Effect of comorbidity assessed by the Charlson Comorbidity Index on the length of stay, costs, and mortality among colorectal cancer patients undergoing colorectal surgery

Abstract Objective Charlson Comorbidity Index (CCI) is a good predictor for hospitalization cost and mortality among patients with chronic disease. However, the impact of CCI on patients after colorectal cancer surgery is unclear. This study aims to investigate the influence of comorbidity assessed by CCI on length of stay, hospitalization costs, and in-hospital mortality in patients with colorectal cancer (CRC) who underwent surgical resection. Methods This historical cohort study collected 10,271 adult inpatients for CRC undergoing resection surgery in 33 tertiary hospitals between January 2018 and December 2019. All patients were categorized by the CCI score into four classes: 0, 1,2, and ≥3. Linear regression was used for outcome indicators as continuous variables and logical regression for categorical variables. EmpowerStats software and R were used for data analysis. Results Of all 10,271 CRC patients, 51.72% had at least one comorbidity. Prevalence of metastatic solid tumor (19.68%, except colorectal cancer) and diabetes without complication (15.01%) were the major comorbidities. The highest average cost of hospitalization (86,761.88 CNY), length of stay (18.13 days), and in-hospital mortality (0.89%) were observed in patients with CCI score ≥3 compared to lower CCI scores (p < .001). Multivariate regression analysis showed that the CCI score was associated with hospitalization costs (β, 7340.46 [95% confidence interval (CI) (5710.06–8970.86)], p < .001), length of stay (β, 1.91[95%CI (1.52–2.30)], p < .001), and in-hospital mortality(odds ratio (OR),16.83[95%CI (2.23–126.88)], p = .0062) after adjusted basic clinical characteristics, especially when CCI score ≥3. Notably, the most specific complication associated with hospitalization costs and length of stay was metastatic solid tumor, while the most notable mortality-specific comorbidity was moderate or severe renal disease. Conclusion The research work has discovered a strong link between CCI and clinical plus economic outcomes in patients with CRC who underwent surgical resection.


Introduction
Colorectal cancer (CRC) is the third most common malignancy and the 16th leading cause of death among all diseases and injuries worldwide 1,2 . There were 1.4 million new incident cases of CRC and almost 700,000 deaths in 2012 3 . However, approximately 1.8 million new cases were found in 2018, with 900,000 deaths annually 4 . No exception, the burden of CRC in China had increased over the past several decades 5 . From 2011 to 2015, the hospitalization expenses of cancer inpatients in China increased by 84.1%, among which the total hospitalization costs reached 177.1 billion CNY in 2015, accounting for 4.3% of the total health expenditure 6 . Furthermore, the hospitalization costs for CRC diagnosis and treatment were substantial and increased rapidly in China 7 .
Comorbidity is common in patients with CRC, ranging from 46% to 62% 8 . A cohort study found that the most frequent comorbidities were congestive heart failure and diabetes based on two Spanish population-based cancer registries and electronic health records with 1061 CRC patients 9 . Similarly, another study indicated that 23.7% CRC patients had diabetes, ranking first among all comorbidities 10 . Thomsen et al. found that the most prevalent diseases were metastatic solid tumor (9.6%), followed by chronic pulmonary disease (7.4%), cerebrovascular disease (6.5%), and diabetes (6.2%) 11 . Despite the common coexistence of comorbidity and CRC, the guidelines and delivery of cancer care generally focus on the management of a single disease 12 . Effectively managing CRC treatment in patients with comorbidities is an essential clinical issue, but there is no reliable treatment standard regarding this 13 . More specialized and individualized care for CRC patients is needed in clinical practice 14 .
Previous research has shown that comorbidity has adverse effects on the prognosis of CRC patients 15,16 . The Charlson Comorbidity Index (CCI), which included 19 medical conditions with corresponding weights, was the most common method to quantify the overall burden of comorbidity conditions in the electronic healthcare database 17 . So far, the CCI has been considered to be a predictor of hospitalization expenses 18,19 . Also, the effect of CCI on length of stay [20][21][22] , and mortality 20,23 has already been reported in populations with different diseases. However, research on the subject has been mostly restricted to limited comparisons of the impact of comorbidity on outcomes in patients with CRC after surgical resection.
While evidence of the influence of comorbidities on CRC outcomes is consistent, 24 . little is known about their association with clinical plus economic outcomes in CRC patients undergoing Colorectal Surgery. Therefore, it is necessary to assess the impact of comorbidity on outcomes in postoperative patients with CRC. This paper describes a system for analyzing the influence of comorbidity assessed by the CCI on length of stay, hospitalization costs, and in-hospital mortality in patients with CRC who underwent surgical resection.

