Correction: Development of an Online Health Care Assessment for Preventive Medicine: A Machine Learning Approach.

[This corrects the article DOI: 10.2196/18585.].

Development of an Online Health Care Assessment for Preventive Medicine: A Machine Learning Approach.

BACKGROUND: In the era of information explosion, the use of the internet to assist with clinical practice and diagnosis has become a cutting-edge area of research. The application of medical informatics allows patients to be aware of their clinical conditions, which may contribute toward the prevention of several chronic diseases and disorders. OBJECTIVE: In this study, we applied machine learning techniques to construct a medical database system from electronic medical records (EMRs) of subjects who have undergone health examination. This system aims to provide online self-health evaluation to clinicians and patients worldwide, enabling personalized health and preventive health. METHODS: We built a medical database system based on the literature, and data preprocessing and cleaning were performed for the database. We utilized both supervised and unsupervised machine learning technology to analyze the EMR data to establish prediction models. The models with EMR databases were then applied to the internet platform. RESULTS: The validation data were used to validate the online diagnosis prediction system. The accuracy of the prediction model for metabolic syndrome reached 91%, and the area under the receiver operating characteristic (ROC) curve was 0.904 in this system. For chronic kidney disease, the prediction accuracy of the model reached 94.7%, and the area under the ROC curve (AUC) was 0.982. In addition, the system also provided disease diagnosis visualization via clustering, allowing users to check their outcome compared with those in the medical database, enabling increased awareness for a healthier lifestyle. CONCLUSIONS: Our web-based health care machine learning system allowed users to access online diagnosis predictions and provided a health examination report. Users could understand and review their health status accordingly. In the future, we aim to connect hospitals worldwide with our platform, so that health care practitioners can make diagnoses or provide patient education to remote patients. This platform can increase the value of preventive medicine and telemedicine.

Predicting Metabolic Syndrome With Machine Learning Models Using a Decision Tree Algorithm: Retrospective Cohort Study.

BACKGROUND: Metabolic syndrome is a cluster of disorders that significantly influence the development and deterioration of numerous diseases. FibroScan is an ultrasound device that was recently shown to predict metabolic syndrome with moderate accuracy. However, previous research regarding prediction of metabolic syndrome in subjects examined with FibroScan has been mainly based on conventional statistical models. Alternatively, machine learning, whereby a computer algorithm learns from prior experience, has better predictive performance over conventional statistical modeling. OBJECTIVE: We aimed to evaluate the accuracy of different decision tree machine learning algorithms to predict the state of metabolic syndrome in self-paid health examination subjects who were examined with FibroScan. METHODS: Multivariate logistic regression was conducted for every known risk factor of metabolic syndrome. Principal components analysis was used to visualize the distribution of metabolic syndrome patients. We further applied various statistical machine learning techniques to visualize and investigate the pattern and relationship between metabolic syndrome and several risk variables. RESULTS: Obesity, serum glutamic-oxalocetic transaminase, serum glutamic pyruvic transaminase, controlled attenuation parameter score, and glycated hemoglobin emerged as significant risk factors in multivariate logistic regression. The area under the receiver operating characteristic curve values for classification and regression trees and for the random forest were 0.831 and 0.904, respectively. CONCLUSIONS: Machine learning technology facilitates the identification of metabolic syndrome in self-paid health examination subjects with high accuracy.

Clustering Heatmap for Visualizing and Exploring Complex and High-dimensional Data Related to Chronic Kidney Disease.

BACKGROUND: Preventive medicine and primary health care are essential for patients with chronic kidney disease (CKD) because the symptoms of CKD may not appear until the renal function is severely compromised. Early identification of the risk factors of CKD is critical for preventing kidney damage and adverse outcomes. Early recognition of rapid progression to advanced CKD in certain high-risk populations is vital. METHODS: This is a retrospective cohort study, the population screened and the site where the study has been performed. Multivariate statistical analysis was used to assess the prediction of CKD as many potential risk factors are involved. The clustering heatmap and random forest provides an interactive visualization for the classification of patients with different CKD stages. RESULTS: uric acid, blood urea nitrogen, waist circumference, serum glutamic oxaloacetic transaminase, and hemoglobin A1c (HbA1c) were significantly associated with CKD. CKD was highly associated with obesity, hyperglycemia, and liver function. Hypertension and HbA1c were in the same cluster with a similar pattern, whereas high-density lipoprotein cholesterol had an opposite pattern, which was also verified using heatmap. Early staged CKD patients who are grouped into the same cluster as advanced staged CKD patients could be at high risk for rapid decline of kidney function and should be closely monitored. CONCLUSIONS: The clustering heatmap provided a new predictive model of health care management for patients at high risk of rapid CKD progression. This model could help physicians make an accurate diagnosis of this progressive and complex disease.

