| Literature DB >> 32664388 |
Diya Li1, Harshita Chaudhary2, Zhe Zhang1.
Abstract
By 29 May 2020, the coronavirus disease (COVID-19) caused by SARS-CoV-2 had spread to 188 countries, infecting more than 5.9 million people, and causing 361,249 deaths. Governments issued travel restrictions, gatherings of institutions were cancelled, and citizens were ordered to socially distance themselves in an effort to limit the spread of the virus. Fear of being infected by the virus and panic over job losses and missed education opportunities have increased people's stress levels. Psychological studies using traditional surveys are time-consuming and contain cognitive and sampling biases, and therefore cannot be used to build large datasets for a real-time depression analysis. In this article, we propose a CorExQ9 algorithm that integrates a Correlation Explanation (CorEx) learning algorithm and clinical Patient Health Questionnaire (PHQ) lexicon to detect COVID-19 related stress symptoms at a spatiotemporal scale in the United States. The proposed algorithm overcomes the common limitations of traditional topic detection models and minimizes the ambiguity that is caused by human interventions in social media data mining. The results show a strong correlation between stress symptoms and the number of increased COVID-19 cases for major U.S. cities such as Chicago, San Francisco, Seattle, New York, and Miami. The results also show that people's risk perception is sensitive to the release of COVID-19 related public news and media messages. Between January and March, fear of infection and unpredictability of the virus caused widespread panic and people began stockpiling supplies, but later in April, concerns shifted as financial worries in western and eastern coastal areas of the U.S. left people uncertain of the long-term effects of COVID-19 on their lives.Entities:
Keywords: Basilisk algorithm; COVID-19 pandemic; Correlation Explanation (CorEx); Patient Health Questionnaire (PHQ); mental health; social media data mining
Mesh:
Year: 2020 PMID: 32664388 PMCID: PMC7400345 DOI: 10.3390/ijerph17144988
Source DB: PubMed Journal: Int J Environ Res Public Health ISSN: 1660-4601 Impact factor: 3.390
Illustration of seed words for the Basilisk algorithm [37].
|
| ||||||
| addiction | boredom | dissatisfaction | grief | insecure | fear | stress |
| tense | burnout | meditation | guilt | irritable | panic | alcoholism |
| anger | conflict | embarrassment | headache | irritated | pressure | tension |
| anxiety | criticism | communication | tired | loneliness | problem | impatience |
| backaches | deadline | frustration | impatient | nervous | sadness | worry |
|
| ||||||
| chill | satisfaction | self-confidence | cure | perfection | heart | prevention |
| enjoy | happiness | self-improvement | distress | overwork | right | self-talk |
| love | productivity | empowerment | wedding | perfectionism | change | tension |
| relax | perfection | self-image | marriage | self-help | family | tired |
| relaxation | well-being | commitment | relax | control | joy | empower |
Description of dependency information.
| Information in Text | Description |
|---|---|
| Index | Index of the word in the sentence |
| Text | Text of the word at the particular index |
| Lemma | Lemmatized value of the word |
| Xpos | Treebank-specific part-of-speech of the word. Example: “NNP” |
| Feats | Morphological features of the word. Example: “Gender = Ferm” |
| Governor | The index of governor of the word, which is 0 for root |
| Dependency relation | Dependency relation of the word with the governor word which is root if governor = 0. Example: “nmod” |
Illustration of the Basilisk algorithm [41].
|
|
| Procedure: |
Performance evaluation of classifiers.
| Model | Validation Accuracy |
|---|---|
| Support Vector Machine (SVM) (Radial basis function kernel) | 0.8218 |
| SVM (Linear kernel) | 0.8698 |
| Logistic Regression | 0.8620 |
| Naïve Bayes | 0.8076 |
| Simple Neural Network | 0.8690 |
Illustration of Correlation Explanation (CorEx)Q9 algorithm.
| Procedure: |
Patient Health Questionnaire (PHQ)-9 lexicon description and examples.
| PHQ-9 Category | Description | Lexicon Examples |
|---|---|---|
| PHQ1 | Little interest or pleasure in doing things | Acedia, anhedonia, bored, boring, ca not be bothered |
| PHQ2 | Feeling down, depressed | Abject, affliction, agony, all torn up, bad day |
| PHQ3 | Trouble falling or staying asleep | Active at night, all nightery, awake, bad sleep |
| PHQ4 | Feeling tired or having little energy | Bushed, debilitate, did nothing, dog tired |
| PHQ5 | Poor appetite or overeating | Abdominals, anorectic, anorexia, as big as a mountain |
| PHQ6 | Feeling bad about yourself | I am a burden, abhorrence, forgotten, give up |
| PHQ7 | Trouble concentrating on things | Absent minded, absorbed, abstracted, addled |
| PHQ8 | Moving or speaking so slowly that other people could have noticed | Adagio, agitated, angry, annoyed, disconcert, furious |
| PHQ9 | Thoughts that you would be better off dead | Belt down, benumb, better be dead, blade, bleed |
Figure 1Fuzzy membership functions of uncertainty evaluation of assigned PHQ category.
Average coherence measure score.
| Model | Average UMass | Average UCI |
|---|---|---|
| CorExQ9 | –3.77 | –2.61 |
| LDA | –4.22 | –2.76 |
| NMF-LK | –3.97 | –2.58 |
| NMF-F | –4.03 | –2.36 |
Abbreviations: UCI = The UCI measure was first introduced by researches in University of California, Irvine; UMass = Umass measure was first introduced by researches in University of Massachusets. Related papers using these measures are just using UCI and Umass directly; LDA = latent Dirichlet allocation; NMF-LK = Kullback-Leibler divergence non-negative matrix factorization; NMF-F = Frobenius normalized NMF.
Figure 2Process undertaken to generate spatiotemporal stress symptom maps and topics.
Illustration of detected stress symptoms based on PHQ-9 category.
| PHQ-9 Category and Description | Top Symptoms and Topics |
|---|---|
| PHQ0: Little interest or pleasure in doing things | |
| PHQ1: Feeling down, depressed | |
| PHQ2: Trouble falling or staying asleep | |
| PHQ3: Feeling tired or having little energy | |
| PHQ4: Poor appetite or overeating | |
| PHQ5: Feeling bad about yourself | |
| PHQ6: Trouble concentrating on things | |
| PHQ7: Moving or speaking so slowly that other people could have noticed | |
| PHQ8: Thoughts that you would be better off dead |
Figure 3Spatiotemporal pattern with fuzzy accuracy assessment for stress symptom analysis result generated by CorExQ9. (a) from 01.26.2020 to 02.09.2020; (b) from 02.09.2020 to 02.23.2020; (c) from 02.23.2020 to 03.08.2020; (d) from 03.08.2020 to 03.22.2020; (e) from 03.22.2020 to 04.05.2020; (f) from 04.05.2020 to 04.19.2020; (g) from 04.19.2020 to 05.03.2020; (h) The legend of for (a)–(g).
Figure 4The number of daily confirmed new cases in the United State (five-day moving average) [63,64].
Notation table.
| Notation | Description |
|---|---|
|
| A subset of the extraction patterns that tend to extract the seed words |
|
| The candidate nouns extracted by |
|
| Total correlation, also called multi-information, it quantifies the redundancy or dependency among a set of |
|
| Kullback–Leibler divergence, also called relative entropy, is a measure of how probability distribution is different from a second, reference probability distribution [ |
|
| Probability densities of |
|
| The mutual information between two random variables |
|
| |
|
| The Kronecker delta, a function of two variables. The function is 1 if the variables are equal, and 0 otherwise. |
|
| A constant used to ensure the normalization of |