Waiting to Exhale: The Feasibility and Appropriateness of Home Blood Oxygen Monitoring in Oncology Patients Post Hospital Discharge

Abstract

Purpose:

Pulse oximetry remote patient monitoring (RPM) post hospital discharge increased during the COVID-19 pandemic as patients and providers sought to limit in-person encounters and provide more care in the home. However, there is limited evidence on the feasibility and appropriateness of pulse oximetry RPM in patients with cancer after hospital discharge.

Methods:

This feasibility study enrolled oncology patients discharged after an unexpected admission at Memorial Sloan Kettering Cancer Center from October 2020 to July 2021. Patients were asked to measure their blood oxygen level daily during the hours of 9 am to 5 pm during a 10-day monitoring period post hospitalization. An automated system alerted clinicians to blood oxygen levels below 93.0%. We evaluated the feasibility (>50.0% of patients providing at least one measurement from home) and appropriateness (>50.0% of alerts leading to a clinically meaningful patient interaction) of pulse oximetry RPM.

Results:

Sixty-two patients were enrolled in the study with 53.2% female patients and a median age of 68 years. The most prevalent malignancy was thoracic (62.9%). The feasibility metric was met with forty-five patients (72.6%, 45/62) providing blood oxygen levels at least once during the 10-day monitoring program. The appropriateness threshold was not met; of the 121 alerts, only 39.7% (48 alerts) were linked to a clinically meaningful interaction.

Conclusions:

This feasibility study showed that while patients with cancer were willing to measure blood oxygen levels at home, most alerts did not result in meaningful clinical interactions. There is a need for improved patient support systems and logistical infrastructure to support appropriate use of RPM at home.

Introduction

Remote patient monitoring (RPM) technologies with pulse oximetry have been used for many years in post hospitalization care of chronic diseases such as chronic obstructive pulmonary disease (COPD), heart failure, and other high-risk medical conditions.^1–4 There is much heterogeneity regarding the implementation of RPM interventions with most having components of either automated or patient report of vital signs from the home along with other symptom reports. There are also differences in the frequency of monitoring, the duration of monitoring, the alert thresholds, and the clinician response to alerts. While calling attention to this variation, a recent systematic review found that RPM interventions employing pulse oximetry may reduce acute care use in patients with conditions including cardiovascular disease and COPD.⁵

The COVID-19 pandemic further accelerated the adoption of home pulse oximetry monitoring given concern for unrecognized hypoxemia as a potential cause of morbidity and mortality.⁶ In addition, home pulse oximetry was seen as a means by which to increase hospital capacity by allowing for earlier patient discharges and to support clinical triage to remotely assess the need for inpatient evaluation or admission.⁷ Early studies during the COVID-19 pandemic demonstrated that home pulse oximetry monitoring improved outcomes including mortality,⁸ decreased acute care utilization,⁹ and were cost effective.¹⁰ Further, these programs had high patient satisfaction scores and were seen as patient-centric as resources were being deployed to keep patients at home and out of the hospital.^11–13 In a cohort of 257 oncology patients diagnosed with COVID-19 either in the outpatient setting or upon hospital discharge enrolled in a pulse oximetry RPM program, 91% felt participation in the program was helpful and 59% perceived the program as preventing emergency department or urgent care center visits.¹¹ During the pandemic, the Centers for Medicaid and Medicare modified RPM billing requirements to allow for RPM reimbursement for a minimum of 2 days of monitoring for all patients rather than the standard 16-day monitoring requirement further enabling the growth of RPM interventions.¹⁴

In addition to COVID-19, dyspnea due to other pulmonary conditions remains a significant cause for acute care visits in patients with cancer.¹⁵ Remote oxygen monitoring after hospital discharge for oncology patients with pulmonary conditions could enable improved symptom management at home, thus preventing readmission, or facilitating triage for emergency room evaluation of patients with deteriorating symptoms.¹⁶ However, few studies have focused on pulse oximetry monitoring in discharged patients with cancer and feasibility and clinical appropriateness post hospital discharge has not been established. Further, there is a need for better understanding of the mechanism for how this technology can lead to clinically meaningful patient interactions, including interventions in the home. We hypothesized that pulse oximetry RPM would be feasible and clinically appropriate in oncology patients after hospital discharge.

