Background
The RPM program was the last leg in our digital COVID-19 screening, testing, and care pathway that we have written about previously in our blog: It included a virtual triage leveraging a symptom checker powered by Buoy Health, asynchronous testing requests and scheduling with built-in clinician triage, optional synchronous on-demand video visits, and for those testing positive, an automated outreach program offered through GetWell with remote symptom and clinical condition monitoring, supported by our centralized virtual care team of experienced nurses. All those components were built into our mobile applications that provided a unified experience to patients.
Summary
Our intuition when we initiated the analysis was that patients who would have tested positive for COVID-19 and were monitored at home using our RPM program, could defer on-site visits if they did not require hospitalization but that RPM would help identify sicker patients and get them to the hospital sooner.
Of the 9,378 patients invited in the study period, 5,364 (57%) ended up enrolling in the program. This is an incredible result knowing that for most patients, this is the first time they receive an invitation from a healthcare system to join a remote patient monitoring program. Patients who enrolled in the program were 47 years old on average, more identified as women (65%), and more likely to be already users of our digital services. An interesting note is that the patients between 35 and 75 years old were more likely to enroll than patients in their 20s or early 30s, going against the notion that the youngest are the most likely to enroll in a digital tool. Of course, early on, it was fairly apparent that age was a risk factor for more severe disease, and it would make sense that more concerned people would be more apt to enroll in such a program.
Since the intervention was deployed as a real world study, we had to account for potential confounding between the different groups when conducting our analysis since we offered the program to everyone, without randomization as a clinical trial might do. For instance, people who enrolled in the program could be younger and healthier than the control group and hence could bias the findings. We used robust statistical methods such as propensity weighting and included in our covariates patient characteristics such as co-morbidities, socio-economic status, and the Area Deprivation Index, a marker for socioeconomic status. Our team spent several weeks conducting extensive validation of the data and the methodology, and augmented the statistical analysis with additional sensitivity analyses including propensity score matching and multivariate survival analysis. Overall, the data was similar across all models and were directionally concordant. However, we are still cautious when reporting the results since we can't adjust or account for non observable confounding.
Results
We identified several findings within the data.
The results showed that when adjusting for observable confounding, the patients who were enrolled in our Remote Patient Monitoring tool were less likely to be admitted to the hospital with an odds ratio of 0.68 (p-value=0.001). In less technical terms, a patient who enrolled in the RPM had lower risk of being admitted at the hospital: the RPM could be interpreted as a protective factor against hospitalization. RPM was associated with lower odds of being admitted by 32% as compared to a non-enrolled patient. Importantly, our analysis shows association not causation, but it's an intriguing finding.
Patients were not admitted earlier if they participated in the monitoring program. This was contrary to our initial hypothesis. Instead, patients who were monitored were admitted on average 1 day later. Importantly, we looked for any signal that this later admission would be related to any worse outcomes, such as a longer hospitalization, more intensive care such as being admitted to the ICU, or even mortality. We did not see any suggestion in the data that this was the case. In fact, despite the time from test to admission being slightly longer, the length of stay was shorter for those who were enrolled in the RPM program prior to admission.
RPM was not associated with higher probability of being admitted for very short-stay admissions defined as lengths of stay of only a day or less. We looked for this signal to see if RPM would result in people being referred to the hospital but quickly discharged, a possible signal that RPM was referring people to the hospital who did not need to be there.
Conclusion
The main lesson of this study is that Healthcare systems should consider a diverse toolset when delivering care services that must blend the physical and the digital. As one of our colleagues puts it: “Remote patient monitoring during COVID should have been considered the standard of care”. The data highly supports that an integrated experience of remote patient monitoring has delivered a better care experience and outcomes to our population who enrolled in the program even when adjusting for potential confounding.
This intervention and the study we published in JAMA Open does provide evidence that there is so much room for improvement in the care delivery model. This intervention which was launched within the first days of the pandemic with little resources provides evidence that the combination of technology, innovation, care redesign, and integrated clinical care team could contribute to make a significant impact on patients’ outcomes. Regardless of the “true” impact, we must recognize that the current evidence highly supports the conclusion that Remote Patient Monitoring is effective and should have become a standard of care in similar situations.
On the flip side of these results, we also must recognize that the current model of healthcare systems suffer high deficiencies and that the results from this study should lead us to question the effectiveness, safety, and scalability of traditional care delivery models. As healthcare evolves over the next few years, we must recognize that traditional brick-and-mortar are not elastic enough to support a changing landscape of population health.
Our study and our intervention can be considered a case study of scaling up modalities of care delivery by combining innovation with rigorous outcome validation. It was in hindsight going back to basics, in a way, but also forward thinking. Back to basics because it was about taking care of people, and doing what was right for people whenever they are. It was forward thinking because we had to answer new questions: How may we care for people outside of our walls, particularly for patients with Covid? How may we safely care for them, look out for early signs of clinical worsening, and also prevent the spread of the contagious disease, especially in the early days of the pandemic when knowledge about transmission was limited, personal protective equipment was scarce, and fear was high. Early on, we realized that we had to quickly identify areas where we can deliver services differently from a traditional brick and mortar model; and these changes had to be implemented in a matter of hours and days; the usual “transformation” that would have taken dozens of committees and months of debates and discovery followed by proof of concepts, vanished. Scale, safety, and reliability became paramount. We had to move fast.
When Froedtert & MCW created Inception Health in 2015, the goal was to create experiences and care models that were several years out into the future. This was a form of investment. When the COVID-19 pandemic hit, the investment paid off. We had an established digital health team and nurses experienced with virtual care. We had partnerships with digital health companies like GetWell that we were able to immediately leverage and apply our platform to COVID-19. We also had been providing tele-health services within primary care, which of course increased exponentially during the pandemic. While the COVID-19 pandemic jostled everyone, and the change pulled patients and clinicians at warp speed into new care experiences, this RPM program enrolled its first patient in late March, 2020, a mere 2-3 weeks after reality set in that COVID-19 was in our community.
Our nurses caring for patients in the program remarked that this was among the most rewarding work they had done in their careers. This is the type of care that we aspire to have — always on, longitudinal, and built around the person, wherever they are.
We have shown that this form of care makes a difference. Technology supports the scale, enabling more people to be cared for by the same number of clinicians, which is what is needed to bring down costs in the system. Each avoided admission saved patients, employers, and/or taxpayers money. Yet it was still personal care that was provided. We need more of this.
