Discussion

In one of the largest observational studies of its kind, we examined recorded psychotropic medication prescribing in a cohort of patients diagnosed with personality disorder, and compared this between groups who did and did not use psychological services. Consistent with previous literature, the most commonly diagnosed personality disorder was the emotionally unstable type,^{Reference Paton, Crawford, Bhatti, Patel and Barnes11,Reference Coid, Yang, Yrer, Roberts and Ullrich19,Reference Cohen, Crawford, Johnson and Kasen20} the majority of the patients in the sample were receiving psychotropic medication from at least one class, just over a half were prescribed medications from two or more classes,^{Reference Baker-Glenn, Steels and Evans4,Reference Paton, Crawford, Bhatti, Patel and Barnes11} and antipsychotics and antidepressants were the most commonly prescribed medications.^{Reference Crawford, Kakad, Rendel, Mansour, Crugel and Liu3,Reference Baker-Glenn, Steels and Evans4,Reference Paton, Crawford, Bhatti, Patel and Barnes11}

Patients with personality disorder who were using psychological services were slightly younger, had less comorbid serious mental illness, more comorbid depression and anxiety, and less functional problems, such as activities of daily living impairment and problems with living conditions and work. They also had lower in-patient stay during the observation window, compared with patients not seen by psychological services. In agreement with Crawford et al,^{Reference Crawford, Kakad, Rendel, Mansour, Crugel and Liu3} we found that patients who had contact with psychological services were less likely to be prescribed psychotropic medication in general, and more specifically, were less likely to receive antipsychotics, benzodiazepines and antidepressants. These associations remained largely unchanged after several sensitivity analyses. One possible explanation for these results could be that patients seen by psychological services are generally more stable, with fewer symptoms. However, we adjusted for severity of functional problems (as measured by HoNOS), and the presence of comorbid diagnoses, with little effect on the size of the associations. We also took account of confounding by indication, in other words the probability that medication prescription (or the reduction in this occurring) is a marker of other clinical features that are triggering referral to psychological services. To test for this possible explanation, we used a propensity score restriction and adjustment, and the results remained unchanged, suggesting that differences in the characteristics between the two groups could not explain why patients with psychological service contact were less likely to receive psychotropic medication. There is a need to understand factors that may influence psychotropic prescribing in this population. It is not known whether differences in psychotropic medication use reflect patient choice between medication and psychological treatment, or whether it is a function of service factors such as fewer prescribers in psychological treatment settings. Considering the observed differences from a service delivery perspective, it is not known whether medication is being withheld from a group of patients that could potentially benefit from it, or whether psychological treatment settings facilitate collaborative, judicious decision-making around medication use. Patient factors, such as lack of psychological mindedness and choosing not to engage in therapy, may preclude access to available psychological treatments,^{Reference Rueve and Correll21} and medication may provide an alternative therapeutic approach in general psychiatry settings. Patients with personality disorder may wish to take psychotropic medication to obtain immediate relief from symptoms such as paranoia, affective dysregulation or anxiety, and may resist attempts to reduce or stop medication. It has also been suggested that the act of prescribing constitutes a way of relating to patients, which may be reflective of the psychopathology of the condition^{Reference Martean and Evans22} such that the act of prescribing itself may symbolise a tangible act of care. In the absence of psychological treatment, there may be an expectation to prescribe for the purposes of providing such care and attenuating distress.

