Remote Pulse Oximetry Detects Silent Hypoxia but Shows No Mortality Benefit in Honduras COVID-19 Trial

Could Remote Monitoring Transform COVID-19 Care in LMICs?

A pragmatic, randomized trial conducted in Honduras has investigated whether adding self-administered pulse oximetry to remote monitoring can reduce morbidity and mortality among non-hospitalized COVID-19 patients at high risk for adverse outcomes. The study, which enrolled 1,821 participants between March 2022 and January 2023, represents the first randomized trial of its kind in a lower-middle income country (LMIC). While the findings revealed important insights about silent hypoxia detection, the trial could not demonstrate a mortality benefit from pulse oximeter use, largely due to the unexpectedly low incidence of severe outcomes among participants.

The trial utilized a cluster-randomized design with temporal clustering, where participants enrolled on a given day were assigned to one of four intervention packages in a predetermined sequence. This approach helped prevent contamination between study arms, as household members often attended triage centers together. Of the total cohort, 924 participants were randomized to receive remote monitoring only, while 897 received remote monitoring plus a pulse oximeter. The median participant age was 61 years, with 60.8% being female and 91.4% reporting at least one high-risk comorbidity. Notably, the study population had high COVID-19 vaccination coverage, with 98.9% having received at least one dose and 90.2% having received three or more doses. Participants were enrolled at an average of 2.9 days after symptom onset.

The remote monitoring protocol involved daily phone calls from study-employed nurses who administered standardized questionnaires about warning signs including blue/grey lips or nails, chest pain, difficulty breathing, new confusion, or rapid breathing. For participants in the pulse oximetry arm, SpO₂ readings of 94% or lower also triggered referrals for clinical evaluation. The primary outcome was the difference in mortality between study arms, with secondary outcomes including risk of referral, SpO₂ percentages upon presentation for additional care, hospitalization rates, and need for intensive interventions. However, the trial observed only one death across both arms, making it impossible to assess the primary mortality endpoint. This unexpectedly low mortality rate likely reflects the changing landscape of COVID-19 during the study period, including high vaccination rates and the shift from Delta to Omicron variants, which are associated with less severe disease.

How Does Pulse Oximetry Influence Patient Outcomes?

Despite limitations in assessing the primary outcome, the trial yielded valuable findings regarding hypoxia detection and referral patterns. Among participants who received pulse oximeters, 35.1% recorded at least one SpO₂ reading of 94% or lower during the monitoring period. Importantly, 59.4% of these participants with low oxygen saturation did not report rapid breathing, suggesting that pulse oximetry enabled detection of silent hypoxia in a substantial proportion of cases. This finding aligns with the theoretical basis for the intervention, as silent hypoxia has been identified as a concerning feature of COVID-19 that can precede clinical deterioration. The pulse oximetry arm experienced significantly more referrals for additional clinical evaluation compared to the remote monitoring-only arm (6.91% vs. 4.23%, odds ratio 1.73, p=0.009), directly attributable to the detection of low oxygen saturation readings. However, this increase in referrals did not translate to a significant difference in hospitalization rates between the two arms (OR 1.6, p=0.40).

Multivariable analyses revealed several factors associated with increased referral risk across the entire study population, including underweight BMI (OR 5.93), chest pain (OR 3.04), chills (OR 2.27), diarrhea (OR 2.08), and shortness of breath at enrollment (OR 3.02). Among participants in the pulse oximetry arm, the risk factors for recording low oxygen saturation included increased age (OR 1.03 per year), male gender (OR 1.75), obesity (OR 2.03), fatigue at enrollment (OR 1.54), and reporting rapid respiratory rate during monitoring (OR 4.09). Interestingly, for each percentage point higher a participant's baseline SpO₂ was at enrollment, their odds of reporting hypoxia during monitoring decreased by 25% (OR 0.75). These findings could help identify which patients might benefit most from pulse oximetry monitoring in resource-limited settings where universal distribution is not feasible.

The study achieved remarkable adherence to the monitoring protocol, with 97.7% of planned daily monitoring calls (12,307 out of 12,602) successfully completed. Participants received an average of 6.9 monitoring calls during their enrollment period. During these calls, chest pain (2.4%) and difficulty breathing (1.3%) were the most commonly reported warning signs. The trial was conducted during sequential waves of transmission, with most participants enrolled between June-August 2022 and December 2022-January 2023, capturing the evolving nature of the pandemic in Honduras. As of August 2023, Honduras had reported 474,566 COVID-19 cases and 11,127 deaths (2.3% case fatality rate) to the World Health Organization.

