Association between cardiometabolic disease multimorbidity and all-cause mortality in 2 million women and men registered in UK general practices

The prevalence of multimorbidity, defined as having two or more long-term conditions [1], has been increasing, a global trend partly driven by increasingly ageing population and improved survival from major causes of mortality [2‐4]. However, the impact of multimorbidity on mortality remains unclear. The combined impact of different conditions on mortality risk could simply reflect the sum of the total effects of each condition, or it could be less than the sum of individuals effects of each condition, particularly when coexisting conditions are concordant, that is, when co-occurring diseases are likely to share aetiology, predisposing factors or clinical management. Alternatively, the presence of a condition might exacerbate the impact of another such that the combined effect of the different conditions is substantially higher (or lower) than would normally be expected from the separate effects of each disease.

Myocardial infarction, diabetes mellitus and stroke are cardiometabolic diseases that are prevalent and among the leading causes of mortality globally [5, 6]. Others have reported increased mortality risk in those with more than one of these cardiometabolic conditions (Additional file 1: Table S1) [7‐20]. However, these studies have been based on small numbers of deaths, did not compare risk with those without cardiometabolic disease and did not explore the impact of other coexisting chronic conditions. Considering that these diseases are largely concordant, we hypothesised that, whilst the mortality risks of patients with two or all three of these conditions will be high, the combined risk will not be more than the excess deaths expected from each individual cardiometabolic condition, which could have implications on managing the conditions of these people with multimorbidity.

In addition, multimorbid individuals have been thought to be at increased risk of mortality particularly those with higher numbers of coexisting conditions [12, 21‐25]. Since people with myocardial infarction, diabetes and stroke tend to also have other long-term conditions [4, 19, 26, 27], the presence of these additional comorbidities could potentially influence mortality risk in people with cardiometabolic disease multimorbidity. We therefore examined the separate and combined associations of having myocardial infarction, diabetes and stroke with all-cause mortality and assessed the impact of having other comorbidities on these associations.

In this cohort of 2,007,731 (51% women), the mean (standard deviation [SD]) baseline age was 51.4 (SD = 17.5) years, and 7.3% (N = 145,910) have had a diagnosis of at least one of the cardiometabolic diseases of interest (Table 1). The proportions of those with a single condition were 3.6%, 1.7% and 1.3% for diabetes, myocardial infarction and stroke, respectively; for those with two conditions, the proportions were 0.4% for diabetes and myocardial infarction, 0.2% for diabetes and stroke and 0.1% for myocardial infarction and stroke; for those with all three conditions, the proportion was 0.05%. Compared to those without cardiometabolic disease at baseline, those with the disease were older, had higher deprivation level and were more likely to be ever smokers. Except for those with only stroke at baseline, the proportions of women were lower than those of men in all other cardiometabolic disease status at baseline. Table 1 also shows that patients with cardiometabolic disease were more likely to have a higher number of additional comorbidities. For example, the proportions of those with five or more comorbidities in patients with all cardiometabolic conditions were 25%; in contrast, among those without any cardiometabolic disease at baseline, the proportion was only 2%.

Figure 1 (and Table S4) shows the risk of death associated with baseline disease status. Adjusting for sex, smoking and deprivation, each single condition was associated with increased risk of mortality, with the highest risk seen in patients with stroke only (hazard ratio = 1.84 [95% CI 1.82 to 1.86]) when compared to those without any of the cardiometabolic condition at baseline. In absolute terms, there were 116 (95% CI 115 to 117) excess deaths per 10,000 per year in the former when compared to the latter group. When two conditions were present, there were doubling of risks relative to those without any of the cardiometabolic diseases. The risk was highest when all three diseases were present (hazard ratio = 3.22 [95% CI 3.15 to 3.30]), which was associated with excess deaths of 306 (95% CI 298 to 314) per 10,000 per year. After additionally adjusting for other comorbidities at baseline, the risks associated with myocardial infarction, stroke and diabetes, when present individually or in combination, were attenuated. In particular, when all three conditions were present, the hazard ratio was attenuated to 1.81 (95% CI 1.74 to 1.89) and the absolute excess risk to 114 (95% 105 to 123) per 10,000 per year when compared to those without any of the condition. The risk estimates remained similar when we additionally adjusted for comorbidities occurring after baseline (Table S4). The results remained consistent in women and men, in attained ages <75 and ≥75 years (Additional file 1: Tables S5 and S6). The impact of adjusting for comorbidities operationalised by the number of coexisting conditions as well as by broadly-related categories of chronic conditions are shown in Table 2.

