The inclusion criteria were adults 18 years of age and older, patients in the emergency department or hospitalized, with acute respiratory failure, receiving ultrasound assessments, and on non-invasive ventilation. The exclusion criteria were studies in languages other than English, pregnant patients, post-surgical patients, patients with neuromuscular disorders, using ultrasound to determine if patient needs to be advanced to NIV, using ultrasound to evaluate if patient can be extubated or weaned from invasive mechanical ventilation, and if the study involved drug interventions. Two investigators (M.K. and V.D.) reviewed the titles and abstracts of all articles that met the inclusion and exclusion criteria, and full-text articles were obtained to verify that the criteria were met.

Several studies employ an ultrasonographic method of scoring lung pathology, namely the lung ultrasound score (LUS) (Table 1). A lung ultrasound score assessment is performed for multiple reasons, such as: perioperative oxygenation (31), predicting disease severity and mortality in ARDS (32), monitoring of lung aeration changes (after proning in ARDS, antimicrobial therapy in ventilator-associated pneumonia, selecting the ideal positive end-expiratory pressure, etc.) (33–35), weaning from mechanical ventilation (36), and as a prognosticator for respiratory failure (15, 18–20, 22, 28). There are multiple variations of the LUS, however arguably the more commonly utilized version of the LUS assessment consists of scanning a predetermined 6 regions per lung. Lung aeration of each region is graded between 0 to 3 depending on the ultrasound pattern visualized. In each region, points are allocated according to the following ultrasound pattern: normal = 0, well-defined B-lines = 1, coalescent B-lines = 2, and consolidation = 3. The total score for the LUS assessment therefore ranges from 0 to 36 (37) (Figure 2). A broader score, known as the integrated lung ultrasound score (i-LUS), was implemented during the coronavirus disease 2019 (COVID-19) pandemic that incorporates other important parameters seen with COVID-19 pneumonia. These additional parameters include: presence of pleural effusion (absent = 0, present = 1), presence of pericardial effusion (absent = 0, present = 1), measurement of IVC respiratory variation (<0–33%; absent = 0, present = 1), and diaphragm excursion (excursion >2 ± 0.5 cm was considered normal = 0, values below = 1) (18). Again, there are several other variations of the standard LUS (most of which are described in this present article) and these lung ultrasound score assessments have demonstrated positively in predicting outcomes for patients on NIV.

Ahmed et al. performed a prospective observational study of 50 ICU patients who presented with acute respiratory failure and were indicated for NIV (including COPD with exacerbation, pneumonia with no or mild secretions, acute lung injury manifested by PaO2/FiO2 < 200, and acute congestive heart failure with pulmonary edema). Lung ultrasound score, HACOR score, and serum lactate were recorded on admission and after 12 h from NIV application. To predict NIV failure, LUS with a cut-off value of 18 on admission showed sensitivity 77.0%, specificity 60.0% (p = 0.001). A cut-off value of 15 after 12 h NIV use showed a sensitivity of 87.0% and a specificity of 75.0% (p < 0.001) (15).

A similar study of 120 patients admitted for COVID-19 pneumonia showed that the optimal cutoff for the baseline LUS score was 18 (sensitivity 62%, specificity 74%). Both mortality and requirement for invasive mechanical ventilation were increased with a baseline LUS score > 18 compared to patients with baseline LUS score between 0 and 18. The hazard ratio of invasive mechanical ventilation or death for LUS score was 1.12 per point (p = 0.0008) (30). Dell’Aquila et al. also examined COVID-19 patients in a prospective study of 143 patients. They noted that in the survivor group, patients had a median i-LUS score of 16, while the score was 20 in the non-survivor group (interquartile range 12–20 vs. 15–24; p = 0.005) (18).

In a prospective cohort study of 85 patients with COVID-19 respiratory failure, Biasucci et al. showed that LUS (denoted as “LUSsc”; median score of 12 for NIV failure and median score of 6 for NIC success) as well as the amount of involved sonographic lung areas (denoted as “LUSq”; median areas involved were 6 for NIV failure and median areas involved were 3 for NIC success) were significantly higher in patients who failed NIV (p < 0.001) (28). This was also seen in a larger study of 280 patients with COVID-19 pneumonia by Ji et al., which demonstrated that patients with a high LUS score (defined as LUS > 12) were more likely to have poor outcomes including requiring invasive mechanical ventilation, higher incidence of ARDS, and higher mortality rates compared to the lowest LUS score group (defined as LUS 0–1), with a specificity and sensitivity of 90.5 and 91.9%, respectively (19). Blair et al. utilized the mean LUS (mLUS) score (ranging from 0 to 3) across 12 lung zones. The study revealed that every increase in mLUS score at enrollment was associated with disease progression to ICU-level of care (adjusted hazard ratio [aHR], 3.61; 95% confidence interval [CI], 1.27–10.2; p = 0.016) as well as 28-day mortality (aHR, 3.10; 95% CI, 1.29–7.50; p = 0.012) (22).

In addition to the lung ultrasound score, point of care ultrasonography can be used to rapidly assess for many pathologies including diaphragm dysfunction (DD) (Table 2). Similar to other ultrasound modalities, diaphragm-related ultrasound is noninvasive, safe, and can serve as a repeatable bedside tool. Patients with significant hypoxia who require NIV have a risk of developing diaphragmatic impairment which may poorly affect outcomes, including the need for invasive mechanical ventilation. Numerous studies have employed ultrasonography to assess functionality of the diaphragm to assist in predicting outcomes for patients on NIV (16, 17, 21, 23–27). These studies assessed diaphragm thickness (DT) or the variation of diaphragmatic thickness fraction (DTF) between end-inspiration and end-expiration. DTF is calculated from the following formula: [(Thickness at end-inspiration − Thickness at end-expiration)/(Thickness at end expiration)] × 100, see Figure 3. A DTF, also known as change in diaphragmatic thickness (ΔTdi), <20% is consistent with diaphragmatic dysfunction (39) (Figure 4). For sake of consistency, this article will use the term DTF with knowledge that it is interchangeable with ΔTdi.

