Author ' s response to reviews Title : Induced abortion , pregnancy loss and intimate partner violence in Tanzania : a population based study

1. The first paragraph combines “induced abortion and pregnancy loss.” Perhaps the authors mean “pregnancy loss from other causes.” We have changed this accordingly. 2. Methods: 1st sentence analyzes We have changed this accordingly. 3. If as many as 41% of the women have IPV there is the possibility of a ceiling effect, which could actually obscure findings We have measured intimate partner violence as ever experienced physical and/or sexual violence by an intimate partner to make it comparable to other studies that have used the same data and that have previously explored the associations with induced abortion and pregnancy loss, for example see Garcia-Moreno 2005 and Fanslow 2008. Despite a fairly high prevalence of intimate partner violence in our sample we did not detect a ceiling effect in this particular study. We do agree though that future studies should investigate whether, for example, more severe forms of physical and sexual violence would lead to even stronger effects. 4. Was any effort made to match the time periods of violence exposure and pregnancy loss? This was unfortunately not possible. In line with other studies on this issue we only looked at ever experiencing intimate partner violence and ever having an induced abortion and/or pregnancy loss. 5. Start sentence with “Twenty-three percent” not the number 23% Introduction We have changed this accordingly. 6. Reference to footnote 11: this was the only study performed in subSaharan Africa and what did they find? Need to include somewhat more detail while citing sources. This was the only study in sub-Saharan Africa using a representative survey design. We have included more details on the studies we are citing. 7. Could authors provide more information about the conditions of obtaining an abortion in Tanzania? Since they are illegal this is critical information. Report on Footnote 14, and expand. This seems pretty important to include. We have expanded this section and provided more on the different unsafe abortion practices and the potential social consequences of illegal abortions in this setting. Discretionary changes recommended: 8. No hypothesis is offered as to why IPV would lead to abortions or miscarriages. Authors should provide some framework for the findings. Also, IPV should differentially result in increased abortion v. miscarriages because with IPV comes unwanted sex and therefore unwanted pregnancy so women would actively seek a solution. It’s unclear how IPV is a catalyst for miscarriages, although there are some researchers who have explored this (see E. Lieberman et al). Further references to IPV in Tanzania and reproductive health are: McCloskey, L.A., Williams, C.M. & Larsen, U. (2005) Gender inequality and IPV among women in Moshi, Tanzania. International Family Planning perspectives, 31, 3. Williams CM, McCloskey LA, U. Larsen (2008) Sexual violence at first intercourse against women in Moshi, northern Tanzania. Population Studies, 3. We have inserted different direct and indirect pathways on how intimate partner violence and pregnancy loss and induced abortions are linked. We appreciate that you directed us to these helpful references, which we have now included. Methods 9. The compliance rate ranges from 97% to 100%, which is unusually high. What was the compensation offered to participants if any? Only 27 household refused to participate, 22 in Dar es Salaam and 7 in Mbeya, which made up less than one percent of the households sampled. Among the individual women sampled from each household 120 women refused, were not available or did not complete the interview, 72 in Dar es Salaam and 48 in Mbeya. No compensation was offered to the participants. We have added this on page 8. 10. Was any attempt made to pilot some of the terms relating to pregnancy loss in particular? How do Tanzanian women talk about this issue? Do they share the same perspective on “ever being pregnant” with women from other nationalities? Of course the data have been collected and it would be too late to find this out from the survey but if there were some qualitative notes in the records it might be worth including. In any event, a topic to address in the Discussion. Unfortunately this survey did not set pilot these terms in particular. As one of the first surveys to measure intimate partner violence and violence against women across several countries, the focus was on those measurements. Measurements for pregnancy loss and induced abortion were taken from the Demographic and Health Survey. Also, the study by Haws et al, which suggests that Tanzanian women have a different perspective on “Ever being pregnant”, “Abortion” and “Pregnancy loss” only came out recently, to challenge the questions used by the Demographic and Health Survey in Tanzania. 11. The authors state that they referred women to domestic violence services however in the area of Tanzania we worked in around the same time (Moshi) there were none. What exact services were available to the women? Did the researchers build capacity in this area by training counselors? In Mbeya there were no domestic violence services per se, but there were women’s organizations that could also deal with issues of domestic violence to which women were referred to. Also, provisions were made in case women reported suicidal thoughts. We have included this information on page 8. Results 12. P. 8 there is a better way to say “while only ever experiencing any form of IPV...” rewrite We rephrased this sentence 13. In the first paragraph of the Results the authors state that 41% of women reported IPV, 21% sexual assault, etc. It is unclear whether they mean “only” IPV excluding sexual assault. That would be preferable; otherwise these groups are overlapping and comparing non-exclusive groups raises problems. We have deleted the reference to any intimate partner violence in Table 2 since it caused too much confusion. Further Minor Essential changes recommended: 14. In Table 1 there is one confusing statistic for “currently married”: For women with an abortion the difference is minuscule 50.3 v 49.1 yet the p value is listed as p=.038, whereas for the other group of women the difference