Similar to previous findings,^{Reference Crawford, Kakad, Rendel, Mansour, Crugel and Liu3} we found low mood stabiliser use in our cohort. Our results suggest that there was no difference in the likelihood of being prescribed mood stabilisers between the patients who were in contact with psychological services and those who were not. However, this lack of effect could have been a result of a lack of statistical power, given the infrequent use of mood stabilisers in the cohort overall. Given the teratogenicity of several mood stabilisers, avoidance of prescribing in women of childbearing age may account for the low use of mood stabilisers in the sample. Alternatively, the low level of mood stabiliser use could be because of the lack of consistent evidence to support their clinical efficacy for patients with personality disorder.^{Reference Crawford, Sanatinia, Barrett, Cunningham, Dale and Ganguli23} Consistent with previous research, we found that a high proportion of the patients in our sample had multi-class psychotropic use^{Reference Zanarini, Frankenburg, Reich, Conkey and Fitzmaurice5,Reference Paton, Crawford, Bhatti, Patel and Barnes11} ; however, we found no difference in the risk of multi-class psychotropic use between patients who were receiving psychological service and those who were not. Therefore, our findings do not support previously proposed hypotheses that patients seen by psychological service are less likely to be offered medication treatment as a result of receiving alternative treatments.^{Reference Crawford, Kakad, Rendel, Mansour, Crugel and Liu3} An alternative hypothesis could be that patients who receive multi-class polypharmacy have either an inherently more severe personality disorder diagnosis or more severe comorbid mental disorders.

This study had several strengths. We examined a large and representative sample of patients seen by secondary mental health services. Although the sample was drawn from a single service provider, the heterogeneity of individual clinicians and teams are likely to outweigh any homogeneity imposed by the organisation, and our findings should be broadly reflective of real-world clinical practice in the UK.^{Reference Perera, Broadbent, Callard, Chang, Downs and Dutta14} Furthermore, utilising information available from de-identified electronic health records allowed us to identify and adjust for rich and diverse contextual information, reducing residual confounding. We further limited the effect of confounding by indication, through applying a propensity score method as a sensitivity analysis. However, several potential limitations also need to be considered. SLAM, like London, has a substantially higher proportion of minority ethnic groups, and higher presentation of people in both the lowest and highest socioeconomic groups, compared with England,^{Reference Perera, Broadbent, Callard, Chang, Downs and Dutta14} and this needs to be considered when drawing inferences. Although we adjusted for multiple confounders, it is possible that some residual confounding may have occurred. We were unable to measure factors such as duration of illness or stages of treatment as patients entered the observation period. Nor could we capture psychotropic prescribing reasons, which may have given information about possible mechanisms underlying our findings. Also, despite this being one of the largest observational studies in this filed, because of low numbers of certain personality disorder subtypes, we were unable to investigate medication use by personality disorder subtype. Nonetheless, our findings remain highly relevant as the field of personality disorder is moving toward the abandonment of subtypes of personality disorder and the retention of a single category of personality disorder.^{Reference Tyrer, Reed and Crawford24} Lastly, clinical symptoms were measured at one point in time. Capturing clinical symptoms over a period and examining changes over time could have provided a better understanding of how symptoms are related to medication use.

This study has important clinical implications. Improving the quality of psychotropic medication use in people with personality disorder is a clinical priority. National Institute for Health and Care Excellence guidelines¹ currently provide no guidance, although there is a clear need for clinicians to be better supported in their decision-making. Although American Psychiatric Association guidelines^{Reference Oldham, Gabbard, Goin, Gunderson, Soloff and Spiegel2} provide some support in relation to emotionally unstable personality disorder, evidence that most prescribing does not clearly target comorbid conditions^{Reference Paton, Crawford, Bhatti, Patel and Barnes11} is important in considering their clinical utility. Development of formal guidance on conducting medication reviews for patients with personality disorder is needed to ensure appropriate and rational prescribing.

We suggest that the broad framework of medication optimisation, which provides a person-centred approach to the safe and effective use of medication by using the best available evidence to guide shared decision-making about treatment,^{Reference Tyrer, Reed and Crawford24} may support this aim. Medication review, a core part of a medication optimisation process, involving the structured and critical evaluation of prescribed medication, should be offered to all patients with personality disorder, including those in psychological treatment services.

There is also a need for clinical trials to ascertain the efficacy of psychotropic medication in this patient group. However, perhaps both more urgently and more practically, research in real-world clinical settings is needed to examine medication effectiveness. Efficacy studies with necessarily strict inclusion and exclusion criteria demand a degree of patient uniformity that is seldom seem in clinical practice, and the complexities of setting up efficacy studies mean that this is perhaps a longer-term goal. Research should not exclusively focus on emotionally unstable personality disorder given that almost half of the patients in our cohort had an undefined personality disorder or another subtype.