This is a proof-of-concept study consisting of three substudies: (1) a survey of the recognition of blockchain medical records changes and the digital literacy of participants; (2) an observational study on participants using the blockchain-based mobile alert app; and (3) a usability survey study. The study was approved by the Institutional Review Board of Yonsei University Health System (Y-2019-0127) and all participants provided informed consent.

Before proceeding with the design and development of the mobile app, since the EHR systems used in each hospital setting differ, the documents used in the EHR systems were profiled using HL7 Fast Healthcare Interoperability Resources (FHIR; see Multimedia Appendix 1) [15,16]. Although it is reasonable that patients are notified of any changes in medical records, patients might be overwhelmed by too many notifications whenever any change occurs. Moreover, to improve the usability of the mobile app, changes should be summarized. Therefore, medical documents were divided into three types according to their importance and impact, and the mobile app was designed so that patients are notified of only high-risk changes in documents according to the following three risk levels: risk 1, medical information and other critical items that should not be changed; risk 2, medical information and other items that are allowed to be changed; and risk 3, nonmedical information. These risk levels were also considered in the profiles created.

The mobile app used in this study leverages the blockchain network MediBloc Panacea [17] that was developed based on the Tendermint blockchain [18]. The blockchain uses the delegated proof-of-stake method implemented by the practical Byzantine fault tolerance algorithm to create blocks. In this system, “delegated” refers to delegated nodes that perform and validate transactions and blocks. Validators are selected through voting. Normally, one block is generated per second, and all transaction history transmitted over the network is stored in the generated block. Further, similar to other blockchains, once created, blocks cannot be reversed. We used four different properties in the blockchain transaction: (1) writer of the transaction, which can only be specific clinics; (2) topic that is assigned per individual patient; (3) key; and (4) value. Key and value have the following five attributes: (1) hash value for medical documents, (2) document URL for the FHIR profile in which the document was transformed, (3) hash value before the medical document was changed, (4) risk and number of documents with changes, and (5) date to represent when the document was created. As soon as the document is created, the health information system spontaneously hashes and transacts the metadata of the document to the blockchain network (Figure 1). The mobile app provides users with logs of changes in medical records. The app has been available on the Google Play Store [19] since October 23, 2019.

Before the proof-of-concept study was deployed, we simulated the app to evaluate how it captures fake medical records caused by fraudulent actions. We created five fake medical documents for the purpose of simulation. All except for one dataset were assumed to have changes in the documents. The risk levels of the changes were set differently for each dataset. We transacted the datasets and checked how the mobile app worked.

The study was conducted at two primary care clinics specializing in pediatrics in Korea. Only outpatients of these two clinics were enrolled in the study. Anyone who visited the selected clinics was eligible to participate in the study. Because young patients were not interested in their own medical records, whereas their guardians were more interested, the guardians of patients enrolled in the study on their behalf.

When patients launch the mobile app from their devices, the user is authenticated according to a username and password. After login, the patient can see their own profile and events, which include when the records were created as well as which and how many items at high risk were changed. This information was obtained by querying blockchain using Owner and Topic values that were stored in the backend server of the mobile app.

To investigate participants’ level of recognition of blockchain, medical records changes, and their digital literacy, questionnaires were completed by all participants (Figure 2). The primary survey consisted of items regarding the recognition of blockchain, thoughts and experiences of changes in medical records, and digital literacy. There were six questions that measured blockchain recognition; these were based on virtual currency questions from a 2014 Survey of Consumer Payment Choice [20]. The second part of the questionnaires consisted of four questions. The first two questions were related to the recognition of medical record changes and the need for a monitoring system for medical record changes. These two questions were answered on a 5-point response scale. The other two questions addressed the participants’ experience of medical record changes. The third part, digital literacy, consisted of 10 questions answered on a 5-point response scale. These questions were taken from other digital literacy questions [21], modified for the mobile app, and condensed to 10 questions. A higher score represents a higher level of digital health literacy, except in the case of two questions. The questionnaire ended with demographic characteristics (age, sex, and occupation). All 5-point questions were answered, with responses ranging from very negative to very positive (ranging between 1 and 5).

Usage patterns of the app were measured by the number of logins, duration that indicated the difference in days between the first and the last logins, and an event log, which indicated how many items were changed and how important they were. These usage patterns were observed in participants who logged in more than once and for a duration of more than 28 days.

