Dynamic learning of individual-level suicidal ideation trajectories to enhance mental health care

Abstract

There has recently been an increase in ongoing patient-report routine outcome monitoring for individuals within clinical care, which has corresponded to increased longitudinal information about an individual. However, many models that are aimed at clinical practice have difficulty fully incorporating this information. This is in part due to the difficulty in dealing with the irregularly time-spaced observations that are common in clinical data. Consequently, we built individual-level continuous-time trajectory models of suicidal ideation for a clinical population (N = 585) with data collected via a digital platform. We demonstrate how such models predict an individual's level and variability of future suicide ideation, with implications for the frequency that individuals may need to be observed. These individual-level predictions provide a more personalised understanding than other predictive methods and have implications for enhanced measurement-based care.

Fig. 1. The joint distribution for the median values of the individual-level parameters.

We show the median predicted baseline observation (${\tilde{y}}_{m,0}$) compared to the diffusion parameter (${\varphi }_{m,2}$) for each individual $m$. This is compared to the prior predictive distribution for each parameter given the population-level parameters, where we show the median and 95% highest density credible intervals (HDI).

Fig. 2. Predictive trajectories for individuals with qualitatively different historical trajectories.

The distribution of predicted trajectories 60 days from the last observation for representative individuals that have a improved, b deteriorated, c had moderate risk and volatility, d high volatility, e no suicidal thoughts, and f high risk but moderate volatility. The distribution of trajectories is summarised using the median (black line), 68% (dark grey) and 95% (light grey) ETI each day. To summarise the individual’s future ideation trajectories, we show the probability that the individual will be in the high-ideation category on any day during the next 60 days (integrated high-ideation probability, IHIP), the future observational variability (V) and a recommended follow-up time (FUT).

Fig. 3. Predictive trajectories for an individual as data is collected.

Predictions for 60 days into the future from a baseline, and then after b three, c five, and d 10 observations. Each set of predictions uses the observations at or prior to that time (filled red circles) which are compared to future observations (red outlined circles). The distribution of trajectories is summarised using the median (black line), 68% (dark grey) and 95% (light grey) ETI for each day. To summarise an individual’s future ideation trajectories, we show the probability that they will be in the high-ideation category on any day during the next 60 days (integrated high-ideation probability, IHIP), the future observational variability (V) and a recommended follow-up time (FUT).