Machine learning-derived model for predicting poor post-treatment quality of life in Korean cancer survivors.

PURPOSE: A substantial number of cancer survivors have poor quality of life (QOL) even after completing cancer treatment. Thus, in this study, we used machine learning (ML) to develop predictive models for poor QOL in post-treatment cancer survivors in South Korea. METHODS: This cross-sectional study used online survey data from 1,005 post-treatment cancer survivors in South Korea. The outcome variable was QOL, which was measured using the global QOL subscale of the European Organization of Cancer and Treatment for Cancer Quality of Life Questionnaire, where a global QOL score < 60.4 was defined as poor QOL. Three ML models (random forest (RF), support vector machine, and extreme gradient boosting) and three deep learning models were used to develop predictive models for poor QOL. Model performance regarding accuracy, area under the receiver operating characteristic curve, F1 score, precision, and recall was evaluated. The SHapely Additive exPlanation (SHAP) method was used to identify important features. RESULTS: Of the 1,005 participants, 65.1% had poor QOL. Among the six models, the RF model had the best performance (accuracy = 0.85, F1 = 0.90). The SHAP method revealed that survivorship concerns (e.g., distress, pain, and fatigue) were the most important factors that affected poor QOL. CONCLUSIONS: The ML-based prediction model developed to predict poor QOL in Korean post-treatment cancer survivors showed good accuracy. The ML model proposed in this study can be used to support clinical decision-making in identifying survivors at risk of poor QOL.

Stronger association of perceived health with socio-economic inequality during COVID-19 pandemic than pre-pandemic era.

OBJECTIVE: The COVID-19 pandemic has changed peoples' routine of daily living and posed major risks to global health and economy. Few studies have examined differential impacts of economic factors on health during pandemic compared to pre-pandemic. We aimed to compare the strength of associations between perceived health and socioeconomic position (household income, educational attainment, and employment) estimated before and during the pandemic. METHODS: Two waves of nationwide survey [on 2018(T1;n = 1200) and 2021(T2;n = 1000)] were done for 2200 community adults. A balanced distribution of confounders (demographics and socioeconomic position) were achieved across the T2 and T1 by use of the inverse probability of treatment weighting. Distributions of perceived health [= (excellent or very good)/(bad, fair, or good)] for physical-mental-social-spiritual subdomains were compared between T1 and T2. Odds of bad/fair/good health for demographics and socioeconomic position were obtained by univariate logistic regression. Adjusted odds (aOR) of bad/fair/good health in lower household income(< 3000 U.S. dollars/month) were retrieved using the multiple hierarchical logistic regression models of T1 and T2. RESULTS: Perceived health of excellent/very good at T2 was higher than T1 for physical(T1 = 36.05%, T2 = 39.13%; P = 0.04), but were lower for mental(T1 = 38.71%, T2 = 35.17%; P = 0.01) and social(T1 = 42.48%, T2 = 35.17%; P < 0.001) subdomains. Odds of bad/fair/good health were significantly increased at T2 than T1 for household income (physical-mental-social; all Ps < 0.001) and educational attainment (social; P = 0.04) but not for employment (all Ps > 0.05). AORs of bad/fair/good health in lower household income were stronger in T2 than T1, for mental [aOR (95% CI) = 2.15(1.68-2.77) in T2, 1.33(1.06-1.68) in T1; aOR difference = 0.82(P < 0.001)], physical [aOR (95% CI) = 2.64(2.05-3.41) in T2, 1.50(1.18-1.90) in T1; aOR difference = 1.14(P < 0.001)] and social [aOR (95% CI) = 2.15(1.68-2.77) in T2, 1.33(1.06-1.68) in T1; aOR difference = 0.35(P = 0.049)] subdomains. CONCLUSIONS: Risks of perceived health worsening for mental and social subdomains in people with lower monthly household income or lower educational attainment became stronger during the COVID-19 pandemic compared to pre-pandemic era. In consideration of the prolonged pandemic as of mid-2022, policies aiming not only to sustain the monthly household income and compulsory education but also to actively enhance the perceived mental-social health status have to be executed and maintained.

RNF213 R4810K Variant in Suspected Unilateral Moyamoya Disease Predicts Contralateral Progression.