After 28 days of using the app, participants were invited via the app to take part in the secondary survey, which included a System Usability Scale (SUS) survey. For the evaluation of learnability and usability, we used the modified SUS, a reliable, low-cost usability scale that can be used for global assessments of systems usability (see Multimedia Appendix 2) [22,23]. Bangor et al [23] described the adjective ratings associated with SUS scores as follows: worst possible (mean SUS score 25), poor (39.17), satisfactory (52.01), good (72.75), excellent (85.58), and best possible (100).

Categorical variables are presented as numbers and percentages and were compared using the Chi-square test and the Fisher exact test. Continuous variables are expressed as mean (SD) and were compared using the Student unpaired t-test, analysis of variance, Wilcoxon signed-rank test, or Mann-Whitney U test as appropriate. Surveys answered by a Likert-type scale and scores are expressed as the mode as well as numbers and percentages. All statistical analyses were performed using R version 3.6.1 (R Foundation for Statistical Computing, Vienna, Austria) and 2-tailed tests. Results with P<.05 were considered to be statistically significant.

This proof-of-concept study applied a blockchain-based medical document monitoring system in two primary care clinics. Although there was a lack of recognition of the concept of blockchain (compared with bitcoin), participants’ trust level in blockchain was higher than that in bitcoin. Moreover, although there were few people who had experienced changes in their medical documents, the participants considered that medical records should not be changed, and therefore that this monitoring system is necessary.

There have been some blockchain-based implementations for managing medical records. Ariel C Ekblaw designed MedRec—a decentralized record management system for EHRs—and implemented the pilot system in the Beth Israel Deaconess Medical Center [24]. Zhang et al [9] designed the architecture of DApp, named FHIRChain, based on 5 key requirements provided by the ONC interoperability roadmap and demonstrated the prototype. Roehrs et al [11,12] designed a distributed architecture model to integrate PHRs, which was called OmniPHR, and showed the performance of the prototypes. However, most of these models focus on data storage and sharing, whereas our study focused on monitoring changes in medical documents themselves.

There are some concerns about creating blockchain-based EHR systems because sensitive data—such as medical records—are protected by the General Data Protection Regulation legislation [25], which ensures that full control of data are given to data owners. Four rule-of-thumb principles that entrepreneurs and innovators can consider when designing blockchain-based apps have been proposed. Among these principles, the second states that personal data should be avoided on blockchain using data obfuscation, encryption, and aggregation techniques [26]. To comply with this principle, we used the hash value of the document using SHA-256—which is considered to be one of the most popular hashing algorithms in the world—to make it difficult to reverse the original.

Some blockchain systems ensure privacy and confidentiality using zero knowledge proofs (ZKPs) such as Zerocash [27]. ZKPs allow data to be verified without revealing the data. Although the app does not have access to the exact medical document contents, it can show whether the medical document has been changed or not. Our system therefore satisfies the concept of ZKPs because the app only identifies whether the item was changed or not, without knowledge of the item’s content.

Medical records may be often falsified to hide medical accidents [28] or to claim insurance by fraud [29]. In Korea, the Medical Service Act states that “Where any medical personnel or the founder of a medical institution makes an addition or revision to electronic medical records, he/she shall separately keep the access logs thereof, as prescribed by Ordinance of the Ministry of Health and Welfare” [30]. This act was implemented in September 2018 and the situation is likely to be similar in other countries. Most of the tertiary hospitals in Korea have well-integrated EHR systems that can manage changes in medical documents. However, primary care clinics use vendor-dependent EHR systems, which are not equipped to track changes in documents. Therefore, we chose a primary care clinic to trace changes in records.

This study has several limitations that should be noted. All logs in the system indicate that there were no changes to medical documents during the study period owing to the short duration and small sample size. In fact, there were not many documents in primary clinics to work with compared with those available at tertiary hospitals; the fewer the documents in primary clinics, the fewer the changes in documents. Nevertheless, this study demonstrated that the app functions as designed when using simulation data. Additionally, the selection of the hospitals was biased toward pediatric clinics. Finally, the SUS scores in this study were low. Because there were no changes in medical documents, the participants barely had a chance to realize the value of the app.

This study introduced an app that notifies patients of changes in medical records using blockchain technology. Blockchain helps the app to inform patients of changes in their documents without uploading the medical record itself to the network. Therefore, blockchain can help ensure the transparency of medical records and advance patient empowerment.