F 3 | Orals | SPHC 2024

Room

1er étage - E130


Theme

Tools and methods for better health


Chair

Oliver Gruebner



Title
Tools and methods for automatic detection of healthcare-related adverse events: a systematic review

Name
Arthur Escher 

Affiliation
Unisanté


Abstract

Background:
Healthcare-related adverse events (AEs) are a prevalent and significant concern, occurring in approximately 10% of hospital stays and accounting for 15% of hospital expenses. Traditional screening methods rely on retrospective manual record reviews, demanding substantial human and temporal resources. The integration of artificial intelligence (AI) offers a promising avenue for more efficient detection of AEs.

Objective:
The objective is to systematically review both scientific and grey literature on automated tools for detecting healthcare-related adverse events.

Methods:
We conducted a comprehensive search using established Cochrane methodologies in Medline, Embase, CINAHL, Cochrane Library, Joanna Briggs Institute, Web of Science, and Google Scholar from 2009 to 2024. This search also included a review of references from selected articles.

Our inclusion criteria encompass all articles in French, English, and German related to healthcare-related adverse events and automated screening, without any exclusion based on study type or patient age.

Two independent reviewers will undertake study selection, data extraction, risk of bias assessment, and GRADE evaluation. Results will include: 1) Methodological characteristics such as the automated screening method, its metrics, comparison with non-automated methods, data sources, and patient selection criteria; 2) Characteristics of AEs like prevalence/incidence, types and classification, validation criteria, and severity; 3) Evaluation for potential inclusion in the Swiss quality framework, international comparisons, cost-benefit analysis, and examination of facilitators and barriers to implementation. Results will be categorized by healthcare setting (acute or long-term hospital stay, outpatient).

Results:
An initial search across various databases indicates a need to screen around 4,000 abstracts (prior to duplicate removal). Results will be presented at the conference.

Conclusions:
The review aims to identify or develop cost-effective and reliable automated methods for screening and monitoring healthcare adverse events across various settings. This will inform Swiss healthcare policy in Quality and Safety.



Title
Entwicklung einer Smartphone Applikation zur benutzerfreundlichen Erfassung des individuellen Gesundheitszustandes

Name
Claudia Huber 

Affiliation
University of Applied Sciences of Western Switzerland

Abstract

Hintergrund:
Junge Erwachsene sind in ihrem Alltag hohen Anforderungen ausgesetzt, welche Auswirkungen auf ihre physische und psychische Gesundheit haben können. Im Rahmen eines Vorprojekts, welches von 06.2021 bis 01.2022 dauerte, wurde zusammen mit Studierenden der Fachhochschule Freiburg eine Smartphone Applikation «MonCoSaMo» (Monitoring de la Consultation Santé Mozaik) zur Erfassung des individuell erlebten Gesundheitszustandes entwickelt. Diese App umfasst auch zusätzliche Unterstützungsangebote. In einer erweiterten Anwendung und in einem Folgeprojekt wird diese App, die sich gleicherweise als Instrument zum Gesundheitsmonitoring von jungen Erwachsenen im Kanton Freiburg eignet, in einer Längsschnittstudie getestet werden. Hier wird die Entwicklung der App vorgestellt.

Methode:
Mit Hilfe von halbstrukturierten Fokusgruppeninterviews und einer Fragebogenerhebung wurden bei einer gezielt ausgewählten Bevölkerungsgruppe von jungen Erwachsenen der individuell erlebte Gesundheitszustand, die Bedürfnisse an Unterstützungsangeboten und die Erwartungen an eine Smartphone Applikation ermittelt.

 
Ergebnisse:
Insgesamt haben 37 von 71 angefragten Studierenden an der Umfrage teilgenommen, was einer Rücklaufquote von 52% entspricht. Davon haben 15 Studierende zusätzlich an einem Fokusgruppeninterview mitgemacht. Dabei wurden Fragen zu folgenden Themen erforscht: Stress, Ernährung, körperliche Aktivität, Suchtmittelkonsum, Internet-Nutzung, Sexualität, Gewalt, Allgemeine Gesundheit und Wohlbefinden. Die Datenanalyse der obigen Parameter und die Rückmeldungen von Anwenderinnen und Anwendern zeigten, dass die zweisprachige App (D/F) ein benutzerfreundliches Instrument zur Erfassung des individuellen Gesundheitszustandes ist. Mit wenigen Anpassungen lässt es sich auch bei einer grösseren, stärker diversifizierten Bevölkerungsgruppe anwenden.

Schlussfolgerung:
Schliesslich ist die bestehende Smartphone Applikation hilfreich für ein zweckmässiges Monitoring der Gesundheit von jungen Erwachsenen im Kanton Freiburg. Daraus können zum einen gezielte und gesundheitsfördernde und zum anderen präventive Massnahmen geplant und umgesetzt werden. Eine regelmässige Datenerhebung ist notwendig, um Risiken und Ressourcen zu erkennen und um gesundheitspolitisch über den Verlauf der Gesundheit junger Erwachsener informieren zu können.



Title
De-Identification of health data in compliance with Swiss legislation 

Name
Julia Maurer 

Affiliation
Swiss Institute of Bioinformatics/SPHN


Abstract

The de-identification of health data postulates an essential approach to protect patient privacy and is, beside other requirements, a prerequisite for data sharing in multi-center research projects. Even though there are international guidelines available concerning the de-identification of data, there is no guidance for the de-identification of health data specifically taking into account the Swiss law and data protection regulations. The U.S.-Health Insurance Portability and Accountability Act (HIPAA) Privacy Rule establishes national standards to protect individuals' medical records and other personal health information, and also Swiss research projects often refer to the HIPAA when documenting the de-identification process. However, the HIPAA Privacy Rule cannot be executed in its form in Switzerland.

