Exploring post-COVID-19 health effects and features with advanced machine learning techniques | Scientific Reports – Nature.com
							May 5, 2024
							    In the last two and a half years, the COVID-19 pandemic has    drastically affected millions worldwide. The impact hammers on    physical and mental health problems in the post-COVID-19    state1. This phenomenon    raises the necessity to investigate the relationship between    post-COVID conditions and mental health2. Primarily, the    investigation shows that coronavirus has a long-term effect of    post-COVID-19 disease on sleep and mental illness, which also    opens the door to detecting possible relationships between the    severity of COVID-19 at the onset and sleep and mental    illness3. Coronavirus    affects the brain by bypassing the blood-brain barrier (BBB) in    blood or via monocytes which could reach brain tissue via    circumventricular organs7. Importantly,    research shows a prominent frequency of impaired performance    across cognitive domains in post-COVID patients with subjective    complaints25. At the same    time, the discovery of inflammatory biomarkers in COVID-19    survivors has come into broad light through MRI samples and    other means4. One out of five    patients hospitalized for COVID-19 was diagnosed with PTSD or    subthreshold PTSD at a 3-month follow-up6. Potential    contributing factors cause post-COVID-19 patients to suffer    from different memory complaints5. Moreover, some    psychiatric issues like depression prevail in COVID recovery    patients, which causes a 25 times greater risk for suicide than    the general population26. A summary of    data from last year about the impacts on physical, cognitive,    and neurological health disorders in COVID-19 survivors    suggests three crucial aspects to manage: nutritional status,    neurological disorders, and physical health28. So, the    impaired cognitive deficits and emotional distress among    COVID-19 patients should be addressed by functional    rehabilitation27. Side by side,    a brief study is to be analyzed on post-COVID-19 pandemic era    mental health issues, vulnerable populations, and risk factors,    as well as recommending a universal approach for mental health    care and services29. Physiological    and Neurological factors have been examined, with 39%    classified as Physiological and 61% as Neurological.    Neurological factors influence the mind and are connected to a    persons mental and emotional state.30. Here anxiety    is a major Neurological factor among post-COVID patients with a    frequency rating of 8 as shown in Table 2. Anxiety is the most    common mental illness in post-COVID1. Physiological    factors deal with the functions of a living organism and its    parts30. Fatigue is one    of the most frequent alterations of post-COVID patients as    shown in Table 2. Over the past three    years, extensive research has explored physiological and    neurological health complications in the aftermath of COVID-19.    We reviewed 23 research articles using keywords like mental    health, cognitive impairment, and post-COVID trauma. From these    studies, we identified 17 health factors associated with COVID    infection, including fatigue, forgetfulness, and anxiety. These    factors were categorized into two groups: Physiological and    Neurological. Notably, 39% are Physiological factors, while 61%    are Neurological factors, impacting the mind and emotional    well-being30. Here anxiety    is a major neurological factor among post-COVID patients with a    frequency rating of 8 as shown in the Table 2.    Anxiety is the most common mental illness in    post-COVID1. Physiological    factors deal with the functions of a living organism and its    parts30. Fatigue is one    of the most frequent alterations of post-COVID patients Table    2.  
    In this way, all revealed health factors are listed in Table    2 along with references    and frequency of presence in those references.  
    Among the 17 factors we have divided them into two categories,    as shown in Table 2;  
        Physiological factors: Physiological factors deal        with the functions of a living organism and its        parts30. For        example, fatigue is one of the most frequent alterations of        post-COVID patients in Table 2. There are 7        physiological factors identified among all post-COVID-19        factors in this study, as shown in Table 2.      
        Neurological factors: Neurological factors are the        one that influences or affects the mind and are related to        the mental and emotional state of a        person30. For        example, anxiety is the most common mental illness in        post-COVID1. There are        10 neurological factors identified among all post-COVID-19        factors in this study, as shown in Table 2.      
    We have given a statistical overview of our data in    Fig.2 to make our data more    understandable. Data statistics, such as count, min, max, mean,    standard deviation, variance, and median, are essential for    understanding a dataset. Count shows dataset size, min/max    indicates its range, mean reflects central tendencies, standard    deviation measures data spread, and variance quantifies overall    variability. The median is a robust central measure. These    stats form the foundation for data summary, with quartiles,    percentiles, skewness, and kurtosis for deeper dataset    analysis.  