Could These Findings Shape Future Clinical Protocols?

The study has several important limitations that warrant consideration. The primary limitation was the low incidence of adverse outcomes, which constrained the ability to draw conclusions about the intervention's impact on mortality and severe morbidity. Additionally, inclusion criteria were slightly modified after the study began, potentially affecting the participant pool. The open-label design may have introduced performance bias, as healthcare providers' knowledge of intervention assignment could have influenced clinical decision-making. Furthermore, pulse oximeters have been found to systematically overestimate oxygen saturation in individuals with darker skin tones, which could have led to misclassification of hypoxia among some participants. This measurement bias represents an important consideration for implementing pulse oximetry-based interventions in diverse populations, potentially requiring population-specific calibration or alternative technologies to ensure equitable healthcare delivery.

The findings from this trial join an inconclusive body of evidence regarding the impact of self-administered pulse oximetry on COVID-19 outcomes. Previous studies have shown mixed results, with both positive and null findings reported in different contexts. The effectiveness of pulse oximetry in reducing morbidity and mortality may depend on various factors, including population characteristics, healthcare system capacity, and the severity of circulating SARS-CoV-2 variants. While this study could not demonstrate a mortality benefit, the detection of silent hypoxia in a substantial proportion of participants suggests potential value in specific contexts. Future research might need to focus on retrospective, population-level analyses to identify significant adverse outcomes and draw more definitive conclusions about the utility of pulse oximetry in remote monitoring protocols for COVID-19 and other respiratory conditions.

Could the inclusion of pulse oximetry in remote monitoring protocols be more beneficial in settings with lower vaccination rates or during waves of more virulent SARS-CoV-2 variants? How might the threshold for clinical referral based on pulse oximeter readings be optimized to balance early intervention with healthcare system capacity, particularly in resource-constrained settings? These questions remain important considerations for healthcare systems developing protocols for managing high-risk patients during respiratory disease outbreaks, especially in contexts where hospital capacity is limited and early intervention could potentially prevent severe outcomes.

How Was the Trial Adapted for Real-World Conditions?

Regarding the geographic context of the study, the trial was conducted at four SESAL-operated COVID-19 triage centers and one hospital-based COVID-19 screening site operated by the Honduran Social Security Institute in Tegucigalpa and Comayagüela, Honduras. This urban setting provided important insights into the feasibility of implementing remote monitoring protocols in an LMIC environment. Interestingly, the study found that residing in Tegucigalpa (rather than Comayagüela) was associated with increased odds of hospitalization (OR 5.18), which led researchers to investigate potential explanations. Further analysis revealed relationships between residing in Comayagüela and both lower educational attainment and lower SpO₂ at enrollment, suggesting potential disparities in baseline health status, healthcare access, or healthcare-seeking behaviors between the two locations.

The study design incorporated several practical elements to enhance its real-world applicability. In addition to the pulse oximetry intervention, all participants received personal protective equipment to prevent disease transmission, including N95 respirators for study participants and surgical masks for caregivers. Half of the participants also received alcohol-based hand rub (ABHR), though analysis revealed no clinical effect associated with ABHR distribution. This two-by-two factorial design allowed researchers to efficiently evaluate multiple interventions simultaneously. Participants assigned to the pulse oximetry arm received either MedLine Soft Touch Fingertip Pulse Oximeter or CuraPlex Fingertip Pulse Oximeter, both FDA-approved devices, along with education on proper use and pictorial instructions for recording measurements.

The monitoring process was comprehensive and well-structured. Study-employed nurses called participants daily, starting the day after enrollment. If warning signs were reported, nurses referred participants for clinical evaluation at the enrollment site. Participants were instructed to seek medical evaluation if warning signs arose outside of daily monitoring calls. Phone-based monitoring continued until ten days post symptom onset, unless the participant was admitted to a hospital for more than 24 hours or reported fever, in which case monitoring continued until they were fever-free for 24 hours. This protocol achieved high completion rates, with 97.1% of participants (1,768 out of 1,821) completing study monitoring per protocol.

What Do the Final Data and Analyses Tell Us?

