Correction: Estimating the Size of Key Populations in Kampala, Uganda: 3-Source Capture-Recapture Study

Population Size Estimation Methods: Searching for the Holy Grail

“…Focused on application of empirical methods, a special collection of papers in JMIR Public Health and Surveillance has been released under the title “Key Population Size Estimations” [ 11 ]. These 9 reports on empirically based PSE include innovative approaches, such as use of social media apps (Vietnam) [ 12 ], a reverse-tracking method (Namibia) [ 13 ], multiple-source capture-recapture (CRC; Uganda) [ 14 ], and successive sampling (SS)-PSE incorporating imputed visibility [ 15 ]. We provide a summary of these papers to highlight advances in the application of empirical methods and persisting gaps in information.…”

“…PSE frequently combines TGW with MSM; because the 2 populations have distinct needs, generating separate estimates will be crucial for advocating for both KPs and appropriately targeted programs and services. Although only 2 papers in this collection reported PWID size estimates [ 14 , 15 ], many countries have included or are planning to include PSE in their BBS among PWID [ 25 , 26 ; personal communication]. Estimating the size of the PWID population may be less challenging in areas with substantial drug-injecting activity, for example, along drug transportation routes and border areas.…”

“…Numerous challenges remain, including the aforementioned need for distinct PSE for TGW and female PWID. If a BBS is deemed impractical because of relatively small population sizes or lack of resources, standalone PSE exercises as described in 4 of the papers [ 12 , 14 , 19 , 20 ] presented in this special-themed issue (described earlier) may be considered to fill these gaps. We continue to seek new, innovative, or improved methods in a search for the holy grail (ie, a gold standard for finding the true population size).…”

“…Almost all the papers (Table 1) presented estimates based on some form of CRC methodology -conventional two-source (2S)-CRC; multiple-source CRC; or service, unique object, unique event, or social app multiplier methods -or used prior estimates that may have been partially based on these methods for analysis (SS-PSE). Of the 9 papers, 4 described PSE embedded within a BBS [13,[16][17][18]], 1 described using datasets from previously conducted BBS [15], and 4 presented results from standalone PSE exercises [12,14,19,20]. Apodaca and colleagues [19] conducted a 2S-CRC in 11 small Ugandan towns with peers distributing unique objects to tag female sex workers (FSW) and MSM in the first capture.…”

“…Authors highlighted the challenges in using these methodsthe wide variation in results and importance in understanding the biases in data collection, including issues with the availability and quality of HIV-service data. [14] demonstrated the feasibility of using 3-source (3S)-CRC as a standalone (ie, not done in conjunction with a survey) method for PSE in a resource-limited setting [14]. One of the benefits of 3S-CRC is the ability to account partially for sample dependencies (thereby relaxing the assumption of independence required by 2S-CRC methods) by allowing sources to be examined pairwise.…”