The epidemiology and clinical features of personality disorders in later life; a study of secondary care data

Abstract Objectives Personality disorders (PDs) are often conceptualised as impacting individuals throughout their life. However, there has been limited study of the disorders in those over the age of 65. We have used the psychiatric secondary care medical records of 21,971 individuals over the age of 65 from Cambridgeshire, UK, who received care between 2014 and 2021 to characterise older patients with a PD diagnosis. Methods The data from all patients >65 with a diagnosis of personality disorder (PD) was extracted (n = 217) along with two comparison groups (n = 2170); patients <65 with a diagnosis of PD and patients >65 with a psychiatric diagnosis other than PD or dementia. Results Compared to younger patients with PD, older patients were more likely to be male, married, suffering from a mixed PD and live in less deprived areas. Compared to patients >65 with diagnoses other than PD, older patients were more likely to be female, single or divorced and had a higher level of social deprivation. Our most striking finding was that older patients with PDs were more likely to experience polypharmacy. A mean of 18.48 different drugs had been prescribed over their lifetime, compared to 9.51 for patients >65 with other mental health diagnoses. Conclusion Here we present the largest ever description of this group of patients and provide insights that could inform clinical practice and future research.

� The number receiving care is much smaller than expected based on studies of the general population, which may indicate unmet need.
� Patients in this group show improvement in their Health of the Nation Outcome Scales scores following treatment in NHS mental health services.
� Older people with a diagnosis of PD received double the number of diagnoses and different prescription medications in their lifetime compared to patients of the same age with other psychiatric diagnoses, raising the possibility that interventions such as a medication review might be helpful.

| INTRODUCTION
Personality disorder (PD) encompasses a group of diagnoses where an individual's thoughts, feelings and emotions deviate significantly from the expectations of society. 1 These characteristics must be present over a prolonged period, independent of other physical or mental health diagnoses. These disorders impair an individual's ability to function, for many leading to significant distress and a concomitant increased likelihood of experiencing further mental disorder and premature mortality. 2 Individuals diagnosed with PD often have above-average levels of interaction with healthcare services which potentially increases their risk of iatrogenic harm and has a substantial economic cost. 3 There is an overlap between the concept of PD and other clinical constructs, including complex emotional needs. In this paper we use the term PD as the data here are based on that diagnostic classification. Similarly, we recognise that diagnostic categorisation of personality has been questioned, for example, in contrast to a dimensional/trait approach but in this paper we have used ICD diagnostic categories to identify patients as that is how the clinical records are categorised.
The prevalence of PDs in the general population of western countries was reported as 12.16% by Volkert and colleagues in a systematic review and meta-analysis. 4 However, the average age of patients in the 10 papers reviewed was 33-51 years old. Few patients over the age of 65 were included meaning this group are relatively poorly described. Whilst some have looked at specific associations, for example, between older people with PD and quality of life, 5 not all subtypes of PD were included, and the majority of patients were not over 65. More recent systematic reviews, for example, by Penders and colleagues, 6 have highlighted the urgent need for more research concerning the epidemiology of this patient group.
Given the paucity of current research into PDs in later life, we sought to characterise this patient group and compare them to younger patients with PDs and older patients with other psychiatric diagnoses using a large clinical database from an NHS clinical service. The patients identified were then compared to two 10 times larger control groups whose data was extracted using the same methodology. These groups were patients under 65 treated by CPFT with a PD diagnosis (n = 2170) and patients over 65 treated by CPFT with a psychiatric diagnosis other than PD or dementia (n = 2170).

| MATERIALS AND METHODS
The groups were both unmatched samples from the overall patient population. Patients with dementia were excluded from all groups.
Demographic features of patients over 65 with PDs were also compared to those over 65 in the general population using publicly available 2011 census data for Cambridgeshire and Peterborough. 7 The index of multiple deprivation (IMD) was used as an indicator of socioeconomic status. This is a national system compiled by dividing England into lower-layer super output areas each with a population of approximately 1500. The areas are ranked based on scores in seven domains from 1, the most deprived, to 32,844, the least.
We used the Health of the Nation Outcome Scales (HoNOS) to indicate the psychiatric health of our patients before and after treatment. This enabled analysis of treatment outcomes using paired initial and discharge assessments. Not all the patients had been discharged during the period of the study and, of those who had been discharged only some had received an initial HoNoS assessment. To enable paired analysis, manual extraction of pairs was completed. A pair was defined as the earliest initial assessment coded 'initial' combined with the earliest HoNoS assessment coded 'discharge' to follow this. The pairs were then analysed using a paired two-sample t-test assuming equal variance.
The descriptive statistics calculated were mean, median, mode and standard deviation. Chi-squared tests were conducted to assess the significance of the differences between groups in categorical data, such as marital status, PD diagnosis and IMD decile. To test the significance of the difference between average HoNOS scores, unpaired t-tests were used for between-group comparisons and paired t-tests for within groups comparisons.

