In people with type 2 diabetes, achievement of target levels for cholesterol, BP and HbA1c 1 year after diagnosis of diabetes differed by SMI status, with differing patterns observed across these cardiovascular risk factors and by individual SMI disorder and sex. Compared to people without a mental illness, those with SMI were less likely to achieve target cholesterol levels but more likely to achieve target BP levels. People with bipolar disorder and men with schizophrenia were more likely to achieve the HbA1c target level, but women with major depression were less likely to achieve this target level. The association between SMI status and statin prescribing differed by prior history of CVD. Among people without CVD, those with an SMI were equally or more likely to be prescribed statins compared to those without mental illness, both at diabetes diagnosis and 1 year later. In people with pre-existing CVD, statin prescribing was less likely among people with an SMI vs those without.

Few studies have investigated achievement of lipid target levels by SMI status among people with type 2 diabetes. Two studies found no difference in lipid target level achievement based on SMI status [14, 15], which contrasts somewhat with our findings. However, direct comparison is difficult as the previous studies did not differentiate between SMI diagnoses. Moreover, these studies defined SMI using both primary and secondary care data, and may therefore have included a group with less severe SMI than in our study. If associations differ by severity of SMI, then this may account for these inconsistent findings. Our study confirms variation in associations by individual SMI disorder, and the importance of analysing conditions separately, as previously demonstrated by a Danish cohort study [12]. Interestingly, a fourth study found no difference in lipid target level achievement among those with hospital-diagnosed depression vs no mental illness, whereas people with depression defined based on antidepressant prescription data were more likely to achieve lipid target levels than those without mental illness [17]. In contrast to the results for lipid target achievement, our findings on BP target achievement do align with the results of two previous studies [14, 15]. Additionally, our findings on achievement of HbA1c target levels are somewhat consistent with results of previous studies, which either reported that target level achievement did not differ by mental illness status [14, 15, 17] or that target level achievement was more likely in people with mental illness [12]. Two studies investigated differences by sex, and, in contrast to our findings, did not find any evidence for effect modification by sex [14, 17]. Only one study analysed SMI disorders separately [12] and one included people with depression only [17], with the remainder analysing SMI as a composite group [14, 15].

Our findings on prescription of lipid-modifying drugs (statins) by SMI status align with the few previous studies that have investigated this [16, 17]. A Danish study reported that having a hospital diagnosis of depression was associated with a greater likelihood of initiation of treatment with lipid-modifying drugs, but they did not stratify by history of CVD [17]. Interestingly, a Finnish cohort study [16] found more marked differences in prescription of lipid-modifying drugs by SMI status for secondary prevention of CVD than for primary prevention of CVD [16]. Qualitative studies involving healthcare professionals and patients provide plausible explanations for these findings [29,30,31,32]. These explanations include overshadowing of physical health problems and challenges of cardiovascular risk management in people with SMI, especially among patients with more complex issues, such as those with multimorbidity, including CVD [29,30,31,32]. Further quantitative and qualitative research is needed to explore the interplay between mental and physical health problems. In particular, studies should investigate how this interplay affects health inequalities in high-risk subgroups such as people with prior CVD or other physical health problems.

It appears paradoxical that people with major depression had the lowest likelihood of achieving total cholesterol target levels whilst being the most likely to have statins prescribed at baseline and 1 year thereafter. This may reflect confounding by indication or a difference in medication adherence, as reported in qualitative studies [30, 31]. Interestingly, Rohde et al used a proxy to assess treatment adherence among people with type 2 diabetes with and without depression, and reported that people with depression appeared to have higher treatment adherence compared to people without a mental illness [17]. This result warrants further investigation to determine whether these findings are replicable in other settings, including Scotland, and to establish whether differences in medication adherence may partly account for our observed findings.

Our study has multiple strengths. We included a representative population-based cohort from the Scottish Diabetes Research Network National Diabetes Dataset, which includes information on >99% of people with diabetes in Scotland. Additionally, our previous research demonstrated that, in Scotland, cardiovascular risk monitoring during the first year after type 2 diabetes diagnosis is similar or better in people with an SMI compared to those without [11], and so we can be confident that differences in cardiovascular risk management are due to differences in the translation of risk monitoring into risk management. Due to the breadth of data included in the Scottish Diabetes Research Network National Diabetes Dataset, we were able to adjust for a wide range of covariates. Finally, the sample size allowed for stratification by sex during primary analyses and history of CVD during secondary analyses, thereby allowing investigation of differences in outcomes within these subgroups, which have rarely been addressed in previous studies. The observed sex differences in achievement of the HbA1c target level are intriguing, and add to the findings of other studies that reported important sex differences in biomedical outcomes among people with diabetes [33].

