This paper is in JAMA psychiatry. It is the final and complete report of the Turnaway project, that followed women who had an abortion and those who were denied (I’m not sure how that happens in the land of Roe v Wade) an abortion.

This is part of the abstract.

This study presents data from the Turnaway Study, a prospective longitudinal study with a quasi-experimental design. Women were recruited from January 1, 2008, to December 31, 2010, from 30 abortion facilities in 21 states throughout the United States, interviewed via telephone 1 week after seeking an abortion, and then interviewed semiannually for 5 years, totaling 11 interview waves. Interviews were completed January 31, 2016. We examined the psychological trajectories of women who received abortions just under the facility’s gestational limit (near-limit group) and compared them with women who sought but were denied an abortion because they were just beyond the facility gestational limit (turnaway group, which includes the turnaway-birth and turnaway-no-birth groups). We used mixed effects linear and logistic regression analyses to assess whether psychological trajectories differed by study group.

We included 6 measures of mental health and well-being: 2 measures of depression and 2 measures of anxiety assessed using the Brief Symptom Inventory, as well as self-esteem, and life satisfaction.

Of the 956 women (mean [SD] age, 24.9 [5.8] years) in the study, at 1 week after seeking an abortion, compared with the near-limit group, women denied an abortion reported more anxiety symptoms (turnaway-births, 0.57; 95% CI, 0.01 to 1.13; turnaway-no-births, 2.29; 95% CI, 1.39 to 3.18), lower self-esteem (turnaway-births, –0.33; 95% CI, –0.56 to –0.09; turnaway-no-births, –0.40; 95% CI, –0.78 to –0.02), lower life satisfaction (turnaway-births, –0.16; 95% CI, –0.38 to 0.06; turnaway-no-births, –0.41; 95% CI, –0.77 to –0.06), and similar levels of depression (turnaway-births, 0.13; 95% CI, –0.46 to 0.72; turnaway-no-births, 0.44; 95% CI, –0.50 to 1.39).

Conclusions and Relevance In this study, compared with having an abortion, being denied an abortion may be associated with greater risk of initially experiencing adverse psychological outcomes. Psychological well-being improved over time so that both groups of women eventually converged. These findings do not support policies that restrict women’s access to abortion on the basis that abortion harms women’s mental health.

The way you get a turnaway group is you take those who want a late termination. From the methods.

These groups were women whose pregnancy was within 2 weeks under a facility’s gestational limit who presented and received an abortion (near-limit group; n?=?452), women whose pregnancy was up to 3 weeks past a facility’s gestational limit who presented and were denied an abortion (turnaway group; n?=?231), and women who received a first-trimester abortion (first-trimester group; n?=?273). The first-trimester group served to assess whether outcomes differed among women who seek abortions earlier vs later in pregnancy. The turnaway group was further divided into those who gave birth (turnaway-birth group; n?=?161) and those who miscarried or later had an abortion elsewhere (turnaway-no-birth group; n?=?70). The gestational age limits of each recruitment site varied, so there was some overlap in gestational ages between study groups.

There is a problem which is mentioned in the methods. Only a third of those eligible were recruited, and about half of those — or a sixth of the original cohort — complete the trial.

A total of 1132 of 3016 eligible participants approached (37.5%) consented to participate. A total of 956 women (84.5%) completed the baseline interview, with an average of 5% lost from wave to wave and 558 (58.4%) retained at the last interview. There was no differential loss to follow-up by whether women had a history of anxiety or depression or by study group through wave 10. By the final interview wave (wave 11), women in the turnaway-birth group (80 of 162 [49.4%]) were less likely to participate than those in the near-limit group (182 of 452 [40.3%]; P?=?.045). The last interview wave was conducted from 5 to 6 years after the women sought an abortion (mean, 5.1 years)

You then have to consider what happened to the five-sixths who did not complete the trial. Do we have a less or more healthy group of participants? Do the small numbers at the end of the trial skew the results?

For at the end there were no differences. Is this a factor of statistical loss of power? The authors think that their statistical modelling deals with this.

Our participation rate of less than 40%, our loss of 42% of our sample by the end of the 5-year period, and our differential loss by study group from wave 10 to wave 11 raise some concerns that attrition could bias our results. Although our loss to follow-up represented good participant retention of about 5% from wave to wave, we cannot rule out the possibility that women with adverse mental health outcomes may have been less likely to participate and/or to be retained. Mitigating concerns of bias include the lack of differential loss to follow-up based on mental health history, as well as our ability to control for history of mental health conditions, child abuse and neglect, and substance use. Furthermore, our use of mixed model regression protects against bias owing to loss to follow up that is predictable from previously measured covariates or outcomes. Concern about bias introduced by low study participation is further lessened by the consistent findings in our sensitivity analyses restricted to sites with more than 50% participation. The lack of change from wave 10 to wave 11 on any of our outcomes lessens our concerns that differential loss by study group at the last interview wave affected our findings.

I am less sure. The surveys I am involved with aim for a 70% participation, and peer reviewers criticise us if the participation rate is below 60%. The prospective studies Otago University runs in Christchurch and Dunedin (I am not involved in them) have followup rates — over 30 and 42 years — over 90%.

I would not make conclusions on policy from a negative. They are too easy to disprove. The abstract is written in a way that agrees with the local narrative. I think the authors will be shredded if it is widely publicized.