Figure 1 presents a framework for estimating the comorbidity burden of inpatient cancer patient, which consists of four steps. The first step is the extraction of HIS system data like demographic data, diagnostic data, medication data and cost data. The inclusion and the exclusion process were stringently restricted. The diagnostic data were coded by trained coders with the 10th revision of the International Classification of Diseases (ICD-10). A total of 4,666 patients were finally included in the analysis. The second step is the identification of basic comorbidity characteristics. Comorbidities in this study were assessed using the NCI Comorbidity Index. Rates, numbers, types and severity of comorbidity for inpatient cancer patients together form the characterization of comorbidities. The third step is the estimation of the comorbidity burden. Inpatient cancer patients were classified based on comorbidities according to the agglomerative hierarchical cluster analysis. All prevalent conditions in this cohort were included in the cluster analysis. The final step is the examination of the associations between comorbidity patterns and outcome measures which include treatment options and medical cost. Treatment options were divided into conventional treatment and targeted therapy. More detailed information of the framework development was shown in the following part.

Fig. 1

Flow chart of the framework for Estimating Comorbidity Burden of Inpatient Cancer Patients

Study site

Shandong province is an important coastal province with over 100 million people in East China. The trend of incident cases and death of cancer in Shandong Province is consistent with that of the. whole country, among which lung cancer, gastric cancer, colorectal cancer, esophageal cancer and liver cancer are the most common types of cancer. The case hospital founded in 1916 was one of the largest municipal hospitals in Shandong province and represented the regional highest level of oncology care.

Step 1: extraction of HIS system data

The entire participant inclusion flowchart is illustrated in Fig. 2. By extracting the information from the hospital information system (HIS) of the case hospital, the data of all potential research objects were obtained. The inclusion criteria included: lung, colon, rectal, breast and gastric cancer cases confirmed by pathological examination; and the time of diagnosis was from January 2017 to October 2019. Cancer patients typically undergo follow-up examinations every 3–6 months. To maintain the integrity of data regarding patients’ treatment experiences, we excluded those who had only one hospital admission within a 3 months period. Furthermore, to account for the confounding effects of diverse pathological characteristics, we also excluded several specific patient groups due to their unique treatment requirements and relatively low incidence rates. These groups include: non-invasive breast cancer and special types of breast cancer including mucinous adenocarcinoma, medullary carcinoma, adenoid cystic carcinoma, and Paget's disease; male breast cancer; cases of colorectal cancer with histological types such as lymphoma, sarcoma, squamous cell carcinoma other than adenocarcinoma; and the small cell lung cancer [19]. According to the cancer type, patients in the departments of oncology, thoracic surgery, breast surgery, general surgery, and colorectal oncology in the case hospital were selected. The data contained more than 300,000 records, including demographic data (such as age and sex), diagnostic data, medication data and cost data. The diagnostic data consisted of one primary diagnosis and up to 18 secondary diagnoses, which were coded by trained coders with the ICD-10. A total of 4,666 patients were finally included in the analysis.

Fig. 2

Flow chart of the included participants from the case hospital

The characteristics of 4,666 cancer patients are presented in Table 1. Among all cancer patients, half of them were male (50.32%) and the average age was 62.42 (SD = 11.62) years. The majority of patients were insured (94.68%) and got married (87.74%). More than one third of patients were diagnosed with lung cancer (37.03%), followed by gastric cancer (21.20%), breast cancer (17.08%), rectal cancer (12.47%) and colon cancer (12.22%). Supplementary Table S1 presents the detailed information by cancer type. Breast cancer patients with the youngest average age (54.40) of all participants, while rectal cancer patients with the oldest (65.34). More gastric cancer (61.78%) and rectal Cancer patients (52.41%) were in the III or IV stage. Except for breast cancer, all cancer types were male dominated.

