Manual evaluation of record linkage algorithm performance in four real-world datasets.

BACKGROUND: Patient data is fragmented across multiple repositories, yielding suboptimal and costly care. Record linkage algorithms are widely accepted solutions for improving completeness of patient records. However, studies often fail to fully describe their linkage techniques. Further, while many frameworks evaluate record linkage methods, few focus on producing gold standard datasets. This highlights a need to assess these frameworks and their real-world performance. OBJECTIVE: We use real-world datasets and expand upon previous frameworks to evaluate a consistent approach to the manual review of gold standard datasets and measure its impact on algorithm performance. METHODS: We applied the framework, which includes elements for data description, reviewer training and adjudication, and software and reviewer descriptions, to four datasets. Record-pairs were formed and 15,000 records were randomly sampled from these pairs. After training, two reviewers determined match status for each record-pair. If reviewers disagreed, a third reviewer was used for final adjudication. RESULTS: Between the four datasets, the percent discordant rate ranged from 1.8-13.6%. While reviewers' discordance rate typically ranged between 1% and 5%, one exhibited a 59% discordance rate, showing the importance of the third reviewer. The original analysis was compared to three sensitivity analyses. The original analysis most often exhibited the highest predictive values compared to the sensitivity analyses. CONCLUSION: Reviewers vary in their assessment of a gold standard, which can lead to variances in estimates for matching performance. Our analysis demonstrates how a multi-reviewer process can be applied to create gold standards, identify reviewer discrepancies, and evaluate algorithm performance.

Concentrations of selenium and mercury in eared grebes (Podiceps nigricollis) from Utah's Great Salt Lake, USA.

We examined selenium and mercury concentrations in eared grebes (Podiceps nigricollis) that spent the fall of 2006 on the Great Salt Lake (UT, U.S.A.), where their diet consisted mainly of brine shrimp (Artemia franciscana). Selenium concentrations in livers varied based on when the grebes were collected (lower in September [mean +/- standard error, 9.4 +/- 0.7 microg/g dry wt] than in November [14.5 +/- 1.4 microg/g]), on where the birds were collected on the Great Salt Lake (Antelope Island, 8.6 +/- 0.5 +/- microg/g; Stansbury Island, 15.2 +/- 1.4 microg/g), and on the grebe's age (juveniles, 8.5 +/- 1.5 mcirog/g; adults, 15.8 +/- 1.3 microg/g), but not by sex. Selenium concentrations in blood differed only by collection site (Antelope Island, 16.8 +/- 2.3 microg/g; Stansbury Island, 25.4 +/- 3.0 microg/g). Mercury concentration in the blood of grebes varied by when the grebes were collected (September, 5.6 +/- 0.5 microg/g; November, 8.4 +/- 1.2 microg/g), where the birds were collected (Antelope Island, 4.3 +/- 0.5 microg/g; Stansbury Island, 10.1 +/- 2.6 microg/g), and the grebe's age (juveniles, 5.5 +/- 0.8 microg/g; adults, 8.4 +/- 1.0 microg/g), but not by sex. Selenium concentrations in blood were correlated with selenium concentrations in the liver and with mercury concentrations in both blood and liver. Body mass of grebes increased dramatically from September (381 +/- 14 g wet wt) to November (591 +/- 11 g). Body, liver, and spleen mass either were not correlated with selenium or mercury concentrations or the relationship was positive. These results suggest that high mercury and selenium levels were not preventing grebes from increasing or maintaining mass.

Selenium and mercury concentrations in California gulls breeding on the Great Salt Lake, Utah, USA.

We examined selenium (Se) and mercury (Hg) concentrations in adult California gulls (Larus californicus) nesting on the Great Salt Lake, Utah, USA, during 2006 and 2007. During 2006, the mean Se concentration (+/- standard error) was 18.1 +/- 1.5 microg/g in blood on a dry-weight basis and 8.1 +/- 0.4 microg/g in liver. During 2007, Se concentrations were 15.7 +/- 1.5 microg/g in blood and 8.3 +/- 0.4 microg/g in liver; Hg concentrations were 2.4 +/- 0.3 microg/g in blood and 4.1 +/- 0.5 microg/g in liver. Gulls collected from a freshwater colony located within the watershed of the Great Salt Lake had similar levels of Se in the blood and liver as gulls collected on the Great Salt Lake but lower Hg concentrations. Body mass of adult gulls was not correlated with Se or Hg concentrations in their blood or liver. Selenium concentration in California gull eggs collected during 2006 was 3.0 +/- 0.10 microg/g. Of 72 eggs randomly collected from Great Salt Lake colonies, only one was infertile, and none of the embryos exhibited signs of malposition or deformities. We examined 100 newly hatched California gull chicks from Great Salt Lake colonies for teratogenesis; all chicks appeared normal. Hence, the elevated Se and Hg concentrations in adult gulls nesting on the Great Salt Lake did not appear to impair gulls' health or reproductive ability.