Methods and Materials

Study Design and Participants

This study enrolled adult patients with cancer who were discharged after an unexpected admission at Memorial Sloan Kettering Cancer Center (MSK) between October 2020 to July 2021. Patients were required to have a clinical indication for blood oxygen monitoring including thoracic disease progression, pneumonia, pneumonitis, or pleural effusion. This program was reviewed by the MSK Institutional Review Board and determined to be an exempt protocol as per 45 CFR 46.104(d)(2),(i),(ii). The present report follows the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cohort studies.¹⁷

Eligible patients were approached at the time of hospital discharge regarding study enrollment. Patients that consented to participation were provided with a pulse oximeter, onboarded by a nurse, and shown an educational video on how to use the pulse oximeter prior to hospital discharge. The components of the nurse onboarding process included a program overview, introduction to the technology interface, and emphasized the role of home oxygen monitoring in patient care. Once onboarded, patients entered the monitoring phase and were required to measure their oxygen saturation at least once a day between 9 AM and 5 PM during a 10-day monitoring period. A 10-day monitoring period was chosen as this is the median time to readmission for medical oncology patients and thus a critical window to provide early clinical intervention as necessary.^18,19

Alert Analysis

Pulse oximetry measurements were captured and transmitted to a central electronic database. An automated system triggered an alert to the patient’s clinical team if the blood oxygen level was less than 93.0%. This cutoff was chosen because blood oxygen levels of 92.0% or below may be a sign of clinical deterioration for patients with underlying pulmonary conditions.²⁰ These alerts interfaced with the institution’s electronic health record and were incorporated into clinical notes related to the management of enrolled patients. The patient’s primary nurse and oncologist would respond to alerts between the hours of 9 AM – 5 PM and monitoring of alerts was provided on nights and weekends by an after-hours triage team.

Retrospective chart review was subsequently performed on all alerts to determine the action taken, which was then classified into six different categories: 1) clinical management at home, 2) in-person or telehealth visit, 3) acute care referral, 4) no action, 5) missed, or 6) technical issue/device trouble shooting (Table 1). These designations were determined via review of chart documentation in the electronic health record and independently confirmed by two reviewers (SL, SDD, JH, BD).

Table 1.

Categorization of Alerts

Category	Description	Documentation Example
Clinical management at home	Nursing or physician led symptom education and adjusting supplemental oxygen; medication or oxygen prescription; clinical status check	Spoke with patient and spouse. His oxygen (O2) was only on 1 liter per minute (LPM) and he had not yet drained his PleurX. O2 elevated to 3LPM and .475ml drained. O2 improved to 92%. Patient wanted to lower O2 back to 1LPM. Requested that he leave it at 3LPM for an hour, then re-check his saturation level.
In-person or telehealth visit	A new visit is planned or scheduled	Spoke with registered nurse (RN) at Visiting Nurse Service (VNS) - She reports crackles to the right upper lobe, 90% blood oxygen reading, blood pressure 148/64, heart rate 99. VNS confirms patient is not in any distress. He will have a telemedicine visit with [oncologist – name redacted] today who will further assess. VNS agrees with plan and will relay to patient’s family.
Acute care referral	Clinical recommendation for a patient to seek medical attention at an urgent care center or emergency department	Called wife and patient is more short of breath today with minimal activity stating blood oxygen level is in low 80’s even with home O2. Patient was just discharged from hospital yesterday. Instructed to come back to urgent care center.
No action	Lack of response from the patient after nursing or physician communication or no change in plan undertaken based on the alert	Called to check in, spoke with patient’s sister. They advised all is stable, no change. They know to call with any concerns.
Missed	An absence of chart documentation on communication from a nurse and/or physician
Technical issue/device trouble shooting	Inaccurate reading due to user error or device malfunctioning	Call due to low O2 alert of 87%. Daughter states machine provided has never worked properly. She has 2 other pulse oximeters that read 94% on 4LPM, denies shortness of breath, and patient shows no signs of distress.

