Decrease of suicide risk with delay of self labeling as gay
Decrease of suicide risk with delay of self labeling
By Neil E. Whitehead, PhD
September 2005
Summary: Remafedi et al. state in a study of attempted suicide among gay and bisexual youth that the risk diminished by 80% per year. I show this is a misprint and should read “diminished to [bolding added] 80% each year”
The paper under consideration is Remafedi, G, Farrow, JA, Deisher, RW, (1991) Risk factors for attempted suicide in gay and bisexual youth. Pediatrics 87:869-875. The 137 male subjects mostly lived in Minnesota and Washington states.
In it the authors state (p74):
For each year’s delay in bisexual or homosexual self-labeling, the odds of a suicide attempt diminished by 80%.
This is a dramatic statistic, but immediately suspicious. The ages involved were 14-21. If he risk is taken as 100% at age 14, it would be 20% at age 15, 4% at age 16, 0.8% at age 17 and about 0.15% at age 18. After that anyone who self labeled as gay or bisexual would essentially never commit suicide. We know from common experience that is not true. We also find it hard to see why one year would make such a dramatic difference. It seems unlikely.
The 80% figure can only be derived from Table 4 which contains the following line:
| Table 4. Summary table: Regression Analysis of Suicide Attempts | | ||||
| Variable | 95% C.I | Reg. Coeff. | Std. Error | P value | |
| Label age | 0.82 | 0.69-0.96 | -0.21 | 0.08 | 0.01 |
| | | | | | |
“Reg. Coeff.” Means the regression coefficient, and means that for each year of age, the probability of an attempt at suicide decreases by 21%. The “Odds ratio” means that for each year the probability of an attempt at suicide diminishes from 1.00 to 0.82, i.e. it diminishes 18%.
The two figures 21% and 18% are not in conflict because of the size of the standard error, making them the same with statistical error. A good round figure would be to say the probability diminishes by 20% per year or diminishes to 80%. If strongly looks like the authors wrote “by” when to be consistent with the more precise numbers in their table, they should have written “to”.
This means that the earlier self-labeling occurs, the higher the risk of suicide attempts (emphasis added). The authors state in the paper that:
….suicide attempts were not explained by experiences with discrimination, violence, loss of friendship, or current personal attitudes towards homosexuality.
This is contrary to frequently and reasonably expressed hypothesis that discrimination against homosexual people causes suicide attempts. However all numerical tests ever done, such as the one in this paper, do not support the link. There are usually other reasons for suicide attempts, and the authors state that family problems were the most frequently cited reason.
The authors stated:
….gender nonconformity and precocious psychosexual development were predictive of self-harm. Compared with peers, suicide attempters recognized homosexual attractions and told other persons at younger ages. First sexual experiences with males and females also occurred at younger ages than peers’.
Early sexual experience was predictive of suicide attempts.
Now correlation is not causation. Although there are links, we have to be cautious saying delaying self-labeling will prevent suicide, or that postponing sexual experience will prevent suicide. It may not. However from the point of view of policy, which must always err on the conservative side when matters of life-and-death like suicide attempts are involved, it would be highly prudent to
(a) Discourage early sexual experience
(b) Discourage early self-labeling
(c) Discourage activists who promote early self-labeling
Those in charge of young people are usually legally responsible for the best possible care, and would probably be legally liable if they failed to take into account the above data and a death resulted, particularly if it were brought to their attention beforehand.
