3.1 Demographics and characteristics upon admission
In the first stage of the study, 916 inpatients with confirmed COVID-19, divided into non-elderly and elderly groups, were included (Figure 1). The clinical characteristics and outcomes of the non-elderly and elderly groups with coinfections are compared in Table 1.
Demographically, the median age was 54.0 (IQR 49.0–59.0) years for the non-elderly group and 79.0 (IQR 73.0–84.0) years for the elderly group, while sex did not differ significantly between the two groups. Of note, in the elderly group, frequencies of non-hyperpyrexia (T < 39℃, 44.2% vs. 25%, P = 0.026) and dyspnoea (53.9% vs. 31.2%, P = 0.009) were significantly higher, whereas the rate of headache (9.09% vs. 25%, P = 0.008) was significantly lower, than these clinical features in the non-elderly group. The proportions of diabetes (62.4% vs. 10.4%, P < 0.001), hypertension (53.9% vs. 35.4%, P = 0.036), chronic cardiovascular diseases (35.8% vs. 14.6%, P = 0.009), and chronic respiratory diseases (33.3% vs. 14.6%, P = 0.019) were significantly higher, while that of haematological disorders (1.82% vs. 8.33%, P = 0.047) was significantly lower in the elderly vs. non-elderly individuals. We found a substantially higher proportion of the elderly group that had undergone invasive procedures (42.4% vs. 16.7%, P = 0.002). In addition, several laboratory indices, including oxygenation index (OI), albumin (ALB), albumin/globulin (A/G), estimated glomerular filtration rate (eGFR), blood potassium (K+), blood sodium (Na+), and procalcitonin (PCT), were significantly different (P < 0.05) between the two groups, as shown in Table 1.
The coinfection rate was significantly higher in the elderly than that in the non-elderly group (26.61% vs. 16.22%, P = 0.001, Table 1). There was no significant difference in the number of body sites with coinfections between the two groups (P > 0.05). We found that the co-occurrence rate of pulmonary coinfection (85.5% vs. 62.5%, P < 0.001) and other coinfections (0.61% vs. 10.4%, P = 0.002) in the elderly group was significantly different from those in the non-elderly group, with significantly higher rates of pulmonary coinfection in the elderly. No significant differences were observed between the two groups in terms of bloodstream coinfection, intestinal coinfection, urinary tract coinfection, or skin and soft tissue coinfection (Table 1).
3.2 Clinical outcomes
Elderly COVID-19 patients with coinfections showed a significantly higher rate of intensive care unit admissions (30.9% vs. 14.6%, P = 0.04) and mortality (40% vs. 22.9%, P = 0.046) than the non-elderly population (Table 1).
Table1. Comparison of clinical characteristics and outcomes between the non-elderly (< 65) and elderly (≥ 65) COVID-19patients (916 in total) with coinfections in the first stage.