Background Early-stage unilateral moyamoya disease (MMD) is difficult to discriminate from isolated intracranial atherosclerotic stenosis, and identification of contralateral progression may aid in the diagnosis of MMD. The RNF213 (ring finger protein 213) R4810K variant is a strong genetic susceptibility factor for MMD; however, the role of contralateral progression in unilateral MMD is unknown. Methods and Results Patients who had undergone RNF213 R4810K genotyping with suspected unilateral MMD between January 2017 and August 2021 from 2 tertiary university hospitals were retrospectively reviewed. We compared the clinical features and radiographic outcomes of patients with and without this variant. The risk factors of contralateral progression in patients with suspected unilateral MMD were evaluated. The RNF213 R4810K variant was observed in 72 of 123 patients with suspected unilateral MMD, all of which were heterozygous. The allele frequency of the R4810K variant was significantly higher in the suspected unilateral MMD group compared with the historical control group (29.3% versus 1.2%; P<0.0001). Family history of MMD was significantly more common in patients with the variant than in those without (17% versus 4%; P=0.003). Eleven of 72 patients with the variant developed contralateral progression, whereas only 1 of 51 patients without the variant developed contralateral progression during a median follow-up period of 28 months (log-rank test; P=0.03). The presence of the RNF213 R4810K variant significantly correlated with contralateral progression (adjusted odds ratio, 6.39 [95% CI, 1.11-36.63]; P=0.04). Conclusions Contralateral progression is more likely to occur in patients with suspected unilateral MMD with the RNF213 R4810K variant than in those without the variant. However, because our study used a small sample size, this finding should be carefully interpreted and requires further studies with more patients and longer follow-up periods.

Clinical Implication of Candida Score in Multidrug-Resistant Pneumonia with Airway Candida Colonization.

BACKGROUND: The growth of Candida in respiratory secretions is usually considered colonization, and antifungal therapy is rarely required. The role of Candida colonization in the progression of bacterial pneumonia remains controversial. The aim of this study was to identify the clinical implication of Candida score by analyzinge the relationship with multidrug-resistant (MDR) pneumonia and prognosis in patients with airway Candida colonization. MATERIALS AND METHODS: This study was a retrospective review of patients with airway Candida colonization by bronchial washing or bronchoalveolar lavage. The Candida score was calculated according to the four factors (severe sepsis, surgery at baseline, total parenteral nutrition, and multifocal Candida colonization). Pneumonia related mortality or hopeless discharge expecting death was defined as a poor outcome. RESULTS: A total of 148 patients were enrolled in the study. In a multivariate analysis model, Candida score was identified as an independent predictor of poor outcomes (odds ratio 2.23; 95% confidential interval 1.57 - 3.17; P <0.001) in pneumonia patients with airway Candida colonization. With a Candida score of three or higher compared with low score group, it was associated with bacterial pneumonia, especially methicillin-resistant Staphylococcus aureus (MRSA) infection (0.0% vs. 15.2%, P = 0.004). In addition, patients with a high Candida score had a longer hospital stay (13 vs. 38 days, P <0.001), longer duration of intensive care (7 vs. 18 days, P <0.001), and higher pneumonia-related mortality (0.0% vs. 45.5%, P <0.001) as compared to the low Candida score group. The Candida score showed a positive correlation with other pneumonia severity scales such as CURB-65 (Confusion, Urea, Respiratory rate, Blood pressure, and age ≥65 years) (r = 0.461, P <0.001), Pneumonia Severity Index (r = 0.397, P <0.001), and predisposition, insult, response, and organ dysfunction (PIRO) score (r = 0.425, P <0.001). CONCLUSION: This study revealed that Candida is no longer a bystander of airway colonization, and that it affects the progression of bacterial pneumonia, including multidrug-resistant pathogens, particularly MRSA infection. Also Candida score can be used to predict the prognosis of patients with pneumonia.

Aggressiveness of Differentiated Thyroid Carcinoma in Pediatric Patients Younger Than 16 years: A Propensity Score-Matched Analysis.

