Cardiovascular disease (CVD) is twice as prevalent among individuals with mental illness compared to the general population. Prevention strategies exist but require accurate risk prediction. This study aimed to develop and validate a machine learning model for predicting incident CVD among patients with mental illness using routine clinical data from electronic health records.

A cohort study was conducted using data from 74,880 patients with 1.6 million psychiatric service contacts in the Central Denmark Region from 2013 to 2021. Two machine learning models (XGBoost and regularised logistic regression) were trained on 85% of the data from six hospitals using 234 potential predictors. The best-performing model was externally validated on the remaining 15% of patients from another three hospitals. CVD was defined as myocardial infarction, stroke, or peripheral arterial disease.

The best-performing model (hyperparameter-tuned XGBoost) demonstrated acceptable discrimination, with an area under the receiver operating characteristic curve of 0.84 on the training set and 0.74 on the validation set. It identified high-risk individuals 2.5 years before CVD events. For the psychiatric service contacts in the top 5% of predicted risk, the positive predictive value was 5%, and the negative predictive value was 99%. The model issued at least one positive prediction for 39% of patients who developed CVD.

A machine learning model can accurately predict CVD risk among patients with mental illness using routinely collected electronic health record data. A decision support system building on this approach may aid primary CVD prevention in this high-risk population.

Involuntary admissions to psychiatric hospitals are on the rise. If patients at elevated risk of involuntary admission could be identified, prevention may be possible. Our aim was to develop and validate a prediction model for involuntary admission of patients receiving care within a psychiatric service system using machine learning trained on routine clinical data from electronic health records (EHRs).

EHR data from all adult patients who had been in contact with the Psychiatric Services of the Central Denmark Region between 2013 and 2021 were retrieved. We derived 694 patient predictors (covering e.g. diagnoses, medication, and coercive measures) and 1134 predictors from free text using term frequency-inverse document frequency and sentence transformers. At every voluntary inpatient discharge (prediction time), without an involuntary admission in the 2 years prior, we predicted involuntary admission 180 days ahead. XGBoost and elastic net models were trained on 85% of the dataset. The models with the highest area under the receiver operating characteristic curve (AUROC) were tested on the remaining 15% of the data.

The model was trained on 50 634 voluntary inpatient discharges among 17 968 patients. The cohort comprised of 1672 voluntary inpatient discharges followed by an involuntary admission. The best XGBoost and elastic net model from the training phase obtained an AUROC of 0.84 and 0.83, respectively, in the test phase.

A machine learning model using routine clinical EHR data can accurately predict involuntary admission. If implemented as a clinical decision support tool, this model may guide interventions aimed at reducing the risk of involuntary admission.

Clinical decision support systems (CDSS) based on machine-learning (ML) models are emerging within psychiatry. If patients do not trust this technology, its implementation may disrupt the patient-clinician relationship. Therefore, the aim was to examine whether receiving basic information about ML-based CDSS increased trust in them.

We conducted an online randomized survey experiment in the Psychiatric Services of the Central Denmark Region. The participating patients were randomized into one of three arms: Intervention = information on clinical decision-making supported by an ML model; Active control = information on a standard clinical decision process, and Blank control = no information. The participants were unaware of the experiment. Subsequently, participants were asked about different aspects of trust and distrust regarding ML-based CDSS. The effect of the intervention was assessed by comparing scores of trust and distrust between the allocation arms.

Out of 5800 invitees, 992 completed the survey experiment. The intervention increased trust in ML-based CDSS when compared to the active control (mean increase in trust: 5% [95% CI: 1%; 9%], p = 0.0096) and the blank control arm (mean increase in trust: 4% [1%; 8%], p = 0.015). Similarly, the intervention reduced distrust in ML-based CDSS when compared to the active control (mean decrease in distrust: −3%[−1%; −5%], p = 0.021) and the blank control arm (mean decrease in distrust: −4% [−1%; −8%], p = 0.022). No statistically significant differences were observed between the active and the blank control arms.

Receiving basic information on ML-based CDSS in hospital psychiatry may increase patient trust in such systems.

Psychiatric research applies statistical methods that can be divided in two frameworks: causal inference and prediction. Recent proposals suggest a down-prioritisation of causal inference and argue that prediction paves the road to ‘precision psychiatry’ (i.e., individualised treatment). In this perspective, we critically appraise these proposals.

We outline strengths and weaknesses of causal inference and prediction frameworks and describe the link between clinical decision-making and counterfactual predictions (i.e., causality). We describe three key causal structures that, if not handled correctly, may cause erroneous interpretations, and three pitfalls in prediction research.

Prediction and causal inference are both needed in psychiatric research and their relative importance is context-dependent. When individualised treatment decisions are needed, causal inference is necessary.

This perspective defends the importance of causal inference for precision psychiatry.

Natural language processing (NLP) methods hold promise for improving clinical prediction by utilising information otherwise hidden in the clinical notes of electronic health records. However, clinical practice – as well as the systems and databases in which clinical notes are recorded and stored – change over time. As a consequence, the content of clinical notes may also change over time, which could degrade the performance of prediction models. Despite its importance, the stability of clinical notes over time has rarely been tested.

The lexical stability of clinical notes from the Psychiatric Services of the Central Denmark Region in the period from January 1, 2011, to November 22, 2021 (a total of 14,811,551 clinical notes describing 129,570 patients) was assessed by quantifying sentence length, readability, syntactic complexity and clinical content. Changepoint detection models were used to estimate potential changes in these metrics.

We find lexical stability of the clinical notes over time, with minor deviations during the COVID-19 pandemic. Out of 2988 data points, 17 possible changepoints (corresponding to 0.6%) were detected. The majority of these were related to the discontinuation of a specific note type.

We find lexical and syntactic stability of clinical notes from psychiatric services over time, which bodes well for the use of NLP for predictive modelling in clinical psychiatry.

The quality of life and lifespan are greatly reduced among individuals with mental illness. To improve prognosis, the nascent field of precision psychiatry aims to provide personalised predictions for the course of illness and response to treatment. Unfortunately, the results of precision psychiatry studies are rarely externally validated, almost never implemented in clinical practice, and tend to focus on a few selected outcomes. To overcome these challenges, we have established the PSYchiatric Clinical Outcome Prediction (PSYCOP) cohort, which will form the basis for extensive studies in the upcoming years.

PSYCOP is a retrospective cohort study that includes all patients with at least one contact with the psychiatric services of the Central Denmark Region in the period from January 1, 2011, to October 28, 2020 (n = 119 291). All data from the electronic health records (EHR) are included, spanning diagnoses, information on treatments, clinical notes, discharge summaries, laboratory tests, etc. Based on these data, machine learning methods will be used to make prediction models for a range of clinical outcomes, such as diagnostic shifts, treatment response, medical comorbidity, and premature mortality, with an explicit focus on clinical feasibility and implementation.

We expect that studies based on the PSYCOP cohort will advance the field of precision psychiatry through the use of state-of-the-art machine learning methods on a large and representative data set. Implementation of prediction models in clinical psychiatry will likely improve treatment and, hopefully, increase the quality of life and lifespan of those with mental illness.