Comparison with previous research
The socio-demographic differences found here between SMI and non-SMI patients are in line with previous studies using comparable data [17, 20]; namely, SMI status was associated with older age, male gender, and increased deprivation. Non-white ethnicity is also an established risk factor for SMI . The prevalence of SMI found in the current study (1.4%), is slightly higher than that reported previously for the Lambeth primary care population (1.0%, Pinto et al. ).
In a comparable study, Smith et al.  compared the physical health of Scottish primary health care patients with schizophrenia (SZ) or related non-organic psychoses to those without and found a stronger association between multimorbidity and SMI status even after adjusting for deprivation. They also reported SMI patients to have 1, 2 and 3+ comorbid health conditions, likely due to the more extensive range of health conditions considered (32 conditions compared to 12 here).
The prevalence of each individual condition was slightly lower among the Scottish study sample compared to that reported here, with the exception of the two most common conditions – hypertension and diabetes – which, alongside CKD, were more prevalent among the SMI group in this study (19% vs 16% and 13% vs 9% respectively). This is likely due to ethnic differences between the samples; for example diabetes and hypertension are more prevalent among African Caribbean, Asian and Black African groups, which constitute a greater proportion of the population of Lambeth compared to Scotland. Other differences may arise as the current study included outcomes that were based on QoF recording. Since QoF outcomes are incentivised, any bias may be expected to increase the recorded prevalence of outcomes in our study. As we found that all except two outcomes were less prevalent in our sample than in the study by Smith et al. based on Read codes, such bias would have led to an underestimation of the difference in the two samples and thus a more conservative interpretation of the difference between the two samples.
Positive associations between SMI status and COPD, thyroid disorders, diabetes and epilepsy were recorded in both studies in regression analyses after adjustment for socio-demographic status - though the associations were stronger in the Lambeth sample despite additional adjustment for ethnicity. Contrary to the current findings, they report no difference in relation to heart failure, stroke or CKD between the two groups; and, unlike the current study, found cancer, CHD, and atrial fibrillation to be significantly less likely among cases. Lastly, hypertension to be less likely after adjustment for socio-demographic characteristics, a finding we replicated only after additional adjustment for BMI.
These differences may be partly due to the differential definition of cases; while they included only SZ and non-organic psychoses, we used a composite measure which also included bipolar affective disorder. Further, we adjusted for ethnicity in addition to age, gender and deprivation. This may have been more important in the current study, given the high proportion of non-white ethnic groups in Lambeth, compared to Scotland, in which according to 2011 Census data, over 95% of the population overall is white. In this study, there was considerable ethnic variation in both SMI status and prevalence of health conditions considered. Ethnicity accounted for much of the association of most health conditions with SMI status, though to a lesser extent than age. Such differences are important to consider in future studies among this population.
Previous research finds those with SMI at greater risk of mortality than unaffected individuals, not accounted for by excess suicide [3, 8]. Recent large scale population based studies find much of this excess mortality to be associated with preventable physical health problems, including cardiovascular diseases and cancer [2, 4, 8], as well as diabetes mellitus [2, 22], infectious and respiratory diseases [2, 8].
The lack of difference found for CHD and atrial fibrillation in patients with and without SMI, and lower levels of hypertension among SMI patients in the current study, run counter to those findings. It may be interpreted that these outcomes are being underreported in primary care . Alternatively, this finding may be due to excess mortality associated with CHD among the SMI group. This explanation cannot be tested in the current study due to the cross-sectional nature of the data, thus can only provide data on prevalence (which is influenced by both the duration of conditions and the incidence of new cases). Cohort studies providing incidence data do support this hypothesis; for example, Crump et al.  report much higher mortality from ischaemic heart disease (and cancer) but no increased risk of diagnoses with these conditions among SMI patients; and, Osborn et al.  report an excess of deaths from CHD and stroke among SMI patients compared to control patients.
This study found that SMI patients were less likely to have missing data for ethnicity, BMI and smoking status – possibly suggesting that they do not have reduced access to GP services. Previous research has found an elevated consultation frequency among SMI patients which may account for this finding . Previous work suggests that care offered to SMI patients may be inferior to those without SMI . This finding may also indicate lower consultation rates among non-SMI patients. Future research may elucidate whether the more comprehensive reporting of BMI, ethnicity and smoking is due to greater frequency of access to primary care among SMI patients in this sample; whether or not this translates into better recording of health outcomes (and might account for some of the elevated risk observed here); and, whether or not care offered for physical problems reflects any greater access.