Observational study designs

There are different types of observational study, which we shall briefly discuss here. The main designs are: descriptive studies of population prevalence, cross-sectional surveys, case–control studies and cohort studies.

Descriptive studies

Prevalence is the proportion of a population who have a specific disease. The point prevalence is the prevalence of the disease measured at a particular time point. The period prevalence is the prevalence during a time interval. The prevalence of dementia increases significantly with age, therefore prevalence for this condition is often expressed for age brackets. Prevalence of dementia according to 5-year age groups can be reviewed in Table 1. Roughly, the prevalence of dementia doubles every 5 years after 65 years of age (Prince Reference Prince, Bryce and Albanese2013).

TABLE 1

The prevalence of late-onset dementia according to 5-year age groups

Source: Prince et al (Reference Prince, Bryce and Albanese2013).

The prevalence is approximately the same as the incidence (Box 2) multiplied by the disease duration; therefore, diseases with short durations have similar prevalence and incidence rates. For dementia, a chronic condition, the prevalence is far greater than the incidence. Incidence is not always a useful way to describe mental illnesses. Most mental illnesses, for example schizophrenia and bipolar affective disorder, have relapsing and remitting courses, which can be hard to capture in a single incidence study. In any case, both prevalence and incidence figures depend on the definition of a case and its operational criteria (Minett Reference Minett, Stephan, Brayne, Dening and Thomas2013).

Case–control studies

Case–control studies are more useful than cross-sectional surveys in studying rare diseases. Rather than studying a whole representative population, cases are chosen as individuals who have the disease of interest. Controls are then chosen as individuals who are disease free. The risk factors between the cases and controls are then examined to determine whether a particular risk factor is associated with developing the disease. In this way an odds ratio can be calculated. An odds ratio determines the odds of the outcome developing in the exposure group, as compared with the odds of the outcome developing in the control group (Box 3).

BOX 3

Relative risk and odds ratios

Relative risk is the ratio of the probability of developing the disease in the exposure group, compared with the non-exposure group:

$${\rm RR} = \lpar {{\rm a}/\lpar {{\rm a} + {\rm b}} \rpar } \rpar /\lpar {{\rm c}/{\rm c} + {\rm d}} \rpar )$$

The odds ratio is the ratio of the odds of the disease in the exposure group compared with the odds of the disease in the control group:

$${\rm OR} = \lpar {{\rm a}/{\rm b}} \rpar /\lpar {{\rm c}/{\rm d}} \rpar $$

(Woodward Reference Woodward2014)

In old age psychiatry, case–control studies have been most commonly used to investigate risk factors for the development of dementia. There are disadvantages to this type of study, for example they are prone to biases. Individuals in the case group may have been more extensively investigated or they may be more likely to recall exposures to risk factors than those in the control group. In addition, reverse causality can be problematic. For example, depression has been identified as a risk factor for dementia; contrariwise, perhaps dementia causes depression; or there may be a common pathophysiological pathway (for example, cerebrovascular disease or inflammation) to cause both (Jorm Reference Jorm2001; Singh-Manoux Reference Singh-Manoux, Dugravot and Fournier2017). Given these limitations, cohort studies where participants are followed up over time may be used to help ascertain causality.

Alzheimer's disease

The leading cause of dementia is Alzheimer's disease. Alois Alzheimer first described the pathological hallmarks of plaques and neurofibrillary tangles in 1907, but even then there were vascular components too (Stelzmann Reference Stelzmann, Norman Schnitzlein and Murtagh1995). There is an overlap between the risk factors for Alzheimer's disease and other forms of dementia, and these are laid out in Table 2.

TABLE 2

Risk factors for dementia and Alzheimer's disease

Identifying these risk factors has led to recommendations to delay and prevent the development of dementia. These include active treatment of hypertension and diabetes, alongside proactive smoking cessation advice and management of hearing loss (Livingstone Reference Livingstone, Sommerland and Orgeta2017).

