Table 3.

Associations of Year, Digital Media Use,^a and Suicidal Behavior^b and Mediation Results, by Sex.^c.

	Females			Males			Test of sex difference,hP Value
	Estimate	95% CI	P Value	Estimate	95% CI	P Value
Association of year with suicidal behavior, OR (95% CI)d
2011 vs. 2009	1.09	0.96, 1.24	0.17	1.27	1.11, 1.44	<0.001*	0.43
2013 vs. 2009	1.24	1.14, 1.36	<0.001*	1.16	1.01, 1.33	0.04*	0.76
2015 vs. 2009	1.41	1.20, 1.66	<0.001*	1.23	1.04, 1.45	0.02*	0.39
2017 vs. 2009	1.23	0.98, 1.54	0.07	1.08	0.88, 1.32	0.46	0.48
Association of year with digital media use, b (95% CI)e
2011 vs. 2009	0.36	0.27, 0.46	<0.001*	0.29	0.20, 0.40	<0.001*	0.16
2013 vs. 2009	0.89	0.72, 1.06	<0.001*	0.54	0.37, 0.71	<0.001*	0.01*
2015 vs. 2009	0.88	0.68, 1.07	<0.001*	0.47	0.34, 0.59	<0.001*	0.005*
2017 vs. 2009	0.89	0.72, 1.06	<0.001*	0.59	0.45, 0.73	<0.001*	0.01*
Association of suicidal behavior with digital media use, OR (95% CI)f
Pooled, 2009–2017	1.12	1.10, 1.14	<0.001*	1.11	1.09, 1.14	<0.001*	0.91
Cross-year difference in suicidal behavior mediated by digital media use, % (95% CI)g
2011 vs. 2009	0.61	0.44, 0.79	<0.001*	0.33	0.21, 0.50	<0.001*	0.14
2013 vs. 2009	1.56	1.24, 1.90	<0.001*	0.67	0.36, 0.99	<0.001*	0.03*
2015 vs. 2009	1.70	1.38, 2.02	<0.001*	0.63	0.34, 0.91	<0.001*	0.02*
2017 vs. 2009	1.76	1.43, 2.11	<0.001*	0.38	0.26, 0.51	<0.001*	0.02*

Abbreviations: OR, odds ratio; 95% CI, 95% confidence interval; b, unstandardized regression coefficient.

^aDigital media use: “On an average school day, how many hours do you play video or computer games or use a computer for something that is not school work?“ - None = 0, <1 h/day = 1, 1 h/day = 2, 2 h/day = 3, 3 h/day = 4, 4 h/day = 5, and ≥5 h/day = 6 (range: 0–6).

^bPast 12-month report of either suicidal thought, plan, or attempt (yes/no).

^cResults from path analysis models adjusted for obesity, school safety concern, and race/ethnicity tested separately in females (N = 36,726) and males (N = 35,877).

^dLogistic regression-based parameter estimate for bivariate association path for model testing total association of year with suicidal behavior (see Supplemental Fig. 1a).

^eLinear regression-based parameter estimate for bivariate association path.

^fLogistic regression-based parameter estimate for bivariate association path additionally adjusted for year from mediation model (Supplemental Fig. 1b).

^gCovariate-adjusted estimated increase in suicidal behavior prevalence since 2009 statistically mediated by concurrent increases in digital media use.

^hCalculated by chi-squared goodness of fit test.

^*Statistically significant after Benjamini-Hochberg correction for multiple testing to control false-discovery at 0.05 (based on 2-tailed corrected P value).