Variables
|
Non-elderly COVID-19
patients with coinfection
(48/296, 16.22%)
|
Elderly COVID-19
patients with coinfection
(165/620, 26.61%)
|
P
|
Demographic data
|
|
|
|
Male
|
28 (58.3%)
|
121 (73.3%)
|
0.069
|
Age
|
54.0 [49.0;59.0]
|
79.0 [73.0;84.0]
|
<0.001
|
Clinical symptoms
|
|
|
|
T <39℃
|
12 (25.0%)
|
73 (44.2%)
|
0.026
|
T ≥39℃
|
20 (41.7%)
|
52 (31.5%)
|
0.256
|
Chill
|
4 (8.33%)
|
4 (2.42%)
|
0.079
|
Cough
|
7 (14.6%)
|
21 (12.7%)
|
0.926
|
Dyspnea
|
15 (31.2%)
|
89 (53.9%)
|
0.009
|
Headache
|
12 (25.0%)
|
15 (9.09%)
|
0.008
|
Muscular soreness
|
8 (16.7%)
|
16 (9.70%)
|
0.278
|
Comorbidities
|
|
|
|
Diabetes
|
5 (10.4%)
|
103 (62.4%)
|
<0.001
|
Hypertension
|
17 (35.4%)
|
89 (53.9%)
|
0.036
|
Chronic cardiovascular diseases
|
7 (14.6%)
|
59 (35.8%)
|
0.009
|
Chronic respiratory disease
|
7 (14.6%)
|
55 (33.3%)
|
0.019
|
Chronic renal disease
|
8 (16.7%)
|
19 (11.5%)
|
0.485
|
Chronic liver disease
|
7 (14.6%)
|
7 (4.24%)
|
0.018
|
Autoimmune disease
|
3 (6.25%)
|
6 (3.64%)
|
0.425
|
Cerebrovascular diseases
|
3 (6.25%)
|
30 (18.2%)
|
0.074
|
Hematological disorder
|
4 (8.33%)
|
3 (1.82%)
|
0.047
|
Tumor
|
9 (18.8%)
|
17 (10.3%)
|
0.186
|
Invasive procedure
|
8 (16.7%)
|
70 (42.4%)
|
0.002
|
Arterial blood gases
|
|
|
|
Lac
|
1.21 [0.82;1.85]
|
1.30 [0.90;1.72]
|
0.719
|
OI
|
311 [212;373]
|
263 [148;352]
|
0.043
|
Blood routine
|
|
|
|
WBC
|
6.70 [4.27;8.67]
|
7.20 [4.90;10.3]
|
0.245
|
HGB
|
110 [87.0;130]
|
118 [107;130]
|
0.053
|
PLT
|
167 [97.0;226]
|
168 [123;242]
|
0.331
|
Percentage of NE
|
81.7 [66.5;89.5]
|
81.9 [71.2;91.0]
|
0.418
|
Percentage of LY
|
10.1 [5.19;20.7]
|
8.70 [4.76;17.1]
|
0.264
|
Percentage of MO
|
6.89 [3.81;9.96]
|
7.40 [4.14;10.6]
|
0.996
|
NLR
|
7.91 [3.05;17.3]
|
9.29 [4.43;17.9]
|
0.246
|
Liver function
|
|
|
|
ALB
|
32.8 [30.6;36.3]
|
31.2 [28.1;34.8]
|
0.007
|
GLB
|
27.8 [24.3;30.5]
|
28.9 [25.1;32.8]
|
0.133
|
A/G
|
1.21 [1.07;1.47]
|
1.11 [0.96;1.31]
|
0.003
|
TBIL
|
9.85 [6.18;17.2]
|
10.8 [7.90;14.2]
|
0.531
|
DBIL
|
3.35 [2.18;7.08]
|
4.10 [3.00;5.70]
|
0.352
|
ALT
|
19.9 [14.9;34.9]
|
25.0 [16.3;40.2]
|
0.092
|
AST
|
29.4 [20.3;38.7]
|
34.1 [23.8;49.9]
|
0.063
|
AST/ALT
|
1.38 [0.90;1.96]
|
1.37 [1.03;1.77]
|
0.967
|
Kidney function
|
|
|
|
eGFR
|
89.1 [34.3;115]
|
71.4 [37.7;86.5]
|
0.024
|
Myocardial enzyme
|
|
|
|
CK
|
68.3 [31.6;153]
|
88.0 [49.9;185]
|
0.075
|
Serum electrolytes
|
|
|
|
K
|
4.28 [4.11;4.51]
|
3.95 [3.56;4.40]
|
<0.001
|
Na
|
138 [135;140]
|
140 [136;142]
|
0.021
|
Coagulation function