Vascular dementia

Vascular dementia is a form of dementia caused by diminished cerebral blood flow. Vascular dementia often presents as cognitive impairment following a stroke, although it can present as cognitive impairment with established cerebrovascular disease and no history of stroke (Smith Reference Smith2017). The classification of dementia is debated – how helpful are such classifications? – and we know that often we are not describing discrete entities (Hughes Reference Hughes2011). Generally, people presenting with cognitive impairment in their 60s and 70s are more likely to have a single aetiology underlying their dementia. Older people, however, as we have already seen, are more likely to have mixed pictures (Wharton Reference Wharton, Brayne and Savva2011).

Vascular dementia is the second most common cause of dementia after Alzheimer's disease. Some degree of cerebrovascular disease is seen in 80% of dementias (Smith Reference Smith2017). Summing up a variety of studies, Kalaria (Reference Kalaria, Ames, O'Brien and Burns2017: p. 645) suggested that ‘the worldwide frequency of [vascular dementia] in autopsy-verified cases is determined to be 10–15%, being marginally less than when clinical criteria alone are used’. Again, the prevalence of vascular dementia rises with age: across 11 cohorts in Europe the prevalence is 1.6% among people 65 years and over; but it rises from 0.1% in females and 0.5% in males between 65 and 69 years to 5.8% in females and 3.6% in males who are 90 years old or above (Lobo Reference Lobo, Launer and Fratiglioni2000).

Lewy body dementia

Dementia with Lewy bodies (DLB) is the most common cause of dementia after Alzheimer's disease and vascular dementia. Lewy bodies are neuronal inclusion bodies and the location in the brain where they aggregate determines the clinical syndrome that predominates. Lewy bodies largely located in the basal ganglia lead to Parkinson's disease. Cerebral Lewy bodies are associated with DLB, although they are not specific to it and are also found at autopsy in patients with Alzheimer's disease (Hamilton Reference Hamilton2000).

Like other dementias, the prevalence of DLB increases with age. The average age at presentation is 75 years old (Barber Reference Barber, Panikkar and McKeith2001). Lewy body pathology is identified more commonly than Lewy body disease is diagnosed clinically. In one systematic review where dementia classification was in accordance with DSM-IV and DLB subtype was diagnosed using the original (1996) or revised (2005) International Consensus Criteria, DLB accounted for 4.2% of all diagnosed dementias in the community, a figure that rose to 7.5% in secondary care (Vann Jones Reference Vann Jones and O'Brien2014). The incidence of DLB was 3.8% of new dementia cases (Vann Jones Reference Vann Jones and O'Brien2014). The exact prevalence of DLB is debated. Elsewhere it has been estimated to be 10–22% of the dementias in people 65 years or over, suggesting that 1% of this population have DLB (Rongve Reference Rongve, Aarsland, Dening and Thomas2013). But the diagnostic rate depends on the diagnostic criteria used (Vann Jones Reference Vann Jones and O'Brien2014). Men have a higher prevalence of DLB than women, with one study showing a male-to-female preponderance in the ratio of 4:1 (Savica Reference Savica, Grossardt and Bower2013).

Behavioural and psychological symptoms of dementia

Behavioural and psychological symptoms of dementia (BPSD) include irritability, apathy, psychotic symptoms, agitation, wandering and depression. A systematic review of longitudinal studies of BPSD, the majority from the USA, Canada and Europe, showed a high prevalence of apathy, depression, agitation and wandering and a low prevalence of anxiety and hallucinations (van der Linde Reference van der Linde, Dening and Stephan2016). Affective symptoms such as apathy and depression tend to persist rather than occur episodically. Included studies used a variety of assessment measures. For example, studies that used the informant-based Behavioural Pathology in Alzheimer's Disease (BEHAVE-AD) rating scale to measure BPSD demonstrated a higher prevalence than studies using the Neuropsychiatric Inventory (NPI) (van der Linde Reference van der Linde, Dening and Stephan2016).