Evaluation of urate-lowering therapy in hyperuricemia patients: a systematic review and Bayesian network meta-analysis of randomized controlled trials.

OBJECTIVE: Hyperuricemia is a strong precursor of gout, which deteriorates patients' health and quality of life. Sustained adherence to urate-lowering therapies (ULTs) is crucial for efficacy and therapeutic cost-effectiveness. Recently, several new ULTs have been proposed. We performed a systematic review and meta-analysis of randomized controlled trials (RCTs) to reassess the efficacy and safety of the current ULTs, focusing on adherence attrition-related adverse event reporting. METHOD: The Bayesian network meta-analysis was applied to compare ULTs. Drug efficacy and safety were measured by whether the target level of serum urate acid was achieved and whether any adverse events occurred. The results were summarized using the pooled estimates of effect sizes (odds ratios), their precisions (95% credible interval), and the ranking probabilities. RESULTS AND CONCLUSIONS: Thirty-nine RCTs were identified, accumulating 19,401 patients. Consistent with previous studies, febuxostat (≥ 40 mg/day) was superior to other monoagent ULTs. The new findings were as follows: (i) dual-agent ULTs were superior to febuxostat alone, and further surveillance on the adverse effects when lesinurad is uptitrated is needed, and (ii) terminalia bellerica 500 mg/day, a novel xanthine oxidase inhibitor (XOI) made of natural fruit extracts, and topiroxostat ≥ 80 mg/day, an XOI used mostly in Japan, could be new effective options for lowering the occurrence of adherence attrition events. Evidence from RCTs regarding second-line agents, such as probenecid and pegloticase, remains insufficient for clinical decision-making.Key Points• Dual-agent ULTs were superior to febuxostat alone, and further surveillance on the adverse-effects when lesinurad is uptitrated is needed.• Terminalia bellerica 500 mg/day, a novel xanthine oxidase inhibitor (XOI) made of natural fruit extracts, and topiroxostat 80 mg/day, an XOI used mostly in Japan, could be new effective options for lowering the occurrence of adherence attrition events.

Noninvasive and Convenient Screening of Metabolic Syndrome Using the Controlled Attenuation Parameter Technology: An Evaluation Based on Self-Paid Health Examination Participants.

BACKGROUND: There is a medical need for an easy, fast, and non-invasive method for metabolic syndrome (MetS) screening. This study aimed to assess the ability of FibroScan to detect MetS, in participants who underwent a self-paid health examination. METHODS: A retrospective cohort study was conducted on all adults who underwent a self-paid health examination comprising of an abdominal transient elastography inspection using FibroScan 502 Touch from March 2015 to February 2019. FibroScan can assess the level of liver fibrosis by using a liver stiffness score, and the level of liver steatosis by using the controlled attenuation parameter (CAP) score. The logistic regression analysis and receiver operating characteristic curve were applied to select significant predictors and assess their predictability. A final model that included all significant predictors that are found by univariate analysis, and a convenient model that excluded all invasive parameters were created. RESULTS: Of 1983 participants, 13.6% had a physical status that fulfilled MetS criteria. The results showed that the CAP score solely could achieve an area under the curve (AUC) of 0.79 (0.76-0.82) in predicting MetS, and the AUC can be improved to 0.88 (0.85-0.90) in the final model. An AUC of 0.85 (0.83-0.88) in predicting MetS was obtained in the convenient model, which includes only 4 parameters (CAP score, gender, age, and BMI). A panel of predictability indices (the ranges of sensitivity, specificity, positive and negative likelihood ratio: 0.78-0.89, 0.66-0.82, 2.64-4.47, and 0.17-0.26) concerning gender- and BMI-specific CAP cut-off values (range: 191.65-564.95) were presented for practical reference. CONCLUSIONS: Two prediction systems were proposed for identifying individuals with a physical status that fulfilled the MetS criteria, and a panel of predictability indices was presented for practical reference. Both systems had moderate predictive performance. The findings suggested that FibroScan evaluation is appropriate as a first-line MetS screening; however, the variation in prediction performance of such systems among groups with varying metabolic derangements warrants further studies in the future.