Table 1 Baseline characteristics of Chinese inpatient cancer patients included in this studyStep 2: identification of basic comorbidity characteristics

In this study, comorbidities were assessed using the NCI Comorbidity Index, which identifies multiple comorbidities. Based on the Charlson Comorbidity Index (CCI) first developed in 1987 by Mary Charlson and colleagues [20], the cancer-specific NCI Comorbidity Index developed by Carrie Klabunde and colleagues [21] excluded solid tumors, leukemias, and lymphomas as comorbid conditions, given that the NCI Comorbidity Index was developed from a cohort of cancer patients. The NCI Comorbidity Index was created to address some limitations of the Charlson Comorbidity Index, especially when applied to cancer patients. The remaining 16 Charlson index conditions were included in the NCI Comorbidity Index, with further consolidation to 14 conditions (Table 2): moderate/severe liver disease, cerebrovascular disease (CVD), peripheral vascular disease (PVD), renal disease, paralysis (hemiplegia or paraplegia), myocardial infarction, peptic ulcer, dementia, AIDS, mild liver disease, congestive heart failure (CHF), COPD, diabetes with complications, and diabetes. Adjusted specifically for cancer-specific NCI Comorbidity Index, the CCI accounts for multiple comorbidities according to the presence of 14 comorbid conditions. Hypertension was also included as it had the highest prevalence rate in the study sample, aside from those included in the NCI Comorbidity Index. For the severity of Charlson’s comorbidity, it was classified into mild, moderate and severe categories based on the CCI weight. Each condition was assigned a weight from 1 to 6, according to the estimated 1 year mortality hazard ratio from a cox proportional-hazards model. These weights were summed to produce the Charlson comorbidity score [22].

Table 2 Frequency of comorbidities in included Chinese inpatient cancer patients

Sociodemographic and cancer characteristics were compared according to the number of comorbidity and the severity of comorbidity. Continuous variables were summarized as mean (SD) and were examined using the Student t test. Categorical variables were presented as the proportion (%) and compared using the Pearson chi-square test. Of the 4,666 participants, there were more patients (76.17%) with comorbidities than those without. Compared with those without comorbidities (Table 1), patients with comorbidities were older (64.53 vs 55.67 years, P < 0.001), more male (P < 0.001), more urban lived (P < 0.001), more lung cancer diagnosed (P < 0.001) and more III-cancer-stage (P < 0.001). In terms of the number of comorbidities (Supplementary Figure S1), among 3554 cancer patients with comorbidities, 97.36% of patients had less than three types of comorbidities, of which 37.03% of patients had one comorbidity, 29.04% had two, and 17.14% had three. Compared with those with less than three types of comorbidities (Supplementary Table S2), patients with more than three types of comorbidities were older (P < 0.001) and in later cancer stage (P < 0.001). The most common comorbidities were hypertension (32.04%), mild liver disease (22.42%), renal diseases (17.60%), diabetes (15.90%) and myocardial infarct (15.65%). For the severity of comorbidity, more than half of patients were with the severe comorbidities (69.11%). Most patients with the severe comorbidities were in the IV cancer stage (81.57%), living in urban areas (67.49%), and diagnosed with lung cancer (56.69%).

Step 3: estimation of the comorbidity burden

Figure 3 demonstrated the rates, numbers, types and severity of comorbidity for inpatient cancer patients by cancer types. The highest comorbidity rate was observed in colon cancer patients (84.21%), followed by lung cancer (83.68%), gastric cancer (79.47%), rectal cancer (78.52%) and breast cancer (48.31%). Most cancer patients had fewer than three types of comorbidities across all cancer types, with the highest proportion observed in breast cancer patients (92.60%). Hypertension and mild liver disease were the most prevalent comorbidities across all five cancer types. Diabetes was the third most common comorbidity in lung (17.36%) and breast (9.91%) cancer patients, while renal disease ranked the third in colon (27.72%) and rectal (28.01%) cancer patients. A significant proportion of lung (56.89%), colon (55.26%), and gastric (51.69%) cancer patients had moderate to severe comorbidities.