| RESULTS
The database contained n = 21,971 records of patients over 65, and Analysis of the distribution of IMD also indicates patients over 65 with a diagnosis of PD are more deprived than the general population in absolute terms. However this difference in distribution did not reach statistical significance (χ 2 = 13.65, d.f = 9, p > 0.1).
Patients received a broad range of diagnoses with at least one from each of the ICD10 PD subtypes. The most frequently coded was F60.3, emotionally unstable personality disorder (EUPD) (n = 90).
Amongst women (n = 156), 44% of PD diagnoses coded were F60.3 but only 23% (n = 90) in men. This is illustrated in Figure 2A The outcome of secondary care treatment was evaluated using HoNoS scores. Eighty five paired records were identified where scores were taken on initial assessment and at discharge. There was a significant difference between these on a paired two-sample test for total HoNoS scores p = 0.0004 with lower scores on discharge suggesting a positive effect of intervention by the mental health services. This is illustrated in Figure 3A,B.

| Comparison 1: Patients under 65 with a diagnosis of F60-62
A 10 times larger sample of n = 2170 of patients under 65 was extracted for comparison. This was 69.5% of the under 65 s in the database with a PD diagnosis (n = 3121). The group had a mean age of 37.4 years old (SD = 11.39).
There was no significant difference between the gender distribution of the patient groups (χ 2 = 2.17, d.f = 2, p > 0.05). In both samples, most patients with a PD diagnosis were female, 69.1% of those under 65% and 65.0% of those over 65.
There was a significant difference in marital status (χ 2 = 417.81, d.f = 6, p < 0.001). As shown in Figure Figure 2C. However, this difference was not found to be significant in a chi-squared test (χ 2 = 16.02, d.     diagnosed. This might be consistent with a longitudinal study of individuals with borderline PD which found older adults had more severe depressive and anxiety symptoms while younger adults had a higher incidence of suicide attempts and aggression. 11 Alternatively, the low numbers could reflect some of the impediments to diagnosing older people with a PD. One contributor to this is the increased likelihood of our older patients having comorbid physical and mental health conditions. 12 Disentangling the signs and symptoms of these conditions increases the challenge of making a definitive diagnosis of a PD. Finally, there may be reluctance in diagnosing PD on the part of clinicians working in old age psychiatry. From a clinical perspective a comprehensive psychological formulation is likely to inform and enhance care planning and may be easier for patients to accept than strict diagnostic categories, but may also contribute to lower rates of diagnostic coding.
The disparity in numbers between the under and over 65 s with a PD diagnosis is also notable. A PD is a diagnosis associated with increased mortality, and it may be some patients do not live long enough to move into old age services. 13 Alternatively, a PD may improve over time meaning individuals do not need secondary care.
However, the disparity could also be explained by there being a large unmet need of patients not being seen by services or individuals who are assessed being given diagnostic codes for other comorbid disorders (for example affective disorder) rather than a PD. These possibilities require further examination.  While the analysis that forms the basis of this paper endeavoured to produce the most comprehensive characterisation of this patient group to date, we acknowledge there are limitations. The first of these arises from only having 217 patients over 65 with a PD diagnosis within our records. To mitigate this, we included patients from the full timespan covered by our database. To advance our work future studies could pool data from mental health trusts to assemble a greater sample size.
The second limitation of our work occurs due to the nature of the database we have used. Text-based extraction allows us to make inferences and form hypotheses based on comparisons to younger patients and those with conditions other than PDs. However, to take this work further interventional studies are needed to test the hypothesises and assess if the differences we have found can be used for the benefit of patients. A good example of this is the significantly higher number of different medications prescribed to individuals with PDs over their lifetime. From this, we infer they may be at a higher Conceptualisation.