Feasibility and Clinical Appropriateness Outcomes

This study evaluated feasibility and clinical appropriateness of pulse oximetry RPM in oncology patients post hospital discharge. The feasibility metric was defined by patient adherence with measuring blood oxygen level at home. Adherence is a well-established metric that has been evaluated in other studies examining RPM technologies in oncology.^21,22 A patient was deemed adherent if they measured their blood oxygen level at least once during the 10-day monitoring period. A priori, we defined feasibility as 50.0% or greater of enrolled patients would measure their blood oxygen level at least once during the 10-day monitoring period. The overall percent of days adherent for each patient is also reported along with the trend in reporting during the 10-day monitoring period. A 50.0% adherence rate was chosen as this has been a threshold for feasibility for other studies evaluating remote symptom monitoring for oncology patients using electronic patient-reported outcomes (ePROs).^19,23 Research on oncology patient engagement with ePROs in ambulatory care clinics find that 35–73% of patients complete at least one ePRO in the monitoring period.^23–27 For example, in the National Cancer Institute funded Electronic Symptom Management (eSyM) study of 25,358 patients, 48% of patients used eSyM at least once to report symptoms.²³ Clinical appropriateness was defined by >50.0% of all alerts leading to a clinically meaningful patient interaction. This threshold was chosen a priori by our clinical team, composed of oncologists and oncology nurses in consultation with our Patient and Family Advisory Council for Quality. A clinically meaningful patient interaction included the sub-classifications of 1) clinical management at home, 2) in-person or telehealth visit, or 3) acute care referral.

Statistical Analysis

Descriptive statistics summarize patients’ characteristics and were reported as frequencies and proportions for categorical variables and as median and range for continuous variables. Data for the feasibility and appropriateness outcomes were analyzed according to frequency and percentage for qualitative variables and by median and range for quantitative variables. Statistical analyses were performed with Graphpad (version 9.5).

Results

Clinical Characteristics

A total of 62 patients were enrolled in this study (Table 2). The approach-to-consent rate was 100%. The median age of respondents was 68 years (range, 18–91 years) and 53.2% were female. The majority of patients were White (69.4%), 12.9% were Asian, and 6.5% were Black. The most prevalent malignancy was thoracic cancer (62.9%), followed by breast cancer (19.4%) and gastrointestinal cancer (14.5%). The most common reasons patients were discharged with pulse oximeters were for progression of disease in the thoracic cavity (30.6%, 19/62), pneumonia or pneumonitis (30.6%, 19/62), and other, including recurrent pleural effusion (38.7%, 24/62).

Table 2.

The Baseline Characteristic of Enrolled Patients

Characteristic	No. (%)
Sex
Male	29 (46.8)
Female	33 (53.2)
Age
Median, Range	68 (18–91)
Race
White	43 (69.4)
Asian	8 (12.9)
Black	4 (6.5)
Unspecified	7 (11.3)
Primary Cancer
Thoracic	39 (62.9)
Breast	12 (19.4)
Gastrointestinal	9 (14.5)
Genitourinary	1 (1.6)
Head and neck	1 (1.6)

Feasibility of Home Oxygen Monitoring

Forty-five patients (72.6%, 45/62) provided blood oxygen levels at least once during the 10-day monitoring program. Of those 45 patients, the median number of days they provided a blood oxygen level was 4 (Range: 1 – 10). Across the 62 patients in the cohort, 22 (35.5%) provided a blood oxygen measure for 5 or more days during the monitoring period (Figure 1). Home blood oxygen reporting declined with time in the program. On the first day of the monitoring period, blood oxygen levels were received from 73% of patients; this rate decreased over time to 22% of patients by the 10th day (Figure 1). At the individual reporting level, 7 patients had daily blood oxygen level measurements throughout the 10-day monitoring program, whereas 13 patients provided measurements for only 1 day. (Figure 2).

Figure 1.

