The Management and Supervision Tool (MaST): an electronic crisis risk prediction tool to support safe and effective mental healthcare

Introduction

The increasing global burden of mental disorders has led to rising demand for mental healthcare services. Effective resource management is essential to ensure safe and timely access to care. Electronic health records (EHRs) provide a real-time source of data on clinical presentation and prognostic factors that could be harnessed to provide clinicians with actionable insights to prioritise mental healthcare delivery. We describe the development and evaluation of MaST, an EHR data visualisation tool that provides information to clinicians on risk of mental health crisis defined as an admission to a psychiatric hospital or acceptance into a community crisis service.

Objectives

(i) To develop an EHR-data driven risk prediction tool for risk of crisis. (ii) To evaluate predictive performance in a real-world clinical setting.

Methods

The risk of crisis algorithm was developed and evaluated with EHR data from six UK NHS mental health providers using Ordered Predictor List propensity scores grouped into 5 quintiles. The predictor variables were clinical and sociodemographic factors including previous mental health service contacts.

Results

Data from 2,620 patients contributed to algorithm development which was subsequently tested on data from 107,879 patients. The risk of crisis algorithm performed well with an overall accuracy for predicting the greatest risk of crisis (top quintile) ranging from 64% to 80%.

Conclusions

The MaST algorithm accurately predicted risk of mental health crisis in UK community mental health services. EHR data visualisation tools can provide actionable insights to clinicians to prioritise mental healthcare delivery in real-world clinical practice.

Disclosure

This study was funded in full by Holmusk.

A Surrogate Modelling Approach Based on Nonlinear Dimension Reduction for Uncertainty Quantification in Groundwater Flow Models

In this paper, we develop a surrogate modelling approach for capturing the output field (e.g. the pressure head) from groundwater flow models involving a stochastic input field (e.g. the hydraulic conductivity). We use a Karhunen–Loève expansion for a log-normally distributed input field and apply manifold learning (local tangent space alignment) to perform Gaussian process Bayesian inference using Hamiltonian Monte Carlo in an abstract feature space, yielding outputs for arbitrary unseen inputs. We also develop a framework for forward uncertainty quantification in such problems, including analytical approximations of the mean of the marginalized distribution (with respect to the inputs). To sample from the distribution, we present Monte Carlo approach. Two examples are presented to demonstrate the accuracy of our approach: a Darcy flow model with contaminant transport in 2-d and a Richards equation model in 3-d.

Impaired insulin-dependent glucose metabolism in granulosa-lutein cells from anovulatory women with polycystic ovaries

BACKGROUND

Insulin resistance and hyperinsulinaemia are well-recognized characteristics of anovulatory women with polycystic ovary syndrome (PCOS) but, paradoxically, steroidogenesis by PCOS granulosa cells remains responsive to insulin. The hypothesis to be tested in this study is that insulin resistance in the ovary is confined to the metabolic effects of insulin (i.e. glucose uptake and metabolism), whereas the steroidogenic action of insulin remains intact.

METHODS

Granulosa-lutein cells were obtained during IVF cycles from seven women with normal ovaries, six ovulatory women with PCO (ovPCO) and seven anovulatory women with PCO (anovPCO). Mean body mass index was in the normal range in all three groups. Granulosa-lutein cells were cultured with insulin (1, 10, 100 and 1000 ng/ml) and LH (1, 2.5 and 5 ng/ml). Media were sampled at 24 and 48 h and analysed for glucose uptake, lactate production and (48 h only) progesterone production.

RESULTS

Insulin-stimulated glucose uptake by cells from anovPCO was attenuated at higher doses of insulin (100 and 1000 ng/ml) compared with that by cells from either ovPCO (P=0.02) or controls (P=0.02). Insulin and LH stimulated lactate production in a dose-dependent manner, but insulin-dependent lactate production was markedly impaired in granulosa-lutein cells from anovPCO compared with either normal (P=0.002) or ovPCO (P<0.0001). By contrast, there was no difference in insulin-stimulated progesterone production between granulosa-lutein cells from the three ovarian types.

CONCLUSIONS

Granulosa-lutein cells from women with anovPCOS are relatively resistant to the effects of insulin-stimulated glucose uptake and utilization compared with those from normal and ovPCO, whilst maintaining normal steroidogenic output in response to physiological doses of insulin. These studies support the probability of a post-receptor, signalling pathway-specific impairment of insulin action in PCOS.