Ascertainment and coding of dementia diagnosis

For each CADDY participant, we received primary care data on the presence or absence of (1) dementia diagnoses, (2) any concerns about cognitive function and (3) referrals to specialist services following cognitive concerns (see Appendix 2 for detail of the CRF). The earliest date and specific clinical code used for each clinical diagnosis, concern or referral were also recorded.

By comparing dates recorded in primary care with the dates of each participant’s CFAS assessments, we encoded the status of each participant at the date of their W1 and W2 assessment into one of four mutually exclusive categories as described as follows:

1. diagnosed (clinical diagnosis) on the interview date if the GP had recorded a clinical diagnosis of dementia before that date
2. referred if no clinical diagnosis had been made but a referral to a specialist service following a cognitive concern had been noted before the interview date, irrespective of whether the patient had attended or the referral result had been noted
3. concern if there was any record of a cognitive concern before the date of interview, but without a clinical diagnosis or referral having been made
4. unknown if there was no report of any cognitive concern, referral or clinical diagnosis up to the interview date (note that unknown is not a missing category but means that the practice reported no cognitive concerns up to the corresponding CFAS assessment date).

Sociodemographic factors

Date of birth, sex and years of completed education were ascertained or confirmed at the baseline interview. Education was coded into three groups: (1) < 10 years of formal education, (2) 10 years of formal education and (3) > 10 years of formal education.

Cambridge Cognitive Examination

The Cambridge Cognitive Examination (CAMCOG)³⁵ is a more comprehensive assessment than the short cognitive assessment that has been in widespread use throughout the NHS until relatively recently, the Mini Mental State Examination (MMSE).³⁶

The CAMCOG has a range of total scores on a 0–103 scale and has subscores in 10 cognitive subdomains. These are:

1. orientation
2. language (comprehension)
3. language (expression)
4. memory (remote)
5. memory (recent)
6. memory (learning)
7. attention and calculation
8. praxis
9. abstract thinking
10. perception.

The CAMCOG was administered as part of the total interview to as many participants as possible, but if the interviewer considered that the participant might refuse to continue the interview, or the participant failed early items in the questionnaire on orientation, a shorter ‘emergency’ assessment was conducted, in which the MMSE but not the CAMCOG was administered.

Any missed questions were coded as failed for the purpose of this study, if the participant attempted any of the CAMCOG questions. Twenty-five CADDY sample participants did not attempt any of the CAMCOG assessments, in which case all subscales and the total score are coded as missing.

A CAMCOG score of 80 has been reported as the optimal threshold for identifying people with dementia,³⁷ whereas CAMCOG scores of 60, 40 and 30 are typical of minimal, mild and moderate/severe dementia, respectively.³⁸

Memory impairment (informant report only)

The informant is asked whether or not the participant has any difficulty with memory (in general) and then specifically with (1) remembering short lists of items when shopping, (2) remembering what happened yesterday, (3) remembering where they put things or (4) remembering the names of people. All positive responses to any of these questions are summed, leading to a score of 1–5.

Changes in personality (informant report only)

This is determined from responses to the binary question ‘Have you noticed any changes in his/her personality, such as the way she/he behaves socially?’.

Changes in behaviour (informant report only)

This is determined from the answers to two questions on whether or not behaviour has changed and participants’ awareness of their behaviour. Responses to each question about change in a specific behaviour were scored as none (0), mild (1) or severe (2). Values were based on the maximum reported score for either question.

Problem with eating or drinking

This is defined as any report of eating unusual things, overeating or any problem that the participant experiences with physically feeding themselves.

Incontinence

This is a binary variable corresponding to any report from the participant or informant of difficulty with getting to the toilet or with bladder control.

Sleep disturbance

This is recorded if either the participant or the informant reports any sleep disturbance, which could include difficulty sleeping/getting to sleep or waking early.

Communication difficulty (carer’s report only)

This is defined as any mention of a problem with communicating. We used five questions with binary responses on word-finding and talkativeness, and three questions on repeating words, which were added to produce a single score (0–5) of communication difficulties.

Blessed Dementia Scale

The Blessed Dementia Scale (BDS)³⁹ was developed as a screening instrument for the detection of mild cognitive impairment and dementia, but is also used as a measure of severity of functional impairments associated with dementia. CFAS uses a modified version of the BDS, including 11 questions on the performance of household tasks, managing money, navigating indoors and outdoors, remembering lists, interpreting surroundings, recall of recent events, dwelling in the past, difficultly eating, dressing and continence. The BDS is administered at the informant interview, and so reflects both the severity of dementia and the informant’s knowledge of these symptoms of dementia in the respondent. It is scored from 0 (not impaired) to 17 (extremely impaired). The BDS cannot be administered in the case of participants with no informant interview. In CADDY, the BDS was used for generating weights, as elements of the scale are included in other predictor variables.

Inpatient hospital stays

The CFAS II does not specifically ask about unplanned hospital admissions, but does ascertain at each wave the number of separate inpatient stays in the year before the assessment, and the number of nights for up to seven visits. These are reported either by the participants themselves or by a proxy.

Anxiety and depression

Anxiety and depression were both coded based on interview data according to the Automated Geriatric Examination for Computer Assisted Taxonomy (AGECAT) algorithm.⁴⁰ Both were coded at each assessment as being absent ‘none’, ‘subthreshold’ or ‘case level’, similar to previous analyses using the MRC CFAS.⁴¹

Sensory impairment

Visual impairment was dichotomised from participant report and interviewer observations as the presence or absence of any degree of visual impairment that interferes with reading, writing or drawing after correction with glasses.

Hearing impairment was dichotomised from the interviewer’s subjective assessment as presence or absence of a hearing impairment that interferes with the questioning or from the informant report of whether or not the respondent wears a hearing aid. A positive response to either question was considered to indicate a hearing impairment.

Social participation

Participant’s or proxy’s report of the number of activities they participate in, from the following list:

• political parties
• trade unions
• environmental groups
• tenants’ or residents’ groups
• Neighbourhood Watch
• evening classes
• University of the Third Age
• other adult learning
• arts, music or singing groups
• charity, voluntary or community groups
• groups for elderly (e.g. lunch club)
• youth groups (e.g. Scouts/Guides/youth club)
• Women’s Institute, Townswomen’s Guild
• social clubs (Rotary, working men’s club)
• sports clubs, gym, exercise/dance groups
• other groups or organisations.

Social contact

This was determined from the participant’s or proxy’s report of the frequency of contact between the participant and friends or family and was coded as ‘daily’, ‘two or three times per week’, ‘weekly’, ‘monthly’, ‘less often’ or ‘never’. This variable is coded only for participants not in a residential care setting.

Comorbidity

Participants and informants were asked whether or not participants had even been diagnosed with any of the following comorbid health conditions:

1. heart attack
2. stroke
3. angina
4. high blood pressure
5. intermittent claudication
6. bronchitis
7. other respiratory disease
8. asthma
9. arthritis
10. pernicious anaemia
11. diabetes
12. thyroid problem.

Each comorbidity was coded as a binary variable, reflecting any prior diagnosis. The comorbidities were then summed into two variables: cardiovascular comorbidity (sum of conditions 1–5) and non-cardiovascular comorbidity (sum of conditions 6–12).