A total of 11,659 hospitalization records were extracted, including those of some patients who had been hospitalized multiple times. Duplicated records were deleted, and 6,816 records of the final hospitalization of each patient were retained. Data with missing records were deleted; therefore, between 2013 and 2020, 6,180 COPD admissions and 163 comorbidities were included in this study.
Patient characteristics
The baseline characteristics of 1,235 high-cost patients and 4,945 low-cost patients are presented in Table 1. There were significant differences in age (p < 0.0001) and smoking status (p = 0.003) between high-cost and low-cost groups. The rate of ICU admission (p < 0.0001) and the use of non-invasive ventilator (p < 0.0001) and invasive ventilator (p < 0.0001) in the high-cost group were higher than those in the low-cost group. To account for the statistical difference in mean age, all comparisons between the high-cost and low-cost groups were stratified using the following age brackets: ≤ 55, 56–64, and ≥ 65 years (Table 1). In addition, Using propensity score matching, we selected 1235 patients in the low-cost group and matched patients in the high-cost group by age, sex, and smoking status and performed the same analysis (Table S3 in supplementary material). The median number of comorbidities per individual differed significantly between the two groups and the difference remained significant when the comparison was adjusted by sex, age, and matched numbers (Table 1 and Table S3 in the supplementary material).
Table 1 Participant characteristics
Demographic
|
High hospitalization costs
|
Low hospitalization costs
|
P-value
|
Subjects (n)
|
1235
|
4945
|
|
Age
|
71.44±9.207
|
70.29±9.461
|
<0.0001
|
Age categories
|
|
|
|
Age<55 years
|
46 (3.72)
|
255(5.16)
|
0.036
|
Age 55–64 years
|
230 (18.62)
|
1127(22.79)
|
0.002
|
Age ≥ 65 years
|
959 (77.65)
|
3563(72.05)
|
<0.0001
|
Male
|
1109 (89.80)
|
4424(89.46)
|
0.732
|
Current smoker
|
373 (30.20)
|
1716(34.70)
|
0.003
|
Respiratory failure
|
536 (43.40)
|
812(16.42)
|
<0.0001
|
ICU
|
184 (14.90)
|
3(0.06)
|
<0.0001
|
Invasive mechanical ventilation
|
181 (14.66)
|
24(0.49)
|
<0.0001
|
Noninvasive ventilation
|
648 (52.47)
|
648(13.10)
|
<0.0001
|
Death
|
22(1.78)
|
14(0.28)
|
<0.0001
|
Length of stay
|
14(10–19; 0–190)
|
7(5–8; 0–26)
|
<0.0001
|
Number of people with comorbidities
|
1004(81.30)
|
3672(74.26)
|
<0.0001
|
Comorbidities per patient#
|
|
|
|
Comorbidities (whole)
|
2 (1–3; 0–9)
|
1(0-2; 0-10)
|
<0.0001
|
Comorbidities by age
|
|
|
|
<55 years
|
1 (0-3; 0-5)
|
1(0-2; 0-7)
|
0.562
|
55–64 years
|
2 (1-3; 0-7)
|
1(0-2; 0-7)
|
<0.0001
|
≥ 65 years
|
2 (1-3; 0-9)
|
1(1-3; 0-10)
|
<0.0001
|
Comorbidities by sex
|
|
|
|
Male
|
2 (1-3; 0-9)
|
1(0-2; 0-10)
|
<0.0001
|
Female
|
2 (1-4; 0-7)
|
2 (1-3; 0-9)
|
0.006
|
Data are presented as mean ± SD, n (%), or median (interquartile range), unless otherwise stated. #: number of comorbidities per patient, presented as median (interquartile range; range).
Figures E1 and E2 in the supplement material show the comparison of the prevalence of the top 30 most frequent comorbidities between high-cost and low-cost patients stratified by age and sex. We observed an increase in the number of comorbidities from 69 in the under 55 age group to 109 and 155 in the 56–64 and 65 and above age groups, respectively. with increasing age groups, and the prevalence was higher in the high-cost patients than in the low-cost patients.
Relationship between the number of comorbidities and cost
Patients with COPD were divided into four groups according to the number of comorbidities, and the COPD hospitalization costs were compared among the groups (Table 2). There was a statistically significant difference in the total hospitalization cost among different groups (H = 176.35, p < 0.05).
Table 2 Relationship between the number of comorbidities in COPD patients and hospitalization costs
Number of comorbidities
|
Total hospitalization costs
|
P-value
|
0
|
14396.9750(11117.32-20043.64)
|
<0.001
|
1~3
|
15485.61(11884.55-22450.75)
|
4~5
|
17843.13(13008.62-27599.42)
|
>5
|
19742.94(13796.21-35017.20)
|
COPD:Chronic obstructive pulmonary disease
Regression analysis of hospitalization costs
Univariate analysis was performed to identify potential factors associated with high hospitalization cost (Table S4). Then, the selected significant factors were included in the final multivariate analysis: hypertension, Parkinson’s disease (PD), pulmonary embolism (PE), non-tuberculous mycobacterial pulmonary disease (NTM-PD), pneumomycosis, coronary artery disease (CAD), heart failure (HF), hyperlipidemia, cerebrovascular disease (CVD), diabetes, sick sinus syndrome (SSS), atrial fibrillation (AF), chronic kidney disease (CKD), acute kidney injury (AKI), asthma, obstructive sleep apnea-hypopnea syndrome (OSAHS), hypothyroidism, fatty liver, mental disorders, lung cancer, venous thromboembolism (VTE), destroyed lung, lymphoma, rheumatoid arthritis, aortic dissection, arteriosclerosis, lower extremity atherosclerotic occlusive disease (LEAOD), aneurysm, malnutrition, and epilepsy. These 30 factors were included in the final multivariable regression model. Table 3 summarizes the results of multivariable regression (F = 1537.628, p < 0.001). PE (OR = 17.66, 95% CI = 6.619–47.099), pneumomycosis (OR = 9.667, 95% CI = 5.952–15.699), HF (OR = 1.735, 95% CI = 1.345–2.238), CVD (OR = 1.906, 95% CI = 1.396–2.601), diabetes (OR = 1.317, 95% CI = 1.042–1.663), SSS (OR = 12.151, 95% CI = 1.104–133.688), CKD (OR = 1.822, 95% CI = 1.134–2.927), OSAHS (OR = 0.522, 95% CI = 0.289–0.946), fatty liver (OR = 0.571, 95% CI = 0.368–0.885), lung cancer (OR = 6.113, 95% CI = 4.009–9.321), VTE (OR = 3.338, 95% CI = 1.841–6.052), destroyed lung (OR = 1.871, 95% CI = 1.013–3.459), lymphoma (OR = 7.8, 95% CI = 1.937–31.405), arteriosclerosis (OR = 1.442, 95% CI = 1.046–1.990), and aneurysm (OR = 2.678, 95% CI = 1.309–5.479) had statistically significant associations with high hospitalization costs.