A polygenic risk score for breast cancer risk in a Taiwanese population.

BACKGROUND: Multiple common variants identified by genome-wide association studies showed limited evidence of the risk of breast cancer in Taiwan. In this study, we analyzed the breast cancer risk in relation to 13 individual single-nucleotide polymorphisms (SNPs) identified by a GWAS in an Asian population. METHODS: In total, 446 breast cancer patients and 514 healthy controls were recruited for this case-control study. In addition, we developed a polygenic risk score (PRS) including those variants significantly associated with breast cancer risk, and also evaluated the contribution of PRS and clinical risk factors to breast cancer using receiver operating characteristic curve (AUC). RESULTS: Logistic regression results showed that nine individual SNPs were significantly associated with breast cancer risk after multiple testing. Among all SNPs, six variants, namely FGFR2 (rs2981582), HCN1 (rs981782), MAP3K1 (rs889312), TOX3 (rs3803662), ZNF365 (rs10822013), and RAD51B (rs3784099), were selected to create PRS model. A dose-response association was observed between breast cancer risk and the PRS. Women in the highest quartile of PRS had a significantly increased risk compared to women in the lowest quartile (odds ratio 2.26; 95% confidence interval 1.51-3.38). The AUC for a model which contained the PRS in addition to clinical risk factors was 66.52%, whereas that for a model which with established risk factors only was 63.38%. CONCLUSIONS: Our data identified a genetic risk predictor of breast cancer in Taiwanese population and suggest that risk models including PRS and clinical risk factors are useful in discriminating women at high risk of breast cancer from those at low risk.

Risk factors correlated with risk of insulin resistance using homeostasis model assessment in adolescents in Taiwan.

The study aims to discover risk factors significantly correlated with insulin resistance among adolescents in Taiwan. A total of 339 study subjects were recruited in this cross-sectional study. A self-administered questionnaire and physical examinations including anthropometrics and biochemistry profiles were collected. Insulin resistance was assessed using homeostasis model assessment for insulin resistance (HOMA-IR). Study subjects had a significantly increased risk of IR for those with abnormal level of body mass index (odds ratio [OR] = 3.54; 95% confidence interval [CI] = 1.81-6.91), body fat (OR = 2.71; 95% CI = 1.25-5.88), and waist circumference (OR = 25.04; 95% CI = 2.93-214.14) when compared with those who have normal values. Furthermore, a significantly joint effect of 10.86-fold risk for HOMA-IR abnormality among body fat, body mass index, and systolic blood pressure was observed. The identification of risk factors significantly correlated with IR will be important to prevent metabolic syndrome-related diseases and complications for adolescents in their future life.

Risk of ovarian cancer in women with pelvic inflammatory disease: a population-based study.

BACKGROUND: Ovarian cancer is commonly fatal and incidence has persistently risen in Taiwan over the past 20 years. Prevention strategies, however, are limited. Pelvic inflammatory disease (PID) has been suggested to increase the risk of developing ovarian cancer, but the results of studies have been inconsistent. Therefore, we investigated whether PID increases the risk of developing ovarian cancer in a large, nationwide cohort. METHODS: From the Longitudinal Health Insurance Database 2005 (LHID2005) in Taiwan, we obtained data for women aged 13-65 years for whom a diagnosis of PID, confirmed by multiple episodes, had been recorded between Jan 1, 2004, and Dec 31, 2005. We also obtained data for two controls per patient, matched for age and the year of first entry into the LHID2005. All patients were followed up from the date of entry in the LHID2005 until they developed ovarian cancer or to the end of 2006, whichever was earlier. We used Cox's regression models to assess the risk of developing ovarian cancer, with adjustment for age, comorbid disorders, and socioeconomic characteristics. FINDINGS: We identified 67,936 women with PID and 135,872 controls. Among these 90 had developed ovarian cancer during the 3-year follow-up period (42 patients with PID and 48 controls, incidence 2·78 and 1·44 per 10,000 person-years, respectively). The adjusted hazard ratio for ovarian cancer in patients with PID was 1·92 (95% CI 1·27-2·92) compared with controls, which rose to 2·46 (1·48-4·09) in women who had had at least five episodes of PID. The adjusted hazard ratio was slightly higher for women aged 35 years or younger with PID than in older women with PID (2·23, 1·02-4·79 vs 1·82, 1·10-3·04). INTERPRETATION: We found an association between PID and ovarian cancer. PID might, therefore, be a useful marker for ovarian cancer, and early treatment could help to improve prognosis. Whether pelvic inflammation itself accelerates the growth of ovarian cancers or affects cancer-cell differentiation in ways that adversely alter prognosis needs to be investigated. FUNDING: None.