|
|
|
|
PTA
|
90.5 [77.0;104]
|
93.2 [78.4;101]
|
0.945
|
INR
|
1.06 [0.94;1.12]
|
1.02 [0.96;1.14]
|
0.887
|
FIB
|
4.38 [3.41;5.37]
|
4.37 [3.34;5.50]
|
0.949
|
D Dimer
|
0.30 [0.16;0.82]
|
0.36 [0.17;0.99]
|
0.327
|
Inflammatory factor
|
|
|
|
PCT
|
0.12 [0.05;0.53]
|
0.44 [0.09;2.80]
|
0.003
|
CRP
|
40.7 [15.1;96.3]
|
57.6 [18.3;102]
|
0.530
|
ESR
|
57.0 [33.0;83.0]
|
61.0 [39.0;82.0]
|
0.784
|
IL 10
|
5.00 [2.97;8.23]
|
5.00 [3.01;7.49]
|
0.890
|
IL 6
|
19.3 [4.14;54.2]
|
15.9 [6.15;42.4]
|
0.664
|
Details of coinfection
|
|
|
|
Numbers of coinfection
|
|
|
|
1
|
7 (14.6%)
|
38 (23.0%)
|
0.289
|
2
|
1 (2.08%)
|
9 (5.45%)
|
0.462
|
3
|
1 (2.08%)
|
3 (1.82%)
|
1.000
|
Position of coinfection
|
|
|
|
Pulmonary coinfection
|
30 (62.5%)
|
141 (85.45%)
|
0.001
|
Bloodstream coinfection
|
8 (16.7%)
|
30 (18.18%)
|
0.978
|
Intestinal coinfection
|
9 (18.8%)
|
28 (16.97%)
|
0.944
|
Urinary tract coinfection
|
4 (8.33%)
|
18 (10.91%)
|
0.789
|
Soft tissue coinfection
|
3 (6.25%)
|
7 (4.24%)
|
0.697
|
Other coinfections
|
5 (10.4%)
|
1 (0.61%)
|
0.002
|
Clinical outcomes
|
|
|
|
ICU admission in hospitalization
|
7 (14.6%)
|
51 (30.9%)
|
0.040
|
Death (while in hospital)
|
11 (22.9%)
|
66 (40.0%)
|
0.046
|
Abbreviations: ALB: albumin; A/G: albumin/ globulin; ALT: alanine aminotransferase; AST: aspartate transaminase; AST/ALT: aspartate-to-alanine transaminase ratio; CK: creatine kinase; CRP: C-reactive protein; DBIL: direct bilirubin; eGFR: estimated glomerular filtration rate; ESR: erythrocyte sedimentation rate; FIB: fibrinogen; GLB: globulin; HGB: hemoglobin; INR: international normalized ratio; IL 10: interleukin 10; IL 6: interleukin 6; ICU: intensive care unit; K+: blood potassium; Lac: lactic acid; LY: percentage of leukomonocyte; MO: percentage of monocyte; NE: percentage of neutrophil; NLR: neutrophil-to-lymphocyte ratio; Na+: blood sodium; OI: oxygenation index; PLT: platelet; PTA: prothrombin activity; PCT: procalcitonin; TBIL: total bilirubin; WBC: white blood cell.
3.3 Characteristics of coinfections in elderly patients with COVID-19
The age distribution of elderly patients with and without coinfections is shown in Figure 2A. The incidence of coinfections increased with age, from 24% to 40%. Notably, pulmonary infection (86%) was the most common coinfections in the elderly group, followed by bloodstream coinfection (19%), intestinal coinfection (15%), urinary tract coinfection (12%), and skin and soft tissue coinfection (2%, Figure 2B).