Table 3. Multivariate analysis of total hospitalization costs
|
Multivariable regression analysis
|
|
OR
|
95% CI
|
P-value
|
Age
|
1.008
|
1.000~1.017
|
0.05
|
Gender
|
1.154
|
0.872~1.527
|
0.317
|
Smoking
Current smokers vs. non-smokers
Former smokers vs. non-smokers
|
1.107
0.827
|
0.881~1.390
0.697~0.980
|
0.025
0.383
0.029
|
Respiratory failure
|
1.287
|
1.062~1.559
|
0.010
|
Invasive mechanical ventilation
|
4.017
|
2.221~7.268
|
<0.001
|
Noninvasive ventilation
|
5.262
|
4.382~6.318
|
<0.001
|
ICU
|
58.149
|
17.360~194.78
|
<0.001
|
Hypertension
|
1.018
|
0.860~1.203
|
0.839
|
PD
|
1.601
|
0.691~3.710
|
0.272
|
PE
|
17.656
|
6.619~47.099
|
<0.001
|
NTM-PD
|
2.770
|
0.131~58.344
|
<0.001
|
Pneumomycosis
|
9.67
|
5.95~15.70
|
<0.001
|
CAD
|
1.032
|
0.790~1.347
|
0.818
|
HF
|
1.735
|
1.345~2.238
|
<0.001
|
Hyperlipidemia
|
0.743
|
0.434~1.273
|
0.279
|
CVD
|
1.906
|
1.396~2.601
|
<0.001
|
Diabetes
|
1.317
|
1.042~1.663
|
0.021
|
SSS
|
12.151
|
1.104~133.688
|
0.041
|
AF
|
1.369
|
0.929~2.016
|
0.113
|
CKD
|
1.822
|
1.134~2.927
|
0.013
|
AKI
|
0.814
|
0.190~3.483
|
0.782
|
Asthma
|
0.870
|
0.623~1.214
|
0.412
|
OSAHS
|
0.522
|
0.289~0.946
|
0.032
|
Hypothyroidism
|
1.083
|
0.331~3.544
|
0.895
|
Fatty liver
|
0.571
|
0.368~0.885
|
0.012
|
Mental disorders
|
7.086
|
0.578~86.879
|
0.126
|
Lung cancer
|
6.1113
|
4.009~9.321
|
0.000
|
VTE
|
3.338
|
1.841~6.052
|
0.000
|
Destroyed lung
|
1.871
|
1.013~3.459
|
0.046
|
lymphoma
|
7.800
|
1.937~31.405
|
0.004
|
Rheumatoid arthritis
|
0.513
|
0.117~2.249
|
0.376
|
Aortic dissection
|
3.394
|
0.760~15.152
|
0.109
|
Arteriosclerosis
|
1.442
|
1.046~1.990
|
0.026
|
LEAOD
|
1.491
|
0.542~4.101
|
0.439
|
Aneurysm
|
2.678
|
1.309~5.479
|
0.007
|
Malnutrition
|
4.414
|
0.975~19.985
|
0.054
|
Epilepsy
|
4.670
|
0.672~32.425
|
0.119
|
OR:odd ratio; CI:confidence interval
Network analysis: correlations between COPD comorbidities and hospitalization costs
The comorbidity network for high-cost and low-cost patients is shown in Figure 1. The high-cost patient network comprised 112 nodes representing the comorbidities and 227 links representing correlations with a p-value of ≤ 0.01. In comparison, the low-cost patient network comprised 138 nodes and 456 links. There was a greater number of nodes and links in the low-cost patient network due to the large number of low cost-patients. However, when we compared the high-cost group (n = 1,235) with a randomly selected low-cost group subset matched by number, age, sex, and current smoking status, the resulting network had 88 nodes and 187 edges (Figure E3 of the online supplement). The prevalence, diversity of diseases, and density of associations (links) were higher in high-cost patients than in low-cost patients.
We used network analysis to identify comorbidities associated with high hospitalization costs. If we use a p-value cut-off of ≤ 0.05, the number of comorbidities associated with high hospitalization costs was 39; however, the number of comorbidities decreased to 23 and 18 if the significance threshold was lowered to p ≤ 0.01 and p ≤ 0.001, respectively (Figure 2). With more stringent p-values, less prevalent and connected nodes were lost, with little changes seen in the highly connected nodes, also referred to as hubs (Figure 3).