Table 1.
Characteristics of the included studies reporting SA solutions for cardiovascular diseases research.
| Authors | Objectives | Data sources | SA methods | Results |
|---|---|---|---|---|
| 1. Detecting emotional risk factors for CVD | ||||
| Eichstaedt et al., (9) | Analyze social-media language to identify community-level psychological correlates of age-adjusted mortality from AHD | Data from 1,347 US counties for which AHD mortality rates, health variables, and 50,000 tweeted words were available | Cross-sectional regression model based on Twitter language | Negativity emerged as significant risk factor (partial rs = 0.06, 95% confidence interval, or CI = [0.00, 0.11], to 0.12, 95% CI = [0.07, 0.17]) for CAD mortality |
| Hemalatha et al., (10) | Identify relevant MI risk factors using Twitter data | Twitter users with a MI history | LR for positive/negative emotion classification, with words weighted using TF.IDF | Not available |
| Medina Sada et al., (11) | Identify the relation between the sentiment of tweets and CVD | Tweets in the counties along Interstate 20 in Texas | Naïve Bayes, Multinomial Naïve Bayes, Bernoulli Naïve Bayes, Support Vector, and Linear Support Vector | High positive-to-negative ratio and positive-to-population ratio tend to associate with counties with low CVD rate |
| 2. Detecting positive/negative attitudes of CV patients toward their disease | ||||
| Verma et al., (12) | Assess public health impact of CVD and patients' adherence and attitudes toward the disease | Tweets in english related to CVD | Not specified | The percentage of positive tweets are 45%, neutral tweets are 30 and 25% are negative tweets |
| Pimenta et al., (13) | Identify which fitness and nutrition apps that support behavior change (which could reduce CVD mortality) elicits a positive response from the users | User store reviews of a sample of fitness and nutrition apps | Text mining with Sketch Engine online app | StepsApp pedometer had the highest percentage of positive tags while VeryFitPro had the lowest |
| 3. Detection of cardiac arrhythmia | ||||
| Behadada et al., (14) | Provides insights into arrhythmia detections from big data information sources | Expert knowledge, data and textual information from Pubmed articles and MIT-BIH database | Semi-automatically fuzzy partition rules and grammar-based text extraction SA | Accuracy of 93% and a high level of interpretability of 0.646 for the detection of cardiac arrhythmia |
| 4. Triage of CV patients | ||||
| Lowres et al., (15) | Assessing the feasibility of using an ML program to triage incoming SMS text messaging replies as requiring health professional review or not | 3,118 SMS text messaging replies received from 2 clinical trials | Naïve Bayes, OneVsRest, Random Forest Decision Trees, Gradient Boosted Trees, Multilayer Perceptron | The multilayer perceptron model achieved the highest accuracy (AUC 0.86) |
| 5. Feedbacks from patients and newspapers: reviews on drugs, therapeutic procedures, or medical devices | ||||
| Pérez et al., (16) | Identify opinions on the drugs prescribed for chronic-degenerative diseases (including hypertension medication) | Blogs and specialized websites in the Spanish language | Hybrid approach (supervised machine learning and use of semantics through a tagged corpus) | The analysis of the sentiments of the opinions on the prescribed drugs is successful and reduces time and effort |
| Austin et al., (17) | Understand patients' attitudes toward LVAD therapy | Posts, comments, and titles from MyLVAD.com | Lexicon-based SA | Positive sentiment words are the most frequent. In comparison to other LVAD complications, “infection” is mentioned disproportionately more times. |
| Emerging Markets, (18) | Assess whether Biotricity (health tech company targeting mainly chronic CVDs) trends positively or not in the media | News media | InfoTrie Financial SA Solutions | Biotricity has been trending positively, achieving a news buzz score of 10 out of 10, with a market sentiment score of 4.0 |
| 6. SA modules integrated in new technological concepts for monitoring CV patients | ||||
| Sharma et al., (19) | Propose a smart conceptual framework for monitoring patients with CV or diabetes | Social media and other online resources (for the SA component) | Hybrid system merging SA techniques, data mining, ML, IoT, bio-sensors, chatbots, contextual entity search, granular computing | Not available |
Cardiovascular disease (CVD); Atherosclerotic heart disease (AHD); United States (US); Coronary arteries diseases (CAD); Myocardial infarction (MI); Logistic regression (LR); Term Frequency * Inverse document frequency (TF.IDF); Left ventricular assist device (LVAD); Machine Learning (ML); Internet of Things (IoT).