Comorbid disease

Of course, dementia tends not to come alone. Older people are prone to a variety of comorbid physical diseases. In the Newcastle 85+ cohort study, people aged 85 years or older had on average five diseases, the most prevalent being hypertension (57.5%) and osteoarthritis (51.8%), with moderate to severe cognitive impairment featuring in 11.7% (Collerton Reference Collerton, Davies and Jagger2009). Data from CFAS I and CFAS II, mentioned above, demonstrated that in the 10 years between the two studies the prevalence of diabetes and visual impairment in diabetes increased, which provides evidence that can be correlated with other possible comorbidities (Bennett Reference Bennett, Norton and Bunn2018).

But it is also possible that more than one mental illness might be present at one time. This is a complicated area because of difficulties ensuring that the same diagnostic criteria have been applied across studies and because of the possibility that one condition leads to or is part of the other. In one systematic review and meta-analysis of comorbid mental illnesses in people with dementia compared with their healthy peers, the mean prevalence of depression ranged from 20% in Alzheimer's disease, to 37% in frontotemporal degeneration and to 13% in healthy controls (Kuring Reference Kuring, Mathias and Ward2018). When all the different types of dementia were combined the prevalence of depression increased to 25%, whereas in the controls it was 13%. In the same study, the prevalence of anxiety in all types of dementia was 14%, but this was not statistically different from the rate in the healthy controls (3%), albeit the anxiety analyses were thought to be underpowered (i.e. the sample sizes were not sufficient to show statistical differences). The review's authors also considered the literature on post-traumatic stress disorder (PTSD) and dementia. They did not find enough (only two studies) to perform a meaningful statistical analysis, but the overall prevalence of PTSD in dementia was 4.4% (Kuring Reference Kuring, Mathias and Ward2018).

Dementia and dying

The majority of people with dementia in the UK die in the community. In a study in the north of England looking at people who had been diagnosed with early-onset dementia (called ‘pre-senile dementia’ in the paper, which referred to people between the ages of 45 and 64 years identified from records between 1985 and 1989) who had died by the end of 1998, 19.3% died at home, 24.5% in nursing or residential homes and 56.3% in hospital (Kay Reference Kay, Forster and Newens2000). In the over-65s dementia is one of the leading causes of death, and for those who have died from another cause with dementia as a contributory factor, stroke, Parkinson's disease and ischaemic heart disease are the most common causes (Public Health England 2016). Indeed, cognitive impairment is strongly associated with an increased mortality; a UK population study gives a median survival time from diagnosis of dementia to death as 4.1 years (Xie Reference Xie, Brayne and Matthews2008).

A large cohort study of nursing-home residents in the USA revealed that, compared with residents with terminal cancer, residents with advanced dementia received more active and non-palliative interventions, less advanced care planning and palliative symptoms were not as well controlled (Mitchell Reference Mitchell, Kiely and Hamel2004). Similar patterns are seen in the UK. For example, in a study in London between 2012 and 2014, of 85 people living with advanced dementia in nursing homes or their own homes, pain (11% at rest, 61% on movement) and significant agitation (54%) were common and persistent (Sampson Reference Sampson, Candy and Davis2018). Aspiration, dyspnoea, septicaemia and pneumonia were more frequent in those who had died by the end of the study (mortality at 9 months was 37%). The researchers found that 76% had ‘do not resuscitate’ orders, 30% had an advance care plan, 40% a documented preferred place of death and 40% a lasting power of attorney (Sampson Reference Sampson, Candy and Davis2018).

Depression

Depression is one of the most common mental health conditions to affect older people, although it is not a normal part of ageing. Depression in older people is associated with impaired functioning (Ormel Reference Ormel, VonKorff and Ustun1994), increased use of healthcare services (Beekman Reference Beekman, Copeland and Prince1999) and an increase in mortality (Djernes Reference Djernes, Gulmann and Foldager2011). Given that depression is a readily treatable disorder, this is a potential target for a significant improvement in morbidity and mortality in older people (Beekman Reference Beekman, Copeland and Prince1999).

