In this trial, providing individuals at high risk of adverse COVID-19 outcomes with at-home COVID tests, on-demand telemedicine, and rapid prescription delivery reduced the number of ICU stays, which reduced the estimated cost for COVID-19 health care per person. This trial also characterizes the ongoing effect of the COVID-19 pandemic on high-risk populations: on average, in the group experiencing standard of care during a one-year study, 5.6% of immunocompromised participants experienced a COVID ICU stay and 3.2% of individuals aged 65 or older did. As a result, the cost of COVID-19 health care during the study was estimated to be an average of $3,718 higher per control participant, compared to intervention participants. At an estimated average of $286 per month ($333 for immunocompromised participants), this is more than the cost of five Cue Health tests per month (which were about $50 each) and is much higher than the cost of supplying high-risk individuals with daily antigen COVID-19 tests (which currently cost about $6 each, or $180 for 30 tests). The expected increased cost of implementing on-demand telemedicine and rapid prescription delivery compared to current practices will vary per health system, but Cue charged $99 for this service. Further, alongside daily antigen testing, we estimate that $1,272 per individual per year would remain for these costs before implementing the intervention became more financially costly than not implementing it. In addition to cost savings at the individual level, there is a potential for the intervention to help reduce the risk of prolonged infections, thereby reducing the emergence of new variants[12–14]. This study adds to the evidence that public health and individual precautions are still warranted and cost-effective in 2024, when both COVID-19 data collection and public health measures have been drastically reduced compared to earlier in the pandemic.
We hypothesize that the reduction in ICU stays is due to intervention individuals accessing a positive test result and treatment earlier in the course of their infection. The hospitalization rate we observed was consistent with previous findings [2, 21]. The intervention did not affect the rate of participants contracting COVID-19, despite that tests could be shared with others in the participant’s household, or which drug was prescribed to treat COVID infections. This could be due to infections coming from outside the household, a lack of willingness for others in the household to test, or the study participant using all the tests themselves. The intervention did not statistically significantly affect hospitalization rates, although there were directionally fewer hospitalizations within the intervention group. One possible explanation is that rapid detection and treatment of an infection is insufficient to prevent hospitalization in this population, indicating that infection prevention efforts should be prioritized.
The test used was an at-home NAAT, which is a molecular equivalent to PCR. While in general, NAAT tests are more sensitive than antigen tests, it is possible that antigen tests could be used when individuals have a fever with minimal loss of sensitivity [22]. When an individual is aware they have been exposed to COVID-19 and they are not febrile, an NAAT or PCR test would improve sensitivity [22–24]. This is one reason it is crucial to maintain community access to NAAT and/or PCR tests. However, accessing NAAT and PCR tests usually require leaving one’s home, and it is unknown how that requirement would affect testing uptake. It is also unknown what level and type of access to testing the control group had; about a quarter of the way through the study, the public health emergency portion of the pandemic ended [25], insurance carriers were no longer required to reimburse for at-home antigen tests, and the number of locations providing PCR testing declined.
It is not possible to break down the three components of the intervention arm—at-home tests, telemedicine, rapid Paxlovid delivery— to determine a driving factor, and lack of availability of data for the latter two components represents a limitation. While the study was closed early due to causes outside of the control of the authors, the data collected before closure was sufficient to drive the conclusions presented in this paper. Our study was also limited by a lower than intended sample size, and underrepresentation of BIPOC participants. Further, we saw higher study engagement in the intervention arm. This only affects survey completion data; the claims and EHR data is without bias between arms. The implication of this biased participation on survey data is not immediately clear; it could be that control data is underreported in the surveys (in which case our results underestimate the intervention’s effect), or that control participants were more likely to complete a survey only when they had a COVID-19 infection. Further, the test, telemedicine, and prescription delivery offering that was used for this study is no longer available, and there is not a clear replacement at-home test with the same level of specificity. It is possible that individuals in the intervention arm changed their behavior as a result of having tests available, for example by doing activities with a higher risk of contracting COVID-19. The cost of the molecular tests used in this trial is high and there is potential for marked reduction of rapid tests in the future, making the intervention even more alluring from a benefit-cost ratio standpoint. Since this intervention did not reduce infection rates, a complementary option is providing high quality masks, e.g. N95s, in particular to those at high risk, and potentially more broadly, to proactively reduce infection rates.
Future research should include a similar study evaluating each component of the intervention (i.e. at-home tests, telemedicine, rapid Paxlovid delivery), along with a comparison of PCR and antigen tests. In particular, it would be helpful to analyze when someone was exposed to COVID-19 (if known), when symptoms began, when they tested positive, and when they began antivirals; this can improve understanding of the role of rapid detection and intervention in improving outcomes. We recommend efforts to increase adoption of testing, enabling earlier detection. For example, tests could be sent to participants on a regular basis by default, to ensure they are readily available when needed. Educating participants that early testing and treatment can reduce their risk of adverse outcomes may improve testing uptake and could improve testing uptake amongst their close contacts. Future studies may include wearable devices, which can detect changes from one’s personalized baseline that suggest a viral infection and suggest that participants test [26, 27]. The optimal number of tests to be provided within a given timeframe is another variable that should be explored. For example, participants may benefit from having a certain size stockpile to feel comfortable sharing tests with others to reduce the COVID-19 case rate. Making Paxlovid and other COVID-19 treatments easier to obtain, e.g. without requiring insurance preauthorization, may also improve outcomes. Based on these results and other data demonstrating the ongoing risk of COVID-19 to high-risk individuals [2, 21], we recommend that payers and public health organizations provide COVID tests and rapid treatment to high-risk individuals, at no cost to the individuals, for as long as the virus continues to circulate.