

(2014) Severity of dementia, anosognosia, and depression in relation to the quality of life of patients with Alzheimer disease: Discrepancies between patients and caregivers. Am J Geriatr Psychiatry 22, 138–147.

35.

Conde-Sala

, Reñé-Ramírez

, Turró-Garriga

, Gascón-Bayarri

, Juncadella-Puig

, Moreno-Cordón

, Viñas-Diez

, Garre-Olmo

(2013) Clinical differences in patients with Alzheimer’s disease according to the presence or absence of anosognosia: Implications for perceived quality of life. J Alzheimers Dis 33, 1105–1116.

36.

Maslow

, Mezey

(2008) Recognition of dementia in hospitalized older adults. Am J Nurs Sci 108, 40.

37.

Bradford

, Kunik

, Schulz

, Williams

, Singh

(2009) Missed and delayed diagnosis of dementia in primary care: Prevalence and contributing factors. Alzheimer Dis Assoc Disord 23, 306–314.

38.

Lin

, O’Connor

, Rossom

, Perdue

, Eckstrom

(2013) Screening for cognitive impairment in older adults: A systematic review for the U.S. Preventive Services Task Force. Ann Intern Med 159, 601–612.

39.

Formiga

, Fort

, Robles

, Riu

, Sabartes

, Barranco

, Catena

(2009) Comorbidity and clinical features in elderly patients with dementia: Differences according to dementia severity. J Nutr Health Aging 13, 423–427.

40.

Dokholyan

, Muhlbaier

, Falletta

, Jacobs

, Shahian

, Haan

, Peterson

(2009) Regulatory and ethical considerations for linking clinical and administrative databases. Am Heart J 157, 971–982.

41.

Fleury

M-J

, Delorme

, Benigeri

, Vanasse

(2019) Utilisation et enjeux des données clinico-administratives dans le domaine de la santé mentale et de la dépendance. Sante Ment Que 43, 21–38.

42.

Institut national de santé publique du Québec [INSPQ] (2017) INSPQ, Québec.

43.

Ponjoan

, Garre-Olmo

, Blanch

, Fages

, Alves-Cabratosa

, Martí-Lluch

, Comas-Cufí

, Parramon

, García-Gil

, Ramos

(2019) How well can electronic health records from primary care identify Alzheimer’s disease cases? Clin Epidemiol 11, 509–518.

44.

Patnode

, Perdue

, Rossom

, Rushkin

, Redmond

, Thomas

, Lin

(2020) Screening for cognitive impairment in older adults: Updated evidence report and systematic review for the US Preventive Services Task Force. JAMA 323, 764–785.

45.

Alzheimer’s Society, Methods for finding out about the experience of people with dementia (2022) https://www.alzheimers.org.uk/dementia-professionals/dementia-experience-toolkit/research-methods. Accessed October 6, 2021.