Using machine learning models to investigate the relationship between corporeal workload and clinical and epidemiological features of patients infected with COVID-19 in Egypt

Objective: To explore if clinical and epidemiological features of patients positive for coronavirus disease-2019 are affected by somatic work stress. Method: The retrospective study was conducted at Kafrelsheik University Hospital, Egypt, and comprised data of patients admitted between April 1, to June 6, 2020, with confirmed coronavirus disease-2019 infection. Health records of healthy subjects who had come to the hospital as part of their routine check-up were also included for comparison, and the researchers were blinded during the gathering and analysis phase. Demographic features, vital signs, infection severity, somatic workload of the patients’ jobs at admission, and detailed discharge profile was noted. The relationship between clinical features and somatic work stress was evaluated. Data was analysed using SPSS 26. Results: Of the 1072 cases, 602(56.2%) were men and 470(43.8%) were women. The overall median age was 43 years (interquartile range: 29 years). The healthy group had 500 random subjects. There were significant differences in all vital signs between the patients and healthy controls ( p <0.05). Among the patients, infection severity was higher in men, but it was not significant (p>0.05). The overall mortality was 69(6.4%); 46(4.3%) men and 23(2.2%) women. There was no significant association between gender and outcome ( p >0.05). There were 816(76.11%) patients with low-intensity physical workload pre-infection, 136(12.68%) moderate and 120(11.19%) high. Infection severity was significantly high in the low-intensity group ( p <0.05). However, the fate of the patients was not significantly associated with their pre-infection work profile ( p >0.05). Conclusion: Coronavirus disease-2019 significantly affected patients’ vital signs, and infection severity was significantly associated with physical work stress. However, mortality and pre-infection somatic workload were not associated.


Introduction
The coronavirus disease-2019 (COVID-) was initially described by the World Health Organisation (WHO) as an unidentified outbreak of viral pneumonia in Wuhan, China, in late December 2019. 1,2The virus subsequently spread rapidly across the country and then to several countries, which caused a huge number of infections and massive social phobia. 3ke the rest of the world, COVID-19 represented for Africa one of the greatest health challenges.With more than 58 million children suffering from malnutrition, more young people were likely to be at risk in Africa than elsewhere in the world. 2 With an incubation period of 1-14 days, mostly 3-7 days, the basic signs of the infection are fever, dry cough and myalgia. 4Patients appear normal during the incubation phase, but can pass the infection to others even without symptoms. 5The main transmission method for COVID-19 was through huge particles when people tossed or sneezed, and this could be confused with a common influenza pestilence. 6Contact with the infected products and devices further spreads the infection. 2Th resulting pneumonia was additionally irresistible. 7re data was found to be necessary about the characteristics of patients in Egypt with COVID-19, especially related to patients who were not seriously ill but whose condition deteriorated rather quickly. 89-positive (COVID-19+) may help optimise resource use and care.
After treatment isolation, numerous patients developed anxiety and sleep disruption.Anxiety as a mental burden can trigger physiological events that may enhance or decrease immunity. 9There is growing evidence, for example, that lifestyle factors, adapting to everyday stress, and dietary behaviour are significant cofactors in the immune system. 10e current study was planned to explore if clinical and epidemiological features of COVID-19+ patients are affected by somatic work stress.

Materials and Methods
The retrospective study was conducted at Kafrelsheik University Hospital, Egypt, and comprised data of patients admitted between April 1, to June 6, 2020, with confirmed COVID-19 infection.The health facility was governmentdesignated to accept COVID-19+ patients.After approval from the institutional ethics committees of Kafrelsheik University Hospital and Faculty of physical therapy-Kafrelsheik University (approval number: P.T/ BAS /1/2021/5, data was retrieved from the electronic health record (ERH) database.Data was related to patients infected with COVID-19, confirmed by reverse transcription polymerase chain reaction (RT-PCR).
Health records of healthy subjects who had come to the hospital as part of their routine check-up were also included for comparison, and the researchers were blinded during the gathering and analysis phase.
Demographic features, like age, gender, body mass index (BMI) and self-reported smoking status, were noted.Also noted were vital signs, such as systolic blood pressure (SBP), diastolic blood pressure (DBP), heart rate (HR), respiratory rate (RR), temperature (T), oxygen saturation (SpO2), and associated diseases present at the time of admission, like cancer, cardiac diseases, paralysis, chronic lung disease, renal insufficiency, liver deficit, pancreatic problems, morbid obesity, and respiratory circulation problem.
Infection severity was categorised as mild, moderate and severe, with mild = no care inside a critical care facility, moderate = intensive care unit (ICU) treatment, but no assistance required for mechanical breathing assistive devices, and severe = needed support for mechanical respirator. 8dical records with negative COVID-19 PCR, comorbidities present on admission, or/and missed data were excluded.Somatic workload of the patients' jobs at admission was classified into three categories as per the International standard classification of occupation-three categories of physical activity (ISCO-3CPA). 11,12The categories were low, moderate and high, with low = low-intensity physical workload (LIPW), moderate = moderate-intensity physical workload (MIPW), and high = high-intensity physical workload (HIPW).Detailed discharge profile was noted.The outcome of each patient was recorded as either dead or alive.Python 3.9 was used for modelling analysis.Data was then further prepped.There were 114 missing values across the variables, and k-Nearest Neighbours (KNN) imputer was used which involved calculating the mean of the five values that were immediately adjacent to the missing value on both sides of the data. 13After that, data was checked for outliers and winsorised to lessen the impact of extreme results 14 .Winsorisation included replacing extreme values with values that fell inside a particular percentile of the original distribution.All research variables were rescaled to recode values above the 95th percentile at the 95th percentile value. 14tients were divided into three categories based on infection severity.The dataset was used to train classifiers on the model after the features were sorted.In order to classify data, models including Logistic Regression, Decision Tree Classifier, Random Forest Classifier, Gradient Boosting Classifier, KNN Classifier, Multi-layer Perceptron (MLP) Classifier, Support Vector Classifier (SVC) model, Gaussian Naive Bayes model, and Bagging Classifier were utilised. 15 some extent, overfitting and selection bias were avoided by training each model using a 10-fold cross-validation procedure 15 .Each of the models was run 10,000 times after being randomly divided into the training set (90%) and the test set (10%).Area under the curve (AUC) of the receiver characteristic operator (ROC) and mean and median "Accuracy" were used to evaluate the classification models.Data was analysed using SPSS 26.Continuous data was expressed mean ± standard deviation (SD), and categorical and nominal data as frequencies and percentages.The connection between categorical and nominal data was found using Pearson's chi-square test.Statistically significant variations in the continuous variables between the two groups were found using t-test.Age was utilised as a co-variate in a 3x2 mixed model analysis of covariance (MANCOVA).Prior to the primary analysis, Shapiro-Wilks test for normality was used with histograms to check all vital sign variables for normality assumptions between the groups.Exponential, power, arcsine, and logarithmic transformations were used for non-normally distributed data.The histograms and Shapiro-Wilks tests did not show Egypt Suppl.: Climate Change and the need for One World One Health

