Absolute Lymphocytes, Ferritin, C-Reactive Protein, and Lactate Dehydrogenase Predict Early Invasive Ventilation in Patients With COVID-19

Salvador Payán-Pernía, MD; Lucía Gómez Pérez, MD; Ángel F. Remacha Sevilla, MD, PhD; Jordi Sierra Gil, MD, PhD; Silvana Novelli Canales, MD, PhD


Lab Med. 2021;52(2):141-145. 

In This Article


Table 1 shows the demographic and clinical characteristics of the 160 patients included in the study. Overall, 58.10% were men. The median age was 57 years (minimum: 23; maximum: 75). A total of 32 patients (20%) required endotracheal intubation and mechanical ventilation, which happened in 96.9% of these patients within 48 hours after the emergency room admission.

Four independent variables were selected for logistic regression model fitting (Table 2): LDH, ALC, ferritin, and CRP. The platelet count was excluded (P >.05). Although each variable in the equation remained statistically significant, CRP and ferritin were close to the limit of significance (P =.046 and P =.045, respectively). Without including independent variables in the model, the probability of not being intubated was 80%. After the inclusion of the independent variables, the model's capacity to predict no intubation improved to 88%.

In the ROC analysis, LDH showed the greater AUC, with a value of 89.1%, followed by CRP (80.5%), ALC (77.6%), and ferritin (77.5%). Using the AUC, we established cutoff values for clinical application, as shown in Table 3. We found that LDH might represent the most useful discrimination variable for clinical application: when it was <219 U/L, patients did not need MIV, with a sensitivity of 100% and a specificity of 11.7%. No patients with CRP <65.65 mg/L or an ALC >1.56 × 10[9]/L needed MIV. Ferritin levels <300 μg/L predicted no MIV with a sensitivity of 93.8%.

On the classification tree analysis (Figure 1), LDH >598 U/L increased the likelihood of MIV, and an <424 U/L defined a category with the lowest probability of MIV.

Figure 1.

Classification tree. The root node (node 0) shows the binary (Yes/No) distribution of the outcome of the dependent variable, MIV. The χ2 test is then applied to ensure that the branch is associated with a statistically significant predictor of MIV—in this case, LDH. As seen in node 1, LDH ≤424.0 U/L could predict that 95.5% of those patients did not need MIV; in addition, 70% of all patients had LDH ≤424.0 U/L. On the other hand, as shown in node 3, LDH >598.0 U/L could anticipate that 81.2% of patients needed MIV; only a minority of all patients (10%) reached those LDH levels. MIV, mechanical invasive ventilation; LDH, lactate dehydrogenase.