Overall compliance rate by day of enrollment during the 10-day monitoring period. Compliance indicates a pulse oximetry measurement between the hours of 9 AM – 5 PM local time.

Figure 2.

The individual reported blood oxygen level throughout the 10-day monitoring program. Patient numbers are denoted along the x-axis (ex. p1 with 80% indicated that patient 1 reported for 8 out of 10 days during the 10-day monitoring period). The dashed line indicates the median compliance rate (40%). Patients who did not measure blood oxygen level are not included in the figure.

Clinical Appropriateness of Home Oxygen Monitoring

Blood oxygen levels lower than 93% led to 121 clinical alerts. There were 277 blood oxygen readings at levels of 93% or higher that appropriately did not lead to clinical alerts. Among these 121 clinical alerts, 48 alerts (39.7%) led to a clinically meaningful patient interaction, with 1 acute care referral, 45 instances of clinical management at home, and 2 instances of a visit with a provider being scheduled (Figure 3). Conversely, 73 alerts (60.3%) did not lead to a clinically meaningful patient interaction. These were either due to a technical issue in 29 (23.9%) instances, were missed in 35 (28.9%) instances, or led to no clinical action in 9 (7.4%) instances. For the alert that required an acute care referral, the patient was referred to a local emergency room and was found to have a worsening pleural effusion and pneumonia requiring inpatient management. Of the forty-five alerts leading to clinical management at home, this included medication prescription and education on adjusting oxygen supplementation and managing pleural effusion drains. Chart review of telephone encounter notes revealed that this form of at home management was sufficient to address the underlying clinical issue without an acute care visit.

Figure 3.

The composition of pulse oximeter alerts. A pie graph illustrates the breakdown of how alerts were addressed.

We next identified which patients had pleural effusion drains at time of discharge and explored clinically appropriate alerts in this subgroup. Of the 62 patient cases, 6 (9.7%) had pleural effusion drains, 4 of which were placed in the index admission. All 6 (100%) of these patients provided blood oxygen levels at least once during the monitoring program. They had a total of 17 clinical alerts, 12 of which (70.6%) led to a clinically meaningful patient interaction.

Discussion

This study assessed the feasibility and appropriateness of remote home oxygen monitoring among discharged patients with advanced cancer. This study did meet its predefined feasibility endpoint; we showed that over half of patients are willing to measure blood oxygen at home with 72.6% of patients providing at least 1 measurement during the 10-day home monitoring period. Standardization of the blood oxygen threshold for alerting at lower than 93% allowed automatic clinical alerts that were incorporated into the EHR with the goal of concentrating care team attention on patients in need. Home oxygen monitoring is thought to help identify silent hypoxia and promote proactive interventions. However, our study did not meet its appropriateness endpoint as we found that fewer than 50% of alerts resulted in a clinically meaningful interaction between the provider and patient. Notably, although a small subgroup of our cohort, patients with pleural effusion drains had a higher rate of clinically meaningful interactions with pulse oximetry RPM.

The findings from this study highlight several challenges with pulse oximetry RPM in the home setting. While onboarding occurred in-person at discharge and we dedicated much time and resources in developing our onboarding materials, patients still struggled with technical challenges at home. These included issues with device functionality, such as inaccurate readings. In the future, developing or partnering with an outside vendor to provide a digital navigator could potentially assist patients and their caregivers with overcoming these technology issues once home.²⁸ There has been significant innovative activity in this space with Best Buy recently announcing a collaboration with Geisinger to launch a pilot program to send Geek Squad members into patients’ homes. Early results have been promising with an 18% decrease in patient reported technical issues since pilot launch.²⁸ In addition, with advances in technology, blood oxygen readings may be provided through other devices with which patients have greater familiarity such as “smart” watches and phones. These devices could serve the purpose of detecting silent hypoxemia without requiring patients to learn how to use a new tool or be prompted to provide a measurement as the device could provide passive monitoring.^29,30 However, the question remains how these devices and services will be reimbursed as they are not traditionally incorporated into the fee-for-service model.

