Patient demographics and baseline characteristics
154 patients were admitted with confirmed SARS-CoV2 infection in the CHU Amiens. These patients were hospitalized in conventional units (n=111, 72%) or ICU (n=43, 28%). Among them, 18 were transferred from conventional care units to ICU. The median age at admission was 77 years (range 23-100), 56% of the patients were male and 44% female. The median follow up was 12 days, ranging 0 to 42 at the time of analysis. The median hospital length of stay in conventional care units and ICU was 12 days. At the time of analysis, 122 patients were alive (79%) and 32 died (21%) from SARS-CoV2 infection.
The baseline characteristics of the patients are summarized in table 1.
Laboratory findings on hospital admission
35 hematological and biochemical parameters routinely measured in the laboratory were retrospectively collected. Characteristics of these analyses performed upon admission to the hospital are summarized in table 2. A small number of tests (<50) was observed for cardiac markers, uric acid, LDH, ferritin and albumin. These tests were then removed from further statistical analysis.
Comparison of laboratory markers between survivor and non-survivor patients
First, we compared the mean levels of the different biological markers in patients who survived (n=122) and patients who died (n=32) from SARS-CoV2 infection (table 3). 4 biological parameters were significantly different between survivors and dead patients, respectively: hemoglobin concentration (12.8 vs 11.9 g/dL, p=0.02), bicarbonate concentration (26.8 vs 25 mmol/L, p=0.03), base excess (0.92 vs -2.11 mmol/L, p=0.02) and potassium concentration (3.96 vs 4.22 mmol/L, p=0.03). We observed a strong trend for lymphocyte count (1.2 vs 0.9 x 109/L, p=0.06) and ALAT concentration (50 vs 30 UI/L, p=0.06).
Comparison of laboratory markers between patients hospitalized in conventional and intensive care units
We compared then major laboratory markers between patients hospitalized in conventional care units (n=111) and in ICU (n=43). Results are summarized in table 4. Among the biological parameters, 9 were statistically different between the two groups : white blood cells count (6.8 vs 8.9 x109/L, p=0.01), lymphocyte count (1.2 vs 0.8 x109/L, p=0.01), neutrophil count (5.2 vs 7.3 x109/L , p=0.001), bicarbonate concentration (27. vs 23.7 mmol/L, p<0.001), base excess (1.27 vs -1.87 mmol/L, p=0.003), lactate concentration (1.45 vs 1.99 mmol/L, p=0.015), sodium concentration (138.1 vs 136.2 mmol/L, p=0.01), calcium concentration (2.27 vs 2.12 mmol/L, p<0.001) and C Reactive Protein concentration (82 vs 199.6 mg/L, p<0.001). Furthermore, there was a strong trend for two additional markers : prothrombin time (15 vs 13.48 seconds, p=0.07) and plasma proteins concentration (67.31 vs 64.51 g/L, p=0.052).
Biological markers to identify high risk patients
We performed then survival analyses to determine if biological parameters at diagnosis could predict clinical outcomes. We evaluated OS as risk of death, and EFS, defined as transfer to ICU or death in conventional care units as risk of clinical worsening. Global OS and EFS were described in our cohort by the Kaplan Meier method (figure 1). At median follow up, EFS and OS were around 85%. We then used univariate analysis to evaluate biological parameters for EFS (table 5). We also tested age in as it was described as a powerful risk factor for mortality (6). Lymphopenia (p=0.048), hyponatremia (p=0.003), high uremia levels (p=0.02), hypocalcemia (p=0.01) were statistically significant. In our multivariate model, hyponatremia (p<0.001, HR=11.7 (3.1-44.2)) and low bicarbonate levels (p=0.02, HR=5.4 (1.4-21)) negatively affected EFS. Age was not associated with worse EFS in our study.
For OS, in univariate analysis (table 6), hyponatremia (p=0.038), hyperkaliemia (p=0.005), high creatinine levels (p=0.049), hyperphosphoremia (p=0.006), and O2 saturation (p=0.01) were statistically significant. There was a trend for prolonged prothrombin time > 16,8 seconds (p=0.1). Age was significant in our cohort (p=0.008, HR=1.04 (1.01-1.07)). Then, in our multivariate model (table 6), age (p=0.002, HR=1.13 (1.04-1.22)), prothrombin time > 16,8 seconds (p=0.03, HR=4.62 (1.19-17.94)), hyponatremia (p=0.005, HR=6.99 (1.83-26.72)) and hyperkaliemia (p=0.01, HR=12.1 (1.66-87.75)) were independent prognostic factors.
Proposal prognostic score to predict early survival in COVID-19 patients
Based on the results of the multivariate analysis above, we proposed a simple prognostic score including sodium, potassium and prothrombin time. The calculation of this score is explained in Table 7. Basically, 0, 1 or 2 points are assigned to each of the 3 parameters, according to the weight of their respective HR, resulting in a score ranging from 0 to 6. Using ROC analysis, we proposed that a cut-off of 2 or more predicted a poor prognosis with a sensitivity of 80,7% and a specificity of 93,3% (table 8). Kaplan-Meier survival analysis of patients with the proposal score < or ≥ 2 showed significative difference in early overall survival (Figure 2, p=0,011; HR=0,17; (0.04- 0.67)).