Table 1 Comparing the characteristics of the ATPR, NSUM and RDS
From: Describing the safety of abortion at the population level using network-based survey approaches
ATPR (applied in this study) | NSUM | RDS (applied in this study) | |
|---|---|---|---|
Results | Counts and detailed information about abortion cases | Counts of abortions cases | Detailed information about abortion cases, no counts |
Recruitment and questionnaire | Respondents in a random sample survey report anonymously on abortions occurring to close female relations 15–49, listed using a network generating question | Respondents in a random sample survey try to estimate the number of abortions which occurred in their entire network of acquaintances, whose size is estimated | A set of initial respondents (seeds) report on their own abortion and recruit other abortion seekers through their networks |
Precision of data collected | Information on a set of finite close relations: information still relatively accurate | Not possible to ask for details on each abortion | Direct information from the abortion seekers: highest precision but still non-clinical knowledge |
Assumptions | Respondent report accurately about abortions of friends (no social desirability bias) Respondents know about friends' abortions (no transmission bias) Abortion seekers have the same networks than general population (no barrier bias) Friends have the same networks than respondents (no selection bias) The network generating question may be biased towards friends with abortions | Respondent report accurately about abortions of entire network (no social desirability bias) Respondents know about abortions in entire network (no transmission bias) Abortion seekers have the same networks than general population (no barrier bias) | Abortion seekers self-disclose in an interview setting whose privacy is guaranteed by a network member (no social desirability bias) Abortion which constitute membership in the population must create connections amongst the members of the population, members must share active ties, and abortion is objectively verifiable (no transmission bias) |
Correcting for biases | Transmission, barrier and selection biases have been corrected for in an ad hoc fashion so far. The social desirability bias cannot be corrected for | A model has been developed to correct for transmission and barrier biases; the social desirability bias cannot be corrected for but is acknowledged.* | If equilibrium is reached (if the characteristics of new recruits are independent of those of the abortion seeker who recruited initially), then assumptions are met and the sample is unbiased |
Representativity of data collected | Theoretically, quasi representative sample of abortion seekers, if biases are inexistent or corrected for | Theoretically, quasi representative count of abortions, if biases are inexistent or corrected for. But estimates are very sensitive to corrections | Theoretically, representative sample of abortion seekers if characteristics of new recruits are independent of those of the seed (i.e. if equilibrium is reached) |
Internal validation | Internal validity check using known quantities | Internal validity check using known quantities | |
Level of application | Can be applied to all levels: local to national | Can be applied to all levels: local to national | Local |
Effort needed | Easy to standardize across times and locations: a module in a population survey | Easy to standardize across times and locations: a module in a population survey | Need for good local knowledge to locate seeds; time and effort intensive |