d-GDM: A mobile diagnostic decision support system for gestational diabetes

ABSTRACT Objective The aim of the study is to describe a portable and convenient software to facilitate the diagnostics of gestational (GDM) and pre-gestational diabetes (PGDM). Materials and methods An open source software, d-GDM, was developed in Java. The integrated development environment Android Studio was used as the Android operational system. The software for GDM diagnosis uses the criteria endorsed by the International Association of Diabetes and Pregnancy Study Group, modified by the World Health Organization. Results GDM diagnosis criteria is not simple to follow, therefore, errors or inconsistencies in diagnosis are expected and could delay the appropriate treatment. The d-GDM, was developed to assist GDM diagnosis with precision and consistency diagnostic reports. The open source software can be manipulated conveniently. The operator requires information regarding the gestational period and selects the appropriate glycaemic marker options from the menu. During operation, pressing the button “diagnosticar” on the screen will present the diagnosis and information for the follow up. d-GDM is available in Portuguese or English and can be downloaded from the Google PlayStore. A responsive web version of d-GDM is also available. The usefulness and accuracy of d-GDM was verify by field tests involving 22 subjects and 5 mobile phone brands. The approval regards user-friendliness and efficiency were 95% or higher. The GDM diagnosis were 100% correct, in this pilot test. d-GDM is a user-friendly, free software for diagnosis that was developed for mobile devices. It has the potential to contribute and facilitate the diagnosis of gestational diabetes for healthcare professionals.


INTRODUCTION
G estational diabetes mellitus (GDM) is the most common complication associated with pregnancy (1,2).Chronic hyperglycemia in pregnancy is associated with several complications in pregnant women and the fetus (3,4).
The definition of gestational diabetes has been changing over time (5).Until recently, a simple definition characterized GDM as "carbohydrate intolerance first diagnosed during gestation" (6,7).Currently, GDM is defined as "a carbohydrate intolerance of variable severity, which begins during current gestation and d-GDM: gestational diabetes diagnostic app Arch Endocrinol Metab.2019;63/5 does not meet the diagnostic criteria for overt diabetes mellitus" (8,9).Therefore, the concept of pre-existing diabetes or pre-gestational diabetes was introduced (10).
GDM diagnosis is based on the blood measurement of glycemia while fasting and/or after an oral glucose load (8,12).Different diagnostic criteria and cutoff values for GDM were proposed (5,13).Figure 1 summarizes the GDM diagnosis as preconized by the International Association of Diabetes and Pregnancy Study Group (IADPSG) with modification by the World Health Organization (WHO) (9).
The described criteria in Figure 1 is based on the findings of the Hyperglycemia and Adverse Pregnancy Outcome (HAPO) study, and the cut-off values proposed by IADPSG (3,14).WHO endorsed the IAPDSG criteria for GDM with minor modifications (9).
The diagnosis of GDM and PGDM (Figure 1) is not simple to follow and requires continuous training of healthcare professionals.
The benefit of information technology (IT) applied to health care is well known (15)(16)(17).Several studies have reported the benefits of IT for interventions, improved clinical decision, and improved training and support for healthcare professionals (18)(19)(20).
The aim of the study is to describe a portable and convenient software to facilitate the diagnosis of GDM and PGDM.

Software development
The software was built using the variables and criteria suggested by WHO.These variables are as follows: fasting glycemia, random glycemia, and glycated hemoglobin (HbA1c) in the first trimester (first evaluation).Subsequently, fasting glycemia and glycemia after 1 hr and 2 hr, after an oral glucose tolerance test (OGTT) with a glucose load of 75 g during a period of 24-28 weeks of gestation (as described in Figure 1).
All informatics tolls that were applied are opensource.Java was used for development.d-GDM was developed for the Android Operating Systems using the Android Studio (https://developer.android.com/studio/) integrated development environment (IDE), in which IntelliJ IDEA from JetBrains is applied.
d-GDM was developed considering the principles of human-computer interaction (HCI).HCI is a field that involves the interaction between people (users) and computational devices, developing methods and tools for the design, creation, and implementation and maintenance of computational systems suitable for human use (21).HCI seeks to minimize the barrier between the human user and the computer system by Figure 1.Flowchart for Gestational and Pregestational diagnosis as recommended by IAPDSG-WHO.During the first trimester, the diagnostic criteria for GDM is briefly based on fasting glycemia.Pregestational diabetes (PGDM), also known as preexisting diabetes, overt diabetes, manifest diabetes, or "diabetes in pregnancy" is characterized in this period using the same criteria applied to nonpregnant women (14).All pregnant women, without previous dysglycemia diagnosis, should be tested using the glucose oral tolerance test (OGTT) with a glucose load of 75 g between 24−28 weeks of gestation (2º/3º trimesters).In OGTT only one value above the cut-off is required for GDM or PGDM diagnosis.offering user-friendly interfaces that can be used with the least effort (22).The flowchart of the programming logic is shown in Figure 2.
The programming logic is described briefly below.A responsive web version (adaptable and flexible to all computer screens) was developed and can be accessed at http://www.cpdm.ufpr.br/dg/.

