Yuncheol Kang

Smart logistics: distributed control of green crowdsourced parcel services

This paper presents the development of an integrated decision-making framework for on-demand parcel delivery services that considers Just-In-Time delivery, fuel consumption and carbon emissions. Optimal policies based on the Markov decision process are established to allow for inclusion of parcel delivery requests. The framework’s integrated dynamic algorithm, based on a continuous variable feedback control, allows for unified processing of delivery requests and route scheduling. Computational experiments show that the integrated approach could increase revenue by 6.4% by reducing fuel and emission costs by 2.5%; however, the approach may incur more cost in terms of timeliness compared to a myopic approach.

Determination of Bicycle Handle Diameters considering Hand Anthropometric Data and User Satisfaction

Ergonomic product design considering both anthropometric variability and user preference is required for harmonizing the target users and products. In this study, bicycle handle diameters for three size categories were determined by considering anthropometric variability and preference. To design the bicycle handles, a four-step process was applied: (1) define anthropometric data, (2) develop size chart, (3) define a design equation, and (4) determine design values. In the first step, the 1988 US Army data was chosen as anthropometric data for the design target population. In the second step, to develop a size chart of bicycle handle, hand length and circumference were selected as key dimensions by principal component analysis on six representative hand dimensions. Next, a size chart of three categories (small: 175.5 mm, medium: 186.7 mm, and large: 196.2 mm) were derived by K-means clustering analysis for hand length and circumference. In the third step, the design equation accounting geometrical relationship between the sizes of two key dimensions and diameters of bicycle handle was adopted from a relevant existing research. In the last step, design values (40.9 mm, 43.5mm, and 45.7 mm) for each size category were calculated by inputting the sizes of the key dimensions to the design equation. To evaluate user satisfaction level of the bicycle handles, a user testing of three handle prototypes was conducted for 17 participants with various hand sizes. The test results showed that satisfaction scores for each hand group were significantly higher at the corresponding size category.

Predictive analytics for delivering prevention services

Logical Analysis of Data (LAD) is applied to prevention research.Real data sets from Baltimore City Public Schools are used for LAD analysis.Data patterns explaining childrens conduct disorder are extracted by LAD.We showed how LAD can reduce the cost of delivering prevention services. Early diagnosis and prevention of problematic behaviors in youth play an important role in reducing treatment costs and decreasing the toll of antisocial behavior. Over the last several years, the science of preventing antisocial behavior in youth has made significant strides, with the development of evidence-based prevention programs (EBP) using randomized clinical trials. In this paper, we use a real case implemented by schools in an urban school district of 80,000 students in a mid-Atlantic state to show how predictive analytics can help to improve the quality of prevention programs and reduce the cost of delivering associated services. Data patterns are extracted from conduct disorder assessments using the Teacher Observation of Classroom Adaptation (TOCA) screening instrument and evaluated using the results of the Diagnostic Interview Schedule for Children (DISC). A mathematical method called Logical Analysis of Data (LAD) is used to analyze data patterns. Experimental results show that up to 91.58% of the cost of administering DISC would be saved by correctly identifying participants without conduct disorder and excluding them from the DISC test.

A Study to Develop the Framework of Estimating the Cost of Replacing Labor Due to Job-Loss Caused by Injuries Based on the Results from Time Study in Agriculture of Korea

Measuring the time for human-driven agricultural work and establishing the standard time for the work are essential for estimating compensation or insurance cost when work-related damage and/or disaster are occurred. Our study aims to investigate the steps required for cultivating crops, and try to estimate the time and amount of job-loss due to the injuries occurring in the agricultural environment. From the analysis we found the work time recorded in the diary was greater than the time measured using GPS by 7% on average. Additionally, we develop the Excel-based macro application that can calculate the costs for the replacing labor forces based upon the data obtained from the time study.

