Table 4. Logistic Regressions Testing Associations Between Presence of Stressors and Any Mental Health Reference in Songs Expressing Negative Emotion^a.

Stressors present	Total No. of songs	Mental health reference, No. (%) (n = 94)		Unadjustedb		Firth adjustedc
		Present	Absent	OR (95% CI)	P value	OR (95% CI)	P value
Social rival (foe)	59	32 (54)	27 (46)	0.3 (0.1-1.2)	.09	0.4 (0.1-1.3)	.11
Environmental conditions	42	32 (76)	10 (24)	11.5 (2.5-51.7)	.001	8.1 (2.1-32.0)	.003
Love life	38	27 (71)	11 (29)	6.2 (1.4-26.8)	.01	4.8 (1.3-18.1)	.02
Work life	18	15 (83)	3 (17)	2.3 (0.5-11.9)	.30	1.9 (0.4-8.5)	.40
Family life	11	11 (100)	0	1.4 × 109 (NC)	.99	23.7 (0.9-602.8)	.06
Authority	10	8 (80)	2 (20)	1.1 (0.1-8.7)	.94	1.1 (0.2-7.4)	.91
Social ally (friend)	7	5 (71)	2 (29)	5.0 (0.4-56.5)	.20	3.9 (0.4-35.6)	.23
Faith	5	5 (100)	0	1.9 × 109 (NC)	.99	21.0 (0.7-683.0)	.09
Societal issues	5	5 (100)	0	2.2 × 108 (NC)	.99	2.2 (0.1-75.8)	.66
Financial strain	1	1 (100)	0	1.4 × 109 (NC)	1.00	27 (0.1-1.5)	.56

Abbreviations: NC, not calculated; OR, odds ratio.

^aAnalysis based on the 94 songs within the total sample that contained cues of negative emotion. Supplemental analyses were also conducted adjusting for year of song as a covariate. No changes to the reported pattern of associations were found when controlling for year. Lack of variability in family life, faith, societal issues, and financial strain (no presence of these stressors in songs without mental health references) generated exponential unadjusted ORs with 95% CIs that could not be calculated (separation phenomenon). These issues were addressed with the Firth bias adjustment.

^bRaw model statistics: χ² = 41.90; P < .001; −2-log likelihood = 84.12; Nagelkerke R² = 0.49.

^cFirth model statistics: χ² = 14.84; P = .14; −2-log likelihood = 86.64; Nagelkerke R² = 0.39.