The most common pathogens isolated from coinfections in elderly patients with COVID-19 were gram-negative bacteria (48%), fungi (38%), and gram-positive bacteria (14%, Figure 2C). The four most common gram-positive bacteria isolated were Enterococcus faecium (15), Staphylococcus aureus (12), Corynebacterium striatum (10), and Coagulase negative staphylococcus (6) (Figure 2D). The five most commonly isolated gram-negative bacteria were Acinetobacter baumanii (95), Klebsiella pneumoniae (72), Pseudomonas aeruginosa (29), Burkholderia cepacian (18), and Escherichia coli (17). The four most commonly detected fungi were Candida albicans (82), Aspergillus spp. (34), Candida glabrata (17), and Candida tropicalis (10). We tested for only two viruses, Cytomegalovirus and Herpes simplex virus (Figure 2D). Four multidrug resistant bacteria were identified, including carbapenem-resistant Klebsiella pneumoniae (9), carbapenem-resistant Acinetobacter baumannii (7), methicillin-resistant Staphylococcus areus (3), and vancomycin-resistant Enterococci (4) (Figure 2E).
The distribution of coinfecting pathogens by age group in elderly patients with COVID-19 is shown in Figure 2F. The most common pathogens isolated in patients aged less than 90 years were gram-negative bacteria, followed by fungi and gram-positive bacteria. In patients aged > 90 years, fungi were the major pathogens detected.
3.3 Nomogram to identify early coinfections in elderly patients with COVID-19
In the first stage of this study, 620 elderly patients with COVID-19 were included in the analysis and randomly divided into training (n = 465) and internal validation (n = 155) cohorts at a ratio of 3:1. The data grouping for the two cohorts was random and matched (P > 0.05, Table S1). Univariable logistic regression analysis showed age, invasive procedure, diabetes, OI, white blood cells (WBC), hemoglobin (HGB), percentage neutrophils, percentage lymphocytes, percentage monocytes (MO), neutrophil-lymphocyte ratio (NLR), ALB, A/G, aspartate-to-alanine transaminase ratio, eGFR, K+, Na+, D dimer, PCT, and C-reactive protein (CRP) were predictors of coinfections in the elderly patient with COVID-19 (Table 2).
Table 2. Predictors to identify elderly COVID-19 patients with coinfections in the training cohort using univariable and multivariate logistic regression analysis
Characteristics
|
Univariable logistic regression
|
Multivariate logistic regression
|
OR
|
CI
|
P
|
OR
|
CI
|
P
|
Male
|
1.51
|
0.97-2.36
|
0.07
|
-
|
-
|
-
|
Age
|
1.03
|
1-1.06
|
0.03
|
-
|
-
|
-
|
Invasive procedure
|
7.71
|
4.75-12.54
|
<0.001
|
5.17
|
2.78-9.69
|
<0.001
|
Diabetes
|
5.14
|
3.33-7.94
|
<0.001
|
5
|
2.86-8.92
|
<0.001
|
Hypertension
|
1.03
|
0.69-1.56
|
0.87
|
-
|
-
|
-
|
Heart disease
|
1
|
0.66-1.52
|
1
|
-
|
-
|
-
|
Chronic respiratory disease
|
1.48
|
0.96-2.3
|
0.08
|
-
|
-
|
-
|
Chronic renal disease
|
1.53
|
0.75-3.11
|
0.25
|
-
|
-
|
-
|
Chronic liver disease
|
0.58
|
0.21-1.55
|
0.28
|
-
|
-
|
-
|
Autoimmune disease
|
1.21
|
0.45-3.26
|
0.7
|
-
|
-
|
-
|
Central nervous system disease
|
1.08
|
0.64-1.85
|
0.77
|
-
|
-
|
-
|
Hematological disorder
|
1.31
|
0.24-7.22
|
0.76
|
-
|
-
|
-
|
Tumor
|