Data overview
The data of this historical cohort study extracted from 33 tertiary hospitals between January 2018 and December 2019 was based on the hospitalization summary reports (HSR) database of the Beijing Municipal Health Commission Information Center. HSR on each hospitalization is required as a routine report which contains information on demographics, diagnosis and treatment, dates of admission and discharge, and hospitalization cost per patient. Among the 33 tertiary hospitals in Beijing, seven are located in Chaoyang District, six are located in Xicheng District, five are located in Dongcheng District, five are located in Haidian District, three are located in Changping District, two are located in Fengtai District, one is located in Pinggu District, one is located in Shunyi District, one is located in Tongzhou District, one is located in Daxing District, and one is located in Fangshan District.

Study population
First, the diagnosis of CRC was defined according to the International Classification of Diseases, tenth Revision, Clinical Modification (ICD-10) code as "C18-C20." Then, patients with CRC who underwent surgical resection were identified using ICD-9-CM codes "45.7x, 45.8, 45.9x, 48.4x, 48.5, and 48.6x". Patients under 18 years were excluded. Finally, 10,271 patients were included for analysis.

Definition of CCI and covariates
Comorbidity was assessed by the CCI, which accounted for 19 different comorbidities by creating a sum score, while each complication was given a certain weight 25 . The corresponding ICD-10 codes for the 19 comorbidities except CRC were listed in Table S1. All patients were categorized into groups according to their CCI score: 0, 1,2, and !3.
Control variables included age, gender, type of insurance, hospital category, surgery type, and histological type of colorectal cancer. The type of insurance included urban employee basic medical insurance for employed workers, urban resident basic medical insurance for urban residents without a stable job, a new cooperative medical system for rural residents, and others. Different insurance varies in terms of payment methods, reimbursement proportion, and coverage 26 . The hospital category comprised Chinese medicine hospitals, western medical hospitals, integrated traditional Chinese and Western medicine hospitals, and tumor hospitals in China, while they had different characteristics concerning treatment features and hospitalization costs. The histological type was categorized into adenocarcinoma, mucinous carcinoma, signet-ring cell carcinoma, and others.

Statistical analysis
Descriptive statistics were performed to summarize the basic characteristics of patients. Continuous variables were presented as mean ± standard deviation (SD) and absolute numbers with percentages for categorical variables. Kruskal Wallis Rank Test for continuous variables and the chi-square test or Fisher exact (with expects <10) for categorical variables. The piece-wise linear regression assessed the independent association between length of stay, hospitalization costs, and CCI, which presented as b with 95% confidence intervals (CIs), while logical regression was used to reveal the relationship between in-hospital mortality and CCI with odds ratio (OR) and 95% CIs. Covariates were included as potential confounders in the final models if they changed the estimates of CCI on outcomes by more than 10% or were significantly associated with outcomes 27,28 . All results of regression analysis were presented using a forest plot, which adjusted age, gender, type of insurance, hospital category, surgery type, and histological type of colorectal cancer. Survival curves for CCI groups and in-hospital mortality were conducted according to Kaplan-Meier and tested by Log-Rank test. To further analyze the specific types of comorbidities, the seven most prevalent types of comorbidities were identified as follows: mild liver disease, moderate or severe renal disease, peripheral vascular disease, chronic pulmonary disease, diabetes without complication, cerebrovascular diseases, and metastatic solid tumor (without metastatic colorectal cancer). In addition, the study also assessed the differences in length of stay, hospital costs, and mortality within CCI groups among hospital categories, surgical types, and histological types.
Moreover, to eliminate the bias of metastatic solid cancer on the results, a separate analysis without metastatic solid tumor was conducted. A two-sided significance level of 0.05 was used to evaluate statistical significance. Data management and analysis were performed using EmpowerStats software and R (version 3.3.2) 29,30 .

