OP11 Ethnicity as a key driver of system-wide health and care service costs in patients with serious mental illness: a linked electronic health record cohort study

Background Improving care for people with serious mental illness (SMI) is a policy priority in England, yet little is known about service use patterns across the broader health and care system outside of acute and mental health care. We sought to investigate individual- and practice-level features associated with health and care system costs (as a comparable marker of use) for patients with SMI using a system-wide linked patient-level dataset that has been developed in Barking and Dagenham (B&D), with the aim of driving improved inter-organisational working and financial sustainability. Data This retrospective cohort study includes patients registered and resident in B&D in financial years 2016/17 and 2017/18. GP electronic health records were linked with administrative data from community, acute and mental health Trusts serving the area for both years, and local authority-funded social care records for 2016/17 (with costs imputed for 2017/18). The study population was adults on the SMI Quality and Outcomes Framework register, with comparator groups (1) matched on age, gender, and number of long term conditions (LTCs), and (2) the general adult population. Cost analyses took three forms: descriptive, regression at individual-level (generalised linear model using a Tweedie distribution), and funnel plots at GP practice-level. Results SMI patients had three-fold greater system costs than matched patients. Increased spending was concentrated in mental health Trust outpatient and social care settings. Increased spending was not observed in primary care and emergency departments. Apart from LTC count, ethnicity was the main driver of spending, particularly within social care. Spending per White British patient per year (£10,181) was almost twice that of Asian or Black patients (£5,442 and £5,505, respectively). Regression analysis demonstrated that this increased spending for White British patients holds after accounting for age, gender, deprivation, LTC count, and primary care network (p=0.0002). Funnel plot analyses did not show significant patterns in GP practice-level variability in social care, emergency departments nor mental health service use. There was no clear individual-level relationship between primary care consumption and use of these other services. Discussion These findings raise questions around ethnicity-based equity of care. Furthermore, they do not support the common narrative around differential access and gate-keeping at a GP practice-level affecting broader system costs. An increased focus on drivers of costs in social care rather than emergency or primary care settings may be needed for patients with SMI. However, the generalisability of these B&D results to the general population has not yet been explored.