JMIR Publications

ABSTRACT

Background:

Current data analysis methods do not effectively support nurses in risk reduction as retrospective reporting lacks real-time insights and precludes proactive care. Analysis of administrative data within Australia’s healthcare sector may have the potential to address this short coming. Predictive analytics can transform this data into meaningful insights, identifying harm risk profiles that benefit the performance of Australian and international clinical programs. Importantly these tools may offer support to nurses in precluding adverse events and predicting high-risk situations. Researchers, in collaboration with local health network staff will develop a proof-of-concept predictive risk algorithm. The ‘Predictive Harm Response Management algorithmic tool to reduce adverse events in healthcare settings’ program (Project #DHCRC-0156) will provide real-time insights via an interactive dashboard, enabling nurses to assess risks and optimise resources in healthcare settings. This protocol details the algorithm development activities for sub-project 1a: ‘Predictive risk model development’; which aims to develop and pilot-test a predictive harm algorithm for two South Australian (SA) Local Health Networks (LHN).

Objective:

1) Identify the clinical harm outcome of interest and relevant data sources per site and build a suitable data solution to model predictors of harm risk. 2) Identify actionable clinical, workforce, and environmental factors that are affecting the harm outcome of interest.

Methods:

This study design includes three phases (i) model generation, (ii) model evaluation, and (iii) prototype development. Data linkage by the SA-NT DataLink can only proceed following receipt of approval from each of the following: SA DHW ethics, UniSA ethics, and Hospital governance committees. The clinical dataset will be divided into a training set, validation set, and test set. Exploratory Data Analysis will be undertaken to ascertain features and classify outcomes from the raw dataset. Initially, modelling packages will be used for feature selection with imputed data. Imputation models will be built using features that have at least a minimal correlation with the variable being imputed. Once feature selection is complete, imputation will be repeated, incorporating additional predictors. Iterative model development will occur over three stages while a dashboard to display these results will be developed.

Results:

The study commenced on 19 July 2021 and will cease on 31 December 2025. Finalised results are expected November 2025.

Conclusions:

This research will conclude with a presentation of the PreHaRM tool, consisting of the algorithm and dashboard, within two SA LHNs. Research activities will be reported in publicly-available reports and manuscripts prepared for peer-reviewed journals and be drafted in accordance with existing and appropriate checklists.