The impact of the first UK COVID‐19 lockdown on presentations with psychosis to mental health services for older adults: An electronic health records study in South London

Abstract Objectives Social distancing restrictions in the COVID‐19 pandemic may have had adverse effects on older adults' mental health. Whereby the impact on mood is well‐described, less is known about psychotic symptoms. The aim of this study was to compare characteristics associated with psychotic symptoms during the first UK lockdown and a pre‐pandemic comparison period. Methods In this retrospective observational study we analysed anonymised records from patients referred to mental health services for older adults in South London in the 16‐week period of the UK lockdown starting in March 2020, and in the comparable pre‐pandemic period in 2019. We used logistic regression models to compare the associations of different patient characteristics with increased odds of presenting with any psychotic symptom (defined as hallucinations and/or delusion), hallucinations, or delusions, during lockdown and the corresponding pre‐pandemic period. Results 1991 referrals were identified. There were fewer referrals during lockdown but a higher proportion of presentations with any psychotic symptom (48.7% vs. 42.8%, p = 0.018), particularly hallucinations (41.0% vs. 27.8%, p < 0.001). Patients of non‐White ethnicity (adjusted odds ratio (OR): 1.83; 95% confidence interval (CI): 1.13–2.99) and patients with dementia (adjusted OR: 3.09; 95% CI: 1.91–4.99) were more likely to be referred with psychotic symptoms during lockdown. While a weaker association between dementia and psychotic symptoms was found in the pre‐COVID period (adjusted OR: 1.55; 95% CI: 1.19–2.03), interaction terms indicated higher odds of patients of non‐White ethnicity or dementia to present with psychosis during the lockdown period. Conclusions During lockdown, referrals to mental health services for adults decreased, but contained a higher proportion with psychotic symptoms. The stronger association with psychotic symptoms in non‐White ethnic groups and patients with dementia during lockdown suggests that barriers in accessing care might have increased during the COVID‐19 pandemic.


| INTRODUCTION
Social distancing restrictions implemented by countries around the world to reduce the spread of coronavirus disease 2019 (COVID -19) have been found to be associated with deteriorating mental health of the general population. 1 The effects might have an even higher impact on older people for a number of reasons, including fear related to higher morbidity and mortality, reduction in already restricted means of socialising (e.g., closed community centres or places of worship), and difficulties in accessing modern technologies to remain in contact with relatives. [1][2][3][4][5] Evidence has already emerged of an increase in affective symptoms in older adults during the pandemic, 2,6,7 particularly in those living alone and therefore perhaps more vulnerable to the impact of social distancing measures. However, little is known about how lockdown affects psychotic symptoms in the older adult population. 8 A number of features of the lockdown may increase the risk of developing or exacerbating a psychotic illness. High perceived stress, as a result of fears generated during the pandemic, can be an important risk factor for both triggering and exacerbating psychotic symptoms. 9 In addition, lockdown has been associated with social isolation and loneliness 10,11 and several case reports have described new onset psychosis in older adults attributed to social isolation. [12][13][14] Given the complexities of diagnosing and treating late-life psychosis, 15 understanding the impact of the pandemic on the older adult population is essential to inform care. Although for many countries social distancing measures have been stopped, for others, they remain a part of life. We sought to understand this subject further through analysing data on routine referrals to mental health of older adult services (MHOA) of a large mental health trust during the first UK lockdown and before the pandemic.

| Data source
We carried out a retrospective observational study using anonymised electronic healthcare records from the South London and Maudsley NHS Foundation Trust (SLaM). SLaM is one of the largest specialist mental healthcare providers in Europe, serving a population of approximately 1.36 million residents across four South London boroughs (Lambeth, Lewisham, Croydon, and Southwark). SLaM's deidentified electronic health records can be accessed for research purposes via the Clinical Record Interactive Search (CRIS) platform. 16 CRIS was developed in 2007-2008, has supported over 250 publications, and is approved by the Oxford Research Ethics Committee C (reference 18/SC/0372) as a database for secondary analysis. 16 Data are extracted both from structured fields and from free text (e.g., clinical case notes, correspondence), the latter through natural language processing (NLP) algorithms developed using General Architecture for Text Engineering (GATE) software. 17,18
Date of referral was defined as index date for definition of patient characteristics.

| Outcomes: Psychotic symptoms
The primary outcome variable was the presence of psychotic symptoms around the index date (maximum of 6 months before or after) in the electronic health record. We ascertained whether patients presented with either hallucinations (regardless of whether patients also had delusions), delusions (regardless of whether patients also had hallucinations), or both (referred to as any psychotic symptom) using NLP to identify documentation of these symptoms from free text. 16 The NLP algorithm has previously been evaluated and in a random sample of 100 health-care documents, whereby high accuracy in identifying delusions (pre-annotated documents: precision = 90%; un-annotated documents: precision = 93%, recall 85%) and hallucinations (pre-annotated documents: precision = 90%; unannotated documents: precision = 84%, recall = 98%) was demonstrated. 19

| Factors potentially associated with psychosis
We additionally extracted data on a number of clinical and sociodemographic characteristics, including age at referral, gender, and ethnicity (dichotomised to White and non-White). We further characterised the presence of the following mental health diagnoses according to ICD-10 criteria 20 closest to the index date: dementia (F00-F03), psychotic illness (F20-29), affective disorder (F30-F39), and delirium (F05).
The presence and severity of clinical symptoms related to the referral, including mental and physical health problems and functional difficulties, were estimated from the Health of the Nation Outcome Scales (HoNOS) which are routinely completed in SLaM patients. 21 HoNOS items we included in our analysis were the mental health problems of agitated behaviour, non-accidental selfinjury, problematic substance or alcohol use, cognitive difficulties, depressed mood, physical illness/disability, as well as functional impairment, reflected in difficulties with activities of daily living (ADL), problems with relationships (social impairment), problems with living condition, and problems with daytime activities. These subscales are rated from 0 (no problem) to 4 (severe or very severe problem). We dichotomised these scores to define binary variables: 0-1, 'minor or no problems' and 2-4, 'mild to severe problems'. 21 Through established NLP algorithms we further identified whether patients were living alone, experiencing disturbed sleep and were prescribed psychotropic medications (antidepressants, antipsychotics, mood stabilisers or sedative medications) within a window from 6 months before or to 6 months after referral. 18

