Analyzing Patient Trajectories With Artificial Intelligence

Abstract

In digital medicine, patient data typically record health events over time (eg, through electronic health records, wearables, or other sensing technologies) and thus form unique patient trajectories. Patient trajectories are highly predictive of the future course of diseases and therefore facilitate effective care. However, digital medicine often uses only limited patient data, consisting of health events from only a single or small number of time points while ignoring additional information encoded in patient trajectories. To analyze such rich longitudinal data, new artificial intelligence (AI) solutions are needed. In this paper, we provide an overview of the recent efforts to develop trajectory-aware AI solutions and provide suggestions for future directions. Specifically, we examine the implications for developing disease models from patient trajectories along the typical workflow in AI: problem definition, data processing, modeling, evaluation, and interpretation. We conclude with a discussion of how such AI solutions will allow the field to build robust models for personalized risk scoring, subtyping, and disease pathway discovery.

Estimating average causal effects from patient trajectories

“…t+1 } T t=1 ) N i=1 . Such patient trajectories are nowadays widely available in electronic health records [2]. For notation, we use a superscript (i) to refer to patients (we omit it unless needed).…”

“…Instead, we propose to leverage results from semi-parametric estimation theory, as this allows us to construct an estimator with better theoretical properties, namely double robustness and asymptotic efficiency. 2 For this purpose, we design our targeting layer so that it estimates the treatment assignments, i. e., the so-called propensity scores…”

“…where we used the convention that Q a T +2 = Y T +1 and where 1(•) denotes the indicator function. 2 For an overview on semi-parametric estimation theory, we refer to [17].…”

“…Therefore, medical practice is increasingly relying on observational data to study causal relationships between treatments and patient outcomes. Nowadays, electronic health records are readily available, capture patient trajectories with high granularity, and thus provide rich observational data in medical practice [2,26].…”

“…To this end, medical research seeks to find increasingly more granular subgroups [25]. These should identify differences in disease dynamics and thus capture different phenotypes [2], so that different treatment guidelines are tailored to each subgroup [10]. To do so, one must again compare causal effects across patient cohorts, that is, average causal effects across patient cohorts.…”