Background: The biological behavior of thyroid cancer in children has been known to be different from that in adults. We sought to understand the differences between DTC presentation in pediatric (<16 years) and adult patients, to guide better prognosis and clinical treatments. Methods: This retrospective study included 48 pediatric patients younger than 16 years who underwent initial thyroid surgery and were diagnosed with DTC between January 1992 and December 2014 at Yonsei University in Seoul, South Korea. For a 1:4 propensity score-matched analysis, adult patients with matched sex and cancer size were included. Results: The mean age was 12.54 ± 3.01 years. Total thyroidectomy (70.8%) without lateral lymph node dissection (47.9%) was the most commonly performed surgery. Central (73.9%) and lateral neck node metastases (62.5%) were common; distant metastasis was observed in 2 (4.2%) patients and recurrence occurred in 11 (22.9%). In propensity score-matched analysis, central lymph node metastasis and lateral neck node metastasis were significantly more frequent in pediatric patients. Symptoms were more common in the pediatric group than in the adult group (p < 0.001). In stratified cox regression, pediatric patients were more likely to experience recurrence [HR 5.339 (1.239-23.007)]. In stratified log-rank analysis, recurrence-free survival was significantly different between the adult and pediatric groups (p = 0.0209). Conclusion: DTC in the pediatric group revealed more aggressive patterns than in the adult group with the same cancer size. Central lymph node metastasis and lateral neck node metastasis were more frequent. Stratified log-rank analysis revealed that recurrence was significantly higher in pediatric patients than in matched adult patients.

RT-induced dynamic changes in the lymphocyte-to-monocyte ratio in patients with breast cancer indicate poor prognosis.

PURPOSE: Radiation-induced anti-tumor responses occur in the immune system, particularly in peripheral blood mononuclear cells (PBMCs), which are overly sensitive to ionizing radiation. Irradiation of PBMCs is associated with inflammation. We assessed the association between radiotherapy (RT)-induced changes in peripheral blood cells, utilizing the lymphocyte-to-monocyte ratio (LMR), and survival outcome in breast cancer patients who underwent breast-conserving surgery followed by adjuvant RT. METHODS: LMR values were obtained from retrospective data, and serial sampling of blood before the first and last RT sessions was performed. The cut-off point was analyzed using the Youden index and receiver operating characteristic curve. Recurrence-free survival (RFS) and overall survival (OS) were the main outcomes. RESULTS: Patients with RT-induced low LMR had significantly shorter RFS (hazard ratio [HR] 2.730; 95% confidence interval [CI], 1.607-4.636, P = 0.0002) and OS (HR 2.890; 95% CI 1.410-5.924, P = 0.0038). The results were more robust in the subgroup of patients who received chemotherapy. Multivariate analysis showed that lymph node metastasis and RT-induced low LMR were associated with poor RFS (HR 1.763; 95% CI, 1.017-3.059, P = 0.044) and OS (HR 2.254; 95% CI, 1.060-4.796, P = 0.035). CONCLUSION: This study demonstrates that RT-induced low LMR is a valid prognostic marker for recurrence and survival in breast cancer patients undergoing RT.

Pulmonary Tuberculosis and Risk of Lung Cancer: A Systematic Review and Meta-Analysis.

Pulmonary tuberculosis (TB) is a known risk factor for lung cancer. However, a detailed analysis of lung cancer type, age, sex, smoking, and TB burden associated with geographic and socioeconomic status has not been performed previously. We systematically appraised relevant observational studies reporting an association between pulmonary TB and lung cancer. All studies were included in the primary analysis, and studies that used robust TB diagnostic methods, such as validated medical diagnostic codes, were included in the secondary analysis. Thirty-two articles were included. The association between the history of pulmonary TB and diagnosis of lung cancer was statistically significant (OR 2.09, 95% CI: 1.62-2.69, p < 0.001). There was a high heterogeneity (I2 = 95%), without any publication bias. The analysis indicated a high association in advanced articles describing stringent pulmonary TB diagnosis (OR 2.26, 95% CI: 1.29-3.94, p = 0.004). The subgroup analyses suggested a significant association in countries with medium or high TB burdens, from East Asia and the Pacific region, and upper-middle income countries. Heterogeneity within the subgroups remained high in a majority of the subgroup analyses. A meta-regression analysis revealed that younger patients showed a significantly higher association between TB and lung cancer (regression coefficient = 0.949, p < 0.001). The history of pulmonary TB is an independent risk factor for lung cancer, especially in younger patients diagnosed with pulmonary TB. Clinicians should be aware of this association while treating young patients with a history of pulmonary TB.

Effect of Prehospital Epinephrine on Out-of-Hospital Cardiac Arrest Outcomes: A Propensity Score-Matched Analysis.

