Towards intelligent complex networks: the space and prediction of information walks

Abstract

In this paper we study the problem of walk-specific information spread in directed complex social networks. Classical models usually analyze the “explosive” spread of information on social networks (e.g., Twitter) – a broadcast or epidemiological model focusing on the dynamics of a given source node “infecting” multiple targets. Less studied, but of equal importance is the case of single-track information flow, wherein the focus is on the node-by-node (and not necessarily a newly visited node) trajectory of information transfer. An important and motivating example is the sequence of physicians visited by a given patient over a presumed course of treatment or health event. This is the so-called a referral sequence which manifests as a path in a network of physicians. In this case the patient (and her health record) is a source of “information" from one physician to the next. With this motivation in mind we build a Bayesian Personalized Ranking (BPR) model to predict the next node on a walk of a given network navigator using network science features. The problem is related to but different from the well-investigated link prediction problem. We present experiments on a dataset of several million nodes derived from several years of U.S. patient referral records, showing that the application of network science measures in the BPR framework boosts hit-rate and mean percentile rank for the task of next-node prediction. We then move beyond the simple information walk to consider the derived network space of all information walks within a period, in which a node represents an information walk and two information walks are connected if have nodes in common from the original (social) network. To evaluate the utility of such a network of information walks, we simulate outliers of information walks and distinguish them with the other normal information walks, using five distance metrics for the derived feature vectors between two information walks. The experimental results of such a proof-of-concept application shows the utility of the derived information walk network for the outlier monitoring of information flow on an intelligent network.

Optimal Physician Shared-Patient Networks and the Diffusion of Medical Technologies

“…In this paper we developed and explored novel unipartite projections of the standard bipartite network that connects physicians and their patients. This work was motivated by the historical disregard shown to the time-order of patient visits with their physicians in the construction of shared-patient networks and makes use of the underlying referral path information (An et al, 2018b(An et al, ,a, 2019 that is lost (or neglected) in the formation of the traditional patient-physician bipartite network. We found that referral paths contain substantial information that can be used to better distinguish the one-mode networks from one another thereby enhancing the potential to discover relationships of network features to important health variables compared to what is possible with networks based on currently-used undirected projections.…”

“…Networks encoding the relationships between physicians and patients, and also between physicians (through shared patients) increasingly are used for research in medicine and health care (An et al, 2018b(An et al, ,a, 2019Barnett et al, 2011). Patient-physician interactions naturally give rise to a bipartite network connecting physicians (on one "side") to the patients that they treat (on the other "side") (see Figure 1 (L)).…”

“…Recently, it has been found that the sequences of (presumably linked) patient-physician visits (herein called "referral paths") may be a more useful encoding of both patient interactions with the health care system as well as a data source for understanding inter-physician interactions. Referral paths can even be used as the basic data of a useful health care-related network construct (An et al, 2018a(An et al, , 2019.…”

Understanding health system operation using networks: A literature review

“…A graph neural network was trained by Lu and Uddin (2021) using a variety of features to predict the development of chronic diseases. An, O'Malley, and Rockmore (2019) used a combination of multiple metrics to develop a predictive model of the next service a patient is likely to access.…”

“…One area that sought to extend the use of network analysis from descriptive to predictive applications was categorised as patient treatment. An et al (2019) used a variety of network metrics to create a predictive model of the next service that a patient was likely to access. A state-of-the-art example of prediction using networks was seen from Lu and Uddin (2021).…”