| Statistical analysis
All statistical analyses were carried out using STATA version 15 (StataCorp. 2017. Stata Statistical Software: Release 15. StataCorp LLC.). Initially, descriptive statistics were generated to compare those referred during lockdown and the corresponding pre-COVID period.
Next, logistic regression models were assembled to investigate associations of sociodemographic characterisitics, diagnosis (present vs.  Table 1 summarises the demographic and clinical characteristics of the full sample and by each referral period. In the full sample the mean age (SD) at referral was 77.9 (�9.5) years and 56.9% of patients were female. While there were no significant differences in age and gender between the referral periods, the proportion from non-White ethnicity backgrounds, with dementia, and living alone were lower during the lockdown period. Patients referred during lockdown more frequently had a diagnosis of affective disorder or delirium. Agitated behaviour, non-accidental self-injury, disturbed sleep, physical health problems as well as impairments of social functioning were more common during lockdown, but no differences in problems with activities of daily living were found. Use of antipsychotics, antidepressants, mood stabilisers and sedative medications were all recorded in a significantly higher proportion of referrals during the lockdown period than the pre-COVID period.  In the full sample, any psychotic symptom was found in 44.4% of referrals. During the lockdown period there was a significantly higher percentage of referrals associated with any psychotic symptom (48.7%) than in the pre-COVID period (42.8%). There was no significant difference between the percentage of referrals with delusions in during the first UK lockdown compared with the pre-COVID period, but the proportion of referrals with hallucinations was about 50% higher in the lockdown compared to the pre-COVID period.
T A B L E 1 Sample characteristics of the full cohort, referrals during lockdown and the corresponding same period pre-COVID

Characteristics
Full cohort (n = 1991) Lockdown period (n = 536) Pre-COVID period (n = 1455) P-value a  Table 2 shows regression models of patient characteristics associated with having any psychotic symptom (hallucinations, delusions, or both) in the lockdown and the comparable pre-COVID period, in unadjusted and adjusted logistic regression models.

| Factors associated with any psychotic symptoms during the lockdown and pre-COVID periods
In the adjusted model (adjusted for age, gender, ethnicity, and diagnosis), clinical characteristics associated with any psychotic symptom regardless of time period referred were: diagnosis of dementia, psychotic illness, affective disorder and delirium, the mental health symptoms of agitated behaviour and disturbed sleep, and the use of antipsychotic and sedative medication.
For three characteristics, interaction terms between lockdown and clinical characteristic were significant, indicating that associations with any psychotic symptom differed between the two time periods. Non-White ethnicity was only associated with any psychotic symptom during the lockdown, but not the pre-COVID period, and a referral with a diagnosis of dementia was more likely to be associated with any psychotic symptom during the lockdown than during the pre-COVID period. While antidepressant prescription was associated with any psychotic symptom during the pre-COVID period, this wasn't the case during lockdown. Table 3 shows patient characteristics associated with being referred to MHOA services with hallucinations. In summary, similar associations were observed to those with any psychotic symptom. Interaction terms with time period indicated that associations of hallucinations with non-White ethnic background and dementia diagnosis were stronger in the lockdown period, while those with affective disorder and antidepressant use were stronger in the pre-COVID period. Absolute referral numbers and the proportion of patients with dementia referred to MHOA services decreased during lockdown.

| Factors associated with delusions during the lockdown and pre-COVID periods
However, any psychotic symptoms, and hallucinations, and delusions were more likely to be associated with dementia during this period, suggesting that these symptoms played an important role in their pathway into care. Psychotic symptoms tend to increase with severity of dementia, though there are disease specific fluctuations. 30 The association demonstrated here may reflect late presentations of dementia, caused by the reduction in access to timely dementia diagnoses and interventions, as memory services activity decreased during the pandemic. 25 In addition, neuropsychiatric symptoms of dementia, including psychotic symptoms, are associated with carer burden and distress, 31 and the presence of these symptoms specifically might have led to carers or patients seeking help from mental health services.
In our regression analyses, non-White ethnicity was more frequently associated with psychotic symptoms during lockdown than during the pre-COVID period, whereby this was more pronounced for any psychotic symptom and delusions, than for hallucinations. Ethnic minority groups have been disproportionally affected by the pandemic in terms of severe illness and mortality, 32 which might also have an impact on their mental health. In the US, people from non-White ethnic groups have reported experiencing significantly worse mental health outcomes, 33 which is echoed in the UK where people from ethnic minorities have reported higher levels of mental distress during the pandemic. 34 These higher levels of distress, in combination with an increased likelihood of a negative life event and financial concerns, are themselves risk factors for developing psychosis, and might explain the stronger associations observed in patients from non-White groups during lockdown compared to pre-pandemic. 9 Additionally, evidence suggests that black older adults have fewer social relationships than white older adults, 35  Other secondary findings from our study include the observation that patients referred during lockdown were less likely to be living alone compared with the pre-COVID period, and that an association of living alone with delusions was less likely during lockdown compared to the pre-COVID period. This is perhaps surprising as other studies have reported living alone as a major risk factor for developing negative mental health outcomes during social isolation. 3