INVESTIGATION OF THE LOWER EXTREMITY STATIC POSTURES PREDICTING VALGUS COLLAPSE DURING DROP LANDING IN KOREAN MALE AND FEMALE – LOGICAL ANALYSIS OF DATA (LAD) APPROACH

Background The result of 2002∼2012 National Health Insurance Sharing Service (NHISS) data analysis demonstrated that the genders odds ratio of ACL was as low as 1.06, which is different from other studies. The static posture measures (SPM) of Asian may not be strongly associated with valgus collapse during landing. Objective To investigate SPMs that account for valgus collapse in Korean male and female using the LAD approach. Design Cross-sectional study. Setting Controlled laboratory. Patients (or Participants) Forty-four active male (age: 22.2±3.6 yrs; Height: 176±6.6 cm; Weight: 73.3±9.7 kg) and 43 female (age: 21.9±2.5 yrs; Height: 163±5.3 cm; Weight: 56.5±5.1 kg) with no history of lower extremity injury were recruited. Interventions (or Assessment of Risk Factors) Gender. Main Outcome Measurements Joint kinematics associated with valgus collapse during drop landing were extracted (Vicon Inc, Oxford, UK) during drop landing task. Thirteen lower extremity SPMs were collected. LAD was implemented using the machine learning software, Weka 3.7.10 (U of Waikato, Hamilton, New Zealand) to generate male and female patterns that characterize features of SPM and its association with valgus collapse. Results A total of eight and 33 distinct male and female patterns respectively could be generated (>10% relative prevalence (RP)). Interestingly, Joint kinematics associated with valgus collapse were not shown in patterns, but only six SPMs were included. The most relevant pattern that represents male (RP 25.00%) was tibiofemoral angle less than or equal to 9.5° and leg length discrepancy greater than −3.5 cm. The most relevant pattern that represents female (RP 34.88%) was Q-angle greater than 17.5°, genu recurvatum greater than −1.5°, femoral anteversion greater than 5.5°, and leg length discrepancy less than or equal to 2.5 cm. Conclusions SPMs of Korean male and female were not associated with valgus collapse during drop jump. It seemed that the Korean female demonstrated similar neuromuscular control characteristics as male to avoid valgus collapse.

Measuring agreement between healthcare survey instruments using mutual information

BackgroundHealthcare researchers often use multiple healthcare survey instruments to examine a particular patient symptom. The use of multiple instruments can pose some interesting research questions, such as whether the outcomes produced by the different instruments are in agreement. We tackle this problem using information theory, focusing on mutual information to compare outcomes from multiple healthcare survey instruments.MethodsWe review existing methods of measuring agreement/disagreement between the instruments and suggest a procedure that utilizes mutual information to quantitatively measure the amount of information shared by outcomes from multiple healthcare survey instruments. We also include worked examples to explain the approach.ResultsAs a case study, we employ the suggested procedure to analyze multiple healthcare survey instruments used for detecting delirium superimposed on dementia (DSD) in community-dwelling older adults. In addition, several examples are used to assess the mutual information technique in comparison with other measures, such as odds ratio and Cohen’s kappa.ConclusionsAnalysis of mutual information can be useful in explaining agreement/disagreement between multiple instruments. The suggested approach provides new insights into and potential improvements for the application of healthcare survey instruments.

Stimulating new and innovative perspectives on old and persistent problems: a commentary on “Attempters adherers and non-adherers: latent profile analysis of CPAP use with correlates” by Wohlgemuth et al.