Baseline characteristics of included patients
A total of 10,271 adult patients with CRC who underwent surgical resection were confirmed. Among all the participants, males accounted for 62.07%. Urban employee basic medical insurance was the primary social health insurance scheme (70.83%), and patients in tumor hospitals accounted for the largest proportion (54.52%). As for surgery and histological types, laparoscopic surgery was the major method of operation (62.39%), while adenocarcinoma was the most common pathological type (88.26%). CRC stage III patients accounted for the largest proportion (37.30%), and the rectum was the main resection site (72.86%). Concerning comorbidities, 48.28% of patients with no comorbidities, 31.20% with one,13.64% with two, and 6.87% with three or more comorbidities. There was a significant difference between the CCI groups in terms of the length of stay, hospitalization costs, and death (p < .001). Details of the baseline characteristics are shown in Table 1 and Table 2. After dividing participants by CCI scores, there were 4959 participants in group 1(CCI ¼ 0), 1639 participants in group 2 (CCI ¼ 1), 752 participants in group 3 (CCI ¼ 2), and 2921 participants in group 4 (CCI ! 3). Baseline characteristics according to CCI groups are shown in Table S2.

Length of stay
There were significant differences in length of stay among CCI score groups, with an average length of stay of 14.63 days in group 1(CCI ¼ 0), 16.05 days in group 2 (CCI ¼ 1), 17.42 days in group 3 (CCI ¼ 2), and 18.13 days in group 4 (CCI !3) ( Table 2). In addition, the average length of stay increased as quartiles increased in different hospital categories, surgical types, and histological types with a maximum length of stay at CCI !3 groups (Figure 1). Among them, integrated traditional Chinese and Western medicine hospitals observed the longest length of stay (31.96 days). Compared to CRC patients with open surgery, a shorter length of stay was found in laparoscopic surgery within different CCI groups (p < .05). Besides, patients with mucinous carcinoma had a longer length of stay than adenocarcinoma and signet-ring cell carcinoma among CCI ¼ 0, CCI ¼1, and CCI !3 groups (p < .05). Figure 2 depicts the b and 95% CI in the length of stay associated with CCI after adjusting for potential confounders. Compared with CCI ¼ 0 group, there were significant differences in CCI ¼ 2 group (b, 1. 13

Hospitalization costs
Significant differences were found in hospitalization costs among different groups of CCI score, with the mean hospitalization costs 80,815.71 CNY in group 1 (CCI ¼ 0), 82,243.97 CNY in group 2 (CCI ¼ 1), 82,403.38 CNY in group 3 (CCI ¼ 2), and 86,761.88 CNY in group 4 (CCI ! 3) ( Table 2). Furthermore, the average hospitalization costs increased as quartiles increased in different hospital categories with a maximum length of stay at CCI ! 3 groups (Figure 1). In terms of surgery type, the highest hospitalization costs (96,209.06 CNY) were observed in patients with open surgery among CCI ! 3 groups. Unlike above, patients with mucinous carcinoma had higher hospitalization costs than adenocarcinoma and signet-ring cell carcinoma among different CCI groups.
Piece-wise linear regression revealed the association between hospitalization costs and CCI after adjusting for control variables (Figure 2). Compared with the reference group of CCI ¼ 0, significant differences were observed in the CCI ¼ 1 group (

