Association Between Psychological Distress and Incident Dementia in a Population-Based Cohort in Finland

Key Points Question What is the association of psychological distress with dementia? Findings In this cohort study of 67 688 individuals and, on average, 25 years of follow-up, symptoms of psychological distress—stress, depressive mood, nervousness, and exhaustion—were associated with a 17% to 24% increased risk of dementia in an etiological Poisson model and with an 8% to 12% increase in the incidence of dementia in the Fine-Gray model. Meaning These findings suggest that psychological distress is likely to be an etiological risk factor for dementia and associated with the incidence of dementia.

*Some individuals received the diagnosis from several registers on the same day; therefore the sum of the cells is greater than the total number of cases. ** If Alzheimer's disease (AD) was diagnosed at some point, the earliest report of dementia was coded as the first occurrence of AD. , age at the end of follow-up (5-year time slots), sex, educational class, body mass index, systolic blood pressure, total cholesterol, smoking, physical activity, and diabetes. CI, confidence interval; IRR, incidence rate ratio.

eTable 9. Associations of psychological distress symptoms with all-cause dementia in fully adjusted cause-specific hazard model (Poisson) with and without covariate of alcohol consumption in the cohort comprising of the FINRISK Study surveys 1982, 1987 and 1992.
Fully adjusted model adjusted for FINRISK survey year, follow-up time (10-year time slots), age at the end of follow-up (5-year time slots), sex, educational class, body mass index, systolic blood pressure, total cholesterol, smoking, physical activity, and diabetes. CI, confidence interval; IRR, incidence rate ratio.

eTable 10. Sensitivity analyses for associations of psychological distress symptoms with Alzheimer's disease in the fully adjusted cause-specific hazard model (Poisson) and subdistribution hazard model (Fine-Gray). (Only individuals with follow-up time ≥ 10 years are included.)
Cause-specific hazard model

Description of all statistical tests
To examine association of psychological distress with dementia, we used in parallel two methods accounting for the competing risk of death, cause-specific hazard model and subdistribution hazard model. As a cause-specific hazard model we used Poisson model and as a subdistribution hazard model Fine-Gray model. Sensitivity analyses for reverse-causation excluded individuals with follow-up time < 10 years. The main analyses were shown for basic model with the most essential covariates and for the fully adjusted model.
Interaction terms with sex in the cause-specific hazard model analyses were examined, and when significant, analyses stratified for sex were performed. In addition, cause-specific hazard models with covariates of insomnia and alcohol consumption were studied. For AD, we performed, similarly to all-cause dementia, in parallel cause-specific hazard model and subdistribution hazard model. Association of the missing data status with dementia in the fully adjusted model was studied in cause-specific hazard model and subdistribution hazard model. In addition, non-responsiveness to psychological distress variable was coded as a separate category, and cause-specific hazard model analysis for dementia was performed. To study differences in the baseline values of the exposures and covariates between those with missing data on the fully adjusted model and those with nonmissing data, we used Pearson's chi-square test for categorical variables and one-way analysis of variance for continuous variables. Correlation of the exposure traits was examined with Spearman's correlation coefficient. Here, the significance threshold was P<.01 The significance threshold of P<.05 or confidence intervals of 95 % were used throughout the study, except for Spearman's correlation.

Study population
Ethnicity was not ascertained in the National FINRISK study surveys 1972-2007. It is also not collected in the Finnish registers.

Outcome
For diagnoses of all-cause dementia, and secondary outcome Alzheimer's disease (AD), we used combined information from the Finnish national registers: Accuracy of the all-cause dementia and Alzheimer's disease (AD) diagnoses based on the Finnish registers is good (96% and 100% positive predictive value) when comparing with clinical diagnostic examination, but some underestimation of occurrence may occur because, when used in combination, the Hospital Discharge Register and the Drug Reimbursement Register led to a sensitivity of 71% for AD 2 . The information from the Causes of Death Register and purchased medication was not, however, used in their definition of set diagnosis as we do; thus the sensitivity is likely to be higher in our study in the later years. However, before the 1990s the rarity of AD diagnostics as well as a lack of AD medication (first described in 1996, eMethods) has surely made the sensitivity lower than that in the later years.
The first diagnoses for AD in the data set appeared in year 1987. Before 1987 and partly after that the cases of AD are likely to be classified as cases of dementia in the Causes of Death Register or in the Hospital Discharge Register. The inaccuracy in the diagnosis of AD in the early years may weaken the associations, because the follow-up for the first FINRISK survey began in 1972 in our study.
The drug-related registers, which are the main source of diagnosis (eTable 1), begin to have information on AD medication only after 1996. After the beginning of the systematic use of dementia drugs, the Drug Reimbursement Register is likely to be a relatively reliable source for the diagnosis of AD, because to receive reimbursement the patient has to be diagnosed for AD by a physician specialized in neurology or geriatrics. In Finland, examination typically includes a medical interview and a clinical examination, brain imaging by magnetic resonance imaging, measures for cognitive decline like those of the Consortium to Establish a Registry for Alzheimer's Disease (CERAD), and laboratory measurements for differential diagnostics 3 . From year 2012, donepezil was not part of the special reimbursement and, from year 2013, neither was memantine. After these changes, the diagnosis based on AD medication was likely to lose some specificity when examination by a doctor specialized in neurology or geriatrics was no longer needed for reimbursement. However, after that, most patients still visit a specialized doctor for diagnosis in Finland, and medication purchases are likely to reflect AD relatively well, because only rarely is the AD medication used for other types of dementias like Lewy body disease or vascular dementia.