RESULTS
The major characteristics of the d-GDM software are shown in Figure 3.
The open-source software shows a screenshot in Portuguese or English.It is easy and simple to use.The user requires information regarding the gestational period and select the appropriate glycemic marker options.Pressing the button "diagnosticar" (or "submit", English version) in this scenario will result in the diagnosis and information regarding the follow up to be presented on the screen.
The usefulness and accuracy application of d-GDM was verified by field tests involving graduate students (n = 22) using five different mobile phone brands.Questions on user-friendliness, language and efficiency showed scores of 95% or higher associated with highest approval of d-GDM (data not shown).Suggestions from the users regarding the interface layout of the app were incorporated in d-GDM final version.
To verify the d-GDM accuracy, the group of postgraduate students received a 30 minutes training on GDM diagnostic with the flowchart described in Figure 1.
Following, the group received a spreadsheet (Microsoft Excel) with 50 results of fasting glycemia and OGTT, randomly containing 25% of positive results for GDM, confirmed by two senior researchers; they In Figure 3, a pregnant woman is considered as an example.In the first screenshot, all variables radio buttons were below the cut-off for diabetes.Tapping the "diagnosticar" button shows the diagnostic of no-diabetes, in Portuguese, in this version.In the second screenshot, the user has two options.Selecting the variables for the 1º trimester or for the 2º/3º trimesters (24−28 weeks of gestation) will cause the appropriate diagnostic to be displayed, such as GDM in the example.In the third screenshot, all radio buttons were in the highest cut-off and the "diagnosticar" bottom will display the diagnostics for preexisting diabetes (overt diabetes).mark (Yes or No) for GDM.This procedure mimics the "usual diagnostic" process.The same spreadsheet was then shuffled, and the group performed the diagnosis with the d-GDM app, after a 5 minutes training, blinded for the first results.
The study group presented a correct diagnosis in 94% with the "usual" procedure and 100% using the d-GDM app.Regardless of the small sample size, this preliminary result for accuracy was considered consistent and satisfactory for d-GDM.

DISCUSSION
The decision to develop a software program to assist the diagnosis of gestational diabetes was motivated by difficulties encountered by healthcare professionals.
The evolution in the definition of GDM and the different options for cut-off values of glycemic biomarkers and glucose loads could be relevant factors affecting diagnosis consistency.
Therefore, errors in diagnosis are expected and could delay the appropriate treatment.The d-GDM, was developed to assist GDM diagnosis with precision and consistency diagnostic reports.
In addition, rapid and continuous changes in the guidelines for the diagnosis of GDM have been observed, requiring constant updating of the professionals and favouring the use of non-standard criteria.
We developed a free application for mobile devices.This application is an intuitive and simple way to diagnose GDM.
The principles of good interaction between the health professionals and the d-GDM application were considered during development.A simplified screen, without the need to advance to subsequent screenshot was developed, thereby reducing unnecessary visualizations.This development yields desirable characteristics, according to the concepts of HCI (21,22).
The software was built with a radio button, also known as an "option button", which is graphical element that allows a user to select a single item from a predefined list of options.This prevents typographic errors associated with concentrations of biomarkers (Figure 3).The program operator selects simple choices and obtains an immediate response.The d-GDM application interface is designed, based on the models provided for the Android-based platform.
A new criterion for the diagnosis of GDM that may be considered in the future can easily be incorporated into d-GDM.The software structure of d-GDM presents a consolidated and flexible logic.
In Brazil, the number of mobiles in operation exceeds the number of inhabitants (23).Health professionals have been expanding the use of multiple mobile-based applications for diagnostic assistance (24,25).
Different technological processes, involving in silico analyses such as the use of artificial intelligence, associated with diabetes, are available and expand rapidly (26).These new technologies, in diabetes, are focused on processes of prevention, monitoring and research, for what has been described as "digital diabetes" (27).
The use of softwares available in mobile phones has been associated with improved glycemic control and self-management in diabetes care in type 2 diabetes, confirming the effectiveness of this technology (28).
Gestational diabetes softwares and apps, identified on the main download platforms, are primarily targeted at pregnant women, addressing glycemic control, weight management, lifestyle modifications, drug control and monitoring, and complications (29).
Different digital mobile health medicine (mHealth) software for gestational diabetes have been associated with reduced risk for the pregnant woman and the fetus (27,30).d-GDM is a software that fits into the emerging mHealth context.
The d-GDM was developed with the interactive process described by Garnweidner and cols. in 2015 (31).Briefly, a multidisciplinary health group, associated with computer and bioinformatics experts, developed the software prototype, which was improved by users.
Healthcare professionals are exposed to constant stress.After long hours of work, misunderstandings and errors are expected, especially in complex diagnoses such as GDM and PGDM.
Therefore, errors in diagnosis are expected and could delay the appropriate treatment.The d-GDM, was developed to assist GDM diagnosis with precise and consistency diagnostic reports.
In addition, rapid and continuous changes in the guidelines for the diagnosis of GDM have been observed, requiring constant updating of the professionals and favouring the use of non-standard criteria.d-GDM was developed for operation in an Android Operating System, which is the most popular in the world.This makes the system accessible to the largest Additionally, any computer can access the d-GDM web version.This option was developed for users who do not use the Android operating system.In the web version, more information such as the logic of the application and the diagnostic criteria, as well as the bibliographic references, are provided.The website also enables better communication between users and the developers of d-GDM, as well as the dissemination of the projects developed in gestational diabetes research.
In this context, the proposed d-GDM software can provide healthcare professionals with a free tool that facilitates the diagnosis of gestational diabetes, which is an issue of significant and increasing prevalence in Brazil and in the world.
To our knowledge d-GDM was the first-of-its-kind that is able to address the GDM diagnosis.
In conclusion, in this study, we developed and described d-GDM, which is free software for GDM diagnosis.d-GDM is user-friendly and has the potential to assist healthcare professionals in the diagnosis of gestational diabetes.
: gestational diabetes diagnostic app Arch Endocrinol Metab.2019;63/5 d-GDM: gestational diabetes diagnostic app Arch Endocrinol Metab.2019;63/5 number of mobile equipment in use.To our knowledge, d-GDM is the first software to address GDM and PGDM diagnosis to mobile devices.
da Propriedade Industrial (National Institute of Industrial Property) under the number BR512018001275-2, linked to the Federal University of Parana, Brazil.