Risk factors for depressed mood amongst a community dwelling older age population in England: cross-sectional survey data from the PRO-AGE study

The analytical samples were derived from the PRO-AGE study, which was the first large-scale randomised controlled trial of health risk appraisal for older people in Europe [13], and used a self-completion multidimensional Health Risk Appraisal instrument for elderly people adapted for use in Europe by the Geriatrics Research Unit of Bern, Switzerland, in collaboration with other institutions [14, 15].

In the PRO-AGE study, subjects were recruited between 2000 and 2001 from three large GP group practices (18 general practitioners) in London participating in a multi-national randomised controlled trial investigating the effect of Health Risk Appraisal for Older persons. Eligibility criteria were patients aged 65 years and over; who were living at home without evidence of need for human assistance in performing basic activities of daily living (BADL); without cognitive impairment, or a terminal illness; who were able to speak English; and who fully completed and returned a postal Probability of Recurrent Admissions questionnaire [16, 17], and a consent to participate form. The Probability of Recurrent Admissions questionnaire is a screening instrument used to identify members of older populations who are at risk for using health services heavily in the future. The general goal of the trial was to evaluate intervention through the use of the Health Risk Appraisal for Older persons (HRA-O) questionnaire, feedback reports to older people and providers, and personal patient education [13]. The details of the design of this trial and questionnaire are given elsewhere [13, 15].

A flow diagram explaining the derivation of the analytical samples used in these analyses using data from the PRO-AGE study are given in Figure 1. Of an initial list of 4466 subjects aged 65 years and older who were initially assessed for eligibility, 4075 (91%) were sent a Probability of Recurrent admissions questionnaire, which also included an additional BADL question to exclude the most disabled. Of the 4075, 117 (3%) were excluded as they were dependent in BADL, 163 (4%) returned incomplete questionnaires, 1292 (32%) were non responders, leaving 2503 (61%) eligible for randomisation. Of these, 1240 (50%) were allocated to the intervention group who were sent a HRA-O questionnaire at baseline (t = 0). The remaining, 1263 subjects were allocated to the control group of usual care and were sent a HRA-O questionnaire one year from baseline (t = 1). The 1090 (88%) subjects who completed and returned this self-administered HRA-O questionnaire at baseline were included in analyses, and used data from this baseline questionnaire.

For confirmatory analyses, a second sample was identified. The sample was derived from the two control arms of this trial who were sent the HRA-O questionnaire one year after baseline (t = 1), but not at t = 0 (see Figure 1). This included data from 1263 subjects who were randomly allocated to the control group of usual care, and 636 subjects who constituted a concurrent London based comparison group (1 large group practice; 8 general practitioners) within this trial.

Data on the outcome of depressed mood, and health, function, and demographic ‘risk’ factors were selected for analyses from the HRA-O questionnaire; age at time of completion of questionnaire, sex, ethnic group, functional change, functional decrease, instrumental activities of daily living (IADL) and health that included chronic health problems, pain, self-perceived general health, vision related difficulties and hearing related difficulties screening scores.

Depressed mood was ascertained from the 5-item Mental Health Inventory Screening test (MHI-5) [18]. It asks questions about how the person felt during the past month. It has been included as one of the subscales of the RAND 36-item health survey 1.0 (distributed by RAND) and the Short Form-36 (SF-36) [19] and have equivalent summary scores of between 0-100 for the subscale [20]; 0 (poor mental health status) to 100 (good mental health status). A score of ≤ 65 was used to indicate depressed mood, and has been found to have a sensitivity of 87% and a specificity of 70% for detecting mood disorders [21]. A further five participants were excluded who had missing data on depressed mood, consequently, reducing the number of subjects contributing to analyses to 1085 subjects.

