This project evaluates the effectiveness of inpatient rehabilitation program in alleviating depression and anxiety in PCS patients using machine learning to analyze large-scale clinical and patient-reported data. The aim is to identify key factors contributing to the improvement or persistence of depression and anxiety post-rehabilitation, providing new insights into managing mental health symptoms associated with PCS.
The study group included a randomized sample from over 3000 post-COVID-19 patients treated between September 2020 and December 2022. Using machine learning models such as Random Forests, Support Vector Machines, and more sophisticated deep learning algorithms, we analyze patient characteristics, clinical histories, questionnaire results (HADS, SF36), and treatment types to identify predictors of rehabilitation success.
A total of 362 patients with a mean age of 63,6 (±9.2) answered the HADS survey before and after the rehabilitation, of which 207 (56,7%) were female and 158 (43,3%) were male. There were 25 patients (6,9%) younger than 50 years old, 91 (24,9%) in range 51-60, 147 (40,3%) in range 61-70, and 102 (27,9%) older than 71. There was a statistically significant change (p0.001) in responses before and after rehabilitation for the first two groups. For every partition (by sex, age group, and number of symptoms), there was a statistically significant change in responses for the SF36 questionnaire. Similarly, for HADS-A and HADS-D, improvements are noticeable; for female patients, HADS-D went from 7.29 mean to 5.57 mean, and HADS-A score decreased from 8.32 to 6.61 while for male patients, scores went from 7.05 to 5.79 and 7.29 to 5.99 respectively. On average, most of the COVID-19 symptoms showed patients aged 51 to 60 years (4.2 symptoms). On average, females under 60 showed more COVID-19 symptoms than males (4.44 to 3.69). Machine learning models revealed that specific clusters of PCS symptoms were linked to persistent anxiety. The analysis found that inpatient rehabilitation significantly improved depression and anxiety levels in most post-COVID-19 syndrome (PCS) patients, particularly those with severe symptoms such as respiratory and cognitive issues.
Patients' quality of life and anxiety/depression symptoms improved after rehabilitation. Machine learning models identified critical predictors of rehabilitation success, such as the severity of initial symptoms, and used rehabilitation models. Clustering methods are best suited to uncover underlying patterns and divisions within the patient's cohort.
Physiotherapy education should focus on the link between PCS symptoms and mental health so practitioners can address psychological factors in treatment plans. AI algorithms can analyze patient characteristics, symptom profiles, treatment regimens, and clinical histories to predict the likelihood of improvement in depression/anxiety following rehabilitation.
Machine learning
Mental health
