Real-world implementation of electronic patient-reported outcome measures (ePROMs) in routine oncology practice: a...

Routine collection of electronic patient-reported outcome measures (ePROMs) enhances symptom monitoring and improves outcomes in cancer care. We evaluated published reports of real-world ePROM implementation in oncology, focusing on approaches to implementation, types of ePROMs used, completion rates, factors associated with ePROM completion, and impacts on patient and health care process/system-level outcomes. We also aimed to propose a classification of ePROM programmes.

We systematically searched Medline, Embase, CINAHL Complete, and Scopus for articles describing routine ePROM implementation and evaluation. Data charting captured characteristics of the study, study population, characteristics of routinely collected ePROMs, and evaluation of their routine use. We also present a new classification of ePROM programmes into three levels: asynchronous, clinician-supervised, and real-time clinician-supervised, based on clinical pathway integration.

Of 10 384 identified manuscripts, 50 met the inclusion criteria, reporting on 39 real-world ePROM programmes. Nearly all ePROM programmes (97.4%) collected data on symptom burden. The median overall completion rate was 61% (interquartile range 47%-76%), with substantial variability over time and across hospitals. Completion was associated with sociodemographic, clinical, and treatment-related factors. Most ePROM programmes (61.5%) were classified as asynchronous models, with five (12.8%) clinician-supervised models and three (7.7%) real-time clinician-supervised models, while seven (17.9%) lacked sufficient detail. The most frequently reported outcomes were fewer hospitalisations or emergency department visits (10.3%) at the patient level and improved symptom identification (17.9%) at the process/system level.

ePROMs are increasingly embedded in routine oncology care, demonstrating benefits at both patient and system levels. However, most current programmes adapt asynchronous models with limited rapid clinician interaction.