Depression and anxiety symptoms post-stroke/TIA: prevalence and associations in cross-sectional data from a regional stroke registry

Mood disorders are common in stroke-survivor cohorts and are associated with increased morbidity and mortality. Meta-analyses of point-prevalence rates suggest one third of stroke-survivors develop post-stroke depression and one quarter develop post-stroke anxiety [1],[2]. More than half of stroke survivors will be affected by depression at some point [3]. These summary data are important and strongly suggestive of a high stable prevalence of post-stroke mood disorder, but meta-analyses are limited by all the caveats that accompany data pooled from various studies and populations.

Stroke-survivor populations are heterogenous and there is still value in describing contemporary patterns of post-stroke mood problems in a suitably large population. A group who are not well represented in the available literature are those whose index event would be classified as a Transient Ischaemic Attack (TIA) or ‘minor stroke’. Although the evidence would suggest that TIA is associated with depression, the literature is characterised by studies of only modest sample size and is far from definitive [4]-[8]. No large study has examined mood problems in both stroke and TIA samples, using an identical assessment procedure, to compare. Further, we are not aware of any suitably large studies of anxiety incidence or prevalence following TIA. It seems plausible that the psychological effects of TIA may differ from stroke and a focussed analysis of mood disorder in TIA patients seems warranted.

Despite their prevalence, post-stroke mood disorders are relatively under researched compared to stroke related physical disabilities. A better understanding of the risk factors for depression and anxiety following stroke and TIA could help inform research and target interventions. Of particular interest are the effects of age; socioeconomic deprivation (SED); smoking and co-morbidity on mood disorder. We know that these factors are associated with mood disorder in unselected non-stroke cohorts [9]-[12] and these variables are also risk factors for stroke itself [13].

A robust population level assessment of post-stroke mood disorder, that describes association with important socio-demographic variables and includes TIA or minor stroke, could provide useful data for clinical practice and service planning. We used a stroke specific, city-wide clinical registry to describe patterns and associations of post stroke depression and anxiety.

Our aims were:

We present a cross-sectional analysis of a city-wide primary care administered data resource. Conduct and reporting of all analyses was in accordance with Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines [14].

To our knowledge, this is the first study to present population level data, describing post-stroke anxiety and depression symptom prevalence in stroke and TIA samples, using an identical assessment procedure (HADS), to compare. Our data add to the literature on mood disorder and stroke [21], in particular post-stroke anxiety and mood problems following TIA have received limited research attention to date and we designed our analyses to focus on those groups.

Our data suggest that depression and anxiety symptoms are common following stroke events, including TIA. Indeed, it is noteworthy that depression and anxiety levels are similar between these two patient groups. The single time point, cross-sectional data presented, does not allow us to look at temporal sequence and so we can make no inferences about causation. However, these data are in keeping with other studies that suggest stroke disease may be responsible for increased rates of mood disorder [1],[2]. Certainly, the prevalence rates we describe are higher than seen in “unselected” community populations assessed with HADS (11.4% depressed) and in patients with coronary heart disease (11.8% depressed) [22],[23].

As noted, we found that those with TIA had similar rates and predictors of mood disorder as those with stroke. This finding has implications for services. Many current guidelines suggest cognitive and mood screening of stroke-survivors but do not make specific comment on those with TIA. Our data would suggest that mood assessment should be similar for both groups and TIA should not be regarded as a “benign” diagnosis. We note recent work showing substantial cognitive deficits in patients with TIA and no stroke [24].

We assessed clinical and demographic associations with post-stroke mood symptoms. Ideally, through describing predictors of an outcome, services can target intervention. Mood disorder was increased in those with SED, a group who are often difficult to engage with chronic disease management. The finding of increased mood disorder in females is in keeping with patterns in the general (non-stroke) population [24],[25], albeit our prevalence of disorder was considerably higher.

Within our mood disorder group will be a number of patients with psychological problems that predate the stroke event and have continued, rather than incident post stroke mood disorder. There is a complex interaction between mood and vascular disease and mood disorder may be a risk factor for vascular events [26]. We are unable to differentiate incident and prevalent mood problems using these cross sectional data. The data available do not allow us to look at time since event as a covariate and this limits interpretation. In terms of service design and access to mood intervention, perhaps the more important matter is the absolute proportion of stroke-survivors with problems rather than the timing or potential aetiology and the data are suited to this purpose.

The strength of our analysis is the large dataset, derived from a population representative of “real world” stroke survivors. We further demonstrate that the LES resources provide a substrate for describing health care and outcomes that can be used to shape practice [27]. Our inclusion of both depression and anxiety should give a better description of post-stroke mood disorder than previous studies that had a focus on depression alone; while including equivalent test data for both classical stroke and TIA allows for comparisons between these groups.

We acknowledge limitations in our study methodology. The main limitation was missing data, particularly HADS data. The missing data is unfortunate but not uncommon in large clinical registries. We offer some description of the groups with and without HADS data but recognise that fundamental differences are likely to exist between the groups and the stroke-survivors included in our analysis may not be representative. We acknowledge cognitive/communication disability may have limited participation and this may impact on prevalence of mood problems recorded. There are assessments that can be used where direct interview is not suitable, for example observational based depression and anxiety measures [28]-[30]. Where patients refused HADS testing, there is no simple solution to improving data capture. For those with data not recorded we note the high proportion who are non-Caucasian. It seems plausible that nurses performing reviews may feel unable to collect data in those with limited spoken English or where socio-cultural factors may impact on the significance of a mood disorder diagnosis. We will use this finding to review how we train nurses in mood assessments. The proportion with missing data had a high prevalence of our independent risk factors for mood disorder (female sex; greater deprivation) and it is possible that our data are an underrepresentation of the true burden of post stroke mood disorder.

Our mood disorder assessment metric was based on HADS. We accept that HADS is not a substitute for expert derived clinical diagnosis and that HADS has complex four-choice response format requiring intact working memory. HADS has been validated for use in stroke, albeit utility is poor in acute stroke settings [31] and has been shown to have good diagnostic accuracy for making diagnosis of depression or anxiety [17]. There remains a lack of consensus regarding optimal sub-scale cut-offs for stroke [32]. LES therefore employs standard cut points (≥8) to detect caseness, following recommendation by the scale authors and consistent with previous stroke research [33]. We were interested in “TIA” patients and looked specifically at this group. We recognise that the time based definition of TIA used in this analysis is becoming obsolete. A tissue based definition of TIA would have been preferable but was not possible in this historical cohort. As a group of patients with no residual neurology at 24 hours, our “TIA” cohort may be better described as TIA or minor stroke.

In conclusion we have described a high prevalence of both depression and anxiety symptoms in a cohort of community dwelling stroke survivors. Rates of depression and anxiety symptoms were similar for TIA and stroke, suggesting that mood assessments and interventions should not be reserved for those with classical stroke only.

NMB acquired the data. NMB, TJQ and AHAR designed the study and analysed the data. AHAR performed the statistical analysis. NMB, TJQ and AHAR interpreted the data, drafted the manuscript and contributed to critical revision of the manuscript for important intellectual content. MRW and JJE performed critical supervision of the manuscript. All authors read and approved the final manuscript.