Across the more recent studies which are highlighted in this article, diaphragm evaluation was performed on patients with an acute exacerbation of COPD (AECOPD) or with acute respiratory failure from SARS-COV2 infection. Marchioni et al. performed a prospective observational study using ultrasound to assess diaphragm dysfunction in patients with AECOPD and reported that patients with diaphragmatic dysfunction had a higher risk for NIV failure than those without DD (risk ratio = 4.4; p < 0.001). Change in diaphragm thickness (DTF) <20% during tidal volume was the predefined cutoff for identifying DD. The study investigated the correlation between ultrasound-assessed DD and the transdiaphragmatic pressure (Pdi) assessed using an invasive sniff maneuver [esophageal pressure (Pes) and gastric pressure (Pga) levels are recorded using balloon catheters before starting NIV; transdiaphragmatic generating pressure capacity (Pdi) was obtained by subtracting Pes from Pga (Pga – Pes) during a sniff maneuver], see Figure 5. It was reported that DTF highly correlated and had similar accuracy in identifying diaphragmatic dysfunction as with Pdi sniff (Pearson’s r = 0.81; p = 0.004). Lastly, the study demonstrated that DTF < 20% had better accuracy in predicting NIV failure than baseline pH value <7.25, as well as early changes in arterial pH and PaCO2 following initiation of NIV (AUCs 0.84, 0.51, 0.56, and 0.54, respectively; p < 0.0001) (17).

Cammarota et al. similarly reported diaphragmatic excursion assessment 2 h after NIV initiation was a better predictor of NIV failure than pH, PaCO2, and left expiratory diaphragmatic thickness. Diaphragmatic excursion was greater in NIV successes than in NIV failures at initiation of NIV (1.92 [1.22–2.54] cm vs. 1.00 [0.60–1.41] cm, p = 0.02) and at 2 h after initiation of NIV (1.99 [1.63–2.54] cm vs. 1.20 [0.79–1.41] cm, p = 0.008), respectively (27). Diaphragmatic dysfunction was also evaluated through diaphragmatic excursion by Barbariol et al. in ICU patients admitted for acute respiratory failure. Data was collected right before starting NIV as well as 1 h after initiating NIV. Diaphragmatic dysfunction was defined as a diaphragmatic excursion of less than 1.00 cm. Out of 47 patients, the patients without diaphragmatic dysfunction had about a 10% increase in NIV success than patients with diaphragmatic dysfunction, which was not statistically significant (p = 0.54). They also performed ROC analysis and noted that when using diaphragmatic excursion as a predictor of NIV response the area under the curve was 0.53; the best sensitivity (58.1%) and specificity (62.5%) was obtained with a diaphragmatic excursion cut-off of 1.37 cm (29).

Elsayed et al. conducted a prospective observational study where patients with acute exacerbation of COPD (AECOPD) were categorized into successful and failed NIV groups and DTF was measured. Readings of DTF in the successful NIV group were ≥ 33–38% (p < 0.001) while failed NIV had DTF values ≤16–18% (p < 0.001). In addition, cut-off value of DTF < 26–29% on both sides (left hemidiaphragm% - right hemidiaphragm%) was associated with NIV failure with 96.67% sensitivity, 80–82.22% specificity, 76.3–78.4% positive predictive value (PPV), and a 97.3–97.4% negative predictive value (NPV) (16). In a similar study by Kocyigit et al. on patients with AECOPD in the emergency department, DD (defined as DTF of <20% during spontaneous breathing) had a high sensitivity of 84.6% (95% CI:54.6–98.1), specificity of 91.5% (95% CI:79.6–97.6), PPV 73.3% (95% CI:51.2–87.8), and NPV 95.6% (95% CI:85.7–98.7) (24). Antenora et al. underwent a pilot study on the prevalence and clinical consequences of diaphragmatic dysfunction diagnosed by ultrasonography during AECOPD and reported that DD (defined as DTF < 20% during spontaneous breathing) was found to be strongly correlated with NIV failure (p < 0.001, R² = 0.27), longer ICU stay (p = 0.02, R² = 0.13), prolonged mechanical ventilation (p = 0.023, R² = 0.15), and need for tracheostomy (p = 0.006, R² = 0.20) (21). In a study by Mercurio et al. on patients with de-novo acute respiratory failure, cut-off values for DTF were explored that would accurately predict NIV failure. It was determined that the cut-off value of DTF < 36.3% significantly predicted NIV failure (p < 0.0001) with sensitivity of 71.7% (95% CI 56.5–84.0) and specificity of 94.3% (95% CI 80.8–99.3) (26).

In SARS-COV2 patients, Corradi et al. underwent a single-center pilot study that reported in their univariate logistic regression analysis that CPAP failure was significantly associated with a low DTF (odds ratio [OR]: 0.673; p < 0.001) and high respiratory rate (OR: 1.572; p < 0.001), but only DTF reached statistical significance at multivariate analysis (OR: 0.681; p < 0.001) (23). The same investigators performed a separate study and noted that patients on NIV who developed adverse outcomes had thinner diaphragms than those who did not (2.0 vs. 2.2 mm, p = 0.001) (25).