Effect of multiple comorbidities on mortality in chronic obstructive pulmonary disease among Korean population: a nationwide cohort study

Chronic obstructive pulmonary disease (COPD) is characterised by chronic irreversible airflow limitation and is usually caused by cigarette smoking. COPD is a major cause of disability and death around the world [1]. Medical comorbidities are prevalent among COPD patients, and previous research has shown that comorbidities affect not only symptom burden and functional performance in patients with COPD but also the risk of exacerbation, hospitalisation and mortality [2, 3]. The majority of COPD patients have more than one comorbidity; however, in most of the studies that have addressed the effects of a comorbidity on COPD clinical outcomes, comorbidities have been studied individually [1, 3].

In the past, a few studies have tried to examine the effect of multiple comorbidities on COPD outcomes more systematically by using or developing the measurement instruments. For example, the Charlson comorbidity index (CCI) is one of the most widely used measurement tools by healthcare professionals to assess the burden of multiple comorbid diseases, and the CCI has been well-validated to predict the mortality in COPD patients [4]. However, CCI is not a disease-specific instrument, and few COPD-specific indices to assess the cumulative burdens of multiple comorbidities on COPD outcomes have been developed. For instance, the COPD-specific comorbidity test (COTE) index and the COMorbidities in COPD (COMCOLD) index were recently developed and validated to predict the mortality and general quality of life in COPD patients, respectively [5, 6]. In addition, Putcha et al. reported a simple scoring system, wherein the number of comorbidities could identify the susceptible COPD patient at the risk of developing exacerbation [7]. Intriguingly, a difference in comorbidity profile by race and ethnicity in COPD patients has been reported [8]. Therefore, it will be of interest to see whether differences in comorbidity profiles according to race and ethnicity lead to differences in disease outcomes.

With this background, this study was conducted to examine the following: first, the comorbidities associated with mortality; second, the number of comorbidities could be an independent risk factor to predict mortality and third, other factors associated with mortality among Korean COPD population using a nationwide population-based cohort.

In this study, we investigated the comorbidities associated with mortality, the effect of multimorbidity on mortality and other factors associated with mortality among physician-diagnosed Korean COPD patients by using two different cohorts derived from a nationwide population-based cohort. Our study has found the following.

Regarding the characteristics of our study participants, most of them had comorbidities, and the average numbers of comorbidities were 2.56 and 2.57 in the entire cohort and health-screening cohort, respectively. Our patients had a relatively fewer comorbidities than compared by other studies [5, 12, 16‐18]. However, a direct comparison between studies is challenging due to the expected heterogeneity of comorbidity definitions used in different studies. The most prevalent comorbidities were hypertension, asthma and dyslipidaemia and GERD in both cohorts (Table 1). These findings are similar to what has been reported previously. Notably, most studies have found cardiovascular comorbidities as the most prevalent comorbidities in COPD patients [17, 19, 20].

Total mortality rates were 38.6 per 1000 person-years (95% CI 37.32–40.01) and 27.4 per 1000 person-years (95% CI 25.68–29.22) in the entire cohort and health-screening cohort, respectively. Although the number of comorbid diseases was similar, the difference in all-cause mortality was quite significant between the two cohorts. A potential selection bias for the national health screening cohort can account for this difference. The participants in this cohort seemed to have been more interested in their health in that they wanted more detailed examinations, and this seems to have decreased the mortality rate. The most common causes of death in both cohorts were the disease progression of COPD, lung cancer and acute myocardial infarction. These findings are in agreement with previous reports (Table 2) [3, 5, 13, 14, 21]. Among the respiratory causes of death, deaths from pneumonia could simply represent deaths from COPD exacerbation associated with disease progression. If so, the most common cause of death in our study participants would be the disease progression of COPD. Some studies have reported that the leading causes of death in mild or moderate COPD are lung cancer and cardiovascular diseases, but respiratory failure from disease progression becomes the predominant cause of death in more advanced COPD [3, 14]. From that standpoint, our study participants may have had advanced COPD. However, due to the lack of detailed clinical data with regard to disease severity, it will not be easy to interpret these findings.

We also examined the factors associated with all-cause and respiratory mortalities in both cohorts. The most robust findings are that male sex, old age and low BMI were universal risk factors associated with all-cause and respiratory mortalities in both cohorts. Further, high BMI was also associated with lower all-cause and respiratory mortalities in both cohorts. Low BMI is reportedly a risk factor of all-cause mortality in people with COPD [22, 23]. Although the role of sex as a determinant of clinical outcome of COPD has been controversial [24, 25], male sex turned out to be an obvious risk factor of death in our study (Figs. 2 and 3).

Among comorbidities, arrhythmia, cerebrovascular disease, heart failure, DM and malignancies, including lung cancer, were associated with all-cause mortality in both cohorts. Intriguingly, unlike all-cause mortality, comorbidities associated with respiratory mortality were quite different between the two cohorts. In addition, in both cohorts, fewer comorbid diseases were associated with respiratory mortality than with all-cause mortalities (Figs. 2 and 3).

