Fig 1.
DFI, diabetic foot infection; DFO, diabetic foot osteomyelitis; SRI, severe renal impairment.
Table 1.
Characteristics of patients with DFI.
Table 2.
Univariate analysis between patients with and without SRI.
Table 3.
Analysis of the interaction between ESR and SRI in diagnosis of DFO.
Fig 2.
Box–Whisker plot of the difference in ESR between DFI patients with and without SRI.
The graph presents the median, quartiles, and range of ESR in DFI patients with and without SRI. The red dots and blue triangles show all individual data points of two groups of patients. The median ESR of the 44 patients with SRI was 70mm/h (IQR, 47–90 mm/h); and the median ESR of the 320 patients without SRI was 40mm/h (IQR, 23–63 mm/h). There was a statistically significant difference between patients with and without SRI (P < 0.001; Kruskal–Wallis H test). DFI, diabetic foot infection; ESR, erythrocyte sedimentation rate; IQR, interquartile range; SRI, severe renal impairment.
Fig 3.
ROC analysis for all of the patients and the two subgroups.
The graph illustrates ROC analysis results for all of the patients (blue solid line), patients without SRI (yellow solid line), and patients with SRI (green solid line). It can be seen that all of the patients and patients without SRI share similar curves, but the curve of patients with SRI is far away from them and close to the expected line representing the performance of random guess (diagonal red dashed line). Moreover, the AUC of patients with SRI is distinctly smaller than the other two groups (patients with SRI, 0.57; patients without SRI, 0.76; all of the patients, 0.74). These indicate ESR has a limited value on predicting DFO in patients with SRI. AUC, area under the curve; DFO, diabetic foot osteomyelitis; ESR, erythrocyte sedimentation rate; ROC, receiver operating characteristic; SRI, severe renal impairment.
Fig 4.
Box–Whisker plot of sensitivity, specificity, PPV, NPV, LR+, and LR–for all of the patients and the two subgroups.
The panels summary the sensitivity (A), specificity (B), PPV (C), NPV (D), LR+ (E), and LR–(F) with their 95% CIs at optimal cutoff points for all of the patients (45 mm/h), patients without SRI (45 mm/h), and patients with SRI (42 mm/h), which were calculated by VassarStats. It is apparent that the sensitivity, specificity, PPV, NPV, LR+, and LR–between all of the patients and patients without SRI are generally similar, with widely overlapping 95% CIs. What is interesting is that, the specificity, PPV, and LR+ in patients with SRI are strikingly lower than the other two groups and 95% CIs do not widely overlap with them. These implies a declining predictive value of DFO by ESR in patients with SRI. CI, confidence interval; DFO, diabetic foot osteomyelitis; ESR, erythrocyte sedimentation rate; LR+, positive likelihood ratio; LR−, negative likelihood ratio; NPV, negative predictive value; PPV, positive predictive value; SRI, severe renal impairment.