Older people suffer bereavements and social role transitions such as retirement or loss of independent living. Despite this, major depression is not common in older people, with a review of the literature by Fiske et al (Reference Fiske, Wetherell and Gatz2009) finding a prevalence of 1–5% (most studies report prevalence figures closer to 1%) in large studies in the USA of community-dwelling older adults. These studies tended to use DSM-IV criteria to define major depressive disorder. Minor depression, however, is common. The prevalence of depression was 8.7% (95% CI 7.3–10.2) in those living in the community in England and Wales (McDougall Reference McDougall, Kvaal and Matthews2007a), and it was as high as 27.1% in a large UK-based cohort study of those living in institutions (McDougall Reference McDougall, Matthews and Kvaal2007b).

The definition of minor depression varies between studies. In some it refers to clinically significant depression that does not meet criteria for major depressive disorder. In others, however, it denotes subclinical depressive symptoms. Which of these definitions is adopted will have a significant effect on prevalence and highlights the importance of the diagnostic criteria adopted by any particular study. Similarly, studies that use depression symptom checklists rather than diagnostic criteria can overestimate rates of depression in older people because positive responses can relate to physical symptoms or bereavement rather than depression (Fiske Reference Fiske, Wetherell and Gatz2009). Data from 47 studies of older adults in acute hospitals show that the prevalence of depression is 29% (RCPsych 2005).

Bipolar affective disorder

Similar to the prevalence of unipolar depression in older people, the prevalence of bipolar affective disorder varies according to the setting in which it is being studied. In the USA, the prevalence of bipolar disorder in the community is 0.1% (Koenig Reference Koenig and Blazer1992), but is higher in nursing homes – between 3% (Tariot Reference Tariot, Podgorski and Blazina1993) and 10% (Koenig Reference Koenig and Blazer1992). A more recent study, albeit less rigorous in its definition of bipolar affective disorder, found a prevalence of 0.5% in those over 65 years in the community in the USA (Hirschfeld Reference Hirschfeld, Calabrese and Weisman2003). Bipolar disorder usually starts in younger adults, with a mean age at onset between 20 and 30 years (Rowland Reference Rowland and Marwaha2018). It has been estimated that 90% of individuals with bipolar disorder will have developed their illness before the age of 50 (Oostervink Reference Oostervink, Boomsma and Nolen2009), which is supported by an Australian study which found that only 8% of patients with bipolar disorder presented for the first time to mental health services once they were over the age of 65 years (Almeida Reference Almeida and Fenner2002). Interestingly, whereas the prevalence for bipolar disorder among men and women is similar in younger adults, in older people a systematic review found a preponderance of women: 69% of individuals presenting with bipolar disorder over the age of 65 are reported to be women (Depp Reference Depp and Jeste2004).

Depp & Jeste (Reference Depp and Jeste2004) also provide a critical review of epidemiological studies of bipolar disorder in older people. They comment that the majority of the studies included used convenience sampling of psychiatric in-patients who were mostly in a manic phase of their illness, whereas the vast majority of people with bipolar disorder live in the community. We know less about those living in the community between episodes or during depressive periods. In addition, a range of different diagnostic classifications were used, including DSM-III, DSM-IV and ICD revisions 8–10.

Late-onset schizophrenia

Individuals with late-onset schizophrenia (Box 4) are more likely to suffer from sensory impairment (Pearlson Reference Pearlson and Rabins1988), particularly hearing loss (Harris Reference Harris and Jeste1988).

The prevalence of late-onset schizophrenia is higher in individuals with a positive family history of schizophrenia, although this association is not as strong as in those with early-onset schizophrenia (Harris Reference Harris and Jeste1988). Similarly to bipolar disorder in older people, there is a female preponderance for late-onset schizophrenia. A systematic review of literature published between 1960 and 2016 found that the pooled incidence of schizophrenia in those over 65 was 7.5 per 100 000 person-years at risk, with an increased risk in women (OR = 1.6, 95% CI 1.0–2.5) (Stafford Reference Stafford, Howard and Kirkbride2018). Included studies used a range of different diagnostic classifications (ICD revisions 8–10, DSM-III and DSM-IV) and organic causes were excluded. One review reported that 23.5% of patients with schizophrenia had symptom onset after the age of 40 years (Harris Reference Harris and Jeste1988).

