Background: In resource-constrained settings, challenges with unique patient identification may limit continuity of care, monitoring and evaluation, and data integrity. Biometrics offer an appealing but understudied potential solution.
Methods: We conducted a mixed-methods study to understand feasibility, acceptability, and adoption of digital fingerprinting for patient identification in a study of household TB contact investigation in Kampala, Uganda. We tested associations between demographic, clinical, and temporal characteristics and failure to capture a digital fingerprint. We used generalized estimating equations and a robust covariance estimator to account for clustering. We evaluated clustering of outcomes by household and community health worker by calculating intra-class correlation coefficients. To understand determinants of intended and actual use of fingerprinting technology, we conducted fifteen in-depth interviews with community health workers and applied a widely used conceptual framework, the Technology Acceptance Model 2.
Results: Digital fingerprints were captured in 74% of participants, with extensive clustering by household (ICC = 0.99) arising from hardware (36%) and software (60%) failures. Clinical and demographic characteristics were not significantly associated with fingerprint capture. Community health workers successfully fingerprinted all contacts in 70% of households, with modest clustering of outcomes by CHW (ICC = 0.18). Fingerprinting success at the household level declined over time (Spearman’s rho = 0.30, P < 0.001). In interviews, CHWs reported that fingerprinting non-capture events lowered their own perception of the quality of the technology, threatened their social image, and made the technology more difficult to use.
Conclusions: We found digital fingerprinting to be feasible and acceptable for indvidual identification, but problems implementing the hardware and software led to a high failure rate. Although CHWs found fingerprinting to be acceptable in principle, their intention to use the technology was tempered by perceptions that it was inconsistent and of questionable value. We emphasize the need for routine process evaluation of biometrics and other digital technologies during implementation in resource-constrained settings.