Next, we investigated whether multimorbidity, assessed by the number of comorbidities, could be a valuable tool to identify the individuals susceptible to death from COPD (Table 3). Consequently, we observed a significant association between the number of comorbid diseases and all-cause mortality of COPD patients in both cohorts. However, the association between the number of comorbidities and respiratory mortality did not seem to be strong in our study participants; this is based on the following. First, there were no significant differences in survival trajectories according to the number of comorbid diseases concerning respiratory mortality in both cohorts (Fig. 4). Second, multivariable Cox regression analysis revealed a significant association between the number of comorbid diseases and respiratory mortality only in the health-screening cohort and not in the entire cohort (Table 3). Therefore, there seems to be a difference with regard to the effect of multimorbidity on mortality between COPD patients who died of disease progression and COPD patients who died of the non-respiratory diseases in our study. We currently do not know how to accurately interpret these findings; however, to speculate, the patients who died of respiratory causes might have had more advanced COPD; consequently, the effect of comorbidities on mortality may mitigate with COPD progression.

Notably, the use of certain medications, particularly LAMA, at the time of diagnosis was associated with increased mortality. Moreover, methylxanthine decreased mortality. We speculate that this is related to the reimbursement criteria of Korean health insurance. LAMA is permitted only in patients with at least moderate airflow limitation. However, there are no strict reimbursement criteria for methylxanthine. Thus, it is likely that LAMA was prescribed in more severe patients and methylxanthine in less severe patients. However, given the lack of pulmonary function data and the detailed history of patient condition, it is difficult to interpret this finding more concretely.

Overall, our study findings can be summarised as follows. First, some comorbid conditions can have a direct impact on COPD mortality; second, comorbid diseases that do not have a direct effect on mortality might still contribute to mortality, probably by intensifying the total burden of comorbidities; third, low BMI, age and male sex are durable risk factors for death and fourth, the number of comorbidities might be an independent risk factor of COPD mortality. And the effect of multimorbidity is more evident on all-cause mortality than on respiratory mortality among Korean COPD population.

We assessed the effect of multimorbidity on COPD mortality by simply counting the number of comorbidities. By doing so, we may have oversimplified the impact of multimorbidity in this study. Previously, a few researchers had developed the measurement instruments to assess the effect of multiple comorbidities on various COPD outcomes [4‐7]. Recently, Divo et al. developed the COPD-specific comorbidity test (COTE) index, a disease-specific comorbidity index to predict mortality [5, 6]. In their study, 12 comorbidities associated with increased mortality were first identified out of 79 comorbidities, and the strength of the association of each comorbidity with COPD mortality was assessed by performing multivariate analyses using Cox proportional hazards regression. Scale value points in the range of 1–6 points were assigned to each comorbidity in proportion to its hazard ratio. By summating the points, they were able to assess multimorbidity effect on COPD death. Needless to say, the COTE index is an invaluable tool; however, its utility needs to be further verified. For instance, the value of CCI, which is not a COPD-specific tool, to predict all‐cause mortality in COPD was higher than that of the COTE index in one study [4]. Recently, Putcha et al. assessed the burden of multimorbidity to predict the exacerbation risk of COPD by employing three different ways: (1) simple count, (2) weighted score and (3) weighted score based on statistical selection procedures. They found that the comorbidity count performs best in terms of quantifying the comorbidity burden [7]. The Putcha’s study, as well as ours, suggests that simple counting might be a reliable measure to assess the burden of multimorbidity on clinical outcomes.

There are a few limitations to our study. Differentiation between asthma and COPD in a real-life practice can sometimes be challenging. It is possible that some patients with asthma may have been just COPD patients, which could explain the high prevalence of asthma in our study population. However, we cannot be certain about this due to the lack of detailed clinical information and pulmonary function data.

It has been shown that some mental illnesses, such as depression or anxiety, are associated with an adverse clinical outcome of COPD [26]. However, these data were not available for this study, which can be another weakness of the study. Our analysis focused on the comorbidities that the patient had during the one-year pre-study period. The number and profile of comorbidities may have changed during the follow-up period. Even if it were possible, we believe that the implications of our study findings would not have changed.

Only physician-diagnosed COPD patients were included in our study. In addition, only physician-diagnosed comorbidities were examined, which could be a considerable strength of this study over many other studies that used self-reported comorbidities [5, 21, 27]. However, it is also possible that some comorbidities that are not routinely examined or monitored at our clinical practice, such as osteoporosis, could have been underestimated in this study.

In general, frailty and disability need to be evaluated separately to address the impact of comorbidities on clinical outcomes. As for the frailty factor, we lacked sufficient clinical information about several parameters (e.g., gait speed, grip strength, physical activity, and cognitive function) to assess frailty that are usually difficult to measure in healthcare claims database like ours [28]. Further, given that there were only 1.8% of participants with disabilities in the national sample cohort, the disability factor was not evaluated as it was expected to have minimal impact on the study results.

This study showed that simple comorbidity counting can be another way to assess comorbidity burden, but the study findings should be carefully interpreted as simple counting can oversimplify and misrepresent the impact of multimorbidity. Although our study findings are intriguing, further research is required to implement this knowledge into standard, patient-centred COPD care. In addition, a probable different effect of multimorbidity on mortality between patients who died of non-respiratory causes and who died of respiratory cause may need to be further explored in future studies.

Our large population-based study showed that some comorbid conditions can have a direct impact on COPD mortality, and comorbid diseases that do not have a direct effect on mortality might still contribute to mortality, probably by intensifying the total burden of comorbidities, given that the number of comorbidities might be an independent risk factor of COPD mortality. In addition, multimorbidity contributes to all-cause mortality in COPD, but the effect of multimorbidity is less evident on respiratory mortality.