Table 2.

Levels of average diagnostic accuracy (ADA) and main variables identified.

Authors, year; country	OSCsa, n	OSC ADA (listing the correct diagnosis first)			ADA range, % (OSC) to % (OSC)		Main variables identified		ADA of additional comparator
		Values, % (95% CI)	Values, mean (SD)					Values, % (95% CI)		Values, mean (SD)
Semigran et al [14], 2015; United States	23	34 (31-37)	—b	5 (MEDoctor) to 50 (DocResponse)		Urgency ↓c Frequency ↑d Demographic data ↔e Maximum number of diagnoses provided ↔ Distributor ↔ Nurse triage protocol ↔		N/Af		N/A
Semigran et al [15], 2016; United States	23	34 (31-37)	—	5 (MEDoctor) to 50 (DocResponse)		None		72.1 (69.5-74.8)g		—
Berry et al [16], 2019; United States	5	—	—	3 (WebMD) to 16.4 (Symcat)		None		N/A		N/A
Gilbert et al [17], 2020; United States	8	—	26.1 (8.9)	18 (Symptomate) to 48 (Ada)		NHSh vignettes (based on transcripts of real calls made to NHS Direct) ↓		—		71.2 (5.6)i
Hill et al [22], 2020; Australia	36	36 (31-42)	—	12 (ePain Assist) to 61 (Symptomate)		Urgency ↑↓j Frequency ↑ AIk ↑ Demographic data ↑ Maximum number of diagnoses provided ↔ Apple vs Google ↑↓		N/A		N/A
Ceney et al [21], 2021; United Kingdom	9l	37.7 (33.6-41.7)	—	22.2 (Caidr) to 72 (Ada)		Urgency ↓ Number of questions ↑ Time to complete ↑		N/A		N/A
Gilbert et al [23], 2021; Australia	1m	65n	—	N/A		Australian-specific vignettes ↓		N/A		N/A

^aOSC: online symptom checker.

^bNot stated.

^cincreases average diagnostic accuracy.

^d↑: decreases average diagnostic accuracy.

^e↓: no substantial influence on average diagnostic accuracy.

^f↔: N/A: not applicable.

^gA total of 234 general practitioners on the Human Diagnosis Project platform.

^hNHS: National Health Service.

ⁱSeven GPs through telephone consultation.

^j↑↓: mixed impact on average diagnostic accuracy.

^kAI: artificial intelligence.

^lOut of 12 online symptom checkers.

^mAda.

ⁿ95% CI values not provided.