Delirium

Delirium is the most common psychiatric condition among general hospital in-patients. The prevalence on medical wards is 20–30% in the UK, and this rises to nearly 50% of patients who have surgery (National Institute for Health and Care Excellence 2010: p. 5). In the community, understandably, the prevalence is lower, at 1–2%. Individuals who develop delirium have more comorbidities and greater polypharmacy (Folstein Reference Folstein, Bassett and Romanoski1991).

One limitation with epidemiological data on delirium is that a diagnosis of delirium relies on a retrospective account of premorbid functioning and this can often be subject to recall bias. This can make it difficult to distinguish between delirium and pre-existing dementia. In addition, studies often exclude participants who have delirium on top of pre-existing cognitive impairment, which leaves a paucity of information on the clinical presentation of delirium in those with dementia (Davis Reference Davis, Kreisei and Terrera2013).

A large population-based cohort study in Finland showed that delirium was strongly associated with a future risk of developing dementia (OR = 8.7, 95% CI 2.1–35) (Davis Reference Davis, Muniz Terrera and Keage2012). The study relied on retrospective recall of episodes of delirium, and it should be noted that hyperactive delirium is more likely to be remembered than hypoactive delirium.

Anxiety-related disorders

The prevalence of new presentations of anxiety-related illnesses reduces with age. This may in part reflect a reduction in care-seeking behaviour by older people, or a reduced rate of detection by clinicians, or it might be that older people with anxiety are inappropriately referred to medical or surgical services for treatment of somatisation (Blazer Reference Blazer, George, Hughes, Salzman and Lebowitz1991). Table 3 summarises the 1-month prevalence of different anxiety-related disorders in people over the age of 65 from the US Epidemiologic Catchment Area (ECA) study (Regier Reference Regier, Boyd and Burke1988). The ECA sampled community residents and those in institutions and used personal interviews in line with the Diagnostic Interview Schedule. Diagnoses were made according to DSM-III.

TABLE 3

Prevalence of anxiety-related disorders in the 65+ age group

Source: Regier et al (Reference Regier, Boyd and Burke1988).

There are few epidemiological studies looking at the prevalence of anxiety-related conditions in older people and The Epidemiological Catchment Area (ECA) study remains one of the most widely used sources for determining how common anxiety disorders are in older people living in the community. A more recent study, the MentDis_ICF65+, is a cross-sectional multicentre survey based on a selection of European countries. Diagnoses were made with an adapted version of the Composite International Diagnostic Interview (CIDI), the CIDI65+. According to this study the overall 12-month prevalence of anxiety disorders were higher than in previous epidemiological studies, with a 12 month prevalence of 3.8% for panic disorder, 4.9% for agoraphobia, 1.4% for post-traumatic stress disorder and 9.2% for phobic disorders (Andreas et al. Reference Andreas, Schulz and Volkert2017).

Eating disorders

Although young women are most at risk of developing eating disorders, eating disorders do affect men and women of all ages. There are very few incidence studies of eating disorders among older people. A review found a prevalence of eating disorders according to DSM-5 of 3–4% in women and 1–2% in men over the age of 40 years (Mangweth-Matzek Reference Mangweth-Matzek and Hoek2017). In women, childhood sexual abuse has been associated with binge eating disorder. In men, excessive exercise is associated with binge eating disorder, which often masks symptoms and makes diagnosis more difficult (Mangweth-Matzek Reference Mangweth-Matzek and Hoek2017).

