Enhancing the ATra Black Box Matching Algorithm: Use of All Names for Deduplication Across Jurisdictions

Hamp, Auntre; Karn, Helen E.; Kwon, Frances Y; Rhodes, Anne; Carrier, James; Bhattacharjee, Reshma; Flynn, Colin; Hsu, Trevor; McNeice,, John; Anderson, Bridget J.; Chicoine, Joyce; Fridge, Jessica; King, Justice; Lum, Garret; Mishra, Tej; Kang, Alisa; Smart, J C

doi:10.1177/00333549211066171

Cited by 3 publications

(1 citation statement)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…One of the more exciting recent developments in empirical social science research is the increasing availability of large administrative databases and the ability to link across them to generate new insights. Indeed linked datasets have allowed researchers working in descriptive, predictive, and causal modalities to generate systematic inferences about large numbers of individuals [1][2][3][4][5][6][7][8][9]. Unfortunately, most administrative datasets are not designed to be linked to others and thus have no common and reliable unique identifiers.…”

Section: Introductionmentioning

confidence: 99%

No ground truth? No problem: Improving administrative data linking using active learning and a little bit of guile

et al. 2023

View full text Add to dashboard Cite

While linking records across large administrative datasets [“big data”] has the potential to revolutionize empirical social science research, many administrative data files do not have common identifiers and are thus not designed to be linked to others. To address this problem, researchers have developed probabilistic record linkage algorithms which use statistical patterns in identifying characteristics to perform linking tasks. Naturally, the accuracy of a candidate linking algorithm can be substantially improved when an algorithm has access to “ground-truth” examples—matches which can be validated using institutional knowledge or auxiliary data. Unfortunately, the cost of obtaining these examples is typically high, often requiring a researcher to manually review pairs of records in order to make an informed judgement about whether they are a match. When a pool of ground-truth information is unavailable, researchers can use “active learning” algorithms for linking, which ask the user to provide ground-truth information for select candidate pairs. In this paper, we investigate the value of providing ground-truth examples via active learning for linking performance. We confirm popular intuition that data linking can be dramatically improved with the availability of ground truth examples. But critically, in many real-world applications, only a relatively small number of tactically-selected ground-truth examples are needed to obtain most of the achievable gains. With a modest investment in ground truth, researchers can approximate the performance of a supervised learning algorithm that has access to a large database of ground truth examples using a readily available off-the-shelf tool.

show abstract

Section: Introductionmentioning

confidence: 99%

No ground truth? No problem: Improving administrative data linking using active learning and a little bit of guile

et al. 2023

View full text Add to dashboard Cite

show abstract

Data to Care Pilot Program in Chicago: Experience, Outcomes, and Direction for the Future

Schmitt,

Devlin,

Mason

et al. 2024

Journal of Public Health Management &Amp; Practice

View full text Add to dashboard Cite

Context: Data to Care (D2C) involves sharing HIV surveillance data between health care facilities and health departments to improve continuity of care for people living with HIV (PLWH). The Chicago Department of Public Health (CDPH) initiated a D2C pilot program at the University of Chicago Medicine (UCM) from June 2016 to September 2019. Objectives: To describe the proportion of patients reported by UCM as not in care who were able to be matched to CDPH enhanced HIV/AIDS Reporting System (eHARS) and to report the individual-level factors associated with matching and viral suppression. Design: Retrospective program evaluation. Setting: UCM, an academic health care center that provides HIV care to adults via a Ryan White clinic on the south side of Chicago. Participants: Adult PLWH who had received care at UCM but did not have current documented HIV care visit(s). Main Outcome Measure: Proportion of matched patients; factors associated with matching and viral suppression. Results: Overall, 72.4% (n = 813/1123) of patients reported by UCM were matched by CDPH to eHARS. Individuals aged 40 to 49 years (odds ratio [OR] = 1.99; 95% confidence interval [CI], 1.10-3.62), 50 to 59 years (OR = 2.47; 95% CI, 1.37-4.47), and 60 years or older (OR = 6.18; 95% CI, 3.18-12.32) were more likely to match in eHARS. People who lived outside of Chicago (OR = 0.09; 95% CI, 0.05-0.15) or with unknown zip codes (OR = 0.08; 95% CI, 0.05-0.12) were less likely to match. Men who have sex with men and persons older than 50 years were more likely to be virally suppressed. Conclusions: D2C is an evidence-based strategy for reengagement of PLWH; however, program implementation relies on successful data matching. We found that a large proportion of patients from UCM were not matched, particularly those who were younger or lived outside of Chicago. Additional research is needed to understand ways to improve data matching to facilitate reengagement in HIV care.

show abstract

Cross-Jurisdictional Data Sharing: Utilization of the ATra Black Box for Deduplicating Cases in the National HIV Surveillance System (Preprint)

Jarris,

Kassaye,

Hamp

et al. 2024

Preprint

View full text Add to dashboard Cite

BACKGROUND As the use of HIV surveillance data becomes a more integral part of national care indicators and ongoing patient engagement strategies, accurate, complete, and timely data are crucial. The ATraTM Black Box is an electronic privacy-assuring system developed by Georgetown University (GU) which allows for the secure and streamlined exchange of data between public health jurisdictions. OBJECTIVE In 2018, GU was awarded a 5-year grant from CDC, PS18-1805 (1805), to de-duplicate persons across HIV surveillance jurisdictions. This paper outlines the processes GU undertook to engage public health jurisdictions and provides results of the Black Box matching sessions through the end of 2023. METHODS GU recruited jurisdictions for participation in the project and developed communication plans and documentation to assist jurisdictions with participating in quarterly matching sessions of the Black Box. GU surveyed jurisdictions to determine technical assistance needs and satisfaction with the project and held virtual and in-person meetings. RESULTS As of October 2023, GU had enrolled 40 public health jurisdictions into 1805 with signed data sharing agreements, and 75% of persons living with diagnosed HIV (PLWDH) in the US reside in these jurisdictions. Through the end of 2023, GU has conducted 21 quarterly matching sessions of the Black Box, processing over 2.1M records in the November 2023 session. CONCLUSIONS The implementation of the Black Box for sharing HIV surveillance data across jurisdictions has resulted in decreased staff time needed to update information on PLWDH. This project has improved the quality of key HIV surveillance data that are needed to measure progress on key HIV indicators at the local and national level.

show abstract

Enhancing the ATra Black Box Matching Algorithm: Use of All Names for Deduplication Across Jurisdictions

Cited by 3 publications

References 10 publications

No ground truth? No problem: Improving administrative data linking using active learning and a little bit of guile

No ground truth? No problem: Improving administrative data linking using active learning and a little bit of guile

Data to Care Pilot Program in Chicago: Experience, Outcomes, and Direction for the Future

Cross-Jurisdictional Data Sharing: Utilization of the ATra Black Box for Deduplicating Cases in the National HIV Surveillance System (Preprint)

Contact Info

Product

Resources

About