Abstract
Purpose :
RETIPROGRAM is a registered clinical decision support system program for diabetic retinopathy (DR) screening. The algorythm calculates the risk of developing DR in the next years attending to medical variables. The initial results were already shown previously in ARVO 2020. Currently the aim is not only to show its accuracy as much as its prediction capability, detecting future DR in short term in patients without DR.
Methods :
The program was developed since a statistical analysis was based on fuzzy random forest methodology, assessing different medical variables.
The sample included 98873 of diabetes mellitus type 2 population. The medical variables registered were age, gender, diabetes mellitus duration, arterial hypertension, HbA1, glomerular filtration rate, microalbuminuria and body mass index. Attending to these variables it is calculated an estimated risk and it is suggested a date for the next control.
Results :
To validate the results consistency the risk percentage were compared with the presence of DR in the retinography.
The test sensibility was 60.36%, specificity 78,61% and 19644 were considered false positive. The false positive group images were reviewed after two years and it was shown that 11.949 had already DR. The progression rate in the false positive group was 60,82% in two years.
The initial sensitivity and specificity values were good enough, but if they are corrected according the results after two years, the sensitivity (84,20%) and specificity (90,37%) values would be even better.
False positive group, is commonly considered as a test error, but in this case, analyzing it carefully it would report some added test value, the capability of prediction in absence of DR in the retinography.
Conclusions :
In general, and also in DR, screening models efficacy means its capability to detect pathologies in early stages and their risk of progression. Currently all patient without DR in the retinography are usually considered into low risk group, ignoring their medical risk and they would be misdiagnosed. On other hand, an isolated retinography offers an accurate information about the current patient status but it has to be completed with medical factors related with the progression risk (RETIPROGRAM).
In our opinion RETIPROGRAM offers a predictive value of DR progression in short term and completes the retinographies information, offering a more personalized and accurate acting plan.
This is a 2021 ARVO Annual Meeting abstract.