            Statistical overview of data.          
    Feature correlation in Figs.3 and 4    gives a statistical measure that assesses the degree of    association or relationship among features (variables) in our    dataset. It quantifies how these features tend to vary    together, providing insights into their dependencies. The    advantages of this feature correlation (pearson) analysis in    Fig. 4 (Full information is    shown in Fig.5) includes its utility    in identifying redundant or highly informative features for    best model performance, detection of multicollinearity in    regression analysis, simplifying data exploration by revealing    hidden patterns and relationships, aiding in model    interpretability, and facilitating feature engineering by    leveraging the knowledge of feature associations to create new    informative variables. Pearson correlation, is a crucial data    science tool. It quantifies the strength and direction of the    linear relationship between two continuous variables, with    values ranging from 1 to 1. This technique is widely employed    in statistics and data analysis to uncover connections,    patterns, and dependencies within complex datasets.  
            Pearson correlation value for all to all input            features.          
            Overview of target classanxiety.          
            TNSE visualization of features for after anxiety.          
    The chi-square test is one of the methods to find out the    association i.e. relationship among the categorical variables.    The relationship can be significant or insignificant. The    standard P-value is considered as 0.05 and any p-value having    less than 0.05 is considered to have a significant association    i.e. relationship among variables as shown in    Fig.3. In this research,    the survey dataset has the responses i.e. level of impact on    various physiological & neurological factors. These factors are    considered categorical variables. The chi-square test is    applyed on all factors and we got P-value for them which    is shown in Fig.3. In the Table    3, calculated p-values    less than 0.05 are marked with Grey color. These values with    corresponding Factors are analysed to possess significant    relationships among them.  
    From the Fig.3, we can see all    comparing factors have an association between them, Some basic    features association as follows: a. Chest Pain & Unhappiness b.    Unhappiness & Forgetfulness c. Depression & vigilance d. Chest    pain & confidence e. Confidence & vigilance f. Energy &    confidence g. Sleep & attentiveness h. Attentiveness &    vigilance i. Sleep & determination j. Determination & vigilance    and k. Fear of COVID & energetic  
    Pearson correlation coefficient is a unit measuring the    strength of the linear relationship between two variables. This    is represented as the r-value. R-value results in the range    from 1 to 1. +1 represents the positive correlations(direct    relationship), 0 shows no relationship & 1 represents the    negative correlations(inverse relationship). In the research,    the physiological & neurological factors of the dataset are    depicted as variables. The Pearson correlation coefficient is    calculated for all factors, and we got the R-value for them    shown in Fig.3. The R-values above    0.05 are considered for positive/direct relation between the    factors. This means an increase in one factor may influence and    increase the degree of another factor. R-values below 0(in the    -ve range) are considered for Inverse relation between factors.    This means a Decrease in one factor may influence and Decrease    another factor. The Pearson correlation revealed a strong    positive relationship between the two variables, with a    correlation coefficient of 0.85, indicating a significant and    direct association.  
    Feature importance analysis shown in Fig.3    using the Ordinary Least Squares (OLS) regression model is a    valuable technique in data analysis and predictive modeling. In    this table, we renamed each feature name and labeled it from 1    to 13. In the context of feature importance, OLS can reveal the    impact of each independent variable on the dependent variable.    Larger coefficient values indicate stronger feature importance,    while coefficients near zero suggest less relevance. This    analysis aids in feature selection, helping us focus on the    most influential variables for building predictive models or    understanding the factors that drive specific outcomes in the    data. Based the outcome shown in Table 3, the most important    feature is 13(with a score of 1.5447) and the less important    feature is 1(with a score -1.0443).  
            Training algorithm for anxiety analysis.          
    Firstly, the compiled dataset is used for Statistical Analysis    to explore whether any impact exists on the factors due to    COVID-19 or not. The dataset possesses the info of both the    Before and After conditions of the factors. The x-axis shows    the categories/responses of people on how much each factor,    like anger, depression, etc is affected. Y-Axis shows the    percentage of how many persons are acknowledged in each    category. In Fig.4b, we present a    comparative view of anxiety before and after COVID-19. The blue    color represents the degree of impact for the factors before    being affected by COVID-19. The red color represents the status    after suffering from the disease.  