0.68
|
0.35-1.33
|
0.26
|
-
|
-
|
-
|
Lac
|
1.05
|
0.84-1.32
|
0.65
|
-
|
-
|
-
|
OI
|
1
|
1-1
|
<0.001
|
-
|
-
|
-
|
WBC
|
1.13
|
1.07-1.2
|
<0.001
|
1.01
|
0.93-1.09
|
0.86
|
HGB
|
0.99
|
0.98-1
|
0.03
|
1
|
0.98-1.01
|
0.6
|
PLT
|
1
|
1-1
|
0.28
|
-
|
-
|
-
|
Percentage of NE
|
1.05
|
1.03-1.07
|
<0.001
|
-
|
-
|
-
|
Percentage of LY
|
0.95
|
0.92-0.97
|
<0.001
|
-
|
-
|
-
|
Percentage of MO
|
0.91
|
0.87-0.96
|
<0.001
|
0.96
|
0.91-1.02
|
0.17
|
NLR
|
1.05
|
1.03-1.07
|
<0.001
|
-
|
-
|
-
|
ALB
|
0.9
|
0.86-0.95
|
<0.001
|
0.97
|
0.89-1.05
|
0.46
|
GLB
|
0.98
|
0.94-1.03
|
0.44
|
-
|
-
|
-
|
A/G
|
0.37
|
0.16-0.87
|
0.02
|
0.71
|
0.16-3.11
|
0.66
|
TBIL
|
1
|
0.98-1.02
|
0.99
|
-
|
-
|
-
|
DBIL
|
1
|
0.98-1.03
|
0.77
|
-
|
-
|
-
|
ALT
|
1
|
1-1.01
|
0.26
|
-
|
-
|
-
|
AST
|
1.01
|
1-1.01
|
0.07
|
-
|
-
|
-
|
AST/ALT
|
1.34
|
1.01-1.79
|
0.04
|
-
|
-
|
-
|
eGFR
|
0.98
|
0.98-0.99
|
<0.001
|
0.99
|
0.98-1
|
0.25
|
CK
|
1
|
1-1
|
0.08
|
-
|
-
|
-
|
K+
|
1.61
|
1.1-2.37
|
0.01
|
1.19
|
0.69-2.07
|
0.53
|
Na+
|
1.06
|
1.02-1.11
|
<0.001
|
1.02
|
0.96-1.09
|
0.46
|
PTA
|
0.99
|
0.98-1
|
0.23
|
-
|
-
|
-
|
INR
|
1.12
|
0.58-2.16
|
0.74
|
-
|
-
|
-
|
FIB
|
1.1
|
0.96-1.25
|
0.16
|
-
|
-
|
-
|
D Dimer
|
1.18
|
1.07-1.31
|
<0.001
|
-
|
-
|
-
|
PCT
|
3.58
|
2.47-5.19
|
<0.001
|
3.48
|
2.47-5.23
|
<0.001
|
CRP
|
1.01
|
1-1.01
|
<0.001
|
1
|
0.99-1
|
0.12
|
ESR
|
1
|
0.99-1.01
|
0.89
|
-
|
-
|
-
|
IL10
|
1
|
0.99-1.02
|
0.81
|
-
|
-
|
-
|
IL6
|
1
|
1-1
|
0.23
|
-
|
-
|
-
|
Abbreviations: ALB: albumin; A/G: albumin/ globulin; ALT: alanine aminotransferase; AST: aspartate transaminase; AST/ALT: aspartate-to-alanine transaminase ratio; CK: creatine kinase; CRP: C-reactive protein; DBIL: direct bilirubin; eGFR: estimated glomerular filtration rate; ESR: erythrocyte sedimentation rate; FIB: fibrinogen; GLB: globulin; HGB: hemoglobin; INR: international normalized ratio; IL 10: interleukin 10; IL 6: interleukin 6; K+: blood potassium; Lac: lactic acid; LY: percentage of leukomonocyte; MO: percentage of monocyte; NE: percentage of neutrophil; NLR: neutrophil-to-lymphocyte ratio; Na+: blood sodium; OI: oxygenation index; PLT: platelet; PTA: prothrombin activity; PCT: procalcitonin; TBIL: total bilirubin; WBC: white blood cell;
LASSO regression was used to further screen the statistically significant parameters in univariable logistic regression analysis (P < 0.05). The coefficients of variables are shown in Figure 3A. A 10-fold cross-validation method was applied to the iterative analysis, and a model with excellent variable performance was obtained when λ was 0.0059 (Figure 3B). The variables screened included invasive procedures, diabetes, WBC count, HGB, MO, ALB, A/G, eGFR, K+, Na+, PCT, and CRP, whose coefficients were not equal to 0 (Table S2). Multivariate logistic regression was further analyzed parameters that were screened by LASSO regression. The results showed diabetes comorbidity (OR = 5, 95% CI = 2.86–8.92, P = < 0.001), previous invasive procedure (OR = 5.17, 95% CI = 2.78–9.69, P < 0.001), and PCT level (OR = 3.48, 95% CI = 2.47–5.23, P < 0.001) were significantly associated with coinfections in elderly COVID-19 patients (Table 2).
