ILI-Feng Lin

I-Feng Lin

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


I-Feng Lin has a total of 75 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Influenza vaccination is associated with a reduced risk of invasive aspergillosis in high-risk individuals in Taiwan: a population-based cohort study

    by Yi-Jyun Chen, I-Feng Lin, Jen-Hsiang Chuang, Hung-Ling Huang, Ta-Chien Chan

    Invasive aspergillosis (IA) has become the emerging life-threatening disease in recent years. Influenza has been identified as an independent risk factor for IA. Vaccination is the most effective way to prevent influenza, while whether it can reduce IA in high-risk population still uncertain. We aimed to investigate the association between influenza vaccination and the risk of IA in high-risk population. We performed a population-based cohort study of people who qualified for government-funded influenza vaccination and were at high risk for IA at the start of the influenza season each year between 2016 and 2019. We utilized Taiwan’s National Health Insurance Research Database to identify the influenza vaccination status and IA diagnosis during the follow-up period. We compared the risk of IA between people with and without vaccination using multivariable logistic regression analysis. Out of total 8,544,451 people who were eligible during the 3 influenza seasons, 3,136,477 (36.7%) were vaccinated. A total of 1179 IA cases with the incidence of 13.8 cases per 100,000 high-risk individuals were identified during the follow-up. Compared to non-vaccinated group, vaccinated individuals had a 21% risk reduction of IA (adjusted odds ratio 0.79, 95% confidence interval 0.70–0.90). Influenza vaccination was associated with a lower risk of IA among males, immunosuppressive conditions, malignancy, diabetes, and those having host factors according to the European Organization for Research and Treatment of Cancer and the Mycoses Study Group Education and Research Consortium. Influenza vaccination is recommended for high-risk population to reduce the risk of IA.


  • Open Access

    Resting Dead Space Fraction as Related to Clinical Characteristics, Lung Function, and Gas Exchange in Male Patients with Chronic Obstructive Pulmonary Disease

    by Ming-Lung Chuang, /sup> Benjamin Yung-Thing Hsieh, I-Feng Lin

    Background: Measures of forced expired volume in one second % predicted (FEV1%), residual volume to total lung capacity ratio (RV/TLC) and diffusing capacity for carbon monoxide measurements (DLCO) are the standard lung function test for evaluating patients with chronic obstructive pulmonary disease (COPD). The dead space fraction (VD/VT) has been shown to be a robust marker of gas exchange abnormality. However, the use of VD/VT has gradually become less common. As VD/VT measured at rest (VD/VTR) has been successfully used in non-COPD conditions, it was hypothesized that in COPD the VD/VTR was more sensitive than the standard lung function test in correlation with clinical characteristics and gas exchange. This study aimed to test the hypothesis and to identify the variables relevant to VD/VTR. Methods: A total of 46 male subjects with COPD were enrolled. Clinical characteristics included demographic data, oxygen-cost diagram (OCD), and image studies for pulmonary hypertension. The standard lung function was obtained. To calculate VD/VT, invasive arterial blood gas and pulmonary gas exchange (PGX) were measured. The variables relevant to VD/VTR were analyzed by multiple linear regression. Results: Compared to lung function, VD/VTR was more frequently and significantly related to smoking, carboxyhemoglobin level, pulmonary hypertension and PaCO2 (all p < 0.05) whereas FEV1% was more related to lung function test, PaO2 and OCD score. VD/VTR and FEV1% were highly related to resting gas exchange but RV/TLC and DLCO% were not. Cigarette consumption, the equivalent for CO2 output, arterial oxyhemoglobin saturation, and the product of tidal volume and inspiratory duty cycle were identified as the parameters relevant to VD/VTR with a power of 0.72. Conclusion: Compared to lung function test, VD/VTR is more related to clinical characteristics and is a comprehensive marker of resting gas exchange. Further studies are warranted to provide a noninvasive measurement of VD/VTR. Registration Number: MOST 106– 2314-B-040-025 and CSH-2019-C-30.


  • Open Access

    Applying Machine Learning Models with An Ensemble Approach for Accurate Real-Time Influenza Forecasting in Taiwan: Development and Validation Study

    by Hao-Yuan Cheng, Yu-Chun Wu, Min-Hau Lin, Yu-Lun Liu, Yue-Yang Tsai, Jo-Hua Wu, Ke-Han Pan, Chih-Jung Ke, Chiu-Mei Chen, Ding-Ping Liu, et al.