PURPOSE: A pilot project using epinephrine at the scene under medical control is currently underway in Korea. This study aimed to determine whether prehospital epinephrine administration is associated with improved survival and neurological outcomes in out-of-hospital cardiac arrest (OHCA) patients who received epinephrine during cardiopulmonary resuscitation (CPR) in the emergency department. MATERIALS AND METHODS: This retrospective observational study used a nationwide multicenter OHCA registry. Patients were classified into two groups according to whether they received epinephrine at the scene or not. The associations between prehospital epinephrine use and outcomes were assessed using propensity score (PS)-matched analysis. Multivariable logistic regression analysis was performed using PS matching. The same analysis was repeated for the subgroup of patients with non-shockable rhythm. RESULTS: PS matching was performed for 1084 patients in each group. Survival to discharge was significantly decreased in the patients who received prehospital epinephrine [odds ratio (OR) 0.415, 95% confidence interval (CI) 0.250-0.670, p<0.001]. However, no statistical significance was observed for good neurological outcome (OR 0.548, 95% CI 0.258-1.123, p=0.105). For the patient subgroup with non-shockable rhythm, prehospital epinephrine was also associated with lower survival to discharge (OR 0.514, 95% CI 0.306-0.844, p=0.010), but not with neurological outcome (OR 0.709, 95% CI 0.323-1.529, p=0.382). CONCLUSION: Prehospital epinephrine administration was associated with decreased survival rates in OHCA patients but not statistically associated with neurological outcome in this PS-matched analysis. Further research is required to investigate the reason for the detrimental effect of epinephrine administered at the scene.

Comparison of the Clinical Features and Outcomes of Gallbladder Neuroendocrine Carcinoma with Those of Adenocarcinoma: A Propensity Score-Matched Analysis.

Neuroendocrine neoplasms (NENs) of the gallbladder (GB) are extremely rare. We aimed to compare the clinical features, disease progression, management, and prognosis of patients with GB-NENs with those of patients with GB-adenocarcinomas (ADCs). A total of 21 patients with GB-NENs and 206 patients with GB-ADCs, treated at three tertiary medical centers between January 2010 and December 2020, were enrolled. Of the 21 patients with GB-NENs, 20 were diagnosed with poorly differentiated small-cell neuroendocrine carcinomas (NECs), and 1 patient had large-cell NEC. All patients presented with advanced stages of cancer with extensive local extension and/or distant metastasis and non-specific symptoms. Tumor-node-metastasis stage IIIB and IV (A/B) tumors were found in 6 and 15 (1/14) patients, respectively. Nine patients with GB-NEC who underwent surgical resection had a significantly better progression-free survival (PFS) than those who did not undergo surgery. After a propensity score matching with a 1:1 ratio using the American Joint Committee on Cancer stage, age, sex, and operation status, 19 pairs of patients were included. Compared with stage-matched patients with GB-ADC, patients with GB-NEC had similar overall survival and PFS. However, as GB-NEC is rarely diagnosed early, further studies investigating methods for the early diagnosis and improvement in the survival of patients with GB-NEC are needed.

Differing Nutrient Intake and Dietary Patterns According to the Presence of Hyper-Low-Density Lipoprotein Cholesterolemia or Hypertriglyceridemia.

Dietary choices may have differing effects on low-density lipoprotein cholesterol or triglyceride levels. The aim of this study was to investigate daily nutrient intake and dietary patterns of individuals with hyper-low-density lipoprotein cholesterolemia (hLDL) and hypertriglyceridemia (hTG) in a large Korean population-based study using propensity score (PS) matching. This study used data from the Korea National Health and Nutrition Examination Survey. Propensity score values for the predicted probability of patients with hLDL or hTG were estimated using logistic regression analysis, with age, sex, body mass index, alcohol consumption, smoking status, physical activity status, hypertension, and diabetes. After PS matching, intake of carbohydrates (%) was significantly lower (p = 0.021), and intake of fats (%) and saturated fatty acids (%) was significantly higher in the hLDL group than in the non-hLDL group (p = 0.025 and p = 0.013, respectively). The percentage of individuals with a high score for the Korean Healthy Eating Index (KHEI) "whole grains" or "saturated fatty acids" components was higher in the non-hLDL group than in the hLDL group (p < 0.05 for both). Dietary sodium/potassium ratio was significantly higher in the hTG than in the non-hTG (p = 0.049). Our results suggest that individualized dietary information and counseling require consideration of a person's specific lipid levels.