The nomogram was developed based on the multivariate logistic regression analysis results with three predictors, diabetes comorbidity, previous invasive procedure, and PCT level (Figure 3C). Based on the ordinary nomogram, we developed a web-based dynamic nomogram application (https://elderly-covid19.shinyapps.io/DynNomapp/) to precisely calculate the probability of coinfections in elderly patients with COVID-19.
3.4 Nomogram validation
3.4.1 Internal validation
The AUCs in the training and internal validation cohorts were 0.86 (95% CI: 0.82–0.90) and 0.82 (0.78–0.87), respectively (Figures 4A and 4B). The Hosmer–Lemeshow test revealed high concordance between the predicted and observed probabilities for the training and internal validation cohorts (P > 0.05, Figures 4D and 4E). The DCA illustrated that the clinical net benefit of the nomogram was higher than the default strategy of "treat all" or "treat none," with a wide threshold probability range in the training and internal validation cohorts for predicting the risk of coinfections in older patients with COVID-19 (Figure 4G and 4H). These results indicated that the proposed model performed well in both the training and internal validation cohorts.
3.4.2 External validation
An independent cohort of 306 elderly inpatients who were diagnosed with COVID-19 between 2 February 2023 and 30 June 2023 at Xiangya Hospital of Central South University in Changsha, Hunan, Jiangxi Provincial People's Hospital in Nanchang, Jiangxi, the Affiliated Nanhua Hospital of University of South China in Hengyang, Hunan, and the First Affiliated Hospital of Xinxiang Medical University in Weihui, Henan were prospectively enrolled for external validation of the nomogram. Comparisons of the baseline characteristics and clinical outcomes between elderly patients with COVID-19 in each of the two stages are shown in Table S3. The independent cohort was used for external validation of the nomogram. The AUC for the external validation cohort was 0.83 (95% CI: 0.78–0.87, Figure 4C). The calibration plot of the nomogram indicated no deviation from the reference line (P > 0.05, Figure 4F). The external validation cohort showed that our nomogram to predict coinfections in elderly patients with COVID-19 achieved positive net clinical benefits over a broad range of threshold probabilities, indicating the high clinical utility of this model (Figure 4I).
3.5 Comparison of the diagnostic performance of the nomogram, PCT alone, and CRP alone
A comparison of the diagnostic performances of the constructed nomogram, PCT alone, and CRP alone to predict the presence of coinfections in elderly patients with COVID-19 is shown in Table 3. The AUC (95% CI) of the nomogram was significantly (P < 0.05) higher than that for PCT alone or CRP alone in all cohorts at 0.86 (95% CI: 0.82–0.90), 0.82 (95% CI: 0.75–0.89), and 0.83 (95% CI: 0.78–0.87), respectively.
