Chronic Kidney Disease (CKD) is a gradual loss of kidney function that can occur due to diabetes, high blood pressure, recurrent infections, and urinary tract obstruction [1]. Unlike acute renal failure, which progresses rapidly and is potentially reversible, CKD is a long-term disease in which kidney damage is permanent and progressive [2]. Based on the 2010 Global Burden of Disease (GBD) investigation, CKD was ranked 27th in the list of the causes of deaths worldwide in 1990, but it was ascended to rank 18th in 2010 report [3]. This disease can progress to ESRD, which is fatal without dialysis or KT [4].
Nowadays, KT is the most effective treatment for advanced and irreversible renal failure [5]. According to the investigation done on the WHO data (2008) in 104 countries accounting for 90% of all transplantations across the world, about 100,800 organ transplants have been performed in these countries, and there have been 69,400 cases of KT (over 68% of all transplants in the world, from which 46% have been carried out with the live donors), [6]. This procedure is remarkably cost-effective and imposes fewer complications on patients compared to dialysis [7]. Kidney graft rejection has many consequences, such as re-transplantation or even death [8]. The problem of graft rejection is mainly attributed to chronic allograft rejection associated with dysfunctional donor-recipient matching [9]. Based on the literature, kidney graft survival prediction is essential for the transplant success, since it both increases the effective use of healthcare system's resources and the utility of available organs [9,10,11].
The increasing use of new data mining techniques, especially to discover unique patterns, has become widespread in the medical industry [12, 13]. These methods could be applied as adjunct tools to predict survival of graft transplantation [14]. Along with the use of data-based algorithms, Information Communication Technology (ICT) has revolutionized the e-health industry. Among the various ICT tools, wireless and smartphone-based technologies offer opportunities to reduce costs and raise access to health services, which improve the effectiveness of healthcare delivery process [15]. Thus, the integration of wireless technologies, data mining and machine learning approaches leads to effective delivery of care and provision of unique diagnostic services to individuals [16].
Previous studies have established several effective prediction data-driven models to identify kidney post-transplantation graft survival rates [9, 10, 17]. Our previous study dealt with data and modeling to predict post-transplantation graft survival among kidney transplant recipients using data mining algorithms [18]. In the present study, C5.0 algorithm with the highest accuracy (96.77%) was the chosen model for predicting patients' kidney transplant survival. In this way, the output of these results can be monitored through a smartphone or a tablet application that is easily accessible. Appropriate medical decisions can also be made based on these results. Thus, the main purpose of this study was to design and evaluate a smartphone-based application to predict the survival of KT for specialists and patients. To evaluate the application's usability, we used a standard questionnaire. In the implementation phase, the beneficiaries of this application included urologists, nephrologists, and kidney transplant patients. The primary use of the smartphone-based application is to help professionals predict the survival of KT. Also, there are facilities in this software through which, patients can use the existing reminders for appointments and set the new version of the reminders for suppressive drugs and post-transplant care instructions.
Implementation
This is an applied-developmental study that was conducted by Tehran University of Medical Sciences. This study was conducted in three main phases:
First phase of the study
The initial parameters in predicting post-transplantation graft survival among kidney transplant recipients were determined by reviewing specialized textbooks and consulting with supervisors and a clinical consultant. Based on the initial review, a researcher-made questionnaire was developed to assess the information needs of the application through urologists and nephrologists. This questionnaire was distributed among specialists (three urologists and four nephrologists) working at Sina Hospital. Participants were asked about data items and the capabilities required by an application to predict kidney transplant survival. The convenience sampling method was used to select the research sample. Each of the required data items was considered essential if an average of 50% of the respondents recognized it as necessary, which was then used in the application design. The reliability of the questionnaire was measured by test–retest method so that, after a short time, the questionnaire was given to the same people to be completed by them again. The scores obtained from the two tests were examined and the correlation coefficient of 92% was obtained. The questionnaire helped to prepare a checklist by which, further extraction of information was performed. Using this checklist, the medical records of 513 kidney transplant patients at Sina Urological Research Center were reviewed and main features were identified.
Second phase of the study
After collecting the input data items related to each patient by checklists, and also modifying the data to reduce the modeling error, useful features were extracted to predict the survival of KT. The predictive models such as C5.0, C&R, and neural network algorithms were used. The details and results of training and testing processes have been provided in our previous study [19]. In order to design and implement a multi-level smartphone application for predicting the survival of KT, the PhoneGap framework was used, which was responsible for communicating with the hardware and smartphone operating system. After installing the PhoneGap framework, Android and iOS platforms were run on it. The coding of programs’ display was done by Hypertext Markup Language (HTML) and Cascading Style Sheets (CSS), which included the formatting and components of the application’s pages. Also, JavaScript and its more advanced codes, such as jQuery, were employed.
Third phase of the study
In the last phase, a standard questionnaire was used to evaluate the usability of the final version of the application [20]. The chosen standard questionnaire was the Questionnaire for User Interface Satisfaction (QUIS), which elicits users’ opinions and evaluates users’ acceptance of the application interface. Validity of the questionnaire was confirmed by face validity and also taking into account the opinions of five specialists, thus its reliability was reported at ɑ = 0.94. Figure 1 shows different phases of the study separately.