Personality disorders

There are numerous methodological concerns and a poverty of information regarding the epidemiology of personality disorders in old age. One common criticism is that the diagnostic criteria for personality disorders are focused on diagnosis for younger people and may not be valid in an older age group, where biological and psychological factors of ageing will affect the behaviour necessary for a diagnosis (van Alphen Reference van Alphen and Engelen2006).

The prevalence varies in different meta-analyses, but is reported to be between 2.8 and 13% of older people living in the community and 5 and 33% of patients in secondary psychiatric settings (van Alphen Reference van Alphen, Derksen and Sadavoy2012). The idea that personality disorders ‘burn out’ in later life may not hold true. A cross-sectional study in The Netherlands showed that the prevalence of personality disorders is similar in old age; however, the prevalence of specific personality disorders changes (Engels Reference Engels, Duijsens and Haringsma2003). This study used the self-report Questionnaire of Personality Traits, which is based on DSM-III and ICD-10. Older people were found to display more schizoid and anankastic personality traits and fewer emotionally unstable and paranoid personality traits (Engels Reference Engels, Duijsens and Haringsma2003).

Substance misuse

Substance misuse among older people is a growing area of concern. Harmful drinking is falling in most age groups in most countries, yet in ‘baby boomers’ (people born between 1946 and 1964) harmful drinking has been increasing (Rao Reference Rao and Roche2017). Alcohol is the most common substance misused, but drug misuse, mainly of prescription drugs, is also increasing in older people (Rao Reference Rao and Roche2017).

Prevalence studies of alcohol misuse in older people are likely to give an underestimate owing to (a) reluctance of older people to disclose this information, (b) less suspicion from clinicians and (c) fewer well-validated screening tools for this age group (Khan Reference Khan, Davis and Wilkinson2002). The prevalence of alcohol misuse in individuals over the age of 65 in the USA as diagnosed by DSM-IV is 1.2%, and is estimated at 0.24% for alcohol dependence (Kuerbis Reference Kuerbis, Sacco and Blazer2014). If this is extended to individuals over the age of 50, the prevalence of alcohol use disorders is 2.98% (Kuerbis Reference Kuerbis, Sacco and Blazer2014).

Older people take more prescribed and over-the-counter medications and, although often taken appropriately, these medications are prone to misuse. One review (Curkovic Reference Curkovic, Dogic-Curkovic and Eric2016) found that psychotropic medications are commonly prescribed to older adults worldwide, particularly benzodiazepines, antipsychotics and antidepressants, namely SSRIs. Polypharmacy is frequent in older people and adverse drug reactions are much more likely when four or more drugs are prescribed (Cadieux Reference Cadieux1989).

The prevalence of illicit drug dependence decreases with age, although rates do vary according to the cohort, with baby boomers using more illicit drugs than previous generations (Rao Reference Rao and Roche2017). In the USA, the prevalence of illicit drug dependence in older people is less than 1%, compared with 17% in 18- to 29-year-olds (Hinkin Reference Hinkin, Castellon and Dickinson-Fuhrman2001).

Suicide

Suicide rates vary according to country and in the UK suicide has been decreasing among older people since the 1950s (Gunnell Reference Gunnell, Middleton and Whitley2003). Recent figures from the Office for National Statistics (2017) show that suicide rates increase with age up to the highest rates among 45- to 49-year-olds. The rates then decrease until the age of 80, when they begin to rise again (Fig. 1); suicide rates in the very old have stayed relatively static over time. Reduction in suicide rates among older people in the UK has been associated with a reduction in access to lethal means and a rise in gross domestic product (GDP) (Gunnell Reference Gunnell, Middleton and Whitley2003).

FIG 1

Age-specific suicide rates per 100 000 population in the UK by gender and 5-year age groups, registered in 2017 (Office for National Statistics 2017: p. 10). Reproduced under the Open Government Licence v3.0.

As the baby boomer generation moves into old age, however, the general fall in suicides in the UK could change. This cohort is associated with higher rates of suicide than other generations and have higher rates of substance misuse than other age groups (Rao Reference Rao and Roche2017), which may predispose to suicide.