    Background: Changeful seasonal influenza activity in subtropical areas such as Taiwan causes problems in epidemic preparedness. The Taiwan Centers for Disease Control has maintained real-time national influenza surveillance systems since 2004. Except for timely monitoring, epidemic forecasting using the national influenza surveillance data can provide pivotal information for public health response. Objective: We aimed to develop predictive models using machine learning to provide real-time influenza-like illness forecasts. Methods: Using surveillance data of influenza-like illness visits from emergency departments (from the Real-Time Outbreak and Disease Surveillance System), outpatient departments (from the National Health Insurance database), and the records of patients with severe influenza with complications (from the National Notifiable Disease Surveillance System), we developed 4 machine learning models (autoregressive integrated moving average, random forest, support vector regression, and extreme gradient boosting) to produce weekly influenza-like illness predictions for a given week and 3 subsequent weeks. We established a framework of the machine learning models and used an ensemble approach called stacking to integrate these predictions. We trained the models using historical data from 2008-2014. We evaluated their predictive ability during 2015-2017 for each of the 4-week time periods using Pearson correlation, mean absolute percentage error (MAPE), and hit rate of trend prediction. A dashboard website was built to visualize the forecasts, and the results of real-world implementation of this forecasting framework in 2018 were evaluated using the same metrics. Results: All models could accurately predict the timing and magnitudes of the seasonal peaks in the then-current week (nowcast) (ρ=0.802-0.965; MAPE: 5.2%-9.2%; hit rate: 0.577-0.756), 1-week (ρ=0.803-0.918; MAPE: 8.3%-11.8%; hit rate: 0.643-0.747), 2-week (ρ=0.783-0.867; MAPE: 10.1%-15.3%; hit rate: 0.669-0.734), and 3-week forecasts (ρ=0.676-0.801; MAPE: 12.0%-18.9%; hit rate: 0.643-0.786), especially the ensemble model. In real-world implementation in 2018, the forecasting performance was still accurate in nowcasts (ρ=0.875-0.969; MAPE: 5.3%-8.0%; hit rate: 0.582-0.782) and remained satisfactory in 3-week forecasts (ρ=0.721-0.908; MAPE: 7.6%-13.5%; hit rate: 0.596-0.904). Conclusions: This machine learning and ensemble approach can make accurate, real-time influenza-like illness forecasts for a 4-week period, and thus, facilitate decision making.


  • Family quality of life among Taiwanese children with developmental delay before and after early intervention

    by S.‐J. Chiu, I.‐F. Lin, Y.‐T. Chou, L.‐Y. Chien

    Background: Although early intervention (EI) practitioners emphasise the importance of individualised family‐centred services for families of children with developmental delay (DD), few empirical studies have evaluated whether EI can improve family quality of life (FQOL). This study aimed to investigate the trajectory of FQOL and its predictors among families of children with DD during the first 12 months of EI.Methods: This study employed a prospective cohort design. Data were collected using structured questionnaires at the placement meeting before the commencement of EI, as well as 3, 6 and 12 months later. We recruited 142 primary caregivers of children with DD in northern Taiwan from March 2015 to August 2016. FQOL was measured using the Mandarin Chinese version of the Beach Centre FQOL Scale. Family resilience (FR) was measured using the Mandarin Chinese version of the FR Assessment Scale. Other independent variables included socio‐demographics, type of DD and EI services. Generalised estimating equations were used to perform multivariate analysis.Results: Family quality of life exhibited a significant quadratic trend in the 12 months surrounding EI. The score was the lowest before EI started (89.85), then increased to peak (94.87) at 6 months and then decreased slightly to 92.34 at 12 months. FR followed a significantly increased linear trend during the period. There were significant and positive correlations between FQOL and FR across all time points. Multivariate analysis showed that employed caregivers, FR, sufficient caregiving manpower and satisfaction with marital quality were positively associated with FQOL. Receiving more types of EI services and having fathers who were not Taiwanese nationals were negatively associated with FQOL.Conclusions: Family quality of life and FR increased significantly after receiving EI, revealing the latter's effectiveness. Unemployment, poor marital quality, father being an immigrant, low FR and insufficient family caregiving manpower were associated with lower FQOL, suggesting that these families require more assistance.


