Table 3.

Causal forest risk differences, RATE, and variable importance.

	n	RD (95% CI)	RATE (SE)	RATE P-value	variable importance
Full population	88,818	3.3 (3.1 to 3.6)	9.4 (1.4)	6.6 · 10−14
Female	57,087	3.9 (3.6 to 4.3)	−6.5 (0.9)	1.6 · 10−12a	0.037a
Male	31,731	2.2 (1.8 to 2.5)
Older than 50 years	40,439	4.3 (3.8 to 4.7)	10.5 (1.3)	1.2 · 10−14b	0.570b
Fibromyalgia	803	10.3 (6.4 to 14.1)	3.0 (0.9)	2.1 · 10−4	0.007
COPD or other lung disease	1197	7.0 (3.9 to 10.0)	2.2 (0.9)	0.016	0.004
Diabetes	2980	6.4 (4.6 to 8.2)	3.6 (1.1)	5.4 · 10−4	0.007
Post-traumatic stress disorder	1757	5.8 (3.4 to 8.2)	2.0 (1.0)	0.036	0.013
High BMI	14,700	5.6 (4.9 to 6.4)	9.2 (1.3)	3.9 · 10−12	0.212
Depression	10,715	5.7 (4.8 to 6.6)	7.0 (1.3)	5.1 · 10−8	0.053
Chronic asthma	6335	5.4 (4.2 to 6.5)	4.2 (1.2)	1.4 · 10−4	0.020
High blood pressure	9817	5.0 (4.1 to 5.9)	4.8 (1.2)	7.3 · 10−5	0.004
Chronic or frequent headaches	3593	4.9 (3.3 to 6.6)	2.2 (1.1)	0.040	0.012
Chronic fatigue syndrome	1345	3.3 (0.7 to 5.9)	−0.03 (0.8)	0.975	0.004
Anxiety	7425	3.4 (2.4 to 4.4)	0.3 (1.1)	0.809	0.007

Risk differences (RDs) and two-sided 95% confidence intervals (CI) obtained using CF and AIPW for full-time sick leave taken 4 weeks to 9 months after the test date between SARS-CoV-2 test-positives and test-negatives for the total study population and possible PCC risk groups. The RATE along with P-values for a RATE based test of treatment effect heterogeneity are reported along each risk factor. A RATE based omnibus test of treatment effect heterogeneity is also provided. All RATEs in the table have been multiplied by 1000.

^aThe RATE P-value tests for a difference in RDs between females and males. The variable importance scores split on sex.

^bThe RATE P-value tests for heterogeneity along the continuous age. The variable importance scores split on the continuous age variable.