In Human reproduction (Oxford, England)
STUDY QUESTION : Can early pregnancies be accurately and cost-effectively diagnosed and managed using a new medical computerized tool?
SUMMARY ANSWER : Compared to the standard clinical approach, retrospective implementation of the new medical software in a gynaecological emergency unit was correlated with more accurate diagnosis and more cost-effective management.
WHAT IS KNOWN ALREADY : Early pregnancy complications are responsible for a large percentage of consultations, mostly in emergency units, with guidelines becoming complex and poorly known/misunderstood by practitioners.
STUDY DESIGN, SIZE, DURATION : A total of 780 gynaecological emergency consultations (446 patients), recorded between November 2018 and June 2019 in a tertiary university hospital, were retrospectively encoded in a new medical computerized tool. The inclusion criteria were a positive hCG test result, ultrasonographical visualization of gestational sac, and/or embryo corresponding to a gestational age of 14 weeks or less. Diagnosis and management suggested by the new computerized tool are named eDiagnoses, while those provided by a gynaecologist member of the emergency department staff are called medDiagnoses.
PARTICIPANTS/MATERIALS, SETTING, METHODS : Usability was the primary endpoint, with accuracy and cost reduction, respectively, as secondary and tertiary endpoints. Identical eDiagnoses/medDiagnoses were considered as accurate. During follow-up visits, if the updated eDiagnoses and medDiagnoses became both identical to a previously discrepant eDiagnosis or medDiagnosis, this previous eDiagnosis or medDiagnosis was also considered as correct. Four double-blinded experts reviewed persistent discrepancies, determining the accurate diagnoses. eDiagnoses/medDiagnoses accuracies were compared using McNemar's Chi square test, sensitivity, specificity, and predictive values.
MAIN RESULTS AND THE ROLE OF CHANCE : Only 1 (0.1%) from 780 registered medical records lacked data for full encoding. Out of the 779 remaining consultations, 675 eDiagnoses were identical to the medDiagnoses (86.6%) and 104 were discrepant (13.4%). From these 104, 60 reached an agreement during follow-up consultations, with 59 medDiagnoses ultimately changing into the initial eDiagnoses (98%) and only one discrepant eDiagnosis turning later into the initial medDiagnosis (2%). Finally, 24 remained discrepant at all subsequent checks and 20 were not re-evaluated. Out of these 44, the majority of experts agreed on 38 eDiagnoses (86%) and 5 medDiagnoses (11%, including four twin pregnancies whose twinness was the only discrepancy). No majority was reached for one discrepant eDiagnosis/medDiagnosis (2%). In total, the accuracy of eDiagnoses was 99.1% (675 + 59 + 38 = 772 eDiagnoses out of 779), versus 87.4% (675 + 1 + 5 = 681) for medDiagnoses (P < 0.0001). Calculating all basic costs of extra consultations, extra-medications, extra-surgeries, and extra-hospitalizations induced by incorrect medDiagnoses versus eDiagnoses, the new medical computerized tool would have saved 3623.75 Euros per month. Retrospectively, the medical computerized tool was usable in almost all the recorded cases (99.9%), globally more accurate (99.1% versus 87.4%), and for all diagnoses except twinning reports, and it was more cost-effective than the standard clinical approach.
LIMITATIONS, REASONS FOR CAUTION : The retrospective study design is a limitation. Some observed improvements with the medical software could derive from the encoding by a rested and/or more experienced physician who had a better ultrasound interpretation. This software cannot replace clinical and ultrasonographical skills but may improve the compliance to published guidelines.
WIDER IMPLICATIONS OF THE FINDINGS : This medical computerized tool is improving. A new version considers diagnosis and management of multiple pregnancies with their specificities (potentially multiple locations, chorioamnionicity). Prospective evaluations will be required. Further developmental steps are planned, including software incorporation into ultrasound devices and integration of previously published predictive/prognostic factors (e.g. serum progesterone, corpus luteum scoring).
STUDY FUNDING/COMPETING INTEREST(S) : No external funding was obtained for this study. F.B. and D.G. created the new medical software.
TRIAL REGISTRATION NUMBER : NCT03993015.
Blavier F, Grobet D, Duflos C, Rayssiguier R, Ranisavljevic N, Duport Percier M, Rodriguez A, Blockeel C, Santos-Ribeiro S, Faron G, Gucciardo L, Fuchs F
2023-Feb-10
artificial intelligence, cost, diagnosis and management tool, pregnancies, software, ultrasound