  • Open Access

    More Impaired Dynamic Ventilatory Muscle Oxygenation in Congestive Heart Failure than in Chronic Obstructive Pulmonary Disease

    by Ming-Lung Chuang, I-Feng Lin, Meng-Jer Hsieh

    Patients with chronic obstructive pulmonary disease (COPD) and congestive heart failure (CHF) often have dyspnea. Despite differences in primary organ derangement and similarities in secondary skeletal muscle changes, both patient groups have prominent functional impairment. With similar daily exercise performance in patients with CHF and COPD, we hypothesized that patients with CHF would have worse ventilatory muscle oxygenation than patients with COPD. This study aimed to compare differences in tissue oxygenation and blood capacity between ventilatory muscles and leg muscles and between the two patient groups. Demographic data, lung function, and maximal cardiopulmonary exercise tests were performed in 134 subjects without acute illnesses. Muscle oxygenation and blood capacity were measured using frequency-domain near-infrared spectroscopy (fd-NIRS). We enrolled normal subjects and patients with COPD and CHF. The two patient groups were matched by oxygen-cost diagram scores, New York Heart Association functional classification scores, and modified Medical Research Council scores. COPD was defined as forced expired volume in one second and forced expired vital capacity ratio ≤0.7. CHF was defined as stable heart failure with an ejection fraction ≤49%. The healthy subjects were defined as those with no obvious history of chronic disease. Age, body mass index, cigarette consumption, lung function, and exercise capacity were different across the three groups. Muscle oxygenation and blood capacity were adjusted accordingly. Leg muscles had higher deoxygenation (HHb) and oxygenation (HbO2) and lower oxygen saturation (SmO2) than ventilatory muscles in all participants. The SmO2 of leg muscles was lower than that of ventilatory muscles because SmO2 was calculated as HbO2/(HHb+HbO2), and the HHb of leg muscles was relatively higher than the HbO2 of leg muscles. The healthy subjects had higher SmO2, the patients with COPD had higher HHb, and the patients with CHF had lower HbO2 in both muscle groups throughout the tests. The patients with CHF had lower SmO2 of ventilatory muscles than the patients with COPD at peak exercise (p < 0.01). We conclud that fd-NIRS can be used to discriminate tissue oxygenation of different musculatures and disease entities. More studies on interventions on ventilatory muscle oxygenation in patients with CHF and COPD are warranted.


  • Open Access

    Variation in the 11-year trajectories of medical care seeking behaviors in diabetes patients under a single payer system: persisting gaps to be filled

    by Tzu-Ho Tsai, Nicole Huang, I-Feng Lin, Yiing-Jenq Chou

    Care-seeking behavior is widely acknowledged to have strong influences on health outcomes among individuals with chronic conditions including diabetes. Despite its dynamic nature, care seeking behavior are often considered as time invariant in most studies. The likelihood of patients changing their regularity and source of chronic care over time is often neglected. This study aimed to determine the long-term trajectories of care-seeking patterns of both care-seeking regularity and health provider choices; and their associated factors among patients with type 2 diabetes under the National Health Insurance (NHI) program in Taiwan. We utilized population-based data from the National Health Insurance Research Database (NHIRD) in Taiwan. Three thousand, nine hundred and eighty-seven adult patients with newly diagnosed type 2 diabetes in 1999 were enrolled in the cohort. We assessed their trajectories of regular care visits and sources of diabetes care from 2000 to 2010. A group-based trajectory model was applied. Seven distinct groups of long-term care-seeking patterns were identified. Only 51.44% of patients with newly diagnosed diabetes had regularly visited their providers over time. Among them, 56.41 and 16.09% had persistently sought care from generalized and specialized providers, respectively. 27.50% had sought care from different levels of providers. Patients who were male, elderly, low-income, and had a higher baseline diabetes severity were significantly more likely to either continue with their irregular care-seeking behavior or fail to maintain their regular care seeking behavior over time. Those who were younger, had a higher socioeconomic status, and lived in an urban area were significantly more likely to persistently seek care from specialized care settings. This study is the first population-based assessment of long-term care-seeking behaviors of type 2 diabetes patients under a single-payer system with a comprehensive benefit coverage. The most alarming finding was that, despite the existence of the comprehensive universal health insurance coverage in Taiwan, almost 50% of patients did not seek or maintain regular visits to providers over time as recommended. Understanding variations in the long-term trajectories of care adherence and sources of care may help to identify gaps in diabetes care management.


  • Open Access

    Applying Machine Learning Models with An Ensemble Approach for Accurate Real-Time Influenza Forecasting in Taiwan: Development and Validation Study (Preprint)

    by Hao-Yuan Cheng, Yu-Chun Wu, Min-Hau Lin, Yu-Lun Liu, Yue-Yang Tsai, Jo-Hua Wu, Ke-Han Pan, Chih-Jung Ke, Chiu-Mei Chen, Ding-Ping Liu, et al.

