Pulse oximetry is commonly used in patients who are critically ill in order to help titrate oxygen delivery. However, there are concerns regarding the accuracy of pulse oximetry in patients who have a greater degree of skin pigmentation and who are critically ill [1, 2]. We have shown previously that pulse oximetry is not affected by ethnicity in patients with COVID-19 who were receiving non-invasive ventilation [3]. We aimed to investigate the accuracy of peripheral pulse oximetry in patients who were critically ill with COVID-19 pneumonitis and were receiving invasive mechanical ventilation and to determine if this was affected by patient ethnicity.

The single-centre retrospective observational study was deemed not to require ethical approval but was registered with the local clinical effectiveness unit. The methodology was identical to that described in our earlier study [3]. In brief, patients were identified by analysis of our critical care clinical information system (MetaVision ICU; iMDsoft^®, Tel Aviv, Israel). We compared paired measurements of arterial oxygen saturations measured by co-oximetry on arterial blood gas analysis (SaO₂) using a RAPIDpoint 500 analyser (Siemens Healthcare GmbH, Erlangen, Germany) and SpO₂ measurements performed as part of the B1x5M/P monitoring system (GE Healthcare, Chicago, IL, USA) using Nellcor™ reusable (Medtronic, Watford, UK) or disposable (Mindray, Huntingdon, UK) probes. We studied adult patients (aged ≥16 y) with COVID-19 pneumonitis who were receiving mechanical ventilation of their lungs. Only readings completed within the first 7 days following tracheal intubation were analysed. We did not study patients who required renal replacement therapy. Limits of agreement and bias between SaO₂ and SpO₂ measurements were assessed using Bland-Altman plots, with the difference between SaO₂ and the mean SpO₂ measurements plotted against the mean saturation measurement. The SpO₂ value used was the mean value over the 4-min period immediately preceding SaO₂ measurement. Bias was calculated as the mean difference between SaO₂ and SpO₂ measurements, limits of agreement were calculated as bias ± 1.96 SD (with correction made for within-subject variation) [4] and accuracy calculated using the following formula: √ (bias²+precison²). Data were analysed using Microsoft Excel (Microsoft Corporation, Redmond, WA, USA), R (version 4.1.0, https://www.r-project.org) and RStudio (v 1.4.1717, rstudio.com).

During the analysis period between 1 March 2020 and 31 October 2021, data from 178 patients, of which 134 (75%) were male, were available for analysis. Baseline patient characteristics were as follows: mean (SD) age 60 (14) y; mean (SD) BMI 30 (6.4) kg.m⁻²; and mean (SD) duration of stay 18 (15.7) days. In total, there were 24,626 paired SaO₂/SpO₂ measurements with mean (SD) SpO₂ of 94.3% (3.1) and mean (SD) SaO₂ of 93.7% (2.7). Patient ethnic groups were pooled as follows: South Asian (Asian British-Indian; Asian British-Pakistani; Asian British-other); Black (Black British-African; Black British-Caribbean; Black British-other); White (White-British; White-Irish; White-other); and other. The bias, limits of agreement and Bland-Altman plots for paired SaO₂ and SpO₂ measurements with respect to ethnicity are shown in Table 1 and Figure 1. Bias was greater in patients of non-White ethnic origin. Overall accuracy of pulse oximetry was 2.43%. Accuracy was 2.30%, 2.56% and 3.07% for patients of White, South Asian and Black ethnic origin, respectively. With respect to the ability of pulse oximetry to detect hypoxaemia, we calculated the proportion of SpO₂ measurements ≥92% that occurred with a paired SaO₂ reading ≤ 90%. Overall, this occurred in 2.16% (95%CI 1.97–2.34%) of measurements. The proportion of such readings was greater in patients of Black and South Asian ethnic origin compared with patients of White ethnicity – 5.33% (95%CI 4.32–6.35%) and 2.90% (95%CI 2.43–3.36%) vs. 3.25% (95%CI 2.98–3.52%), respectively.

Compared with our previous work involving patients with COVID-19 pneumonitis receiving non-invasive ventilation [3], we have shown that pulse oximetry is less accurate in patients who are receiving mechanical ventilation. This could be due to a greater severity of illness or the administration of vasoactive drugs such as propofol sedation and vasopressor therapy, all of which are more common in patients who require invasive ventilation. The small number of patients of Black ethnic origin in our cohort limit our ability to draw strong conclusions regarding the accuracy of SpO₂ measurement for the detection of hypoxaemia; our findings are, however, in line with previous work [2] and suggest an agreement with co-oximetry outside the ≤ 3.0% threshold recommended by the US Food and Drug Administration [5]. In addition, ethnicity was determined by patient self-identification and due to the retrospective nature of our study, we were not able to measure the melanin index of our patients’ skin [6]. We also only investigated a single device made by one manufacturer, and this limits the ability to extrapolate our findings to other monitors. Further work is needed to determine the precise impact of ethnicity on the accuracy of pulse oximetry [7]. In the interim, clinicians should have a low threshold for the direct measurement of SaO₂ by co-oximetry in patients with COVID-19 pneumonitis who are receiving mechanical ventilation.

