Research in ‘older people’ often arbitrarily focuses on people aged 65 years and over, that is, above the usual retirement age in Western nations. This ‘65+’ range includes very varied populations: those within a decade after retirement who can reasonably expect to be in good health, through to more advanced decades where physical and mental decline become much more common and where profound social changes are more frequent, such as bereavement and challenges to independent living. The ‘65+’ cut-off already has little social meaning in many nations and will become steadily less relevant as retirement ages become more flexible in the West. For anxiety and depression, at least two age-group targets may be more appropriate. Retirement itself is a major life-event and may exert a substantial impact on later health; for example, whether the retirement was planned, the retiree’s earlier attitude to financial planning and the resources they have put aside, their ability to find alternative occupation, and the quality of interpersonal relationships as life becomes more home centred. Research around this life-transition should have a broader age focus that includes the decade or more leading up to retirement, in order to understand better what happens afterwards. At the other extreme, more ‘traditional’ late-life research into the impact of physical ill health and social stressors should shift its focus towards older age groups where these have highest salience.

Constructs of Anxiety and Depressive Disorders

Anxiety and depression are applied to mental states where there are no clearly identifiable boundaries between normality and abnormality. The way in which a syndrome is defined is an important consideration. Instruments which define case-level symptoms in a broad and inclusive way will give rise to higher prevalence estimates than those which define in a more selective manner. More restricted syndromes have a higher individual impact, but more ‘mild’ cases may have a stronger societal impact, because of their substantially higher prevalence. Restrictive criteria are often favoured by researchers because it is easier to achieve comparability between studies. However, they frequently encompass only a small proportion of cases seen routinely by clinical services and underestimate substantially the syndrome’s wider impact. It is best not to view one approach as ‘better’ or ‘worse’ than another; they are merely different, and these differences should be taken into account when between-study comparisons are made.

Underlying Dimensions

Hypertension is an artificial category applied to a known physiological parameter (blood pressure), and it is likely that the same can be said for anxiety and depression. However, the underlying parameters are much less well understood, may not be unitary, and may not perfectly reflect traditional ‘diagnosis’ constructs. Criteria for depression as a diagnosis do not simply reflect the pervasiveness of depressed mood but also take into account its manifestations (tearfulness, anhedonia etc.) as well as the presence of other symptoms (e.g. appetite disturbance, insomnia), the quality of other symptoms (e.g. early morning wakening rather than initial insomnia), the manifestation of the overall syndrome (e.g. duration, pervasiveness, diurnal fluctuation) and the degree to which this affects daily life. The concept of the ‘syndrome’ is core to psychiatric nosology. However, a substantial part of this lies in tradition rather than empirical evidence, and it may be preferable for research to clarify underlying symptoms before assuming the validity of syndromes. For example, the EURO-D instrument generates two underlying dimensions (motivation and affective distress) which have different correlates (lower motivation associated with increased age, affective distress associated with female gender¹). It is not yet established whether these findings apply beyond this instrument; however, they at least suggest that a more flexible attitude is needed towards categories and dimensions.

Are Anxiety and Depression Different Disorders?

Although anxiety and depression have traditionally been defined as separate conditions, this tradition has evolved in secondary and tertiary care settings. In community samples of any age group, mixed syndromes predominate. Furthermore, the impact of common mental disorders appears to depend principally on the number of symptoms rather than whether they are anxious or depressive². An exception to this is mortality as an outcome, where U-shaped associations with anxiety symptom load have been found compared to more linear associations with number of depressive symptoms³, which may reflect differences in health service access and use. For example, delay in breast cancer presentation was found to be positively associated with previous depressive disorder but negatively associated with previous phobic disorder⁴. In this chapter, we shall consider depressive and anxiety disorders separately for the most part, but limitations in this approach should be borne in mind.

‘Depression’ as a diagnosis generally requires persistent low mood which is accompanied by other symptoms (such as disturbance of sleep and appetite), and which is causing significant distress and/or disablement. However, mood states vary across a spectrum of normality to abnormality and the prevalence of depression (i.e. how common it is in a given population) depends substantially on the cut-off severity applied. As discussed earlier, severe depressive syndromes have a high individual impact but are rare, while milder syndromes have a lower individual impact but affect many more people.

Prevalence studies using more restrictive criteria include the Epidemiologic Catchment Area (ECA) surveys in five areas of the USA, in which the Diagnostic Interview Schedule (DIS) was administered, and which found major depression in 0.4% of men and 1.4% of women aged 65 years and over⁵. Other estimates include those from the Cache County Study in Utah, also using the DIS, finding major depression in 2.7% of men and 4.4% of women aged 65 years and over⁶, as well as a survey in the Netherlands finding major depression in 2.0% of residents aged 65 years and over⁷.