Fig. 3

Comorbidity Burden of included Chinese inpatient cancer patients by Cancer Type (N = 3554)

Agglomerative hierarchical cluster analysis [23, 24] was used to classify individuals into groups based on comorbidities, and this was a commonly used bottom-up clustering method which clustered from individual patients to a final group containing all patients. Pearson correlation coefficient was used as a measure of distance between data points, and the data were standardized to convert the correlation into a distance measure, where data points with higher correlation were closer together and those with lower correlation were farther apart [25]. All prevalent conditions in this cohort were included in the cluster analysis. Patient characteristics of each comorbidity cluster were described. Among all 15 identified comorbidities, only 13 conditions were included in the comorbidity pattern analysis. Myocardial infarction (n = 0) and dementia (n = 2) were excluded due to the limited number of reported inpatient cancer patients. The diagram from cluster analysis of the remaining 13 conditions by five types of cancer is shown in Fig. 4. For lung cancer patients, four categories of comorbidities were identified: C6 CHF—C8 COPD (n = 14), C1 hypertension—C2 CVD—C7 PVD (n = 74), C9 mild liver disease—C12 renal disease (n = 56), and C10 diabetes—C13 diabetes with complications (n = 20). C1 hypertension—C10 diabetes cluster (n = 45) and C2 CVD—C7 PVD cluster (n = 30) were identified for female breast cancer patients. Gastric cancer patients also identified four comorbidity groups: C1 hypertension—C10 diabetes (n = 82), C2 CVD—C7 PVD (n = 61), C9 mild liver disease—C12 renal disease (n = 116), and C6 CHF—C8 COPD (n = 15). Three comorbidity clusters of rectal cancer patients were similar to those in patients with gastric cancer, which included: C1 hypertension—C10 diabetes cluster (n = 60), C2 CVD—C7 PVD cluster (n = 80), and C9 mild liver disease—C12 renal disease cluster (n = 81). C2 CVD—C7 PVD cluster (n = 57) and C9 mild liver disease—C12 renal disease cluster (n = 82) were for colon cancer patients.

Fig. 4

Comorbidity Pattern of included Chinese inpatient cancer patients (N = 3554)

Step 4: examination of the associations between comorbidity patterns and outcome measures

Primary outcome measures of interest were treatment option and medical cost. Treatment options were divided into conventional treatment and targeted therapy according to the guidelines for the diagnosis and treatment of lung cancer, gastric cancer and breast cancer (2018 edition) [26] and colorectal cancer (2015 edition) [27] issued by the National Health Commission in China. Conventional treatment included surgical treatment, chemotherapy and radiotherapy. Surgical treatment for lung, gastric and colorectal cancer includes palliative and radical surgery, surgical treatment for invasive breast cancer includes modified radical surgery and breast-conserving surgery, and chemoradiotherapy includes (new) adjuvant chemoradiotherapy. Because targeted therapy is generally not the primary treatment plan for early and middle stage cancer patients, it was regarded as an optional treatment plan alone in this study. All treatment options were recoded as binary classification problems to determine whether patients received this type of therapy. Thirdly, the data of medical cost was directly extracted from the HIS system of the case hospital. The mean interpolation method was adopted, and the average total cost was 87,369.84 RMB.

Each cluster to patients without identified comorbidities was then compared using binary logistic regression with treatment options as the dependent variable, and linear regression models with medical cost as the dependent variable. All the models were adjusted for age, sex, education level, living area, and marriage status. IBM SPSS Statistics 26 was adopted for the cluster analysis and RStudio version 3.5.1 (Lucent Technologies, Murray Hill, NJ, USA) was used for the descriptive analysis and regression analysis in this study. Table 3 shows the multivariate-adjusted ORs and 95% CIs for treatment options and medical cost according to whether cancer patients with identified comorbidity cluster. For cancer treatment option, gastric cancer patients in the CHF—COPD (C6—C8) cluster chose less conventional treatment (OR = − 1.38; 95% CI − 2.46, − 0.17), lung cancer patients in the diabetes—diabetes with complications (C10—C13) cluster (OR = − 1.01; 95% CI − 1.89, − 0.30) and hypertension—CVD—PVD (C1—C2—C7) cluster (OR = − 0.87; 95% CI − 1.76, − 0.14) chose less targeted therapy. Rectal cancer patients in the hypertension—diabetes (C1—C10) cluster (OR = 17,347; 95% CI 665.8434028.15) had significantly higher medical cost during the treatment.

Table 3 Outcome measures of each comorbidity cluster compared against Chinese inpatient cancer patients with no comorbidity