In addition, this study illuminates the challenge of better understanding the clinical value that biometric devices provide to patients and care teams. A randomized study of over 2000 patients with COVID-19 assigned to text-based symptom reporting of subjective dyspnea versus text-based symptom reporting and home pulse oximetry found no significant between group differences in the primary outcomes of days alive and out of the hospital at 30 days. Similar to our study in which 72.6% of patients reported at least one reading, 77.7% of patients in the pulse oximetry intervention reported at least one pulse oximetry reading during the monitoring period.³¹ Though this study did not report on the clinical interventions prompted by the pulse oximeter, our study raises the question of whether, due to technical barriers, pulse oximetry might not add more than simple ePROs for symptom monitoring.³¹ Crucial to honing in on the clinical value is identifying the patient populations where the device data is most critical at preventing acute events such as those being monitored for conditions such as neutropenic fever, cytokine release syndrome, or post procedure, like chest tube/pleurX placement. With these narrower clinical focuses, the return-on-investment (ROI) for the enterprise of RPM is more likely to be realized.

Finally, from a resource perspective, physicians and nurses are under increasing pressure to manage new data streams related to their patients. These include ePROs, portal messages, and data provided from mobile phones and consumer health devices. As this data flows into clinic, there is a need for further study on how the data can be best screened, synthesized, incorporated into clinical workflows, and presented in the right clinical context to the appropriate provider to optimize clinical care. In our study, 28.9% of alerts were categorized as missed. We hypothesize that a reason for these missed alerts may include nursing clinical judgement not to contact these patients as they had already responded to prior alerts and the patient was stable not requiring further intervention. There is an opportunity to learn from other nursing models such as the Mayo Clinic, which has developed a robust RPM nursing structure across its enterprise with RPM RN’s operating at a 110:1 patient-to-nurse ratio managing patients in multiple programs simultaneously thus optimizing scalability and cost.³² This type of staffing resources and coordination fits well at larger enterprises to capture the ROI of these programs but may be more difficult to implement for centers with limited RPM interventions where the monitoring is not able to be centralized or they face financial barriers to outsourcing monitoring to a vendor. Artificial intelligence (AI) applications will also likely play a key role here in enhancing how we use RPM technologies to care for our patients, though workflow considerations are important in AI implementation as well.

This prospective study has several limitations. It was conducted at a single National Cancer Institute (NCI)-designated cancer center and thus may lack generalizability to other sites. However, the enrolled patients were cared for in a range of geographic settings from Manhattan as well as Regional Care Network sites in New Jersey, Westchester, and Long Island. Second, the small sample size limits broad conclusions around RPM, but as a feasibility study, we were focused on specific outcomes that could be answered with this limited patient sample prior to considering a larger implementation. In addition, we wanted the findings of this study to be pragmatic, so we were inclusive of disease types and had few exclusion criteria.

Conclusion

The American Society of Clinical Oncology (ASCO) and the NCI have called for more studies around the use of telemedicine and RPM that focus on patient-centered outcomes.³³ This study demonstrates that while most oncology patients are willing to engage with these technologies, meaningful clinical care can be hampered if there are not appropriate technical teams in place to help patients utilize these devices in the home. As future studies evaluate RPM technologies in new applications, such as for early detection of immunotherapy induced toxicity or detection of cytokine release syndrome,³⁴ there is a need for better logistical infrastructure to support clinical and patient teams in the home.

Supplementary Material

PV Appendix Figure 1

Appendix Figure 1: MSK Remote Monitoring. How does it work?

Context Summary:

Key Objective:

Can pulse oximetry remote patient monitoring be effectively used for oncology patients post-discharge?

Knowledge Generated:

The study found that 72.6% of patients successfully provided at least one blood oxygen measurement at home. However, only 39.7% of alerts led to clinically meaningful interactions, indicating limitations in translating monitoring data into actionable clinical responses.

Relevance:

This moderately sized single institution study performed during the COVID-19 pandemic era demonstrates the feasibility of home remote O2 monitoring, but the majority of alerts were not clinically actionable. Further efforts to improve the signal-to-noise ratio while also addressing the costs of equipment and personnel are indicated.