An alternative approach has sought to define ‘depression of clinical significance’; that is, a syndrome severity which would warrant clinical intervention. This construct in older people tends to be found 5-10-times more commonly than major depression, with prevalence ranging from 10 to 20% in most studies. These include studies which have measured ‘clinically significant depressive symptoms’ using brief instruments such as the Centre for Epidemiologic Studies Depression scale (CES-D); for example 16% and 9% of older residents scoring at case level (16+) in two US studies^8,9. A more diagnostic approach is taken by the structured Geriatric Mental State (GMS) interview schedule with its accompanying AGECAT computerized diagnostic algorithm^10,11, which has probably the most wide international use including the EURODEP collaboration of late-life depression surveys in Western Europe¹², the UK Medical Research Council Cognitive Function and Ageing Study¹³ and the 10/66 surveys of older people in low- and middle-income countries¹⁴. The EURO-D instrument was derived from the GMS and has been applied to EURODEP data^1,12,15 as well as used more recently in the Study of Health and Retirement in Europe (SHARE) surveys^16,17. Principal findings from the EURODEP collaboration were GMS depression prevalence ranging from 8.8% (Iceland) to 23.6% (Munich) with a weighted average of 12.3% in total: 14.1% in women and 8.6% in men¹². The SHARE surveys using the EURO-D reported lowest prevalence in Denmark (18.1%) and Germany (18.8%) and highest prevalence in the three Western Mediterranean sites – France (33.3%), Italy (33.7%) and Spain (36.8%). These differences largely persisted following standardization for demographic profiles¹⁷. Prevalence estimates for similar definitions of late-life depression in other world regions have been broadly comparable – for example a 14.8% prevalence of depressed mood in a meta-analysis of studies in China¹⁸. In three Latin American sites (Peru, Mexico and Venezuela) of the 10/66 population-based survey programme, prevalence ranges were 1.3-2.8% for DSM-IV major depression, 4.5-5.1% for ICD-10 depressive episode, 30.0-35.9% for GMS-AGECAT ‘clinically significant’ depression, and 26.1-31.2% for EURO-D caseness¹⁹.

Prevalence of depression is a product of both new cases arising (incidence) and duration of case-level disorder. A higher prevalence in one site or group compared to another may be because of higher incidence or delayed recovery. For late-life depression it could also be accounted for by delayed case mortality, although mortality has been found to be higher in older people with depression²⁰. Incidence of late-life depression has been less commonly measured. Major depression incidence was 0.15% per year in the ECA studies and, in a Swedish cohort, 0.12% per year in men and 0.30% in women²¹. One-year incidence of a broader depression syndrome (17% baseline prevalence) in the north London Gospel Oak study was 12%²², and 2.5 year incidence in Korea using the GMS (baseline prevalence 14%) was also 12%²³. Persistence of depression in the Gospel Oak cohort after a one-year interval was 63%²². However, two-point surveys give relatively crude information on clinical course. Incidence is underestimated (because depression episodes in the intervening period which have recovered by the follow-up point will be missed), and persistence is over-estimated (since some of those with depression at both times may have had an intervening period of recovery).

The course of depression, a classically fluctuating condition, is therefore poorly represented by binary concepts such as ‘incidence’ and ‘maintenance’, particularly in two-wave studies. An alternative approach has described clinical course in a more qualitative manner. The Longitudinal Ageing Study of Amsterdam investigated the six- year outcome of depression defined both by the CES-D and DIS. Only 14% of episodes were found to be short lived. Remission occurred in 23%, an unfavourable but fluctuating course was followed by 44%, and a severe, chronic course by 33%²⁴. These are similar to findings from a very early study of late-life depression where, over six years, 31% recovered, 28% had one or more relapse, 23% had a partial recovery and 17% remained depressed²⁵. In another early study with a one-year follow-up, 35% had a good recovery, 48% had had a recurrence or continuation of the syndrome, 3% had developed dementia, and 14% had died²⁶. In a more recent review of depression prognosis, no clinically significant differences were found in response to pharmacotherapy or ECT by older/younger age of onset, although older people had a higher risk of future episodes. Medical co-morbidity was a potential confounder, being more common in older people and associated with a worse prognosis. Older people with early onset depression had a larger number of previous episodes, which also contributed to a worse prognosis²⁷.

