The present study was a secondary analysis of the Canadian Study of Health and Aging (CSHA), a national, population-based, longitudinal, multicentre cohort study whose original objectives were to determine and investigate the prevalence, incidence and risk factors for dementia and the earlier stages of cognitive impairment. Supplementary objectives covered broader aspects of health and disability among elderly Canadians (11
The CSHA began in 1991 (CSHA-1), when representative samples of Canadians ≥65 years of age were randomly drawn in 39 urban and surrounding rural areas in the 10 Canadian provinces. Subjects were ≥65 years of age at the time of the first survey. Sampling was stratified according to age group, with over-sampling of those 75 to 84 years of age (two times) and ≥85 years of age (2.5 times). Those with life-threatening illnesses (eg, on life support measures or with terminal cancer) were excluded. Individuals who declined to participate or who could not be contacted were replaced by others of the same sex, age group and region.
To make population estimates, sample weights were derived to correct for the different populations in each sampling area and for the oversampling of the older groups (12
The full CSHA study involved 9008 individuals from the community and 1255 from long-term care institutions. The same individuals were contacted again in 1996 (CSHA-2) and 2001 (CSHA-3). The community sampling frame was based on the Canadian provincial universal health insurance plans, with the exception of Ontario, where technical limitations with the health insurance plan list prevented its use at the time. Here, the Enumeration Composite Record was used, which was a composite list of all citizens in Ontario based on electoral lists, updated between elections from information such as property sales. The response rate in the community was 72.1%. The present study included the 8949 community dwellers who were administered the relevant section of the screening questionnaire.
The CSHA screening interview included questions regarding demographics, general health and psychometric tests for cognitive impairment. The interviewers were instructed to collect collateral information from family members or proxy respondents when recording the data, particularly in cases of cognitively impaired respondents in whom personal accounts may have been unreliable.
The question used to determine the presence of digestive symptoms was: “In the past year have you had troubles with your stomach or digestive system? You can just answer Yes or No.” Cognitive function was based on the score from the Modified Mini-Mental Status (3MS) examination (13
). The 3MS is a validated cognitive screening instrument with a possible score of 0 to 100 that tests orientation, immediate and remote memory, attention and concentration, language and naming, verbal fluency and executive function. The respondents were categorized as cognitively normal (3MS score ≥78) or impaired (3MS score <78).
Self-rated health was assessed using a 5-point scale ranging from very good to poor; this was taken from the Older American Resources Utilization Study (14
For these analyses, self-rated health was dichotomized as either good (combining ‘very good’ or ‘good’ responses) or poor (in which ‘fair’, ‘poor’ and ‘very poor’ responses were combined). Self-rated health status was missing for 18 individuals (0.002%).
Functional status was based on self-report and was assessed using the activities of daily living (ADL) scale from the Older American Resources Utilization Study (14
). For our analyses, each of five functional items (eating, dressing, grooming, bathing and toileting) was scored either as 1 (can perform without help) or 0 (needs assistance or unable to perform.)
Participants were categorized according to the number of impairments they reported: no impairments in ADL, one impairment, and two or more impairments in ADL. Nineteen individuals were missing data for one of the five ADL; these individuals were included in the analysis according to the sum of their nonmissing impairments.
Marital status was classified as either married (subjects who were currently married or living in common-law relationships) or unmarried (widowed, divorced, separated and previously married). Marital status data were missing for two individuals.
The CSHA was approved by local ethics committees at all participating study sites.
For participants who died during the 10-year follow-up, survival time was calculated as the time between the baseline interview at CSHA-1 and the date of death. Ten-year mortality was coded as either 1 (deceased by the end of CSHA-3) or 0 (alive by the end of CSHA-3). Similarly, institutionalization was coded as 1 (institutionalized by the end of CSHA-3) or 0 (remained in the community). In Canada, it is very rare for institutionalized elderly people to return to the community.
All subjects for whom complete data were available were included in the analysis. In all analyses, proportional weights were used to account for sampling methodology (15
The baseline characteristics of the subjects with digestive symptoms and the asymptomatic group were compared using the χ2 test for proportions and Student’s t test for continuous variables. In all analyses, having digestive symptoms was specified as the independent variable, with mortality or institutionalization as dependent variables. The potential confounders (age, sex, self-rated health, ADL dependence and cognitive impairment) were included as covariates in the regression models.
Cox proportional hazard regression was used to investigate whether having digestive symptoms was associated with survival, adjusting for potential confounding factors including age, sex, 3MS score, self-rated health and number of impairments in ADL. The results of the Cox regression were reported as HRs with 95% CIs.
Multivariable logistic regression models were used to control for potential confounding variables and to determine the independent association between digestive symptoms and incident institutionalization.
All analyses were performed using STATA version 8.1 (Stata Corp, USA) and were weighted to account for the sampling methodology.