Ethnic group was self-reported according to one of ten categories. For analyses, these were grouped into ‘White UK’ and ‘Other’. Chronic health problems were determined from ‘yes’ or ‘no’ responses to the survey question ‘Has a doctor ever told you that you have…’ high blood pressure; coronary heart disease, heart attack, or heart failure; stroke; chronic bronchitis or emphysema; asthma; arthritis or rheumatism; osteoporosis; diabetes; depression; emotional or mental illness other than depression; glaucoma; irreversible/untreatable retinal disease; and cataracts. Pain was self-reported and categorised as either ‘no pain’ or ‘pain’ over the last 4 weeks. The self-perceived general health of participants was self-reported as one of four categories; excellent, good, fair or poor [22]. The categories excellent and good, were combined, as were, fair and poor for analyses. Functional change and functional decrease were each based on responses to three questions; whether for health or physical reasons, in the past 12 months, the participant either ‘changed the way’ (for functional change) or ‘decreased how often’ (for functional decrease) they walk ½ mile, climb 10 steps or get into or out of a car or bus, a positive response to any of the three questions was recorded as a functional change/functional decrease [23]. IADL [24] categories were based on whether or not there was a difficulty or need for human assistance in ≥ 1 IADL items. Vision related difficulties screen scores were ascertained from a shortened version [15, 25] of the National Eye Institute Visual Function Questionnaire (VFQ) [26]. Eight items were included and scored and included items on general health and vision, difficulty with activities, and responses to vision problems [13]. Subjects were classified as having ‘VFQ-poor vision’ or ‘VFQ-not poor vision’. Poor vision included those describing their eyesight as ≥ very poor eyesight, having ≥ moderate difficulty with an activity, limited because of eyesight most or all of the time, or having given up driving mainly because of eyesight or both eyesight and other reasons. Hearing related difficulties screen results were scored and based on the Hearing Handicap Inventory for the Elderly with a score ≥ 10 indicating a hearing difficulty [27], or a response of ‘poor’, very poor’, or ‘deaf’, to an additional question of ‘How would you rate your hearing (with your hearing aid on, if applicable)?’.

This study used data from the HRA-O questionnaire, which is a comprehensive questionnaire that uses standardized and validated instruments. Rich data from a large community-dwelling non-disabled population of older adults in England were utilised, and comprehensive analyses adjusting for potential explanatory variables was undertaken. The MHI-5 screens for those with depressed mood, and although it is widely used to measure quality of life, it has received less attention within the mental health literature. However, it performs well in detecting a variety of psychiatric disorders, including mood disorders or major depression [21, 34], and has been recommended for use to measure psychological distress within a European framework [35]. The findings are based on data from a limited geographical area, and participants could be considered to be the ‘healthy’ older population. Generalisability to wider areas beyond those of our London based sample and settings requires consideration.

The study uses survey data which were self-reported, and participants are those who self- completed and returned the health risk appraisal questionnaire, and is subject to bias. Responses to self-reported measures of vision, hearing, pain, and self-reported general health might be influenced by ‘mood’. Self-rating of general health reflects a global assessment of health incorporating physical, mental and social factors, albeit the strength of association may be greater for physical functioning [36]. Whilst the study can investigate the strengths of associations, it cannot investigate the temporal association of health, functional, and demographic variables with depressed mood, and therefore, the directionality of association. However, there are potentially aspects of these variables that are distinct from one another as implied by their independence within the multivariable model.

The questionnaire determines self-reported lifetime prevalence of a condition, and whilst previously diagnosed conditions investigated in this study might be considered as chronic conditions, the possibility of diminishing symptomatology from the time of the last episode or correction of a condition, such as, surgery for cataracts, cannot be excluded, and might complicate comparisons with other data investigating the effects of co-morbidity. Multiple testing and therefore the risk of a significant result arising by chance within the current analyses also needs to be considered and interpretation through consideration to the p-values advisable. However, additional confirmatory analyses of the importance of the variables within the multivariable models identified utilising baseline data was undertaken using data from the control arms of this trial.

Approval was obtained from Brent Medical Ethics Committee (BEC 745) and King’s College Hospital Research Ethics Committee (01-010).