Voice could be considered a new vital sign in the future, being collected in routine care and complement other medical assessments.

Vocal biomarkers (VB) are voice characteristics that are specific to a disease or symptom. They can be utilized in clinical practice for remote disease monitoring, screening for diseases, or in clinical research as secondary endpoints to evaluate the efficacy of a treatment or intervention. They can also be used during enhanced consultations to screen for diseases or monitor the progression of a chronic disease, in services to identify at-risk patients, and at home to track patients' symptoms and general health status between consultations.

Despite the potential of VB and many progresses, few have reached the stage of use in clinical routine.

At LIH, the Deep Digital Phenotyping research lab gained experience in the voice-related research field since 2019, with the implementation of methodological tools to collect and pre-process voice recordings, and with the development of vocal biomarker candidates in several therapeutic domains or symptoms (Type 2 diabetes, fatigue, COVID-19 and Long COVID, respiratory health, mental health, etc.) VB is an early-stage research topic; one needs to accelerate the development, validation and integration of vocal biomarkers for personalized medicine and innovative tools - both for clinical practice and remote patient monitoring.

Until now, VB candidates have been developed disease by disease, and there is a lack of data on the influence of concomitant diseases on voice signatures. Therefore, the next vital step to improve vocal biomarkers and assess their specificity is to compare them with both control groups, those without the disease, and those with other diseases.

The study is a cohort of adult people attending one of the hospitals in Luxembourg, either for an hospitalization, a consultation, or a simple visit, regardless of their health conditions. Study will be conducted in French, English, German, and Portuguese which are the most frequent languages in Luxembourg.

Study participants will be followed up over a 5 years period of time, with on-site study visits in study booths installed in the different hospitals in Luxembourg and at-home monitoring in-between visits, using a smartphone app to collect study data.

During both study visits and at-home monitoring voice recordings will be collected together with health data. This will allow to understand how voice evolves with time in a given individual, and to identify VB specific to diseases like diabetes, or to symptoms common to several chronic diseases like fatigue or mental health problems.

The use of voice has significant potential for monitoring health. Specific health conditions have the potential to manifest as perceptible alterations in vocal quality. With state-of-the-art voice processing and machine learning technologies, we can identify these alterations. A vocal biomarker (VB) is a feature or a combination of features from the audio signal of the voice that is associated with a clinical outcome. Despite their innovative potential, vocal biomarkers (VB) are still an early-stage research topic; one needs to accelerate the development, validation, and integration of vocal biomarkers for personalized medicine and innovative tools - both for clinical practice and remote patient monitoring. Until now, VB candidates have been developed disease by disease, and there is a lack of data of the influence of concomitant diseases on voice signatures and their specificity. For these reasons, this study aims at creating a longitudinal cohort of people with different pathologies and with no pathologies with voice being collected longitudinally, to position voice as a new vital parameter.

The potential of Voice AI We hear more and more that "voice is the new blood". This means that voice could be considered a new vital sign in the future, which could be collected in routine care and complement other medical assessments. Indeed, voice reflects our health and the audio signal of our voice is modified, directly or indirectly, by the different symptoms or diseases. Voice is also effortless and inexpensive to collect, which could have the potential to reduce medical on-site visits and ultimately reduce the burden of diseases for patients. VB can be utilized in clinical practice for remote patient monitoring, screening for diseases, or in clinical research as secondary endpoints to evaluate the efficacy of a treatment or intervention. They can be used during enhanced consultations, to screen for diseases or monitor the evolution of a chronic disease, in emergency services to screen for at-risk patients, and at-home to monitor the patients in-between visits.

Despite the potential of VBs and significant progress in the field, few of them have reached the stage of use in clinical routine. Reasons are linked to the relative youthfulness of this research domain, including methodological issues such as the lack of standardized protocols for voice collection and VB development, which renders it difficult to compare and reproduce results, as well as the scarcity of longitudinal studies. One needs to accelerate the development, validation and integration of vocal biomarkers for personalized medicine and innovative tools - both for clinical practice and remote patient monitoring.

It is of the highest importance to have very high diversity in the datasets, in terms of socio-economic factors, mother tongues, age, and conditions to develop relevant VB. Moreover, until now VB candidates have been developed disease by disease and there is a lack of data of the influence of concomitant diseases on voice signatures and their specificity. Therefore, the next important step to improve vocal biomarkers of a given disease is to compare them to control groups without any diseases and with other diseases. It is also important to work with voice recordings with different audio quality, to compare voice recordings collected in very standardized conditions with recordings collected in real-life. Finally, to train algorithms to develop vocal biomarkers, one has to label the voice recordings with clinical data, ideally with results from gold-standard tools.

This study aims at creating a cohort with a longitudinal open adaptative design of people attending the different hospitals in Luxembourg, regardless of their health condition, with voice being collected as a new vital parameter, in controlled settings in booths located in the different hospitals and in real-life settings.

The study is a prospective longitudinal cohort of adult people attending one of the hospitals in Luxembourg. At time of inclusion participants can have one or several specific diseases or no particular disease, and be hospitalized or come to the hospital for a consultation, or be a visitor.

All participants will be included during an on-site visit and will be followed for an initial duration of 60 months. The follow-up will combine regular on-site visits and digital at-home monitoring. An adaptive design is used to tailor the study procedures to the days when participants visit the hospitals, thereby minimizing the burden of study participation as much as possible.

The study will be conducted in the 4 main languages in Luxembourg (French, English, German, and Portuguese).

A smartphone app will be provided to each study participant to collect all the study data that will either be collected during the visits by study nurses directly from participants, or self-reported by the participants themselves.

Inclusion Criteria:

• Adults ≥ 18 years.
• Able to understand the study, provide informed consent and comply with study procedures (with reasonable accommodations as needed).
• Speaks one of the study languages (FR/EN/DE/PT).
• Owns a smartphone compatible with the study app.

Exclusion Criteria:

• - Acute clinical instability at inclusion (e.g., need for immediate medical care, severe respiratory distress, hospitalized patients not allowed to go out of their room) where study procedures could interfere with care.
• Any other condition that, in the judgement of the Medical PI or delegated qualified staff, would place the participant at unreasonable risk or prevent meaningful participation, despite accommodations.

The study population will consist of adult people, with no specific pathologies (can have a chronic disease or have no specific disease), hospitalized or not at inclusion time.