    Before COVID-19 state, no people strongly agreed on having    Anxiety over their COVID issue, but the percentage jumped to    16.67% who strongly agreed after suffering from it. The graph    follows the same pattern in the subsequent remarks. Comparing    the before & after situations, it can be concluded that after    suffering from COVID-19, a large number of people got the new    problem whereas the people having previous Anxiety issues    remained the same/more. In Fig. 4a, we present a    complete view of anxiety amount before and after COVID-19.  
    It is such a factor that shows most of the patients are    suffering from depression more after COVID. 23.33% and 36.67%    patients either strongly agreed or agreed respectively on this    matter. This figure has risen from 16.67% and 20.00% before    COVID. While 36.67% disagreed on this matter before COVID the    figure came down to only 10.00% after COVID. Depression, in    human life, has increased after COVID-19  
    On the factor of unhappiness, 33.33%, and 26.67% people agreed    on their unhappy life before and after COVID respectively.    However we see an almost inverse trend on the neutral point of    view among the patients. Thus comparing the before & after    situation, it can be visualized that after suffering from    COVID-19, unhappiness has decreased among the patients.  
    The degree of confidence before and after the COVID-19 era    shows a drastic change in peoples mentality. Before COVID-19    state, 56.67% of people agreed on their degree of confidence    but COVID had hit hard on their lifestyle shifting down to 20%    confidence degree after COVID. The same trend was seen in the    disagreement chart. Comparing the before & after situation, it    can be concluded that after suffering from COVID-19, the    majority of the peoples confidence in themselves was    shattered.  
    Regarding forgetfulness, double the number of patients either    agreed or strongly agreed that they forgot things now more    after suffering from it. Thus, COVID has fatally affected the    patients memory, resulting in curbing their brains.  
    Before suffering from COVID, about 60% people agreed that they    were more patient in life, but the percentage abruptly dropped    to half who decided to be after suffering from COVID. But none    Strongly Disagreed in this regard, neither before nor after.    Thus comparing the before & after situation, it can be    visualized that after suffering from COVID-19, vigilance has    decreased by almost half or beyond among the patients.  
    Before the COVID-19 state, most people (56.67%) agreed about    being more energetic, whereas the percentage increased in favor    disagreement (36.67% disagree, 10% strongly disagree) in the    post-COVID state. Comparing the before & after situations, it    can be depicted that after suffering from COVID-19, people are    becoming significantly less energetic.  
    Before COVID-19 state, no people strongly agreed about having    chest pain, but the percentage jumped to 23.33% who strongly    agreed after suffering from COVID. Comparing the before & after    situations, it can be concluded that after suffering from    COVID-19, a large number of people got the new problem, whereas    the people having previous chest pain history remained the    same/more.  
    Before COVID-19 state, about 36.67% of people agreed that they    experienced more sleep, but the percentage decreased to 33.33%    who agreed after suffering from COVID. Comparing the before &    after situations, it can be concluded that after suffering from    COVID-19, experiencing sound sleep conditions shows a    sight-decreasing tendency.  
    Before COVID-19 state, about 43% of people were NEUTRAL about    their anger problem, whereas 40% people agreed about the    problem. Comparing the before & after situations, it can be    concluded that after suffering from COVID-19, most people    agreed that their anger has increased.  
    Before the COVID-19 state, most people (50%) disagreed about    having dizziness problems, but the percentage is rising in    favor of strongly agree (16.67%) and agree (36.67) in the    post-COVID state. Comparing the before & after situations, it    can be concluded that after suffering from COVID-19, dizziness    is slowly increasing among people after COVID.  
    Before the COVID-19 state, a few people (3.33%) strongly agreed    that they had been impulsive, but the percentage increased to    20% who strongly agreed after suffering from COVID. Comparing    the before & after situations, it can be concluded that after    suffering from COVID-19, people show a sight-increasing    impulsiveness tendency.  
    Before suffering from COVID, about 60% of people agreed that    they were more vigilant, but the percentage abruptly fell to    16.67% who agreed after suffering from COVID. At the same time,    disagreement degrees increased in the post-COVID situation.    Comparing the before & after situations, it can be visualized    that after suffering from COVID-19, vigilance has decreased    dramatically among the patients.  
See the article here: 
Exploring post-COVID-19 health effects and features with advanced machine learning techniques | Scientific Reports - Nature.com