Our findings highlight that individuals with bipolar disorder have excess physical health comorbidities compared to the general population. Similar to our recent findings for schizophrenia , the majority of individuals with bipolar disorder (63.9%) had at least one chronic physical health comorbidity. Compared to patients without bipolar disorder, individuals with bipolar disorder were more likely to have a primary care record of both single and multiple physical health problems, even after taking into account age, gender and deprivation status. In general, our findings for bipolar disorder are similar to previous findings for schizophrenia, that is, for both conditions there appear to be high rates of multiple comorbidities and a systematic under-recording of cardiovascular conditions . However, in the current paper we have extended this work by also identifying that individuals with bipolar disorder and a primary care record of coronary heart disease or hypertension tend to experience much less intensive prescribing for these conditions.
The three conditions with the highest prevalence within the bipolar group were viral hepatitis, constipation and Parkinson’s disease (which includes Parkinsonism). There were also high rates of diabetes, chronic pain and COPD. Perhaps at least partly as a consequence of known treatment side effects (for example, lithium therapy), thyroid disease and chronic kidney disease both also had higher prevalence, again after age/sex adjustment, compared to controls.
The relatively high rates of viral hepatitis, constipation and Parkinson’s disease may have a number of explanations. There are high rates of drug abuse in bipolar disorder  and within Scotland there is a national programme of testing for blood borne viruses in those diagnosed with drug dependence. This may explain the relatively high rates of viral hepatitis recorded in the bipolar group. Features of Parkinson’s disease can be a direct result of antipsychotic treatment. Although there is no clear mechanism for increased rates of constipation within bipolar disorder, psychological symptoms including those experienced as part of significant mood disorder are related to symptoms of irritable bowel syndrome including constipation [21, 22]. Furthermore, constipation is an under-recognised side effect of antipsychotic medication that can significantly impact on quality of life, experienced by 20% to 30% of all patients prescribed such drugs .
Recording of cardiovascular comorbidity
Rates of coronary heart disease, heart failure, stroke, and transient ischaemic attack (TIA) and peripheral vascular disease were not significantly elevated in the bipolar group. This is of note because rates of proatherosclerotic conditions including smoking and diabetes mellitus were found to be very significantly elevated in patients with bipolar disorder. Furthermore, individuals from the bipolar group displayed lower recorded rates of hypertension and atrial fibrillation compared to those without bipolar disorder. These findings might be unexpected given existing knowledge on physical comorbidity in bipolar disorder [11, 12, 15, 16], but, as noted above, recent reports have highlighted that physical comorbidity in major mental illness is often underdiagnosed and undertreated [15–17, 24].
There is strong evidence of under-recognition of cardiovascular disease in people with major mental illness. A recent Swedish national cohort study found that individuals with schizophrenia were more likely to die prematurely than the general population (15 years earlier for men and 12 years earlier for women) and the leading causes of death were cardiovascular disease and cancer. However, rates of recording of cardiovascular disease and cancer were only slightly increased in people with schizophrenia, even though these individuals had more healthcare system contacts, suggesting that cardiovascular disease and cancer are significantly underdiagnosed and/or under-recorded in this population . Furthermore, using a similar approach to the one used in this study, we have identified an under-recording of cardiovascular illness in individuals with schizophrenia compared to expectations .
Our findings suggest a strong likelihood of under-recording of cardiovascular disease in bipolar disorder. There are a number of possible reasons for this health inequality. People with bipolar disorder may be less likely to seek help from their general practitioner (GP) with symptoms of cardiovascular disease because of low awareness of cardiovascular risk factors and associated symptoms . This could be due to mental state abnormalities, social isolation and in some cases low levels of education. Cardiovascular risk estimation tools such as Framingham, ASSIGN and QRISK have not been validated in individuals with major mental illness  and may underestimate cardiovascular risk in bipolar patients, who are typically younger, smoke more and have higher blood pressure than the population traditionally used to generate the cardiovascular risk. This too may compound problems with under-recognition and undertreatment. People with bipolar disorder may also be prone to less regular attendance at clinics because of depressive and manic or hypomanic episodes .