In-hospital mortality
The death between CCI score groups was significantly different according to Table 2, with mortality of 0.02% in group 1 (CCI ¼ 0), 0.12% in group 2 (CCI ¼ 1), 0.80% in group 3 (CCI ¼ 2), and 0.89% in group 4 (CCI ! 3). Regarding hospital categories, the highest mortality (12.00%) was observed in the integrated traditional Chinese and Western medicine hospitals in CCI ! 3 groups. Compared to patients with open surgery, lower mortality was found in laparoscopic surgery within different CCI groups. Furthermore, patients with adenocarcinoma and signet-ring cell carcinoma had lower mortality than mucinous carcinoma among CCI groups.
Logistic regression showed that difference did not reach statistical significance in CCI ¼1 group ( were associated with a significant likelihood of increased mortality with a remarkable trend across the quartiles (p for trend <.00001). In particular, the most notable mortality-specific  (Figure 3).

Sensitivity analysis
When excluding patients with metastatic solid tumor (N ¼ 8250), the data reported here appear to support the assumption that there is a significant association between CCI and outcomes, which proved the stability of our results (Table S3).

Discussion
To the best of our knowledge, this is the first evidence to investigate the influence of comorbidity assessed by the CCI on length of stay, hospitalization costs, and in-hospital mortality in patients with CRC who underwent surgical resection. Abbreviations. CCI, Charlson Comorbidity Index; AIDS, acquired immune deficiency syndrome. In summary, 51.72% of patients had at least one comorbidity, among which metastatic solid tumor and diabetes without complication were the major comorbidities. In addition, a higher CCI score (CCI ¼ 1,2 and CCI ! 3) was associated with an increased average length of stay, higher hospitalization costs, and a greater likelihood of death than CCI ¼ 0.
More obviously, the most specific complication associated with hospitalization costs and length of stay was metastatic solid tumor, while the most notable mortality-specific comorbidity was moderate or severe renal disease. The evidence from this study suggests that the most frequent comorbid condition in CRC was metastatic solid tumor (19.68%) and diabetes without complication (15.01%). Unfortunately, the causality of CRC and metastatic solid tumor cannot be inferred due to a retrospective post-hoc analysis study design. Moreover, the condition of CRC patients with metastatic cancer was more complicated. So far, the underlying mechanism for this situation is incompletely revealed, which may be related to susceptibility, heredity, and lifestyle, to our knowledge. Similar to our findings, other large cohort studies have identified other malignancies and diabetes as common comorbidities in colorectal cancer 14 . Indeed, these results are predictable, as increased comorbidities in colorectal cancer patients are due to multiple factors. Firstly, as the aging of the global population continues, the demographic structure is gradually changing, and the proportion of elderly people in colorectal cancer patients is increasing. Besides, improving the quality of care for people with chronic diseases may also help them live to an older age when they are at risk of developing colorectal cancer. Finally, unfavorable lifestyles, low-quality diets, physical inactivity, smoking, alcohol consumption, and obesity all contribute to increased nutritional disease. There is evidence suggesting that type 2 diabetes diagnosis could be associated with increased cancer risk and worse prognosis 31 . Therefore, it was not hard to understand the extension of hospitalization costs and length of stay.
Additionally, CRC complicated with abnormal kidney function is frequently encountered in clinical practice. A recent meta-analysis found that chronic kidney disease was associated with a significantly higher risk of developing early-onset CRC 32 . Not alone, another study indicated that the risk of Figure 2. b/odds ratio and 95% CI in length of stay, hospital cost, and mortality for colorectal cancer patients associated with comorbid conditions, adjusted age, gender, type of insurance, hospital category, surgery type, and histological type of colorectal cancer. developing CRC in patients with chronic kidney disease is twice that of the general population 33 . A possible explanation for this might be that the reduction of kidney function and uremic environment are associated with oxidative stress, chronic inflammation, and dendritic cell dysfunction, which may reduce cancer immune surveillance and cancer development 34,35 . This study found an interesting finding that the most notable mortality-specific comorbidities were moderate or severe renal disease, which accords