A limitation of our study is that we defined SMI based on acute and psychiatric hospital records. Therefore, our findings may not be generalisable to people with SMI who have not been admitted to hospital, and whose condition may be less severe. Additionally, we did not account for the development of SMI during the follow-up period. Although unlikely to have affected the findings for schizophrenia and bipolar disorder, which are usually diagnosed at a younger age than type 2 diabetes, this may have affected results for major depression, given the bidirectional association between type 2 diabetes and depression [34]. We also did not have information on other lifestyle-related factors such as physical activity and diet, and so could not explore whether these factors may account for some of the observed associations. We did not adjust for ethnicity due to this information being missing in approximately 9% of patients. Although the general population and subgroup with diabetes in Scotland is predominantly white [35], lack of adjustment for ethnicity may have resulted in residual confounding. Limited data on HDL and LDL levels prohibited analyses of these outcome measures in addition to total cholesterol. We only had information on medication prescribing and not adherence, and so could not determine how statin prescription translates into adherence. It was beyond the scope of the current study to examine the role of psychotropic medication use. Whilst antipsychotic medication and antidepressants are known to have metabolic side-effects, their association with cardiovascular risk factor levels once diabetes has developed is less clear [36]. However, lack of adjustment for psychotropic medication may have led to residual confounding of effect estimates. Finally, almost 40% of individuals had missing data for at least one covariate. We addressed this using multiple imputation. We could not assess target level achievement in approximately 10% of individuals for each risk factor. As using imputed outcome data in analyses is reported to mainly add noise, we excluded these individuals after multiple imputation [37].

Our study makes a valuable contribution to an important but notably under-studied area, and extends our previous work on receipt of diabetes processes of care in Scotland [11]. The findings demonstrate that, in addition to being equally or more likely to receive routine diabetes monitoring (delivered in the primary care setting) [11], people with an SMI are, overall, equally or more likely to achieve BP and HbA1c target levels, but not necessarily cholesterol target levels in the short term following diabetes diagnosis. Our study highlights the importance of investigating these associations by individual SMI disorder and sex, as it reveals novel insights into how the association between SMI and achievement of cardiovascular risk factor targets may vary by these factors. For example, it is interesting that women, but not men, with major depression were slightly less likely to achieve HbA1c target levels within the first year after diabetes diagnosis. As existing data in this area are so limited, further studies need to examine this finding in other populations and settings to confirm or refute these patterns. Moreover, future research should look beyond the short term to examine whether associations between SMI and achievement of cardiovascular risk factor target levels change in the medium and long term, and to explore whether cardiovascular risk management varies between individuals with SMI who do or do not experience subsequent psychiatric admissions during follow-up. This has rarely been investigated, with just one study reporting that HbA1c levels, but not cholesterol levels, increased more in those with an SMI vs those without an SMI in the 1–4 year period after diabetes diagnosis [18]. As also found in a previous study [38], the proportion of all people with type 2 diabetes receiving statins is lower than expected based on Scottish clinical care guidelines [38]. Our findings that statin treatment differences exist among people with SMI and a history of CVD reveal a worrying disparity in this high-risk complex-needs subgroup. This reinforces the importance of clinical review of statin prescribing, particularly for secondary prevention of CVD among people with SMI. Further research to determine the underlying reasons for these disparities is necessary to inform appropriate interventions. In particular, concerns about polypharmacy, which may contribute to the pattern of findings for the high-risk complex-needs subgroup, merit investigation. Future research should also examine, in depth, the role of antipsychotic and antidepressant medication, given their potential for metabolic side-effects, and the role of glucose-lowering treatment, which may vary by SMI status. Additional research steps include investigation of the management of other cardiovascular risk factors, including HbA1c and BP. It is somewhat paradoxical that, in the short term at least, cardiovascular risk factor control is generally similar if not better in people with an SMI vs those without, and yet this vulnerable group have a higher risk of CVD and cardiovascular death [10, 39, 40]. The recommended next research steps will illuminate this complex picture further, particularly the extent to which sub-optimal cardiovascular risk factor management plays a role in SMI disparities in diabetes outcomes.