Prescribing of cardiovascular medications
A recent meta-analysis of prescribing data for patients with and without major mental illness found that individuals with a history of major mental illness had significantly lower prescription rates for cardiovascular medications . Our findings indicate a similar prescribing disparity for both cholesterol lowering and antihypertensive medications. People with bipolar disorder and both coronary heart disease and hypertension had lower rates of statin prescription, as well as less intensive treatment with antihypertensive drugs. Although there was no evidence of different rates of poor blood pressure control and high cholesterol in a comparison between bipolar patients with CHD/hypertension comorbidity versus non-bipolar patients with CHD/hypertension, it is important to note that for people with CHD statins are recommended for all patients irrespective of pretreatment cholesterol level, so lack of prescription is indicative of a missed opportunity to deliver an evidence-based intervention. For people with hypertension, lower treatment intensity in the face of similar blood pressure or cholesterol control is more ambiguous. However, it is not possible to fully explore the relationship between treatment and outcome in cross-sectional data, particularly given the substantial premature cardiovascular mortality in bipolar disorder [1, 12], consistent with people who were undertreated and with poor intermediate outcome control not surviving to be analysed in this kind of cross-sectional analysis. Our study cannot fully examine this issue, but highlights the importance of assessing treatment, intermediate outcome, and cardiovascular events and mortality within longitudinal research.
Although there are no universally accepted explanations for differences in prescribing for those with major mental illness, a number of possibilities exist. Mental health professionals may fail to appropriately diagnose physical health problems in their patients [28, 29], they may carry out incomplete physical examinations  and a proportion may not feel confident with prescribing physical health (non-psychotropic) medications . Despite this, most pharmacological treatments for cardiovascular illness take place in primary care. General practitioners who do not feel confident in managing complex and severe mental illness may be less likely to follow-up patients with major mental illness and comorbid physical health problems . There are, however, conflicting reports within the literature that suggest that a higher frequency of healthcare visits in the context of mental illness may result in increases in guideline-recommended care for comorbid physical conditions . Moreover, health care professionals working with individuals with major mental illness may be cautious when prescribing for this group due to concerns about suicide risk, as some cardiovascular medications, such as beta blockers, may be dangerous in overdose [16, 33]. As noted above, it is also possible that concordance with healthcare provision, including attendance at GP and outpatient clinics, may be reduced in those with major mental illness . Our results in relation to prescribing are potentially of concern because the adherence to cardiovascular medication in individuals with major mental illness may be even lower than the prescribing data suggests.
Strengths and limitations
Strengths of this study include a very large sample (almost 1.8 million individuals) which is representative of the Scottish population and which has been selected from 314 primary care practices. It is worth noting that we have used routine data from primary care in this study rather conducting an epidemiological study. We believe that this large sample of ‘real world’ patients has some important advantages over an epidemiological study which, although more systematic with regards to diagnosis, will inevitably be smaller and almost certainly less representative because it will depend upon the referral of patients or volunteering patients.
Our use of Read Codes to identify both the index condition (bipolar disorder) and multiple medical comorbidities is potentially imperfect but represents the trade off between diagnostic accuracy and real-world representativeness and utility that is implicit in choosing between epidemiological studies and routine data studies. Clearly some estimates for some conditions may have been prone to bias. For example, as noted above, higher recorded rates of viral hepatitis and thyroid disease for bipolar patients may be related to more frequent blood monitoring within this group.
The rate of a recorded diagnosis of bipolar disorder of 0.2% is lower than might be expected, with most estimates of the prevalence of bipolar disorder being around 1% . This likely reflects both underdiagnosis compared to epidemiological estimates, and under-recording reflecting the considerable debate and variation in the definitions and prevalence of the condition. However, this was not an epidemiological study but rather used routine clinical data on a very large number of participants. Our sample therefore represents the more severe end of the bipolar spectrum and we can be reasonably confident that most individuals recorded as having bipolar disorder did indeed have the condition.
It is possible that some patients with bipolar disorder are known to secondary care services but are not recorded within primary care and that a small additional proportion may not be in contact with either primary or secondary care. Further, these are routine data from 314 primary care practices and there may be some variability of diagnostic coding for major mental illnesses across these practices.
These data were cross-sectional, therefore no directionality or temporality can be attributed to the association between conditions, nor can any association confirm a causal relationship. It is possible that some of the comorbid medical conditions assessed may have predated or precipitated a diagnosis of bipolar disorder (and conversely that bipolar disorder was an antecedent for some of the medical conditions) but this is not something that can be adequately assessed using this cross-sectional dataset. All diagnoses were based on administrative data, and we did not have structured interviews to confirm their accuracy.
It should also be noted that an unknown (likely small) number of patients in our bipolar group had a read code of ‘severe/psychotic depression’. The prevalence of severe/psychotic depression in the general population is considerably lower than that of bipolar disorder and psychotic features are present in only a minority of cases of major depressive disorder . There are also reports that indicate that severe or psychotic depression may be an early manifestation of bipolar disorder . Unfortunately, due to limitations in the way these data were extracted it was not possible to separate out patients with psychotic depression from the bipolar group, although we expect that the proportion of patients in this sample with psychotic depression would have been very low, given that 11 out of the 13 Read Codes were for bipolar disorder diagnoses.