O-123 FMEA analysis of an automatic integration of time-lapse incubators into electronic medical records using CHLOE(Fairtility) shows risk reduction through automation of data capture and processing

Abstract

Abstract Study question Can integrations and automatic data processing between time-lapse incubators and EMRs reduce the risks associated with manual moving of data from time-lapse incubators to EMRs Summary answer Redesigning data workflow using CHLOE(Fairtility) decreased risk occurrence and increased risk detection possibilities associated with embryo classification and selection to freeze, biopsy, transfer and discard. What is known already Decisions are made from information derived from time-lapse incubators. Clinically, embryologists decide which embryos (and when) are suitable for transfer, cryopreservation, biopsy, or discarding based on data derived from time-lapse incubators, manually annotated and summarised into the electronic medical record (EMR) where further information useful for embryo selection is stored. Manual movement of data from time-lapse incubators to EMRs is time-consuming, administrative, reduces the granularity of the data available and incurs risk of human-error inaccuracies. These challenges limit the possibilities of how this data can be used to optimise clinical decisions, improve the patient experience and proactively detect operational anomalies. Study design, size, duration Failure mode effects analysis (FMEA) analysis was carried out on the workflow integration into a large (>5000 cycles per annum) IVF centre following ESHRE guidelines for laboratory and time-lapse practice (ESHRE,2015,2020), comparing before and after the introduction of CHLOE(Fairtility). The FMEA analysis evaluated the possible data capture, processing and associated clinical decision risks from embryos entering to leaving the time-lapse incubator. The Risk Priority Number (RPN=likelihoodxseverityxdetection of incidence) was calculated for each failure mode (Rienzi,2015). Participants/materials, setting, methods Through authenticated REST API calls according to the OpenAPI standard, CHLOE(Fairtility) linked the treatment unique identifier from the EMR(LIVO, inhouse developed) to the time-lapse incubator, automatically processed the time-lapse data, captured quantitative and qualitative information (such as morphokinetic time points, PNs, cleavage and blastocyst morphological grades, unusual embryo developmental anomalies and prediction scores for blastulation and implantation) and automatically loaded into the EMR. Main results and the role of chance Before CHLOE(Fairtility), 8 process phases were identified, with 81 associated failure modes, among which 45 risks were given a moderate RPN [RPN>15, i.e. data entry error into the EMR; image feature detection missed (i.e. 2PNs, Inner Cell Mass, incorrectly diagnosing fragments as cells and vice versa; incorrectly diagnosing vacuoles as a PNs, asynchronous PNs missed),with consequences including inaccurate KPI monitoring (n = 20, RPN=4); reduced patient experience and increased stress (n = 18, RPN range 3-16); wrong embryo being selected (n = 42, RPN range 8-36). Wrong embryo selection had three possible consequences: viable embryo discarded leading to a reduction in efficacy of treatment; viable embryo not prioritised for transfer causing reduced chance of pregnancy, or increased time to pregnancy, increasing cost and emotional burden); euploid embryo not prioritised for biopsy, increasing cost. Overall, RPN ranged from 3 to 36. After CHLOE(Fairtility), 51 failure modes were eliminated completely, including quantitative and qualitative morphokinetic annotations, entering data into the EMR for daily embryo grades, and embryo fate decisions. A further 22 failure modes had reduced RPN, including blastocyst morphological grading, number of PNs, identification of unusual embryo cleavages; with 30 low RPNs and 6 moderate RPNs. Implementation of CHLOE(Fairtility) reduced the highest RPN from 36 to 16. Limitations, reasons for caution FMEA is a proactive method to identify potential incidents in order to develop strategies to mitigate risks, forming part of a framework for responsible innovation. The likelihood of incidences were estimated based on a PUBMED literature review, personal experience and the experience of colleagues. Wider implications of the findings CHLOE(Fairtility) has the potential to eliminate risks that exist when manually moving data from time-lapse incubators to EMRs: time-consuming, administrative, reduced data granularity and human-error-based inaccuracies. CHLOE(Fairtility) optimises clinical decisions, providing an opportunity for personalised patient care, improved patient engagement, and the potential to detect operational non-conformities before impacting clinically. Trial registration number NA