Periodic Reporting for period 2 - DCUBATION (Redefining the term 'Incubation Period' using large-scale digital data)
Período documentado: 2022-05-01 hasta 2023-10-31
My first sentence in the DCUBATION grant (submitted before the COVID-19 pandemic) was “Judging by the past, infectious diseases pose the greatest risk for a global catastrophe.” Thus, I proposed that “An infection starts silently, and gradually progresses until clear clinical symptoms appear. Thus, improving our understanding with respect to the incubation period is pivotal for prevention, early detection, and control of infectious diseases.” For example, previous studies suggested that up to 65% of COVID-19 transmissions occur a day prior to symptom onset. Thus, if we could detect infections just one day before the current practice, the basic reproductive ratio of the SARS-COV-2 first identified in Wuhan, would have fallen below 1, meaning the coronavirus pandemic would have been avoided.
I hypothesized based that the actual onset of clinical symptoms “occurs earlier than previously known, can be identified more accurately using data from wearables, and augmented by data from medical records.” Our overall objective is to develop a comprehensive methodology for real-time evaluation of the incubation period for early detection of infectious diseases. Specifically, I aimed to evaluate prior risks for respiratory infections, identify visible and invisible clinical symptoms by analyzing digital sensors from mobile phones and wearables, detect respiratory infections (and deteriorations) earlier than current, and evaluate the population-level effectiveness of earlier intervention/treatment.
I hypothesized based that the actual onset of clinical symptoms “occurs earlier than previously known, can be identified more accurately using data from wearables, and augmented by data from medical records.” Our overall objective is to develop a comprehensive methodology for real-time evaluation of the incubation period for early detection of infectious diseases. Specifically, I aimed to evaluate prior risks for respiratory infections, identify visible and invisible clinical symptoms by analyzing digital sensors from mobile phones and wearables, detect respiratory infections (and deteriorations) earlier than current, and evaluate the population-level effectiveness of earlier intervention/treatment.
Shortly after the ERC project was granted, I received all the regulatory approvals and conducted a Prospective cohort: a large-scale clinical trial for two years in which >4,600 participants received smartwatches, downloaded a dedicated mobile application we developed, and permitted access to their medical records. The smartwatches continuously monitored several physiological measures, including heart rate. For those interested, we provided home testing kits (to test respiratory or bacterial infections including - Group A streptococcus, influenza (A and B), and Covid-19). The mobile application collected daily self-reported questionnaires on well-being (sleeping habits, general mood, sports activities, etc.), reported symptoms, and home-test results. As of June 2023, 2,430 participants completed two years in the trials, and ~2,300 participants continue in the project. Retrospective cohort: A cohort of 250,000 members for 15 years (2010 - 2025) that serve as a representative sample of the Israeli population demographically. We received access to the electronic medical records of these members in a pseudonymized fashion.
As of June 2023, 2,430 participants completed the two years period, 2,369 are active participants. Overall, we have ~820,000 days in which participants completed the daily questionnaire and ~1,950,000 days of smartwatch data. This big and rich detailed, and labeled data is pivotal to the success of the project. Analyzing the data from the perspective and the retrospective cohorts (or part of it), we used ML models to detect COVID-19 test results in the period before antigen kits were available [1]. We developed a machine-learning model for COVID-19 detection that uses four layers of information: (i) sociodemographic characteristics of the individual, (ii) spatio-temporal patterns of the disease, (iii) medical condition and general health consumption of the individual, and (iv) information reported by the individual during the testing episode. We evaluated our model on 140,682 members of Maccabi Health Services who were tested for COVID-19 at least once between February and October 2020 (period before home kits were available and it took 2-5 days to receive PCR test results). Our ability to predict early on the outcomes of COVID-19 tests is pivotal for breaking transmission chains, and can be used for a more efficient testing policy.
We also predict COVID-19 deteriorations in hospitals [2], and evaluate the effectiveness of proactive surveillance programs in reducing infections [3]. Moving toward the epidemic stage of COVID-19, we analyze medical records to evaluate the effectiveness of routinely boosting sub-populations at least every 6 months may be warranted, depending on age and immune status. Our study provides evidence for the potential benefit of a routine 6-month cadence for Covid-19 boosters for the highest-risk groups, even during relatively lower Covid-19 prevalence. We also demonstrated the considerably higher sensitivity of wearable sensors in detecting physiological reactions after COVID-19 vaccinations and COVID-19 infection [4–7]. We are now finalizing a paper that introduces a new term to the medical community – the DCUBATION period— the digital incubation period. Analyzing episodes of COVID-19, Group A step., Influenza, and other ILI, we identified a clear anomaly in physiological measures 6-48 hours before the participant identifies a symptom, and 48-72 before the test is conducted.