A preliminary investigation of the association between serum uric acid and impaired renal function.

BACKGROUND: Hemodialysis for end-stage renal disease (ESRD) incurs huge medical costs in Taiwan. We set out to determine if it is possible to help control chronic renal disease with early treatment of hyperuricemia. METHODS: Data from Taipei Medical University Hospital (TMUH) health center from January 2004 to December 2006 were analyzed to correlate renal function and blood uric acid concentration. Patients were divided into 5 groups according to their serum uric acid concentration (< 4; 4 approximately 5.9; 6 approximately 7.9; 8 approximately 9.9, and > 10 mg/dl). According to our laboratory data, elevated serum creatinine levels (> 1.3 mg/dL) indicated impaired renal function. RESULTS: In total, there were 5722 patients, including 2816 (49.2%) men and 2906 (50.8%) women, with a median age of 67. Impaired renal function was noted in 307 (5.4%) cases. Serum uric acid was significantly correlated with blood urea nitrogen and serum creatinine. Groups with a higher serum uric acid level had an increased risk of impaired renal function. CONCLUSION: Our purpose in this preliminary observation was to try to define a starting point for the early control of serum uric acid, in order to avoid the development of impaired renal function. We found that serum uric acid level to < 6 mg/dl seemed to be associated with less renal function impairment.

Weekly pattern of stroke onset in an Asian country: a nationwide population-based study.

This study used a nationwide population-based dataset to explore the variation among the days of week of stroke onset within population subgroups defined by age, sex, and stroke type. We used ambulatory care data from the 2002 Taiwan National Health Insurance Research Database, focusing on 42,779 emergency room (ER) visits for stroke that year. All analyses were stratified by sex, age (<60 and > or =60 yrs), and type of stroke. Auto-Regressive Integrated Moving Average (ARIMA) was performed to investigate the relationship between daily number of stroke events and holidays and days of the week after adjusting for the effects of seasonality and trends. One-way ANOVA revealed significant differences in stroke ER admissions according to day of week according to age <60 (p<0.01), age > or =60 (p<0.001), male (p<0.001), female (p<0.001), ischemic stroke (IS) (p<0.001), and unspecified stroke (UNSP) (p<0.001). However, the analysis by type-subarachnoid hemorrhage and intracerebral hemorrhage-did not show significant relationships between daily emergency room stroke admissions, holidays, or day of the week. The ARIMA regression analyses also showed that Mondays had the highest rate of emergency room admissions for stroke regardless of sex, age, or IS and UNSP types of stroke, after adjusting for seasonality and trends. We conclude that stroke occurs more frequently on Mondays than on the other days of the week, which might be associated with short-term changes in lifestyle or due to the sudden return of stress on the first working day of the week, and on holidays.

Urbanization and the likelihood of a cesarean section.

OBJECTIVE: This study examines the association between the likelihood of cesarean section (CS) and the degree of urbanization in Taiwan, exploring possible explanations for the difference. STUDY DESIGN: The database used in this study was the Taiwan 2004 National Health Insurance Research Database. A total of 200,207 singleton deliveries fulfilled our criteria and were included in our study. The urbanization level of cities/towns where parturients resided at the time of delivery was stratified into seven categories. A multilevel logistic regression model was applied to examine the relative likelihood of CS by urbanization level after adjusting for parturient, physician and hospital characteristics. RESULTS: There was an upward trend in the CS rate with advancing urbanization level; the CS rates for urbanization level 1 (most urbanized) through 7 (least urbanized) were 33.7, 32.3, 30.4, 30.2 29.7, 29.5, and 28.6%, respectively. Compared with participants living at the highest urbanization level, the adjusted odds of a CS were 0.91 (95% CI=0.85-0.98, p=0.014), 0.84 (95% CI=0.78-0.91, p<0.001), 0.83 (95% CI=0.68-0.88, p<0.001), 0.79 (95% CI=0.72-0.86, p<0.001), and 0.70 (95% CI=0.62-0.80, p<0.001) times, respectively, for those living in cities/towns ranked from the third highest to the lowest levels of urbanization. CONCLUSIONS: We conclude that higher urbanization levels were associated with higher odds of CS. Highly urbanized communities could therefore be targeted for policy intervention aimed at reducing the unnecessary CS rate.

