COVID-19 Symptoms Questionnaire

Can Machine Learning be used to identify key COVID-19 symptoms?

COVID-19 symptoms remain unpredictable in their variety as well as in their sequence during the disease. Some people experience flu-like symptoms followed by gastroenteritis-like symptoms. Others go through extreme fatigue, body aches, and headaches before losing taste and smell. COVID-19 spread is high and difficult to anticipate. This difficulty is due in part to the significant false-negative rate of available tests, leading to potentially thousands of people, with or without COVID-19 symptoms, unknowingly infected, and infectious. As a consequence, clinical diagnosis remains an essential line of defense against disease spread, in particular as long as direct, reliable tests remain uncertain and not widely available. This diagnosis, based on COVID-19 symptoms analysis, can either be performed by a clinician or through a smart, automated questionnaire. 

Problems to solve

  • What are the most informative symptoms indicating COVID-19?
  • Are atypical symptoms worth considering, or are they genuine outliers?
  • Can we, in a couple of questions, identify which patients are sick?
  • Benefits of TADA for
    identifying Key Covid-19 Symptoms

    The disease has generated at least 26 observable COVID-19 symptoms. Among the most common ones are cough, fever, sore throat, fatigue. Yet other symptoms, such as dizziness, joint pain, anosmia, or chills, have been reported and might be even more reliable indications of the disease. In other words, they are weak signals, i.e., infrequent symptoms, said, for instance, in existing databases of COVID-19 patients. Weak signals can provide strong indications. TADA, the Small Data Machine Learning solution from MyDataModels, is well suited to identify these weak signals in the context of investigating such a new disease. It can provide excellent results on small data sets, i.e., a few thousand. Typically, it can identify which criteria/field have the most impact: cough? Fever? No need to be a Data Scientist to use it successfully. No need to code. No need to undergo any specific training. TADA can be used by anybody to identify the most meaningful weak signals indicating the disease.

    Furthermore, it is possible, thanks to TADA, to find the underlying relationships between symptoms. For instance, chills combined with pneumonia might be a reliable indicator of the COVID-19 for the data set mentioned above.

    Eventually, TADA can gather patient records automatically, thus creating patient groups with severe common symptoms and potentially similar behaviors with regards to the disease. For instance, it seems that headaches and body aches are highly informative when they appear in men.

    MyDataModels brings a self-service, easy to use solution, to the healthcare professionals who have Small Data to analyze.

    TADA brings new possibilities for healthcare professionals

    TADA can extract the most significant symptoms to search for among a cohort of studied patients, may it be for COVID-19 or for new diseases to come. Having a short, pertinent list of questions to ask patients is useful for identifying quickly and efficiently sick ones.

    TADA can rapidly spot the COVID-19 symptoms that indicate the disease.”

    This questionnaire can be administered either by nurses and doctors or online as a self-checking questionnaire. This data-driven approach enables us to define the ranking of symptoms to get a reliable prediction of the disease presence based on the specific combination of symptoms.