We examined for any synergistic or antagonistic effects on mortality when two or more of the cardiometabolic conditions were present (Fig. 2). There was no evidence that the presence of two or more conditions resulted in more deaths than expected from mortality risk attributable to each condition. For example, if an interaction between diabetes and stroke results in a combined risk higher than expected from the excess deaths associated with each condition separately, the excess risk should be higher than 393 deaths per 10,000 per year; yet the excess deaths in this patient group was 348 (95% CI 345 to 351) deaths per 10,000 per year (Fig. 2—left panel). We observed similar patterns when we additionally adjusted for baseline comorbidities (Fig. 2—right panel). There is some evidence of an interaction on an additive scale in the presence of both diabetes and stroke (Additional file 1: Fig. S1—left panel), which was attenuated, but not eliminated, after taking into account other baseline comorbidities (Additional file 1: Fig. S1—right panel). Interestingly, there was departure from multiplicativity associated with the presence of both myocardial infarction and stroke (Fig. 2), whether or not other comorbidities were adjusted for. The mortality risk was lower than expected from the risks associated separately with either disease, suggesting an ‘antagonistic’ pattern of interaction. We did not see any significant relative excess risk when all three conditions were present; if anything, the excess risk was lower than expected from the sum of the risks of each condition after adjusting for the additional comorbidities (Additional file 1: Fig. S1—right panel).

In this cohort of over two million women and men, people with a history of myocardial infarction, diabetes or stroke had an increased risk of mortality. These increased risks were, in part, explained by the presence of additional comorbidities. Those with more than one of the cardiometabolic diseases had a higher risk than those with one or no cardiometabolic condition. Whilst the co-occurrence of two or more of these conditions resulted in substantially high excess deaths, the risks were not generally higher than would have been expected from the impact of the individual disease.

Comparison of risks with previous studies is difficult due to varying cardiometabolic conditions being considered (Additional file 1: Table S1) although our study broadly supports earlier findings in showing that mortality risk is high in patients with a history of cardiometabolic disease, particularly when these diseases co-occur. However, findings showing the combined impact of myocardial infarction, diabetes and stroke are limited as previous studies have been based on small numbers of deaths [12, 17], lacked data for comparison with people without any of the cardiometabolic conditions [9] or did not take into account other coexisting comorbidities [9, 12, 15, 17]. Unlike most of these studies, we specifically explored the impact of co-occurring cardiometabolic conditions on mortality risk and examined the importance of other coexisting long-term conditions in contributing to the risk of death.

A previous report showed a multiplicative effect, suggesting an amplification of mortality risk associated with cardiometabolic disease multimorbidity [15]. By pooling data from several prospective cohort studies, mortality associated with a history of heart disease, diabetes and stroke—separately for those with one, two or all of these conditions. The investigators suggest that participants with two or more conditions have substantially increased mortality risk than would have been expected from the mortality rates associated with each disease. Their seemingly discrepant findings from those in our study could be due to differences in the study population. In their investigation, the study population included cohort participants who were younger and more likely to have healthier background risk than our study population identified from general practices. Some of the cohorts in the collaborative study collected baseline data in the 1960s and 1970s, whereas our study population was based on registration with general practices in more recent decades. As mortality risk associated with cardiometabolic disease, when compared to those without the condition, has been decreasing in recent years [13, 14, 26], our estimated mortality risks are likely to reflect those of patients in relatively more contemporary settings whose clinical management differed from those in earlier generations. Our study also differed in that we were able to take into account other comorbidities that patients with myocardial infarction, diabetes and stroke also had, and we have demonstrated that in such patients these additional long-term conditions have a substantial impact on their mortality risk.

