Background. Pooled testing has been implemented on a limited scale, mainly for screening and surveillance in populations with a low prevalence of COVID-19 to save on limited resources.
Objective. To determine the diagnostic accuracy of pooled compared with individual RT-PCR testing for SARS-CoV-2 in individuals suspected of COVID-19.
Methods. We searched websites of living CPGs on COVID-19 (Australian COVID-19, COVID NMA, CEBM Oxford), Philippine DOH HTA, databases (PubMed, CENTRAL, medRXIV/bioRXIV), and Clinicaltrials.gov for studies that used pooled testing on individuals suspected of COVID-19. When appropriate, we pooled data for sensitivity and specificity and obtained the range and median of other data, such as positive predictive value and resource savings. We did a priori subgroup analysis for pool size, presence or absence of symptoms and use case, type of specimen, cutoff for positivity, type of kit, and post hoc subgroup analysis for method of pooling and timing of processing.
Results. We included 21 studies: 6 diagnostic accuracy studies, and 15 clinical validation studies. Studies had varying populations, index test kit and performance characteristics, positivity rate (0.02 to 15%), and pool size (5 to 16). There was moderate pooled sensitivity, 81% (95% CI 72, 88; I2=73.6%; 6 studies, 776 pools) and high pooled specificity, 99% (95% CI, 98 to 100; I2=1.84%; 5 studies, 666 pools). Positive predictive value based on 21 studies ranged from 67% to 100%. Resource savings in the number of test kits used ranged from 49 to 89%. Identified harms of pooled testing were delayed turnaround time for positive samples and laboratory errors.
Conclusion. There is moderate sensitivity and high specificity with pooled testing for the screening of individuals with suspected COVID-19. We recommend further studies to validate the utility based on community prevalence and other test variables.