S-243
J Pak Med Assoc (Suppl.4)     any differences.Where necessary, a pair-wise comparison using Tukey's post-hoc analysis was performed with 0.05 as the threshold.All analyses underwent a Bonferroni correction. 16,17The level of statistical significance was set as p<0.05.
The code for the current project is presented in the following GitHub link: Dr AhmedTorad/COVID_19_and_Corporeal_workload (github.com)
Of the 1072 cases, 602(56.2%)men and 470(43.8%)were women.The overall median age was 43 years (IQR: 29 years).The healthy group had 500 random subjects.There were significant differences in all vital signs between the patients and healthy controls (p<0.05)(Table 1).

Discussion
To the best of our knowledge, the current study is the first to evaluate the association between somatic workload and COVID-19+ patients' outcomes.Most of the patients were admitted from different areas in Egypt.Among the COVID-19+ patients, a spike was seen in infected individuals aged 43 years, and the percentage of infection was much higher in males than in female patients.Men also had higher severity and mortality rate than women.According to the Egyptian census 2020, there are 13.6%Egyptians aged <5 years, 11.1% aged 5-10 years, and only 19% aged >45 years, 18 which means that Egypt has high youth percentage.Also, the current results were in agreement with Sijia et al., who stated that 47.7% of cases were aged 13-45 years. 19An analysis of 425 COVID-19+ patients revealed that 56% were males. 20An analysis of 140 cases revealed that 50.7% of the cases were men. 21Male COVID-19+ prevalence was 2.4 times that of females, and although males and females had the same vulnerability, men were more likely to die. 22 the current study's general population group, there was much better values related to vital signs compared to the COVID-19+ patients.This was in line with earlier direct or indirect findings related to association of COVID-19 with vital signs, including BP, RR, HR, SpO2 etc. [23][24][25][26][27][28] An interesting finding of the current study is the detection of an increased rate of COVID-19 patients and odds of mortality in LIPW than in MIPW and HIPW groups.These may be due to the similarity of MIPW and HIPW with moderate exercises.It has been revealed that physical activity may reduce the occurrence and severity of infection. 29Also, the antibodies or white blood cells (WBCs) circulate more quickly. 30A link has been suggested between moderate activity and immunity boost. 29e current study has its limitations.In line with every reverse analysis of real-world information, the disparity in local practice and the fullness of the data collection constrained the analysis and decreased the study's sample size.EHR was used to retrieve data from the central data store.The study was conducted during a pandemic when standards might have been compromised.Additional efforts were made to ensure that missing data was missing on a spontaneous basis and not due to systematic installation or patient disparities.The current analysis is applicable to Kafrelsheik University Hospital, Egypt, and generalisation should be done with caution.Also, the sample size was not calculated which could have affected the power of the study.

Conclusion
COVID-19 significantly affected patients' vital signs, and infection severity was significantly associated with physical work stress.However, mortality and pre-infection somatic workload were not significantly associated.

Table - 3
Egypt Suppl.: Climate Change and the need for One World One Health Vol.73, No. 4 (Suppl.4) April 2023 COVID-19: Coronavirus disease-2019, SBP: Systolic blood pressure, DBP: Diastolic blood pressure, RR: Respiratory rate, T: Temperature, HR: Heart rate, P: Probability value.: The relationship of physical workload with severity of infection and patient outcome.

Table - 4
: Effect of physical workload on vital signs of COVID-19 patients.