Our observation that the excess deaths were not amplified in those with multiple cardiometabolic conditions is consistent with our hypothesis that, when concordant conditions coexist, the overall risk is unlikely to be greater than the sum of the individual effects of each condition. Clinical factors, such as elevated blood pressure and dyslipidaemia, are common in these patients, and treatment of these risk factors typically forms part of their clinical care. Our findings could reflect the concordance in the pathophysiology, clinical features and management of vascular and metabolic disease and are consistent with research indicating that people with multimorbidity are more likely to receive evidence-based treatment than expected [45‐47]. These results highlight the importance of managing and treating these risk factors to modify mortality outcomes in patients with coexisting vascular and metabolic diseases.

Furthermore, our study also showed the impact of other additional comorbidities on the mortality of patients with cardiometabolic disease. These additional comorbidities could be indicative of the disease burden these patients have, either in terms of their number or type of comorbidity (e.g. cancer-related condition) [21, 24, 25]. It is also plausible that patients with these additional comorbidities have conditions that are ‘discordant’ (e.g. different aetiology), for which they may have received suboptimal care [48]. Indeed, the presence of multiple conditions could potentially affect the quality of care patients with cardiovascular disease or diabetes receive [49‐51]. Yet, there is also evidence that when these discordant conditions are identified, such as during clinical assessment to identify these conditions when this evaluation forms part of routine care for multimorbid patients or opportunistically during clinical encounters, these patients could actually receive better care [12, 45‐47]. Whilst we found some evidence that of an interaction between myocardial infarction and stroke when assessed on an additive scale, the excess risk was largely attenuated when other comorbidities were taken into account. Our findings therefore highlight the importance of other coexisting chronic conditions in influencing mortality risk of people with cardiometabolic disease multimorbidity.

There are a number of considerations when interpreting our results. We used routine practice data, which could be prone to recording errors and biases that include differential recording of diagnoses. We addressed these potential issues with appropriate considerations in designing the study, such as limiting analyses to records flagged to be of research quality standards and imposing restrictions on the minimum duration of registration with the general practice [35]. We neither validated nor adjudicated diagnoses but recorded diagnoses of cardiometabolic disease in the CPRD database have been shown to have high validity [33]. We were unable to distinguish type 1 from type 2 diabetes as the specific type is frequently not recorded in the database. The duration of the cardiometabolic diseases and the timing and sequence of the occurrence of the comorbidities could not be precisely determined as there could be delays between making the diagnoses of multiple chronic conditions and their recording. We also did not have information on other lifestyle factors, such as physical activity level, alcohol consumption and diet, which are not routinely recorded in healthcare databases. Although our study population is predominantly of white ethnicity, patients included in the CPRD database are representative of the UK general population in terms of age, sex and ethnicity [29]. Our findings may not be applicable in acute settings, as our study population would include those who would have had survived at least the first disease event for a duration long enough to be considered in our analysis. Finally, the extent to which our findings are applicable to people with other chronic conditions and discordant comorbidities, or other healthcare settings, requires further investigation. Nevertheless, an important strength of our study is the scale, volume and size of the data, including sufficient duration of follow-up to accrue sufficient numbers of events across the different multimorbidity subgroups.

In this large population of women and men, myocardial infarction, stroke and diabetes, separately and in combination, were associated with excess mortality which was partly due to associated additional comorbidities. We found no evidence that the co-occurrence of all three cardiometabolic conditions contributed to a higher excess mortality than expected from each of them separately, indicating that mortality risk is not necessarily amplified in cardiometabolic disease multimorbidity but rather modifiable. The excess risk was largely related to the presence of other comorbidities, underscoring the need for a wholistic approach when assessing and evaluating risks in people with multimorbidity.