Optimizing the maximum reported cluster size in the spatial scan statistic for survival data.

BACKGROUND: The spatial scan statistic is a useful tool for cluster detection analysis in geographical disease surveillance. The method requires users to specify the maximum scanning window size or the maximum reported cluster size (MRCS), which is often set to 50% of the total population. It is important to optimize the maximum reported cluster size, keeping the maximum scanning window size at as large as 50% of the total population, to obtain valid and meaningful results. RESULTS: We developed a measure, a Gini coefficient, to optimize the maximum reported cluster size for the exponential-based spatial scan statistic. The simulation study showed that the proposed method mostly selected the optimal MRCS, similar to the true cluster size. The detection accuracy was higher for the best chosen MRCS than at the default setting. The application of the method to the Korea Community Health Survey data supported that the proposed method can optimize the MRCS in spatial cluster detection analysis for survival data. CONCLUSIONS: Using the Gini coefficient in the exponential-based spatial scan statistic can be very helpful for reporting more refined and informative clusters for survival data.

Development and validation of the Health-Friendly Activity Index: an assessment tool to comprehensively measure health-friendly activities of corporations or organisations.

OBJECTIVES: We developed the Health-Friendly Activity Index (HFAI) to comprehensively measure the health-friendly activities of corporations or organisations. We validated the developed tool and reported on its use as an assessment tool to improve consumers' health-related outcomes. DESIGN: This was a cross-sectional study. SETTING: Development of the HFAI questionnaire followed a three-phase process: item generation, item construction and validation with field testing. Using relevance and feasibility criteria, we developed a 105-item questionnaire with six domains (Governance and Infrastructure, Needs Assessment, Planning, Implementation, Monitoring and Feedback, and Outcomes). PARTICIPANTS: To assess the sensitivity and validity of the questionnaire, the HFAI and Contribution Assessment Tool for Consumer's Health (CATCH) were administered to 302 participants (151 employers and 151 employees) from 151 Korean companies. PRIMARY OUTCOME MEASURES: The CATCH measured the contribution of each company to the physical, mental, social and spiritual health of its consumers. To estimate the reliability and validity of all six HFAI domains and their respective scales, Cronbach's α coefficients and correlation coefficients were used. RESULTS: Each domain and scale of the HFAI exhibited a Cronbach's α coefficient between 0.80 and 0.98 for the employers and employees. The overall HFAI and its six domains correlated significantly and positively with all health outcomes such as physical, mental, social and spiritual status scores evaluated using the CATCH (Spearman's correlation range: 0.37-0.68). CONCLUSION: The HFAI, a unique assessment tool with acceptable psychometric properties, can help corporate managers assess their health-friendly activities.

Clinical and genetic spectra in patients with dystrophinopathy in Korea: A single-center study.

Dystrophinopathy is a group of inherited phenotypes arising from pathogenic variants in DMD. We evaluated the clinical and genetic characteristics of Korean patients with genetically confirmed dystrophinopathy. We retrospectively reviewed medical records (January 2004-September 2020) from the myopathy database maintained at the study hospital and found 227 patients from 218 unrelated families with dystrophinopathy. Clinical phenotypes included 120 (53%) Duchenne muscular dystrophy (DMD) cases, 20 (9%) intermediate phenotype muscular dystrophy (IMD) cases, 65 (29%) Becker muscular dystrophy (BMD) cases, 18 (8%) undetermined phenotypes, and 4 (2%) symptomatic carriers. The median ages at symptom onset and diagnosis were 5.0 years (interquartile range [IQR]: 3.8-8.0) and 12.0 years (IQR: 7.0-21.0), respectively. Total manual muscle test (MMT) scores decreased annually in patients with DMD, IMD, and BMD. Overall, when age increased by 1 year, total MMT scores decreased on average by -1.978, -1.681, and -1.303 in patients with DMD (p<0.001), IMD (p<0.001), and BMD (p<0.001), respectively. Exonic deletion and duplication were reported in 147 (67%) and 31 (14%) of the 218 unrelated probands, respectively. A total of 37 different small sequence variants were found in 40 (18%) of the 218 probands. The reading frame rule was applicable to 142 (94%) of the 151 probands. The present results highlight the long-term natural history and genetic spectrum of dystrophinopathy in a large-scale Korean cohort.

A Markov chain model for analysis of physician workflow in primary care clinics.