    BACKGROUND Changeful seasonal influenza activity in subtropical areas such as Taiwan causes problems in epidemic preparedness. The Taiwan Centers for Disease Control has maintained real-time national influenza surveillance systems since 2004. Except for timely monitoring, epidemic forecasting using the national influenza surveillance data can provide pivotal information for public health response. OBJECTIVE We aimed to develop predictive models using machine learning to provide real-time influenza-like illness forecasts. METHODS Using surveillance data of influenza-like illness visits from emergency departments (from the Real-Time Outbreak and Disease Surveillance System), outpatient departments (from the National Health Insurance database), and the records of patients with severe influenza with complications (from the National Notifiable Disease Surveillance System), we developed 4 machine learning models (autoregressive integrated moving average, random forest, support vector regression, and extreme gradient boosting) to produce weekly influenza-like illness predictions for a given week and 3 subsequent weeks. We established a framework of the machine learning models and used an ensemble approach called stacking to integrate these predictions. We trained the models using historical data from 2008-2014. We evaluated their predictive ability during 2015-2017 for each of the 4-week time periods using Pearson correlation, mean absolute percentage error (MAPE), and hit rate of trend prediction. A dashboard website was built to visualize the forecasts, and the results of real-world implementation of this forecasting framework in 2018 were evaluated using the same metrics. RESULTS All models could accurately predict the timing and magnitudes of the seasonal peaks in the then-current week (nowcast) (ρ=0.802-0.965; MAPE: 5.2%-9.2%; hit rate: 0.577-0.756), 1-week (ρ=0.803-0.918; MAPE: 8.3%-11.8%; hit rate: 0.643-0.747), 2-week (ρ=0.783-0.867; MAPE: 10.1%-15.3%; hit rate: 0.669-0.734), and 3-week forecasts (ρ=0.676-0.801; MAPE: 12.0%-18.9%; hit rate: 0.643-0.786), especially the ensemble model. In real-world implementation in 2018, the forecasting performance was still accurate in nowcasts (ρ=0.875-0.969; MAPE: 5.3%-8.0%; hit rate: 0.582-0.782) and remained satisfactory in 3-week forecasts (ρ=0.721-0.908; MAPE: 7.6%-13.5%; hit rate: 0.596-0.904). CONCLUSIONS This machine learning and ensemble approach can make accurate, real-time influenza-like illness forecasts for a 4-week period, and thus, facilitate decision making.

  • Open Access

    Effort-reward imbalance and self-rated health among Gambian healthcare professionals

    by Amadou Darboe, I-Feng Lin, Hsien-Wen Kuo

    The effort-reward imbalance (ERI) model of work stress has been widely applied in investigating association between psychosocial factors at work and health. This study examined associations between perceived psychosocial work stress as measured by the ERI model and self-rated health (SRH) among nurses and environmental health officers (EHOs) working in secondary public healthcare facilities in the Gambia. A cross-sectional study on a random sample of 287 health care professionals (201 nurses and 86 EHOs). A 22-item ERI questionnaire was used to collect data on the psychosocial factors defined by the ERI model. SRH was assessed using a single item measure. The distribution of subjective health was not statistically different between nurses and EHOs. However, our study uncovered significant associations between perceived psychosocial work stress and subjective health. Specifically, we found that a perceived high effort-reward imbalance (ER-ratio > 1) is a significant risk factor for poor SRH, in both occupational groups. However, over-commitment was not significantly associated with poor SRH in the two groups. When efforts and rewards were considered as separate variables in the analysis, rewards were inversely associated with poor SRH in both groups. Because of the high perceived Effort-Reward Imbalance among healthcare professionals at secondary public healthcare facilities, it is necessary to modify working conditions through improvement of psychosocial work environment, such as reasonable allocation of resources to increase pay, incentives or other forms of rewards from government. Interventions that could mitigate and prevent stress at work are worth considering in future healthcare policies.


  • Managing nonignorable missing data with clustered multinomial responses

    by Mu-Cyun Wang, I-Feng Lin

    Clustered multinomial responses are common in public health studies. In this situation, the baseline logit random effects model is usually suggested as a general modelling approach. When nonignorable missing outcomes exist, naïve methods such as complete case analysis or likelihood methods ignoring missing information may distort the conclusions that are drawn. While methods to deal with binary and ordinal outcomes have been proposed, no easily implementable method is specifically available for missing clustered nominal responses. Joint modelling is usually one of the available choices but has high complexity in terms of likelihood. The numerical integration of both missing data and random effects is challenging. In this study, we have derived a closed form of likelihood. A simplified likelihood is also proposed, which is an extension of a previous study. One advantage is that both methods are easily implemented with commonly used software. We illustrate our proposed methods using the Global Youth Tobacco Survey and compare the results obtained by naïve methods that ignore missing data with the results obtained using the proposed methods. Our approaches restore the parameter estimates and predicted probability of each category to an acceptable extent. Analysis guidelines for the use of our methods are provided.