Outcomes of Pre-Existing Diabetes in People With/without New Onset Severe Mental Illness: A Primary-Secondary Mental Healthcare Linkage in South London, United Kingdom

AIMS: To compare people with diabetes developing severe mental illness (SMI) to those with diabetes alone with respect to risk status, diabetes care receipt, and diabetes-relevant outcomes in primary care. METHODS: Data from mental health care (Clinical Record Interactive Search; CRIS) linked to primary care (Lambeth DataNet; LDN) were used. From patients with a type 2 diabetes mellitus (T2DM) diagnosis in primary care, those with a new SMI diagnosis were matched (by age, gender, and practice) with up to five randomly selected controls. Mixed models were used to estimate associations with trajectories of recorded HbA1c levels; Poisson regression models compared total and cardiovascular comorbidity levels and number of diabetes complications; linear regression models compared BMI and total cholesterol levels; conditional logistic regression models investigated microalbuminuria, receipt of a foot or retinal examination, use of statins and receipt of insulin; Cox proportional hazards were used to model incident microvascular and macrovascular events, foot morbidity and mortality. RESULTS: In a cohort of 693 cases with SMI (122 bipolar disorder, 571 schizophrenia and related) and T2DM compared to 3366 controls, all-cause mortality was increased substantially in the cohort with SMI (adjusted hazard ratio 4.52, 95% CI 3.73–5.47; for bipolar 5.59, 3.37–9.28; for schizophrenia 4.42, 3.60–5.44). However, for all the other outcome comparisons, the only significant findings were of reduced foot examination (adjusted odds ratio 0.75, 0.54–0.98) and reduced retinal screening (0.77, 0.61–0.96). CONCLUSION: Higher mortality suggests increased risk of adverse outcomes for people with pre-existing T2DM who develop SMI, and reduced foot/retinal examinations suggest disadvantaged healthcare receipt. However, other potential explanations for the mortality difference could not be identified from the outcomes analysed, so further investigation is needed into underlying causal pathways.