With a fixed specificity of 90%, the nomogram had higher sensitivity, accuracy, PPV, and NPV for detecting the presence of coinfections in elderly patients with COVID-19 than PCT alone and CRP alone in all three cohorts (Table 3). Similarly, the test performance of the nomogram at 90% sensitivity in the two validation cohorts was consistent across training validation cohorts. However, close to a sensitivity of 90%, the nomogram had a specificity of 47%, accuracy of 59%, and PPV of 72% in the training cohort, which was slightly lower than those for PCT alone but higher than those for CRP alone. At the maximum Youden's index, the nomogram had an optimal cutoff of -0.30 with a specificity, sensitivity, accuracy, PPV, and NPV of 98%, 64%, 64%, 91%, and 87%, respectively, in the training cohort. All indices of diagnostic performance of the nomogram in the training cohort were higher than those for PCT alone and CRP alone, which is similar to results from the internal and external validation cohorts (Table 3).
Table3. Diagnostic performance of the nomogram, PCT alone, and CRP alone for detection of coinfection in all elderly COVID-19 patients from the three cohorts.
Cohorts
|
Training cohort
(n=465)
|
Internal validation cohort
(n=155)
|
External validation cohort
(n=306)
|
Variables
|
Nomogram
|
PCT
|
CRP
|
Nomogram
|
PCT
|
CRP
|
Nomogram
|
PCT
|
CRP
|
AUC
(95%CI)
|
0.86
(0.82-0.90)
|
0.82
(0.78-0.87)
|
0.61
(0.55-0.67)
|
0.82
(0.75-0.89)
|
0.68
(0.60-0.77)
|
0.55
(0.45-0.65)
|
0.83
( 0.78-0.87)
|
0.70
( 0.64-0.76)
|
0.64
(0.58-0.70)
|
P*
|
NA.
|
0.04
|
<0.001
|
NA.
|
0.005
|
<0.001
|
NA.
|
<0.001
|
<0.001
|
Specificity Fixed at 90%§
|
Sen.
|
70%
|
56%
|
15%
|
33%
|
6%
|
6%
|
44%
|
29%
|
21%
|
Accuracy
|
85%
|
81%
|
70%
|
77%
|
76%
|
71%
|
70%
|
63%
|
59%
|
PPV
|
73%
|
69%
|
38%
|
52%
|
40%
|
15%
|
79%
|
71%
|
64%
|
NPV
|
89%
|
84%
|
73%
|
82%
|
77%
|
76%
|
67%
|
61%
|
59%
|
Sensitivity Fixed at 90%§
|
Spe.
|
47%
|
52%
|
19%
|
53%
|
49%
|
19%
|
45%
|
33%
|
22%
|
Accuracy
|
59%
|
61%
|
39%
|
62%
|
56%
|
36%
|
65%
|
58%
|
53%
|
PPV
|
40%
|
41%
|
30%
|
37%
|
32%
|
26%
|
57%
|
52%
|
48%
|
NPV
|
93%
|
91%
|
84%
|
95%
|
89%
|
88%
|
85%
|
78%
|
74%
|
At maximum Youden's index
|
Optimal cutoff
|
0.30
|
0.23
|
53.45
|
-2.20
|
0.07
|
97.70
|
-0.72
|
0.18
|
30.55
|
Spe.
|
98%
|
87%
|
60%
|
65%
|
55%
|
81%
|
77%
|
73%
|
57%
|
Sen.
|
64%
|
64%
|
57%
|
92%
|
75%
|
33%
|
80%
|
59%
|
64%
|
Accuracy
|
88%
|
81%
|
59%
|
71%
|
60%
|
70%
|
78%
|
67%
|
60%
|
PPV
|
91%
|
65%
|
35%
|
44%
|
34%
|
34%
|
74%
|
64%
|
55%
|
NPV
|
87%
|
86%
|
78%
|
96%
|
88%
|
80%
|
83%
|
69%
|
66%
|
* Refers to the comparison of C-statistic and its 95%CI with that of the nomogram in the same cohort.
§ Fixed at specificity or sensitivity closest to 90%.
Abbreviations: CI, confidence interval; NPV, negative predictive value; PPV, positive predictive value; Sens., sensitivity; Spe., Specificity.