Suicide Attempts of Muslims Compared With Other Religious Groups in the US (2024)

View LargeDownload

View LargeDownload

Health Disparities Psychiatry and Behavioral Health Suicide Equity, Diversity, and Inclusion

Comment

2 Comments for this article

EXPAND ALL

June 7, 2022

Study Findings Appear to be the Result of a Stastictical/Methodological Artifact - Suppressor Effect

Osman Umarji, PhD | University of California, Irvine

I reanalyzed the data and came to a different conclusion. My analyses show that Muslims are NOT more likely to attempt suicide than other faiths. The results of the authors appear to be due to a statistical artifact, which I suggest is a suppressor effect, due to including variables that were highly correlated with each (Arab and Muslim) but not correlated with attempted suicide. I summarize my analyses below:
•I ran a t-test comparing attempted suicide for Muslims vs others and comparing Muslims vs Protestants. Neither t-test was significant (t=-1.34/t=1.66, p>.05). The proportion was 8% for Muslims (n=802) and 6.5% for others (n=1555).
•The correlation between Muslim and attempted suicide (r=.028) was NOT significant.
These analyses demonstrate Muslims did not attempt suicide at a significantly higher rate than non-Muslims. I next ran a series of logistic regressions to replicate the results. I left the variables continuous, rather than make the variables categorical like the authors. This did not change the results. I ran step-wise models to identify where the Muslim coefficient became significant.
•No logistic regression model found Muslim associated with attempted suicide until race was added with all other predictors. Models that included gender, age, income, education, religious importance, location of birth, and religion showed that being Muslim was not significantly related to attempted suicide.
•Adding race to the model led to replicating the results. Muslim was significantly associated (OR=2.09, p<.05) with an increase in attempted suicide, relative to Protestants, while controlling for ALL other variables.
•This model violates the logic of variable-centered models (e.g., logistic regression). The goal is to find the unique effect of an independent variable, while statistically controlling other variables at mean levels (or in the dummy group). We typically expect the effect size of variables to decrease as you add more variables due to their shared variance.
•However, the model that I replicated finds an increase of 69% in the odds ratio (OR) of Muslim between models with and without race. The coefficient for Muslim was not significant (OR=1.4, p>.05) when race was not included but the coefficient becomes larger and significant after adding race (OR=2.09, p<.05). This means that something statistically strange is happening in the model (e.g., confounding) causing a spurious result.
•Looking at the race variables closely uncovers the problem. Race and religion are correlated with each other, especially in Muslims. However, race (White, Black, Asian, Arab) and attempted suicide are not correlated, except in the Other race category (r=.04, p<.05).
•Looking closely, we find that ALL Arabs in the sample are Muslim. There were 107 Arabs and zero were non-Muslim. This means that including Muslim and Arab in a model presents a major statistical violation because being an Arab means you are Muslim in this sample. You cannot isolate the effects of Arab and Muslim cleanly because they overlap. Looking at the prevalence of Arabs who attempted suicide, we find the number is only 3 out of 107 (2.8%). In fact, the OR for Arab in the logistical regression model is .13, p<.01.
•This is a classic suppressor effect where an IV has a partial correlation larger than the zero-order correlation with the DV. Arab and Muslim have no real world association with attempted suicide yet statistically appear to show an association due to a methodological artifact.

The findings of the authors do not appear to be accurate.

CONFLICT OF INTEREST: None Reported

READ MORE

June 14, 2022

Author Reply

Rania Awaad, MD | Stanford University

In our Research Letter, we describe the 7.9% of U.S. Muslims, 5.1% of U.S. Protestants, 6.1% of U.S. Catholics, and 3.6% of U.S. Jewish participants report a lifetime history of suicide attempt in a community-based sample. Our findings are novel in that the few suicide studies on American Muslims are very small and rely on convenience samples. In contrast, this study was a national, community-based study that better reflects the general American Muslim population. Without adjusting for any variables we found that U.S. Muslims tended to be more likely to report lifetime suicide attempts compared to other faith groups:

/>Suicide Attempts

Religion OR (95 %CI) p value

Muslim vs. Protestant 1.604 (0.913-2.820) 0.050
Muslim vs. Catholic 1.330 (0.743 -2.378) 0.168
Muslim vs. Jewish 2.261 (1.271 -4.019) 0.003
Muslim vs. No Religion 1.028 (0.563 -1.878) 0.464

Fortunately, our data are publicly available, and investigators are welcome to replicate the analyses. Dr. Osman Umarji in his comment chose to only report selected t-tests, correlations, and logistic regressions that support his assertion that “Muslims are NOT (sic) more likely to attempt suicide than other faiths.” Dr. Umarji suggests that the findings are not “accurate” and argues that race is a classical suppressor variable. Even if there is a statistical “suppression effect”, including race in the model improves the predictive validity of the model and may provide a more accurate representation of the relationship between religion and suicide attempt. Race is a fundamental control variable that must be considered when studying any suicide epidemiologic study. We can’t simply remove the variable because Arab Christians were underrepresented in this sample. Perhaps the role of race in improving the model suggests that racial minorities are more vulnerable to suicide behavior. We agree that these findings are important to further explore with a larger sample size and more extensive assessments of mental health suicide behavior. But to dismiss the model as “not accurate” is erroneous.

Our Research Letter is not intended to be the seminal paper that comprehensively describes the risk of suicide across American Muslims. In fact, our initial hypothesis when designing the study is that Muslims would report less suicide behavior than other groups; we were surprised by the data. The Research Letter was written to draw attention to the important finding that a significant proportion of American Muslims report a lifetime history of suicide attempts and to promote further research in this understudied population. Unfortunately, the topic of suicide and mental health is still taboo in many segments of the American Muslim community. The results of this study have evoked much discomfort and disbelief in some who may have a hard time coming to terms with the alarmingly high number of reported suicidal attempts. We encourage researchers across disciplines to further study the nature of suicide ideation and behavior as well as help develop effective and culturally appropriate interventions to prevent suicide among American Muslims.

On behalf of,
Rania Awaad, MD; Osama El-Gabalawy, MD, MS; Ebony Jackson-Shaheed, MPH,4; Belal Zia, MA; Hooman Keshavarzi, PsyD; Dalia Mogahed, MBA; Hamada Altalib, DO, MPH

CONFLICT OF INTEREST: None Reported

READ MORE

Views 29,575

View Metrics

1.

Hedegaard H, Curtin SC, Warner M. Increase in suicide mortality in the United States, 1999-2018. NCHS Data Brief. 2020;(362):1-8.Google Scholar

2.

3.

4.

Posner K, Brent D, Lucas C, et al Columbia-Suicide Severity Rating Scale (C-SSRS). Columbia University Medical Center. 2008;10.

5.

6.

Eskin M, Baydar N, El-Nayal M, et al. Associations of religiosity, attitudes towards suicide and religious coping with suicidal ideation and suicide attempts in 11 Muslim countries. Soc Sci Med. 2020;265:113390. doi:10.1016/j.socscimed.2020.113390PubMedGoogle Scholar