Contact of mental and nonmental health care providers prior to suicide in Taiwan: a population-based study.

OBJECTIVE: Higher rates of health care service use prior to suicide were previously reported in Western countries; however, these studies have tended to suffer from small sample sizes. This nationwide, population-based study examines the distribution and patterns of health care service use among suicide victims in Taiwan. METHOD: A retrospective cohort study was conducted using linked population-based data to determine the proportion of health care service use among suicide victims aged 15 years and older within the 1-year and 1-month period prior to their deaths. After adjusting for demographic, socioeconomic and health care indices, the differences in health care service use patterns were assessed for age and sex. RESULTS: Among the 19 426 suicide victims in the sample, 83.1% had used nonmental health care services within the 1-year period prior to their death, while only 22.2% had used mental health care services. Men, and suicide victims aged 55 years and older, were less likely to have had any contact with mental health care professionals prior to their deaths (P < 0.001). CONCLUSIONS: In line with prior studies, similarly high rates and distinct patterns of health care service use were found in Taiwan prior to suicide. These findings will be of practical interest and should support designing appropriate methods of suicide intervention and effective preventive strategies.

Increased risk of stroke in patients who undergo cesarean section delivery: a nationwide population-based study.

OBJECTIVE: This study used a population-based dataset to determine whether (compared with vaginal deliveries), cesarean section deliveries increase the risk of postpartum stroke during the 3-, 6-, or 12-month period after delivery. STUDY DESIGN: This study used 1998-2003 records from the Taiwan National Health Insurance Research Database for 987,010 women with singleton deliveries from 1998-2002. Cox proportional hazard regressions were carried out to compute stroke-free survival rates between the 2 delivery modes. RESULTS: The regression model indicated that, compared with patients who delivered vaginally, the hazard ratio for postpartum stroke among those who delivered by cesarean section was 1.67 times greater within 3 months of delivery (95% CI, 1.29-2.16), was 1.61 times greater within 6 months of delivery (95% CI, 1.31-1.98), and was 1.49 times greater within 12 months of delivery (95% CI, 1.27-1.76). CONCLUSION: Our data indicates that cesarean section delivery is an independent risk factor for stroke.

Effectiveness of acupuncture in the treatment of fibromyalgia.

CONTEXT: Fibromyalgia syndrome (FMS) is a prevalent musculoskeletal disorder associated with pain, mood state alteration, and disability. A structured and effective treatment plan for palliative care has not been established. The genesis of FMS is not clear. FMS occurs primarily in adult women. DESIGN: Using a quasi-experimental clinical design and following the criteria of the American College of Rheumatology (ACR), for FMS, 21 participants completed the study. The mean age was 53.6 years. The data were collected at baseline and at 1 and 2 months. Acupuncture treatments included 17 points for FMS symptoms, and 8 outcome measures were collected. RESULTS: The Fibromyalgia Impact Questionnaire (FIQ) showed significant differences at 1 and 2 months. For the SF-12, 3 subscales showed significant differences between baseline and 2 months. Four of 6 items were significantly changed. The mean number of general health symptoms was significantly decreased by 2 months. For the Catastrophe Index, significant differences were found for baseline vs 2 months. Pain threshold scores were significantly different at end of treatment for 5 bilateral tender points. There was significant improvement in Beck Depression items for both 1- and 2-month periods. In a multivariate regression model, 5 covariates were included--age, number of weeks in treatment, number of doctors treating, number of general symptoms, and baseline FIQ score. The results indicated significant age effect. This analysis showed that the higher the FIQ score, the more positive the change experienced by study participants. Number of weeks in treatment, number of doctors who treated, and total number of general health symptoms did not have a significant effect on outcomes. CONCLUSIONS: Significant improvement was experienced by participants at 8 weeks of treatment. Acupuncture treatment as delivered was effective at reducing FMS symptoms in this outcome study.