Table 1.

Data classification process for official statistics on abortion incidence

Question from Flowchart	Process
Is legal abortion broadly available?	If legal abortion is not broadly available, the datum is classified as a minimum estimate of abortion incidence. If legal abortion is broadly available, then it is possible that the data are complete, and we proceed to the next question.
Does the official report acknowledge that statistics are less than 90% complete?	If the government acknowledges that an abortion datum is incomplete, counting fewer than 90% of abortions, in their official report or in their response to the questionnaire we distribute to its statistics office, then the datum is classified as a minimum. If the government claims that its statistics are complete, then, we proceed to the next question.
Is the official statistic below the estimate from a survey of women?	If the officially reported number of abortions is smaller than the number of abortions estimated in a published study based on women’s reports, then, all years of official statistics from that country are coded as minimum estimates of abortion incidence unless it can be determined that the quality of official statistics was poor in a specific period. If the official report exceeds the number estimated from women’s reports or such a study is unavailable, then we proceed to the next question.
Do a sizable portion of abortions occur outside the formal health sector?	A datum may be classified as a minimum estimate in light of evidence that a sizable portion of abortions occur outside the formal health sector. For example, even if abortions are legally available in the public sector, medical abortions may occur in the private sector that are not counted in official statistics. If not, we proceed to the next question.
After a review of outlying data for implausible levels or trends, can it be ascertained that a country’s data is incomplete?	Outlying data will be reviewed by the study team and the technical advisory group. If information on levels or trends is ascertained to be implausible, such data will be classified as a minimum estimate of abortion incidence. Otherwise the datum will be treated as a point estimate with an error term.
Does the datum include spontaneous abortions?	If the datum includes spontaneous abortions in addition to induced abortions, then, the model revises this number downward based on the formula, no. of pregnancies = 1.2 no. of live births + 1.1 no. of induced abortions, and this datum is treated as a minimum. This downwardly-revised datum is treated as a minimum based on the expectation that this adjustment subtracts abortions as well as miscarriages. Additionally, the unadjusted statistic is treated as a maximum. I.e., where b represents the number of births and m represents the number of abortions including miscarriages, the abortion rate lies along the interval [(b - 1.1 m) ÷ 1.2, m]. Note that this issue may also be relevant to a datum which is classified as a minimum during a previous step in this process; in such cases, the minimum is also adjusted as per this equation.