Designers are increasingly collaborating with data scientists to apply smart data technologies to understand large-scale user behavior during their design research. This is useful in specific impact domains with vulnerable users and unfamiliar contexts, such as healthcare design. Patient journey mapping is the most common design tool for developing and communicating patient-centred perspectives in healthcare design. However, creating a traditional patient journey map is labor intensive. Consequently, they often represent the experiences of a limited number of patients and, therefore, have limitations in including an extensive group patient experience. To overcome these challenges, we present a new data-driven and hybrid intelligent design approach that utilizes tens of thousands of online patient stories and machine-learning techniques through collaboration with data scientists. We set up two studies in the field of oncology and demonstrate that combining the two machine-learning techniques allows for quantifying the experiences of a wide range of patients, detecting relationships between co-occurring experiences within the journey, and detecting new design opportunities/directions. In these studies, designers gained a large-scale, yet qualitative and inspiring, understanding of a complex context in healthcare with reduced time and cost investments@en

Background: Heart rate (HR) monitored by a wearable device (WD) has demonstrated its clinical feasibility for thyrotoxicosis subjects. However, the association of HR monitored by wearables with hypothyroidism has not been examined. We assessed the association between serum thyroid hormone concentration and three WD-HR parameters in hypothyroid subjects. Methods: Forty-four subjects scheduled for radioactive iodine therapy (RAI Tx) after thyroid cancer surgery were included. Thirty subjects were prepared for RAI Tx by thyroid hormone withdrawal (hypothyroidism group) and 14 subjects by recombinant human thyrotropin (control group). Three WD-HR parameters were calculated from the HR data collected during rest, during sleep, and from 2:00 AM to 6:00 AM, respectively. We analyzed the changes in conventionally measured resting HR (On-site rHR) and WD-HR parameters relative to thyroid hormone levels. Results: Serum free thyroxine (T4) levels, On-site rHR, and WD-HR parameters were lower in the hypothyroid group than in the control group at the time of RAI Tx. WD-HR parameters also reflected minute changes in free T4 levels. A decrease in On-site rHR and WD-HR parameters by one standard deviation (On-site rHR, approximately 12 bpm; WD-HR parameters, approximately 8 bpm) was associated with a 0.2 ng/dL decrease in free T4 levels (P<0.01) and a 2-fold increase of the odds ratio of hypothyroidism (P<0.01). WD-HR parameters displayed a better goodness-of-fit measure (lower quasi-information criterion value) than On-site rHR in predicting the hypothyroidism. Conclusion: This study identified WD-HR parameters as informative and easy-to-measure biomarkers to predict hypothyroidism.

@en

Monitoring of dual service failures (e.g., trends in service failures and consecutive service failures) in business is emphasized for service quality management. Previous studies analyzing negative online reviews to conduct dual service failure monitoring from a managerial perspective are scarce. Numerous negative online reviews are useful sources for dual service failure monitoring because they can be easily collected at a low cost. This article proposes a data-driven approach to monitor service failure trends and consecutive service failures from negative online reviews. In the proposed approach, first a classifier is developed to categorize newly collected negative reviews into service failures by Latent Dirichlet allocation. Subsequently, a threshold value is provided to identify a new type of service failure, which was not achieved previously using a control chart. Finally, the probability of consecutive service failures is obtained by association rule mining. A case study of Uber is conducted to validate the proposed approach. The results exhibit that the proposed approach can perform dual service failure monitoring. This study can increase marketing intelligence for dynamic management of service failure and allow rapid responses to service failures.

@en

A multistage manufacturing process (MMP) consists of several consecutive process stages, each of which has multiple machines performing the same functions in parallel. A manufacturing path (simply referred to as path) is defined as an ordered set indicating a record of machines assigned to a product at each process stage of an MMP. An MMP usually produces products through various paths. In practice, multiple machines in a process stage have different operational performances, which accumulate during production and affect the quality of products. This study proposes a heuristic approach to derive the golden paths that produce products whose quality exceeds the desired level. The proposed approach consists of the searching phase and the merging phase. The searching phase extracts two types of machine sequence patterns (MSPs) from a path dataset in an MMP. An MSP is a subset of the path that is defined as an ordered set of assigned machines from several process stages. The two extracted types of MSPs are: (1) superior MSP, which affects the production of superior-quality products, and (2) inferior MSP, which affects the production of inferior-quality products, called inferior MSP. The merging phase derives the golden paths by combining superior MSPs and excluding inferior MSPs. The proposed approach is verified by applying it to a hypothetical path dataset and the semiconductor tool fault isolation (SETFI) dataset. This verification shows that the proposed approach derives the golden paths that exceed the predefined product quality level. This outcome demonstrates the practical viability of the proposed approach in an MMP.

@en

Background: A lifelogs-based wellness index (LWI) is a function for calculating wellness scores based on health behavior lifelogs (eg, daily walking steps and sleep times collected via a smartwatch). A wellness score intuitively shows the users of smart wellness services the overall condition of their health behaviors. LWI development includes estimation (ie, estimating coefficients in LWI with data). A panel data set comprising health behavior lifelogs allows LWI estimation to control for unobserved variables, thereby resulting in less bias. However, these data sets typically have missing data due to events that occur in daily life (eg, smart devices stop collecting data when batteries are depleted), which can introduce biases into LWI coefficients. Thus, the appropriate choice of method to handle missing data is important for reducing biases in LWI estimations with panel data. However, there is a lack of research in this area. Objective: This study aims to identify a suitable missing-data handling method for LWI estimation with panel data. Methods: Listwise deletion, mean imputation, expectation maximization-based multiple imputation, predictive-mean matching-based multiple imputation, k-nearest neighbors-based imputation, and low-rank approximation-based imputation were comparatively evaluated by simulating an existing case of LWI development. A panel data set comprising health behavior lifelogs of 41 college students over 4 weeks was transformed into a reference data set without any missing data. Then, 200 simulated data sets were generated by randomly introducing missing data at proportions from 1% to 80%. The missing-data handling methods were each applied to transform the simulated data sets into complete data sets, and coefficients in a linear LWI were estimated for each complete data set. For each proportion for each method, a bias measure was calculated by comparing the estimated coefficient values with values estimated from the reference data set. Results: Methods performed differently depending on the proportion of missing data. For 1% to 30% proportions, low-rank approximation-based imputation, predictive-mean matching-based multiple imputation, and expectation maximization-based multiple imputation were superior. For 31% to 60% proportions, low-rank approximation-based imputation and predictive-mean matching-based multiple imputation performed best. For over 60% proportions, only low-rank approximation-based imputation performed acceptably. Conclusions: Low-rank approximation-based imputation was the best of the 6 data-handling methods regardless of the proportion of missing data. This superiority is generalizable to other panel data sets comprising health behavior lifelogs given their verified low-rank nature, for which low-rank approximation-based imputation is known to perform effectively. This result will guide missing-data handling in reducing coefficient biases in new development cases of linear LWIs with panel data.

@en