The clinical guidelines for the evaluation, management, and longterm care of obstructive sleep apnea (OSA) in adults emphasize a multidisciplinary, chronic care approach to the treatment and management of OSA [1]. The guidelines, consistent with earlier practice parameters [2], include consensus standards to monitor early continuous positive airway pressure (CPAP) compliance (ie, adherence) for utilization patterns/rates. The guidelines also recommend early support for treatment and for those with poor or low utilization of CPAP, “prompt and intensive efforts to remediate low use” or alternative treatments should be pursued [1]. While evidence abounds for the diagnosis and initiation of CPAP treatment for OSA, there is far less evidence that directly addresses the everyday practice implementation of the standards and recommendations of these guidelines. Beyond a timely review of objective CPAP use records in the clinical setting and trouble-shooting for common complaints about the treatment, there is a paucity of evidence-based recommendations for using objective CPAP use data to guide clinical decision-making in concert with patients. Evidence suggests that simplistic overt monitoring of CPAP use as an “awareness” intervention does not improve CPAP adherence [3]. When CPAP use data is shared with patients, there is emerging evidence that early CPAP adherence is improved [4]. So how can the field move beyond simplistic monitoring of CPAP use to truly harness the power of such data for shared clinical decision-making? In this issue of Sleep Medicine, Wohlgemuth and colleagues [5] report the results of a retrospective cohort study wherein patterns of CPAP use were systematically analyzed using latent profile analysis. The objective of the research was to deduce CPAP user profiles from an existing cohort of CPAP-treated OSA veteran patients with at least seven days of CPAP usage data available and identify predictors of CPAP use profile membership. Using latent profile analysis and standard evaluative criteria, a three-cluster solution (ie, three profiles) was identified as significant and efficient. Predictive factors for profile membership in one of the three profiles, named by the investigators as adherers, non-adherers, and attempters, included self-efficacy, insomnia, apnea-hypopnea index, CPAP pressure, and duration of treatment; unique and modifiable factors were identified for each profile [5]. Though not without limitations that include a retrospective design and limited generalizability, this study [5] addresses two opportunities for advancing the research agenda to support everyday clinical decision-making for the management of CPAP-treated OSA: (1) risk stratification, or profiling, for treatment non-adherence; and (2) moving beyond traditional statistical modeling approaches of readily-available clinical data, specifically objective CPAP use, to effectively use this complex and comprehensive data to potentially support clinical decision-making. The identification of CPAP-treated OSA patients who are likely to be non-adherent, or those likely to struggle with CPAP is important for responsible allocation of resources in the clinical management of OSA. Several previously published studies have reported CPAP non-adherence risk stratification methods [6–8] and have similarly advocated such an approach be empirically advanced and translated. Risk stratification approaches are indeed important so that CPAP-treated OSA patients that require adherence promotion interventions or early referral to alternative treatment options are readily identified. As early CPAP use predicts long-term patterns of use [9–11], permitting non-adherent patterns of CPAP use to go unaddressed for any length of time risks unremitting CPAP non-adherence and treatment failure. Developing evidence that addresses early risk stratification for nonadherence, or user profiles, is likely to translate in the future to the allocation of timely and intensive resources to those most likely to benefit from the investment (ie, non-adherers and attempters). To further extend the potential usefulness of Wohlgemuth and colleagues’ findings [5] as a risk stratification method, future research will need to examine the necessary minimum data set required for CPAP user profiles to emerge and examine the dynamic patterns of usage states. The approach to the discovery of usage profiles by Wohlegemuth et al. [5] limits profile definitions to a singlepoint in time, as is consistent with a cross-sectional perspective of profiles. This perspective does not account for the dynamic, or timevarying, nature of CPAP use as a behavior. Complementing latent profile analysis with latent profile transition analysis may reveal if membership in any one CPAP profile changes with time and/or intervention. To fully capitalize on available CPAP use data with dynamic characteristics (ie, time varying) and clinically respond to such dynamical data, or behaviors, employing data mining analytic approaches such as probabilistic suffix trees has the potential to provide “early warning signals” to accurately identify non-adherence and transitions between CPAP use states (eg, intermediate to low; attempter to non-adherer) [12]. This approach can provide important insights into the timing of interventions by individual patients in order to “shift” profiles/states.

An approach for dynamic optimization of prevention program implementation in stochastic environments

The science of preventing youth problems has significantly advanced in developing evidence-based prevention program (EBP) by using randomized clinical trials. Effective EBP can reduce delinquency, aggression, violence, bullying and substance abuse among youth. Unfortunately the outcomes of EBP implemented in natural settings usually tend to be lower than in clinical trials, which has motivated the need to study EBP implementations. In this paper we propose to model EBP implementations in natural settings as stochastic dynamic processes. Specifically, we propose Markov Decision Process (MDP) for modeling and dynamic optimization of such EBP implementations. We illustrate these concepts using simple numerical examples and discuss potential challenges in using such approaches in practice.