Table 3.

Summary of findings: telemedicine compared to usual care for oral anticoagulation management in adult outpatients.

Study design		Studies, n	Certainty assessment					Patients, n/N (%)		Effect		Certainty
			Risk of bias	Inconsistency	Indirectness	Imprecision	Other considerations	Telemedicine	Usual care	Relative (95% CI)	Absolute (95% CI)
Total thromboembolic events
	Randomized trials	13	Seriousa,b,c	Not seriousd	Not serious	Seriouse	None	204/9657 (2.1)	256/9566 (2.7)	0.75 (0.53-1.07)	7 fewer per 1.000 (from 13 fewer to 2 more)	Low
Major bleeding
	Randomized trials	11	Seriousa,b,c	Not serious	Not serious	Not seriousf	None	349/10,085 (3.5)	371/9877 (3.8)	0.94 (0.82-1.07)	2 fewer per 1.000 (from 7 fewer to 3 more)	Moderate
Death
	Randomized trials	12	Seriousa,b,c	Not serious	Not serious	Not seriousf	Noneg	271/9965 (2.7)	275/9729 (2.8)	0.96 (0.78-1.20)	1 fewer per 1.000 (from 6 fewer to 6 more)	Moderate
TTRh
	Randomized trials	16	Seriousa,b,c	Seriousi	Not serious	Not serious	None	9813	9796	—j	MDk 3.38 higher (1.12 higher to 5.65 higher)	Low

^aA significant number of trials were not adequately masked. However, this is due to the nature of the intervention, and we judged that it would not significantly impact objective outcomes such as death, thromboembolic and hemorrhagic events, or TTR.

^bDowngraded for unclear or inadequate randomization process.

^cDowngraded for high or unclear risk of missing outcome data.

^dAlthough I² suggested serious heterogeneity, we decided not to downgrade for inconsistency because this is completely explained by the inclusion of 1 study [18].

^eThe CI includes an important benefit but also a small harm, since it slightly crosses the null effect.

^fWe decided not to downgrade for imprecision although 95% CI includes the null effect because the intervals are very narrow and centralized in the null effect, which corroborate similarity between telemedicine and usual care.

^gFunnel plot shows an asymmetrical distribution of studies, with significantly more studies published in favor of intervention. Egger’s test resulted in a nonsignificant P value (.135) and the adjusted odds ratio (OR), including the 5 missing studies estimated by Fill and Trim method, indicated that our conclusion would not be significantly altered by a potential publication bias (OR 0.99, 95% CI 0.83-1.19). Therefore, we decided not to downgrade for publication bias.

^hTTR: time in therapeutic range.

ⁱDespite I² of 90%, all but one trial results range from a null effect to a positive effect of telemedicine on TTR. Therefore, we decided to consider it only serious.

^jNot available.

^kMD: mean difference.