B96-06 The “Streetlight Effect” in Occult Hypoxemia: Racial Disparities Are Most Striking in Patients That Are Typically Excluded From Regulatory Studies of Oximeters

Rationale We previously validated a noninvasive method to estimate P/F ratios (ePFRs) and used it to reveal a signature of occult hypoxemia. Black patients appeared better oxygenated (right-shifted cumulative distribution of ePFRs) (Fig.1A), but clinically deteriorated as if they were worse oxygenated (Fig.1B). As the only method that can estimate occult hypoxemia across the full inpatient population, this approach enables the exploration of factors that exacerbate racially disparities in occult hypoxemia. Methods We retrospectively analyzed all COVID-19 hospital encounters at one academic medical center in the first year of the pandemic (n = 1100). We measured the composite outcome (death or ICU transfer within 24 hours), demographics (age, sex, race), comorbidities (Charlson Comorbidity Index, baseline Sequential Organ Failure Assessment), oxygenation (ePFR), and two key factors that can affect the photoplethysmography waveform (mean arterial pressure [MAP] and hemoglobin [HGB]). Racial disparities (i.e., shifts in cumulative distributions of ePFR) were quantified using the Kolmogorov-Smirnov distance (KSD) and the impact of ePFR and race on the primary outcome was modeled using multivariable logistic regression, with 95% confidence intervals calculated via 1000-fold bootstrap resampling by hospital encounter. Results In patients with below-median hemoglobin (< 11g/dL) and blood pressure (MAP < 89mmHg), the signature of racially disparate occult hypoxemia was markedly stronger: ePFR distributions were widely right-shifted (KSD 0.264 [0.259-0.269] for low-HGB; 0.334 [0.330-0.339] for low MAP vs 0.170 [0.166-0.173] overall); race was significantly associated with outcomes (aOR 3.2 [1.5-7.1] for low-HGB and 2.5 [1.1-5.4] for low-MAP). In contrast, among those with above-median hemoglobin and blood pressure, we did not detect racially disparate occult hypoxemia: oxygenation distributions were narrowly right-shifted (KSD 0.089 [0.086-0.095] for high-HGB and 0.087 [0.084-0.091] for high-MAP), and race was not significantly associated with outcomes (aOR 1.4 [0.8-2.6] for high-HGB and 1.5 [0.8-2.9] for high-MAP). Conclusions Our findings show that low hemoglobin and low blood pressure interact with skin tone to amplify the racial disparities in occult hypoxemia. These conditions are common in the sick hospital populations where oximeters are frequently utilized; but they are absent in the healthy regulatory-study populations where oximeters are frequently scrutinized. This reflects the “streetlight effect”; studies conducted where it’s easiest - under the figurative streetlight - rather than the challenging places where the problem truly lies. Occult hypoxemia research must shift toward real-world hospital settings, where complex webs of physiological derangements are prevalent and the consequences of missed hypoxemia are most severe.