脉搏血氧仪通常用于危重患者，以帮助滴定氧气输送。然而，对于皮肤色素沉着程度较高且病情危重的患者，脉搏血氧饱和度测定的准确性存在问题 [ 1, 2 ]。我们之前已经表明，在接受无创通气的 COVID-19 患者中，脉搏血氧饱和度不受种族影响 [ 3 ]。我们旨在调查重症 COVID-19 肺炎并正在接受有创机械通气的患者外周脉搏血氧饱和度的准确性，并确定这是否受患者种族的影响。

单中心回顾性观察研究被认为不需要伦理批准，但已在当地临床有效性部门注册。该方法与我们早期研究 [ 3 ] 中描述的方法相同。简而言之，通过分析我们的重症监护临床信息系统（MetaVision ICU；iMDsoft ^®，特拉维夫，以色列）确定了患者。我们比较了使用 RAPIDpoint 500 分析仪（Siemens Healthcare GmbH，Erlangen，Germany）和 SpO 2 对动脉血气分析 (SaO 2 )_测量的动脉血氧饱和度的配对测量值作为 B1x5M/P 监测系统（GE Healthcare，Chicago，IL，USA）的一部分，使用 Nellcor™ 可重复使用（Medtronic，Watford，UK）或一次性（Mindray，Huntingdon，UK）探头进行的测量。我们研究了接受肺部机械通气的 COVID-19 肺炎成年患者（年龄≥16 岁）。仅分析气管插管后前 7 天内完成的读数。我们没有研究需要肾脏替代治疗的患者。_{使用 Bland-Altman 图评估 SaO 2}和 SpO ₂测量值之间的一致性和偏差限制，其中 SaO ₂和平均 SpO ₂测量值之间的差异针对平均饱和度测量值绘制。血氧饱和度_{2使用的值是SaO 2}测量前4分钟的平均值。_{偏差计算为 SaO 2}和 SpO ₂测量值之间的平均差异，一致性限制计算为偏差 ± 1.96 SD（对受试者内部变化进行了校正）[ 4 ]，精度使用以下公式计算：√（偏差² +精度² )。使用 Microsoft Excel（Microsoft Corporation，Redmond，WA，USA）、R（版本 4.1.0，https://www.r-project.org）和 RStudio（v 1.4.1717，rstudio.com）分析数据。

在 2020 年 3 月 1 日至 2021 年 10 月 31 日的分析期间，来自 178 名患者的数据可供分析，其中 134 名（75%）为男性。基线患者特征如下：平均 (SD) 年龄 60 (14) 岁；平均 (SD) BMI 30 (6.4) kg.m ^-2；平均 (SD) 停留时间 18 (15.7) 天。总共有 24,626 对 SaO ₂ / SpO ₂测量，平均 (SD) SpO ₂为 94.3% (3.1) 和平均 (SD) SaO ₂93.7% (2.7)。患者种族分组如下：南亚裔（亚裔英裔印度裔；亚裔英裔巴基斯坦裔；亚裔英裔-其他亚裔）；黑人（黑人英属非洲人；黑人英属加勒比人；黑人英属其他人）；白色（White-British；White-Irish；White-other）；和别的。_{关于种族的配对 SaO 2}和 SpO ₂测量的偏差、一致性限制和 Bland-Altman 图如表 1 和图 1 所示。非白人种族血统的患者的偏差更大。脉搏血氧仪的总体准确率为 2.43%。白人、南亚和黑人种族患者的准确度分别为 2.30%、2.56% 和 3.07%。关于脉搏血氧仪检测低氧血症的能力，我们计算了 SpO _2的比例测量值≥92%，配对 SaO ₂读数≤90%。总体而言，这发生在 2.16% (95%CI 1.97–2.34%) 的测量中。与白人患者相比，黑人和南亚裔患者的此类读数比例更高——5.33% (95%CI 4.32–6.35%) 和 2.90% (95%CI 2.43–3.36%) vs. 3.25% （95%CI 2.98–3.52%）。

与我们之前涉及接受无创通气的 COVID-19 肺炎患者的工作相比 [ 3 ]，我们发现脉搏血氧仪在接受机械通气的患者中不太准确。这可能是由于疾病更严重或使用了血管活性药物（例如丙泊酚镇静剂和血管加压剂治疗），所有这些在需要有创通气的患者中更为常见。在我们的队列中，少数黑人族裔患者限制了我们就 SpO ₂测量检测低氧血症的准确性得出有力结论的能力；然而，我们的发现与之前的工作一致 [ 2] 并建议与美国食品和药物管理局推荐的 ≤ 3.0% 阈值之外的血氧饱和度达成协议 [ 5 ]。此外，种族是由患者的自我认同决定的，由于我们研究的回顾性，我们无法测量患者皮肤的黑色素指数 [ 6 ]。我们还只调查了一家制造商制造的单一设备，这限制了将我们的发现外推到其他显示器的能力。需要进一步的工作来确定种族对脉搏血氧仪准确性的精确影响 [ 7 ]。在此期间，临床医生应该对 SaO _{2的直接测量有一个较低的阈值}通过对接受机械通气的 COVID-19 肺炎患者进行血氧饱和度测定。