This paper studies physician workflow management in primary care clinics using terminating Markov chain models. The physician workload is characterized by face-to-face encounters with patients and documentation of electronic health record (EHR) data. Three workflow management policies are considered: preemptive priority (stop ongoing documentation tasks if a new patient arrives); non-preemptive priority (finish ongoing documentation even if a new patient arrives); and batch documentation (start and finish documentation when the desired number of tasks is reached). Analytical formulas are derived to quantify the performance measures of three management policies, such as physician's daily working time, patient's waiting time, and documentation waiting time. A comparison of the results under three policies is carried out. Finally, a case study in a primary care clinic is carried out to illustrate model applicability. Such a work provides a quantitative tool for primary care physicians to design and manage their workflow to improve care quality.

Classification of Opioid Usage Through Semi-Supervised Learning for Total Joint Replacement Patients.

Opioid misuse and overdose have become a public health hazard and caused drug addiction and death in the United States due to rapid increase in prescribed and non-prescribed opioid usage. The misuse and overdose are highly related to opioid over-prescription for chronic and acute pain treatment, where a one-size-fits-all prescription plan is often adopted but can lead to substantial leftovers for patients who only consume a few. To reduce over-prescription and opioid overdose, each patient's opioid usage pattern should be taken into account. As opioids are often prescribed for patients after total joint replacement surgeries, this study introduces a machine learning model to predict each patient's opioid usage level in the first 2 weeks after discharge. Specifically, the electronic health records, patient prescription history, and consumption survey data are collected to investigate the level of short-term opioid usage after joint replacement surgeries. However, there are a considerable number of answers missing in the surveys, which degrades data quality. To overcome this difficulty, a semi-supervised learning model that assigns pseudo labels via Bayesian regression is proposed. Using this model, the missing survey answers of opioids amount taken by the patients are predicted first. Then, based on the prediction, pseudo labels are assigned to those patients to improve classification performance. Extensive experiments indicate that such a semi-supervised learning model has shown a better performance in the resulting patients classification. It is expected that by using such a model the providers can adjust the amount of prescribed opioids to meet each patient's actual need, which can benefit the management of opioid prescription and pain intervention.

Consumers' consciousness of health-friendly products and services and its association with sociodemographic characteristics and health status: a cross-sectional survey of the South Korean population.

OBJECTIVES: To identify consumers' consciousness of health-friendly products and services (consumer reaction, purchase intention and willingness to pay more) and its association with sociodemographic characteristics and multidimensional health status. METHODS: From March to May 2018, we administered questionnaires to 1200 individuals from the general Korean population asking about their perception of health-friendly labels, and if they would purchase such labelled products (foods, pharmaceuticals, etc) and services (purifying water, preventing air pollution, etc) at extra cost. RESULTS: The participants placed a high value on the importance of mental, social, spiritual and physical health factors in terms of the company's products and services with a score of about 8 out of 10 (range, 7.74-8.33). Most respondents (72.4%) said that they were interested in adopting health-friendly labels. When a health-friendly label is introduced (such as one by the Business for Social Responsiveness), 65.1% of the respondents said that they intended to purchase the product or service, while 6.8% said that they did not and 75.0% said that they were willing to pay extra for the health-friendly product or service. Multivariate logistic regression models showed urban residence, high education level and good social health to be significantly associated with positive attitudes towards health-friendly labels. People with high income, no religion or normal weight were more likely to say that they intend to purchase products and services with health-friendly labels. They also had a more positive attitude towards paying more for such products and services, as did people with good spiritual health. CONCLUSION: This study provides data that illustrate the importance of health-friendly products and services to the general population and companies.

Reducing COPD readmissions through predictive modeling and incentive-based interventions.

This paper introduces a case study at a community hospital to develop a predictive model to quantify readmission risks for patients with chronic obstructive pulmonary disease (COPD), and use it to support decision making for appropriate incentive-based interventions. Data collected from the community hospital's database are analyzed to identify risk factors and a logistic regression model is developed to predict the readmission risk within 30 days post-discharge of an individual COPD patient. By targeting on the high-risk patients, we investigate the implementability of the incentive policy which encourages patients to take interventions and helps them to overcome the compliance barrier. Specifically, the conditions and scenarios are identified for either achieving the desired readmission rate while minimizing the total cost, or reaching the lowest readmission rate under incentive budget constraint. Currently, such models are under consideration for a pilot study at the community hospital.