with earlier observations. A recent study also supported this finding, demonstrating that patients with chronic kidney disease had an increased risk of 30-day mortality with nonmetastatic colorectal adenocarcinoma who underwent surgical resection 36 . Moreover, a retrospective cohort study revealed that chronic kidney disease stage 5 showed a poorer overall survival in patients with surgical resection for primary CRC 37 . In agreement, a low estimated glomerular filtration rate (eGFR) was an independent risk factor for stage III CRC according to an observational study 38 . Little evidence is available on the impact of CCI on length of stay, hospitalization costs, and in-hospital mortality in patients with CRC who underwent surgical resection. Nevertheless, there were still some studies on the relationship between CCI and outcomes in other populations. As noted above, reports have begun to emerge that CCI was a strong negative prognostic factor for the survival of patients with colorectal cancer, 24,39 . which was consistent with the result of us. In terms of length of stay, a nationwide cohort study revealed a significant difference in length of stay among different CCI scores in patients with elective surgery for CRC 21 . Likewise, a high CCI score was considered to be related to higher hospitalization costs, length of stay, and mortality in patients with acute stroke 20 . and immobile patients after ischemic stroke 22 .
There are four types of hospitals in China: Chinese medicine hospitals, western medical hospitals, integrated traditional Chinese and Western medicine hospitals, and tumor hospitals, with differences in treatment features and hospitalization costs, to some degree. Differences among different CCI groups showed that the mean hospital cost, the average length of stay, and mortality increased as quartiles increased in hospital categories, which proved the robustness of the results. Most importantly, mean hospital costs, the average length of stay, and mortality varied from hospital category. Our findings provide supporting evidence to adjust the hospital category as a potential confounding factor in similar studies.
Surgery is the primary treatment for CRC, usually including open surgery and laparoscopic surgery. Notably, compared with open surgery, it is generally accepted that laparoscopic surgery has developed rapidly worldwide with the advantages of smaller incisions, less pain, fewer complications, faster recovery, and a shorter length of stay 40,41 . Similarly, a shorter length of stay was observed in laparoscopic surgery within different CCI groups (p < .05) compared to CRC patients with open surgery. Furthermore, our study established that laparoscopic surgery has lower hospitalization costs and mortality among different CCI groups.
Several limitations in this study need to discuss briefly. First, although this was a large multicenter study based on 33 tertiary hospitals in Beijing. Anyhow, the level of evidence is not comparable to that of a prospective study. Second, despite adjustment for several potential risk factors, it was not possible to consider variables that were not routinely recorded in the claims database, including specific causes of death, laboratory data, and clinical severity of CRC. Third, due to the insufficient number of cases of some comorbidities involved in CCI calculation, there was no in-depth analysis of the relationship between these comorbidities and outcomes. Last but not least, the work presented in this paper focuses on patients with CRC undergoing colorectal surgery, these findings cannot be extrapolated to all patients due to only focus on postoperative patients with CRC.

Conclusion
In the current study, we tested the hypothesis that comorbidity assessed by the CCI showed a positive graded relationship with hospitalization costs, length of stay, and in-hospital mortality in postoperative patients with CRC. More importantly, a higher CCI score was associated with an increased average length of stay, higher hospitalization costs, and a greater likelihood of death than a lower CCI. With the increased incidence of CRC and population aging, CRC patients should pay more attention to the management of comorbidity.

Declaration of funding
This paper was not funded.

Declaration of financial/other relationships
The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.
Peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Author contributions
Xuexue Zhang, Xujie Wang, and Miaoran Wang designed research. Qiuyan Li and Huijun Guo performed study and analyzed data. Xuexue Zhang, Xujie Wang, and Miaoran Wang extracted the data and wrote the main manuscript text. Yufei Yang, Jian Liu, and Jiyu Gu edited and revised manuscript. All authors reviewed and approved the manuscript.