is large but non-significant 61.4 v 38.6. Was there a notation mistake here? We have run the analysis again and found the same results. While the difference in marital status seems to be larger for women who reported a pregnancy loss than women who reported an abortion, we assume that the reason why the former is insignificant while the later is significant is due to the fact that by far more women reported a pregnancy loss than an induced abortion. Furthermore , when swapping showing the percentages when investigating how many married women reported a pregnancy loss compared to not married women the difference between marital status and pregnancy loss does not appear as stark, showing that 24 percent of currently married women and 21 percent of the not currently married women report a pregnancy loss. Among women who reported an abortion the difference is larger, with 24 women reporting to be currently married versus 21 women who do not report being currently married. However, due to the request of reviewer 1 we do not report the p-value anymore and have inserted crude odds ratios in Table 3 instead. 15. What percent of women report both wanted and unwanted lost pregnancies? We have added this information in the Tables in addition to the results section. 16. Table 2 is a bit confusing to read, but the main problem is that the types of abuse are all divided for comparison when they actually overlap substantially as discussed in the descriptives. With 41% of the women having physical abuse and 49% having either physical or sexual it appears that there was significant overlap. There is no theory per se that would separate the effects of sexual assault (within an intimate relationship) from physical abuse, but perhaps the authors could offer a reason to separate them. I would recommend dropping the comparisons and sticking with one measure. We hope that by adding the different pathways on how intimate partner violence effects pregnancy loss and induced abortion in the introduction the reason for the distinction between physical and sexual violence becomes clearer. The reviewer Is of course correct in stating that there is a substantial overlap, which made us decide to not differentiate in the multivariate analysis. Also, given the confusion this table caused we have removed the variable “any intimate partner violence”. 17. In Table 3authors need to clarify the referent (REF) for at least some of the variables. Also when there are 3 values for a variable and an OR is presented which value is being compared? (For instance, “number of live born children”) We have included the referent category for variable where we have overseen it. Number of children were measured as a continuous variable since there is no clarity on whether a cut off should be made at 1, 2,3 or 6 children. 18. Since Table 1 reveals so few differences between women who have had an abortion and women who have lost a pregnancy accidentally (presumably) would it make sense to collapse the groups of women? The reasons for their pregnancy loss might be quite different, however. We have considered this as well, but since the pathways between intimate partner violence and pregnancy loss and induced abortion are different, we want to keep them separate. 19. The model in Table 3 does not appear theoretically motivated. What is the rationale for comparing these two groups on the SES variables after they’ve shown to be comparable in Table 1. We hope this is clearer now, after we outlined the potential pathways in the introduction, by providing more background to other studies which investigated this issue and by moving the crude odds ratios to Table 3 Further Discretionary changes recommended Discussion 20. The paper would be more helpful if it included some information on the access to abortion services in these regions of Tanzania. The authors indicate that many of the abortions are illegal and unsafe: is there any evidence for this, even anecdotal? Any national statistics? We have provided more information on abortions in Tanzania in the introduction (page 6 and 7). 21. While the authors note that women in Tanzania have substantial rates of abortion, the prevalence rate is much lower than rates in countries where it is legal (e.g., Ukraine @ 50%; @15% South Africa; @25% Sweden). It is correct that the rates of abortions are higher in countries where abortions are legal. Given that abortions are illegal in Tanzania, it is highly likely that the rates are biased by underreporting. Upon suggestion from Sia Msuya we have added information on other African countries in which the WHO multi-country study was conducted. We hope that this provides sufficient context to why we are stating that the numbers are substantial. 22. What do the authors mean by “The association between induced abortion and socio-economic status in particular has to be interpreted cautiously”? Poor women report both miscarriages and abortions more often, but it’s unclear what the authors mean by a cautious interpretation. Lack of transportation may be one disadvantage of women living in poverty, but it is also part of lower SES resources. We have re-phrased this paragraph to address the questions it has raised. 23. On p. 10 the authors state that miscarriages were more common among women kicked in abdomen. This piece of data was not presented in the Results, however, and the Discussion is not the place to introduce new findings. Need to provide the statistic in the Results or delete sentence. This information is not part of the results of this study and is now referenced accordingly. 24. Towards the end of the Discussion the authors state that women in Tanzania have poor contraception options. This issue is overlooked earlier in the paper. There should be some mention of this topic as one overarching rationale for seeking abortion in the Intro. Also, how are the authors certain that miscarriages are not actually intended losses. Are there home remedies for pregnancy (e.g., abortifacients) that some women might take? We have added the information on contraception into the pathways on how intimate partner violence is associated with pregnancy loss and abortion, which we have now inserted into the introduction. Also, we included the possibility of pregnancy losses being intended into the limitation section.