One notable aspect of the trial was the high prevalence of comorbidities among participants. The most frequently reported conditions were hypertension (63.7%), obesity (39.3%), and diabetes (30.2%). These high-risk conditions were part of the inclusion criteria, which initially targeted individuals aged 60 years or older, or those aged 45-59 years with at least one CDC-defined high-risk condition. The inclusion criteria were later expanded to add hypertension as a high-risk condition, aligning with updated CDC recommendations. This focus on high-risk individuals was based on assessments indicating greater likelihood of adverse COVID-19 outcomes among older adults, people with certain underlying medical conditions, and those experiencing barriers to healthcare access.

The statistical approach employed an intention-to-treat analysis, with all randomized participants included regardless of protocol adherence. Sample size calculations were based on expected event rates from the Delta variant wave, targeting detection of a 3.3% absolute difference in mortality with 80% power. The largest calculated sample size (n=1,862) was selected as the trial target. Fisher's exact test compared proportions between study arms, while logistic regression estimated odds ratios for outcomes. Multivariable models included variables with p<0.10 significance in univariable analysis, with p<0.05 considered significant in the final models.

Safety monitoring was an integral component of the trial design. No adverse events or harms related to remote monitoring or pulse oximeter use were reported throughout the study period. The intervention was considered low risk, and the team implementing the trial continuously supervised safety while conducting several interim analyses. The study received approval from appropriate Institutional Review Boards, including the Massachusetts General Brigham Institutional Review Board and the Autonomous University of Honduras, with secondary data analysis approved by the Johns Hopkins Bloomberg School of Public Health Institutional Review Board.

The study's findings regarding silent hypoxia deserve particular attention. Of the 315 participants who recorded at least one SpO₂ reading ≤94%, 128 (40.6%) had elevated respiratory rates at enrollment or reported rapid breathing during monitoring. This means that pulse oximeters enabled the identification of silent hypoxia—low oxygen levels without corresponding respiratory symptoms—in 187 participants (59.4% of those who ever experienced hypoxia). However, the degree of hypoxia was generally mild, with most readings between 90-93%, which may explain why this detection did not translate to improved clinical outcomes.

In summary, while this pragmatic trial could not demonstrate a mortality benefit from adding pulse oximetry to remote monitoring of high-risk COVID-19 patients, it provides valuable insights into the feasibility and potential utility of such interventions in LMIC settings. The high adherence to the monitoring protocol, successful detection of silent hypoxia, and identification of risk factors for hypoxia and referral could inform future approaches to pandemic response, particularly in resource-constrained environments. As we continue to navigate COVID-19 and prepare for future respiratory disease outbreaks, the lessons from this trial may help optimize remote monitoring strategies to better identify and support patients at greatest risk for adverse outcomes.

Summary

A pragmatic randomized trial conducted in Honduras between March 2022 and January 2023 investigated whether adding self-administered pulse oximetry to remote monitoring could reduce morbidity and mortality among 1,821 non-hospitalized COVID-19 patients at high risk for adverse outcomes. This first-of-its-kind study in a lower-middle income country used a cluster-randomized design, assigning 924 participants to remote monitoring only and 897 to remote monitoring plus pulse oximetry. The trial population had a median age of 61 years, high vaccination coverage (98.9% with at least one dose), and prevalent comorbidities including hypertension (63.7%), obesity (39.3%), and diabetes (30.2%). Daily phone monitoring by study nurses achieved remarkable 97.7% adherence rates. While the study could not demonstrate mortality benefits due to unexpectedly low adverse outcomes (only one death across both arms), it revealed important findings about silent hypoxia detection. Among pulse oximetry users, 35.1% recorded oxygen saturation levels of 94% or lower, with 59.4% of these cases representing silent hypoxia without accompanying rapid breathing symptoms. The pulse oximetry arm experienced significantly more clinical referrals (6.91% vs. 4.23%), though this did not translate to significantly different hospitalization rates. Risk factors for hypoxia included increased age, male gender, obesity, and baseline fatigue. The low mortality likely reflected high vaccination rates and the shift from Delta to Omicron variants during the study period. Despite limitations including low adverse event incidence and potential measurement bias in individuals with darker skin tones, the study demonstrates the feasibility of implementing remote monitoring protocols in resource-limited settings and provides valuable insights for optimizing pandemic response strategies in lower-middle income countries.