The prevalence of depressive disorders increases from young to mid- adult age groups, often followed by a fall in prevalence for older people within a decade of the retirement age²⁸. However, studies that have focused upon depressive symptoms and broader depressive syndromes indicate either an increase in their frequency, or stability with increasing age²⁹. Blazer concluded that, if a symptom factor structure was examined, there were no age, race or gender differences in scores²¹. Where positive associations between age and depression prevalence are found, the high prevalence of physical and social stressors in these more advanced age groups is most likely to be responsible. The exclusion, in DSM diagnostic criteria³⁰, of symptoms that are considered to be primarily attributable to bereavement and physical illness may account in part for the apparent lower prevalence of thus-defined depressive disorders in older people. The expression of depression may also be qualitatively different in older people: for example, a higher age of onset for melancholic symptoms such as non-interaction, psychomotor retardation, and agitation³¹, concurring with the positive association between increased age and diminished motivation in the EURO-D scale, both in EURODEP¹ and SHARE centres¹⁷.

Cross-sectional associations with age may reflect birth cohort effects as well as chronological age, but these have been poorly captured to date. On the one hand, rapid improvements in health and economic prosperity for the ‘young elderly’ (e.g. 65-75 years) have been a particular feature of demographic ageing. On the other hand, there may be a reduction in ‘stoicism’ and ‘resilience’ as well as possible differences in willingness to admit to depressive symptoms. Birth cohort differences can only really be evaluated by repeated surveys covering the same population using similar measurements and case definitions; data which have not so far been available for late-life depression.

As well as the age distribution, some research has attempted to estimate the age of onset of late-life depression; that is, the proportion which occurs as new episodes. For example, an Australian study estimated that 52% of cases had their onset at age 60 or above³², while a US study of people receiving home care estimated that 71% of those who were depressed were experiencing their first episode³³. However, this approach depends on accurate recall of previous episodes, an assumption which is questioned by a prospective study in Sweden of people who had been hospitalized 25 years earlier for major depression. Of those traced, only 50% recalled sufficient details for diagnostic criteria to be applied and 30% did not recall the episode at all³⁴. Recollections of more mild forms of depression are likely to be substantially more inaccurate.

Gender and Marital Status

Higher prevalence of depression is usually found in women compared to men, but this difference is substantially less than that found in younger age groups³⁵. The EURODEP consortium¹ reported an excess of depression symptoms in older women in 13 out of 14 European centres. This association was consistently modified by marital status, with marriage being protective for men but associated with higher risk for women. These findings are consistent with the observation from several studies that married older men cite their wife as their main confidant, whereas women more often cite a friend outside the home, and with the association between married status and relatively low mortality or good health being stronger for men than for women.

Social Support/Activity

A consistent finding has been the salience to mood and morale in older people of contact with friends, in particular intimate and confiding relationships. While older people typically receive instrumental support from spouses and relatives, they value friends for the companionship and emotional support which they can provide. However, this generalization does not take into account the fact that some individuals may prefer a relatively isolated existence and find higher levels of social contact stressful. In the longitudinal Gospel Oak study, no contact with friends was the only social support variable prospectively associated with the onset of depression²². In this study, lack of social support and social participation were associated with maintenance rather than onset of depression, and the same has been found in a prospective French study where both social activity and disablement were associated with CES-D depressive symptoms at baseline but, for those with case-level symptoms at baseline, only higher levels of social activity predicted a reduction in symptoms at follow-up³⁶. Physical health, age and gender may modify or explain to some extent the association between social support and late-life depression. There are large gender differences in late-life social support, with women typically having more supportive and extensive networks of friends than men. These may deteriorate following bereavement. Social engagement, such as visiting friends, is impaired by disability, indicating complex relationships between these exposures.

Social Adversity and Life Events

Many surveys find associations between late-life depression and relative disadvantage in income; poverty, for example, was one of the top five correlates of late-life depression along with disability, illness, isolation and bereavement³⁷. These are highly correlated characteristics, and it will always be difficult to determine the effect of one independent of the others. At least some of the effect of poverty may be confounded by and/or mediated through physical health. Environmental measures have received less attention, although one study found that worse quality of the internal home environment was an independent predictor of incident depression one year later³⁸, particularly for cohabiting residents. Poverty, isolation and poor housing represent chronic stressors. The importance of acute stressors has also been investigated, with most studies finding some association^39,40 although varying in exposure and outcome measurement. Specific life events may have particular salience, such as bereavement, threats to health and interpersonal conflicts^41,42