Outcomes
A flow chart of the outcomes of 378 patients with hypoxemic respiratory failure who were treated with HFNC is shown in Fig. 1. The HFNC success rate, defined as patients who did not require IMV, was 71.4% (n = 270; 95% CI 66.6–75.8) compared with 28.6% (n = 108; 95% CI 24.2–33.4) of patients who required IMV. Of the 270 patients who were successfully treated with HFNC, 262 patients (97.0%) were discharged, seven patients (2.6%) were referred, and one patient (0.4%) died. Among those 108 patients who required IMV, 61.1% (n = 66; 95% CI 51.5–69.9) were successfully extubated. Of those 66 patients who required IMV and were successfully extubated, 64 patients (97.0%) were discharged and one patient (1.5%) died. Furthermore, the proportion of patients with HFNC success increased over time from when HFNC treatment was started (Fig. 2). The median (interquartile range) number of days of HFNC administration was 13 (10–18) days in patients who had success compared with 2 (1–3) days in patients who then also required IMV.
When comparing patients who required IMV and those who did not require IMV, age and the presence of hypertension were significantly different (p < 0.0001 and p = 0.03, respectively). The Charlson comorbidity index categories also differed significantly between these two groups (p < 0.0001) (Table 1). Patients who received HNFC only were treated for a significantly longer period of time versus those who eventually required IMV (13 vs 2 days, respectively; p < 0.0001). The intensive care unit admission rate (7.0% vs 96.3%) and length of hospital stay (15 vs 26.5 days) were also significantly different between patients who did not require IMV and those who required IMV.
Laboratory results at admission and when HFNC was commenced are summarized in Table 1 and Online Resource 3 in the online data supplement. Patients who required IMV had less favorable laboratory parameters versus those who did not require IMV. Of note, there were significant differences at admission and when commencing HFNC in absolute lymphocyte counts (p < 0.01 and p < 0.0001, respectively). Although markers of inflammation (median D-dimer and ferritin levels) were not significantly different between patients who only required HFNC and those who also required IMV at admission (p = 0.32 and p = 0.69, respectively), they were significantly different when commencing HFNC (p < 0.01 and p = 0.03, respectively).
Rox index values were significantly higher at each time point in patients who did not require IMV versus those who required IMV (Fig. 3). In patients with HFNC success, Rox index values increased from 5.98 at baseline to 6.41, 6.83, 7.02, 7.37, 7.87, and 8.20 after 1, 2, 4, 6, 12, and 16 hours, respectively. In contrast, in patients with HFNC failure, Rox index values remained low from 5.40 at baseline to 5.70, 5.88, 5.76, 5.93, 5.74, and 5.62 after 1, 2, 4, 6, 12, and 16 hours, respectively. When comparing Rox index values between 2 hours and 16 hours, the difference was significant in the HFNC only group, but not in the HFNC and IMV group (p < 0.0001 and p = 0.91, respectively).
The changes in arterial blood gas parameters from 2 hours after HFNC to 16 hours after HFNC are summarized in Online Resource 4. In patients with HFNC success, the median (interquartile range) SPFI ratio increased from 135.7 (115.3, 160.0) after 2 hours to 158.4 (127.2, 192.2) after 16 hours of HFNC. In patients with HFNC failure, the median (interquartile range) SPFI ratio decreased from 115.0 (98.0, 140.0) after 2 hours to 110.4 (96.5, 134.9) after 16 hours of HFNC. When comparing SPFI ratios between 2 hours and 16 hours, the difference was only significant in the HFNC only group and not in the HFNC and IMV group (p < 0.0001 and p = 0.34, respectively).
Finally, we show that the CALL score at admission (adjusted hazard ratio 1.27; 95% CI 1.09–1.47; p < 0.01), Rox index at 1 hour (adjusted hazard ratio 0.82; 95% CI 0.70–0.96; p = 0.02), and absence of treatment with steroids (adjusted hazard ratio 0.34; 95% CI 0.19–0.62; p < 0.0001) were all significant predictors of HFNC failure (Table 2).
Table 2
Predictors of HFNC failure (n = 238)
Variable | Adjusted HR (95% CI) | p-value |
Age (per year increase) | 1.003 (0.99–1.01) | 0.53 |
Male vs female | 0.73 (0.41–1.32) | 0.28 |
CALL score (per 1-point increase) | 1.27 (1.09–1.47) | < 0.01 |
No diabetes (vs uncontrolled) | 0.81 (0.43–1.51) | 0.50 |
Diabetes (vs uncontrolled) | 0.46 (0.16–1.29) | 0.14 |
Hypertension (vs no hypertension) | 1.08 (0.47–2.50) | 0.85 |
Overweight (vs normal weight) | 1.05 (0.38–2.93) | 0.92 |
Obesity (vs normal weight) | 1.49 (0.55–4.01) | 0.43 |
Rox index 1 hour (per 1-point increase) | 0.82 (0.70–0.96) | 0.02 |
Treatment with steroids (vs no treatment) | 0.34 (0.19–0.62) | < 0.0001 |
Absolute lymphocytes (per unit increase) | 0.99 (0.99–1.00) | 0.24 |
D-dimer 550–1000 (vs normal levels) | 1.23 (0.66–2.27) | 0.51 |
D-dimer > 1000–1500 (vs normal levels) | 1.52 (0.61–3.76) | 0.37 |
D-dimer > 1500 (vs normal levels) | 0.31 (0.09–1.09) | 0.07 |
PAFI (per unit increase) | 0.99 (0.99–1.00) | 0.94 |
Lactate–mmol/L (per unit increase) | 1.25 (0.80–1.97) | 0.33 |
PaCO2 (per unit increase) | 1.04 (0.98–1.11) | 0.18 |
CALL, comorbidity–age–lymphocyte–and LDH; CI, confidence interval; HFNC, high-flow nasal cannula; HR, hazard ratio; PAFI, ratio of PaO2 over FiO2 |
Supplementary material |
Online Resource 1. Clinical protocol for the management of hypoxemic respiratory failure in the Temporary COVID-19 Hospital |
Abbreviations: FiO2, fractional inspired oxygen; HFNC, high-flow nasal cannula; ICU, intensive care unit; RR–respiratory rate; RX, radiography; NEWS, National Early Warning Score; PAFI, ratio of PaO2 over FIO2; SpO2, oxygen saturation |
Online Resource 2. Algorithms for HFNC follow up and admission to the intensive care unit |
Abbreviations: HFNC, high-flow nasal cannula; ICU, intensive care unit; SpO2, oxygen saturation; FiO2, fractional inspired oxygen; PAFI, ratio of PaO2 over FIO2 |
Online Resource 3. Disease severity risk markers at admission and at start of HFNC |