Search for: All records

Advanced machine learning models have received wide attention in assisting medical decision making due to the greater accuracy they can achieve. However, their limited interpretability imposes barriers for practitioners to adopt them. Recent advancements in interpretable machine learning tools allow us to look inside the black box of advanced prediction methods to extract interpretable models while maintaining similar prediction accuracy, but few studies have investigated the specific hospital readmission prediction problem with this spirit.

Our goal is to develop a machine-learning (ML) algorithm that can predict 30- and 90- day hospital readmissions as accurately as black box algorithms while providing medically interpretable insights into readmission risk factors. Leveraging a state-of-art interpretable ML model, we use a two-step Extracted Regression Tree approach to achieve this goal. In the first step, we train a black box prediction algorithm. In the second step, we extract a regression tree from the output of the black box algorithm that allows direct interpretation of medically relevant risk factors. We use data from a large teaching hospital in Asia to learn the ML model and verify our two-step approach.

The two-step method can obtain similar prediction performance as the best black box model, such as Neural Networks, measured by three metrics: accuracy, the Area Under the Curve (AUC) and the Area Under the Precision-Recall Curve (AUPRC), while maintaining interpretability. Further, to examine whether the prediction results match the known medical insights (i.e., the model is truly interpretable and produces reasonable results), we show that key readmission risk factors extracted by the two-step approach are consistent with those found in the medical literature.

The proposed two-step approach yields meaningful prediction results that are both accurate and interpretable. This study suggests a viable means to improve the trust of machine learning based models in clinical practice for predicting readmissions through the two-step approach.

 

Healthcare capacity shortage contributes to poor access in many countries. Moreover, rapid urbanization often occurring in these countries has exacerbated the imbalance between healthcare capacity and need. One way to address the above challenge is expanding the total capacity and redistributing the capacity spatially. In this research, we studied the problem of locating new hospitals in a two-tier outpatient care system comprising multiple central and district hospitals, and upgrading existing district hospitals to central hospitals. We formulated the problem with a discrete location optimization model. To parameterize the optimization model, we used a multinomial logit model to characterize individual patients’ diverse hospital choice and to quantify the patient arrival rates at each hospital accordingly. To solve the hard nonlinear combinatorial optimization problem, we developed a queueing network model to approximate the impact of hospital locations on patient flows. We then proposed a multi-fidelity optimization approach, which involves both the aforementioned queuing network model as a surrogate and a self-developed stochastic simulation as the high-fidelity model. With a real-world case study of Shanghai, we demonstrated the changes in the care network and examined the impacts on the network design by population center emergence, governmental budget change and considering patients with different age groups or income levels. Note to Practitioners —Our work focuses on improving system-wide care access in a two-tier care network. We believe that our work can lead to effective development of a location analytics tool for city-wide healthcare system planners. We also think the importance of this study is further strengthened by the case studies based on real-world hospital choice experimental data from Shanghai, China, a region suffering from the imbalance between healthcare capacity and need. Our case studies are expected to make recommendations on care facility expansion and dispersion to better align with the spatial distribution of residential communities and patient hospital choice behavior. 

While prior studies have designed incentive mechanisms to attract the public to share their collected data, they tend to ignore information asymmetry between data requesters and collectors. In reality, the sensing costs information (time cost, battery drainage, bandwidth occupation of mobile devices, and so on) is the private information of collectors, which is unknown by the data requester. In this article, we model the strategic interactions between health-data requester and collectors using a bilevel optimization model. Considering that the crowdsensing market is open and the participants are equal, we propose a Walrasian equilibrium-based pricing mechanism to coordinate the interest conflicts between health-data requesters and collectors. Specifically, based on the exchange economic theory, we transform the bilevel optimization problem into a social welfare maximization problem with the constraint condition that the balance between supply and demand, and dual decomposition is then employed to divide the social welfare maximization problem into a set of subproblems that can be solved by health-data requesters and collectors. We prove that the optimal task price is equal to the marginal utility generated by the collector's health data. To avoid obtaining the collector's private information, a distributed iterative algorithm is then designed to obtain the optimal task pricing strategy. Furthermore, we conduct computational experiments to evaluate the performance of the proposed pricing mechanism and analyze the effects of intrinsic rewards, sensing costs on optimal task prices, and collectors' health-data supplies.