Impact of linkage quality on inferences drawn from analyses using data with high rates of linkage errors in rural Tanzania

Table 3 Comparison of regression model diagnostics by match score threshold

Sample	n	β	SE	χ²	p	HR (95% CI)	PPV
Gold standard	405	1.61	0.2033	62.4	<.0001	4.98 (3.34, 7.42)	–
Probabilistic linkage threshold, by match score threshold
minimum	405	1.02	0.2383	18.2	<.0001	2.76 (1.73, 4.41)	0.612
low	359	1.20	0.2579	21.7	<.0001	3.32 (2.00, 5.51)	0.649
medium	235	0.86	0.4621	3.5	0.0615	2.37 (0.96, 5.87)	0.745
high	106	0.53	1.1707	0.2	0.6501	1.70 (0.17, 16.87)	0.896

Abbreviations: n - sample size; β - primary exposure coefficient; SE - standard error; χ² - chi-square; p - p-value; HR - hazard ratio; CI - confidence interval; PPV - automated linkage algorithm’s positive predictive value
Note: All models adjusted for age, sex, sub-village, and distance from household to CTC

ISSN: 1471-2288