Scholarlink

Background <p>Duplicate patient records pose a significant challenge to healthcare registries and electronic medical record (EMR) systems in Uganda, primarily due to the absence of a national unique patient identifier. These duplicates lead to fragmented patient care, misallocation of resources, and inaccuracies in data reporting, which hinder effective monitoring of disease progression, disrupt continuity of care, and complicate efforts to track patient outcomes. This study evaluated the performance of three classification algorithms in identifying duplicate records of people living with HIV (PLHIV) and examined which demographic variables support reliable patient matching.</p> Methods <p>The study used a six-step deduplication process involving dataset extraction, preprocessing, indexing, comparison, classification, and performance evaluation. Records of PLHIV who were active in care between June and November 2022 were extracted from the UgandaEMR system - an EMR installed at 15 public health facilities in six districts in the Rwenzori Region. The dataset included demographic variables, i.e. given name, middle name, surname, sex, age, date of birth, address, and phone number. Three classification algorithms were evaluated: a threshold-based algorithm, a weighted average score-based method in which variable weights were assigned based on discriminative importance and refined iteratively, and a decision tree trained on rule-based labels and evaluated using a synthetic reference dataset of 1,000 records (500 original and 500 duplicates) due to the absence of a fully labeled real-world dataset. The algorithms’ performance was assessed using sensitivity, specificity, and F-score metrics.</p> Results <p>A total of 44,717 records for PLHIV active in care in the Rwenzori region from June to November 2022 were extracted. At the pair level (156,700 candidate pairs), the weighted average score-based algorithm identified 447 matches and 2,996 potential matches, while the threshold-based algorithm identified 118 matches and 8,560 potential matches. At the record level, the weighted average score-based algorithm flagged 2,459 records (5.8%) as duplicates, while the threshold-based algorithm flagged 223 records (0.5%) as duplicates. On the synthetic reference dataset, the weighted average score-based algorithm achieved the highest performance: sensitivity (99.0%), specificity (98.8%), and F-score (98.9%), followed by the decision tree and threshold-based methods.</p> Conclusions <p>The weighted average score-based algorithm achieved the best performance. Findings highlight that a combination of a few demographic variables can be employed to differentiate PLHIV. However, in the absence of a fully adjudicated gold standard, these performance metrics should be interpreted as algorithmic benchmarks under controlled conditions rather than estimates of real-world effectiveness. Improving duplicate record detection at scale will require validation on a larger, independently labeled dataset representative of the PLHIV population in Uganda.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Patient deduplication in Uganda’s electronic medical records system: a comparison of three classification algorithms

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年6月2日

查看全文
  • Alex Mirugwe,
  • Arthur G. Fitzmaurice,
  • Alice Namale,
  • Evelyn Akello,
  • Simon Muhumuza,
  • Milton Kaye,
  • Samuel Lubwama,
  • Jonathan Mpango,
  • Paul Katongole,
  • Solomon Ssevvume,
  • Paul Mbaka,
  • Clare Ashaba,
  • Enos Sande,
  • Kenneth Musenge

摘要

Background

Duplicate patient records pose a significant challenge to healthcare registries and electronic medical record (EMR) systems in Uganda, primarily due to the absence of a national unique patient identifier. These duplicates lead to fragmented patient care, misallocation of resources, and inaccuracies in data reporting, which hinder effective monitoring of disease progression, disrupt continuity of care, and complicate efforts to track patient outcomes. This study evaluated the performance of three classification algorithms in identifying duplicate records of people living with HIV (PLHIV) and examined which demographic variables support reliable patient matching.

Methods

The study used a six-step deduplication process involving dataset extraction, preprocessing, indexing, comparison, classification, and performance evaluation. Records of PLHIV who were active in care between June and November 2022 were extracted from the UgandaEMR system - an EMR installed at 15 public health facilities in six districts in the Rwenzori Region. The dataset included demographic variables, i.e. given name, middle name, surname, sex, age, date of birth, address, and phone number. Three classification algorithms were evaluated: a threshold-based algorithm, a weighted average score-based method in which variable weights were assigned based on discriminative importance and refined iteratively, and a decision tree trained on rule-based labels and evaluated using a synthetic reference dataset of 1,000 records (500 original and 500 duplicates) due to the absence of a fully labeled real-world dataset. The algorithms’ performance was assessed using sensitivity, specificity, and F-score metrics.

Results

A total of 44,717 records for PLHIV active in care in the Rwenzori region from June to November 2022 were extracted. At the pair level (156,700 candidate pairs), the weighted average score-based algorithm identified 447 matches and 2,996 potential matches, while the threshold-based algorithm identified 118 matches and 8,560 potential matches. At the record level, the weighted average score-based algorithm flagged 2,459 records (5.8%) as duplicates, while the threshold-based algorithm flagged 223 records (0.5%) as duplicates. On the synthetic reference dataset, the weighted average score-based algorithm achieved the highest performance: sensitivity (99.0%), specificity (98.8%), and F-score (98.9%), followed by the decision tree and threshold-based methods.

Conclusions

The weighted average score-based algorithm achieved the best performance. Findings highlight that a combination of a few demographic variables can be employed to differentiate PLHIV. However, in the absence of a fully adjudicated gold standard, these performance metrics should be interpreted as algorithmic benchmarks under controlled conditions rather than estimates of real-world effectiveness. Improving duplicate record detection at scale will require validation on a larger, independently labeled dataset representative of the PLHIV population in Uganda.