For this reason, the Swiss Personalized Health Network (SPHN) launched the “Data de-identification Project” in 2021 and developed a first set of recommendations for the de-identification of health data. A revised version of the recommendations in line with the SPHN Interoperability Framework will be published soon, enabling data sharing in accordance with current Swiss legislation and data protection regulations.

The developed phased approach not only aims to cover the risk mitigation for re-identification, but also considers the management of a de-identification process that foresees the verification and the periodic review of the risk assessment performed. Hands-on support for defining project specific de-identification and re-identification risk is provided with a template to evaluate the use case and assess the risk according to the project specifications.

Most importantly, the recommendations demonstrate that the simple removal of the direct identifiers does not necessarily mean that the data can only be re-identified with disproportionate effort, as the risk from combination or other remaining risks are not taken into account. Therefore, any rule-based approach will have to be combined with a risk assessment, in order to satisfy Swiss law requirements.

First SPHN research projects recently successfully applied the de-identification approach and submitted it with the proposal to the ethics committees. The goal is to enhance the adoption of de-identification methods by providing training to the Swiss research community in collaboration with Swiss hospitals, thereby fostering its broad implementation.



Title
Détection des hémorragies associées aux antithrombotiques chez les patients âgés hospitalisés à partir du dossier médical informatisé : étude transversale dans trois hôpitaux universitaires suisses.

Name
Marie-Annick Le Pogam  

Affiliation
Centre de Recherche et d’Innovation en Sciences Pharmaceutiques Cliniques, Hôpital Universitaire de Lausanne, Lausanne, Suisse. Ecole des Sciences Pharmaceutiques, Université de Genève, Genève, Suisse.


Abstract

Introduction:
Les hémorragies liées aux antithrombotiques chez les patients âgés, souvent sous-déclarées, représentent des enjeux sanitaires et économiques majeurs. Le développement d'un système de détection automatisé via le dossier médical informatisé (DMI) peut renforcer la surveillance et la gestion des risques liés à ces médicaments fréquemment utilisés chez les séniors.

Ce projet, soutenu par le PNR74, visait à développer et valider un outil automatisé pour détecter les hémorragies sévères (HS) et cliniquement significatives (HCS) associées aux médicaments antithrombotiques chez les patients de 65 ans et plus, hospitalisés en soins aigus.

Méthode:
L'étude a inclus tous les séjours en soins aigus de patients de 65 ans et plus traités par antithrombotique en 2015/2016 dans trois hôpitaux universitaires suisses. Nous avons élaboré des algorithmes pour identifier les HS et HCS, basés sur les critères de l'International Society of Thrombosis and Haemostasis, en utilisant des données structurées (diagnostiques, procédures, résultats de laboratoire, médicaments) et non structurées (lettres de sortie) des DMI. Ces algorithmes combinaient des règles spécifiques et des techniques de traitement du langage naturel. Leur performance a été évaluée par rapport à une revue manuelle de 800 séjours aléatoires.

Résultats:
Parmi 36,039 séjours, les taux d'incidence des HS et HCS étaient de 8.3% et 15% respectivement, avec une mortalité intra-hospitalière de 1% pour les HS. Les contributions des diverses sources de données étaient : laboratoire (66.9% HS/62.9% HCS), codes CIM-10-GM (28.5% HS/35.4% HCS), transfusions sanguines (3.2% HS/1.7% HCS), et prescriptions antihémorragiques (1.4% HS). La validation des algorithmes se poursuit, avec des résultats attendus pour la conférence.

Conclusion:
L'approche automatisée combinant diverses sources de données du DMI représente une innovation majeure, promettant d'améliorer la sécurité des patients en facilitant la détection des effets indésirables des médicaments.



Title
Predicting patients with high insurance expenditures for home care services using machine learning

Name
Flurina Meier Schwarzer

Affiliation
Zurich University of Applied Sciences, Swiss Tropical and Public Health Institute, University of Basel


Abstract

Background:
The number of persons using home care services increased in Switzerland in the past years and is expected to increase further. According to Swiss home care service organisations especially the number of cases with high health care use increased and they assert that the current Swiss tariff system poorly reflects high users. Moreover, it remains poorly understood which patient characteristics are linked to higher service use or high insurance expenditures and home care service organisations have a poor basis to predict which patient might be part of this group. The objective of this study was to identify patient characteristics that predict insurance expenditures above 1000 CHF per month and to determine which patient characteristics are most important to predict expenditures.

Methods:
This observational study was based on data from eight Swiss home care service organisations. All patients undergoing routine assessments with the interRAI Home-Care Switzerland (HCS) questionnaire during the inclusion phase were included in the study. Besides data from the interRAI-HCS further characteristics covering mostly the case environment were assessed. Administrative data on service use in the three months following the assessment was the basis to calculate insurance expenditures. 168 patient features served as input variables and insurance expenditures above/below 1000 CHF per month as endpoint in our predictive classification models. The aim was to find the best prediction model by evaluating six machine learning methods (including random forests, neural networks and XGBoost). Explainable-AI approach SHAP was used to explain the contribution of each feature to the model prediction and to assess global feature importance. 

Results:
1035 patients were included in our study (mean age: 80 years, SD: 11.6). Best predictions were found with the XGBoost model that resulted in a ROC AUC value of 0.857 in the holdout set. The three patient characteristics most influential for expenditure prediction were: strength of limitations for dressing below the waist, use of meal services and needs for help with medication management.

Discussion:
Machine learning is useful to predict insurance expenditures in the home care services setting. However, the generalisability of our models needs to be further evaluated in broader contexts. Our study can provide guidance on which patient characteristics should be routinely collected by organisations to predict insurance expenditures.