As of July 2023, we published 7 publications in leading journals in medicine and multidisciplinary fields including the Lancet Respiratory Medicine (Impact factor 102.6) Cell Reports Medicine (IF 17.0) the formal journal of the US CDC -Emerging Infectious Diseases (IF 11.8) publications from the Nature portfolio, including npj digital medicine – (IF 15.3) and a publication from journal of the Journal of the Royal Society Interface (IF 4.3). Furthermore, 3 publications are currently under review in leading journals, and 1 more is finalized before submission.
As of June 2023, 2,430 participants completed the two years period, 2,369 are active participants. Overall, we have ~820,000 days in which participants completed the daily questionnaire and ~1,950,000 days of smartwatch data. This big and rich detailed, and labeled data is pivotal to the success of the project. Analyzing the data from the perspective and the retrospective cohorts (or part of it), we used ML models to detect COVID-19 test results in the period before antigen kits were available [1]. We developed a machine-learning model for COVID-19 detection that uses four layers of information: (i) sociodemographic characteristics of the individual, (ii) spatio-temporal patterns of the disease, (iii) medical condition and general health consumption of the individual, and (iv) information reported by the individual during the testing episode. We evaluated our model on 140,682 members of Maccabi Health Services who were tested for COVID-19 at least once between February and October 2020 (period before home kits were available and it took 2-5 days to receive PCR test results). Our ability to predict early on the outcomes of COVID-19 tests is pivotal for breaking transmission chains, and can be used for a more efficient testing policy.
We also predict COVID-19 deteriorations in hospitals [2], and evaluate the effectiveness of proactive surveillance programs in reducing infections [3]. Moving toward the epidemic stage of COVID-19, we analyze medical records to evaluate the effectiveness of routinely boosting sub-populations at least every 6 months may be warranted, depending on age and immune status. Our study provides evidence for the potential benefit of a routine 6-month cadence for Covid-19 boosters for the highest-risk groups, even during relatively lower Covid-19 prevalence. We also demonstrated the considerably higher sensitivity of wearable sensors in detecting physiological reactions after COVID-19 vaccinations and COVID-19 infection [4–7]. We are now finalizing a paper that introduces a new term to the medical community – the DCUBATION period— the digital incubation period. Analyzing episodes of COVID-19, Group A step., Influenza, and other ILI, we identified a clear anomaly in physiological measures 6-48 hours before the participant identifies a symptom, and 48-72 before the test is conducted.
As of July 2023, we published 7 publications in leading journals in medicine and multidisciplinary fields including the Lancet Respiratory Medicine (Impact factor 102.6) Cell Reports Medicine (IF 17.0) the formal journal of the US CDC -Emerging Infectious Diseases (IF 11.8) publications from the Nature portfolio, including npj digital medicine – (IF 15.3) and a publication from journal of the Journal of the Royal Society Interface (IF 4.3). Furthermore, 3 publications are currently under review in leading journals, and 1 more is finalized before submission.
Vaccine hesitancy is considered by the WHO as one of the top ten threats to public health. Thanks to the ERC-funded project, we had a fairly large amount of individuals participating in our clinical trial during the vaccination campaigns. We were the first in the world to show that smartwatches and wearables can be more sensitive than humans in detecting physiological changes following vaccination ( e.g. we detected changes even in individuals that reported - "no symptoms" after the vaccine). In the short run, our studies provided safety assurances to the global population who were eligible to receive COVID-19 boosters. These assurances may have helped increase the number of high-risk individuals who opt to receive booster vaccines and thereby prevent severe outcomes associated with COVID-19. Currently, vaccine safety in clinical trials is determined primarily by subjective self-reporting documentation.
These studies were pivotal in determining the safety of COVID-19 vaccines as our findings reflect physicians’ diagnoses, patients’ objective physiological measures, and patients’ subjective reactions. Specifically, the higher sensitivity of smartwatches compared to humans in detecting abnormalities and doing so earlier than current can revolutionize clinical trials by enabling earlier identification of abnormal reactions with fewer subjects. If data permits, we will analyze the association between physiological reactions after vaccinations and the effectiveness of vaccines (e.g. extent and duration of protection).
These studies were pivotal in determining the safety of COVID-19 vaccines as our findings reflect physicians’ diagnoses, patients’ objective physiological measures, and patients’ subjective reactions. Specifically, the higher sensitivity of smartwatches compared to humans in detecting abnormalities and doing so earlier than current can revolutionize clinical trials by enabling earlier identification of abnormal reactions with fewer subjects. If data permits, we will analyze the association between physiological reactions after vaccinations and the effectiveness of vaccines (e.g. extent and duration of protection).