Data Use Community, August 2022
Last updated
The Data Use Community (DUC) is an open global community passionate about improving health and healthcare data sharing and use. It is a forum of virtual meetings and an online discussion board for sharing and learning from peers and country experiences. On August 24, 2022, Frank Agah, Software Engineer with SEJEN-Côte d’Ivoire and Nicolas de Kerorguen, Contractor for Centers for Disease Control and Prevention (CDC)/Center for Global Health (CGH)/Division of Global HIV and TB (DGHT)/Data Management, & Statistics Branch (HIDMSB)/Health Informatics Team, shared Côte d’Ivoire’s experience with using biometrics as a way to implement unique patient identification for patients receiving HIV antiretroviral therapy (ART). Below is a summary written by the DUC Secretariat of what was understood at the time of sharing.
Background
Duplication of patient records can occur when there are fragmented processes in collecting patient information such as when systems are not in place to link records across multiple facilities. This can lead to challenges when a patient presents for treatment, as the most up-to-date patient record might not be available. Depending on the health records infrastructure, different identifiers and information are collected at a facility upon patient registration. An identifier labeled as unique patient identification (UPID) can be used to match patient records and to address this concern with duplication.
Wanting to address the issue of duplicate patient records in the context of not having a UPID or existing national identifiers that could serve as UPIDs, Côte d’Ivoire looked into implementing a UPID system that incorporates fingerprint biometrics for patients receiving HIV antiretroviral therapy (ART) to an already existing enrollment process.
With support from the United States President’s Emergency Plan for AIDS Relief (PEPFAR), Côte d’Ivoire set up activities to work on this initiative by subcontracting with the organization SEJEN starting in 2019. A technical working group consisting of various contractors and partners was also established to support the management of these activities.
The goals of the UPID implementation work include:
Creating UPID for ART patients to support the registration and identification process;
De-duplicating patients seen at several ART facilities to allow for more accurate active data, optimize drug dispensing efforts, and improve patient services; and
Matching records across disparate databases to extend the use case beyond ART patients to improve delivery of health services.
Technical Approach
The UPID is installed at the facility level and connects with the central database via the national electronic medical record (EMR). The diagram below illustrates where this fits within the overall Health Informatics System (HIS) landscape for HIV (Diagram 1).
Diagram 1: HIV Health Information System Landscape*
At an ART facility, a patient will have his/her fingerprint captured. A desktop application is used to extract a fingerprint template file and then communicates with the central system (UPID server housed on Amazon Web Services), which contains secured encounter and template databases used for registration and identification. The fingerprint template file is not a fingerprint image, but rather another type of file that is used based on the fingerprint device and software (SecuGen) and International Organization of Standardization (ISO) format. After this, the desktop application then communicates with the national EMR, SIGDEP-2, using the UPID. The goal is to have SIGDEP-2 communicate with a Master Patient Index (MPI), which is currently not established yet. The diagram below illustrates the UPID Architecture (Diagram 2).
Diagram 2: UPID Architecture
If a patient does not have a fingerprint that can be captured, then the previously established method of collecting demographic information for patient registration and identification will be used.
The technical components and processes used to collect fingerprint biometrics as UPID are highlighted in the table below (Table 1).
Table 1: Technology Components
Impact & Challenges
Since its pilot in 2020, the UPID system has been scaled from 10 ART facilities to 100 ART facilities (out of 400+ that have EMRs) in 2022. Despite these impacts, in 2021, it was observed that there were false positives being generated due to errors with the facility code, which hid the problem. To further investigate the reasons for these results, several actions were taken including: reviewing codes and configurations, analyzing some of the known false positives and true positive duplicates, developing scripts for additional analysis, reviewing logs and literature, and discussing findings with another country. Based on the investigation, it was decided to merge EMRs from the different IP supported facilities and data elements (UPID; Facility code; Facility name; ART Patient code; Sex; Age; UPID creation date). Overall, it was found that there were mostly false positives with few duplicates. And currently, UPID goals are not met. Moreover, it was found that there were false negatives due to UPIDs being generated for existing patients when the system would not recognize the fingerprint.
In addition to the uncovered false positives and false negatives, the following challenges were noted:
Limited literature available on fingerprint biometrics and troubleshooting false positives;
Resistance and hesitancy by service providers at ART facilities in implementing UPID with biometrics; and
Issues with the process of the physical act of capturing fingerprints at the facilities.
Lessons Learned
Investigations on the observed false positives and duplications concluded the primary issue to be stored templates of poor quality; the secondary issue to be the low threshold for matching, and the tertiary issue to be how the algorithm generated new UPID when fingerprints were not matched or not of good quality. Additionally, there was flexibility in troubleshooting techniques for ART facilities in standard operating procedure (SOP).
As a result of the issues with the observed false positives and false negatives, the following were identified:
Need to conduct trainings on the processes and topics outlined in the SOP before implementation: 1) How to capture fingerprints including covering the sequence of fingers, 2) How to position and pressure of the fingers on the device, 3) How to clean the optical reader, and 4) Procedures for finger cleaning, moisturizing and drying.
Work to fine tune parameters and assess results, including incorporating datasets that will serve as references and automated tests, before or early in the implementation. This can help provide a sense of data to expect.
Interact with partners, colleagues, or other countries to gain insights into troubleshooting techniques to use when issues arise. This can serve as an alternative if limited literature is published.
While patients may not be hesitant with biometrics implementation, facilities may be reluctant due to the new processes. Therefore, it would be helpful to have open communication, training, and reminders before and during implementation to ensure everyone is aware of the overall goals.
Looking Ahead
Due to the challenges encountered with false positives and false negatives, the next steps for implementing UPID with biometrics in Côte d’Ivoire focuses on the discussion and implementation of the following identified potential solutions:
Increasing the fingerprint capture threshold to 80;
Having a visualization of the fingerprint image during capture;
Explore whether replacing a finger a few times on the reader until complete will help improve accuracy;
Updating the SOP and conducting training;
Increasing the matching threshold to seven; and
Retaking fingerprints of patients who are returning to a facility.
Finally, there are discussions on establishing an MPI capable of communicating with SIGDEP-2 as part of the UPID architecture.
For more information on the experiences in Côte d’Ivoire, please visit the DUC presentation here.
References
DUC Meeting August 24, 2022: Presentation by Frank Agah, Software Engineer with SEJEN-Côte d’Ivoire and Nicolas de Kerorguen, Contractor for Centers for Disease Control and Prevention (CDC)/Center for Global Health (CGH)/Division of Global HIV and TB (DGHT)/Data Management, & Statistics Branch (HIDMSB)/Health Informatics Team
Resources from Presenters
