Table. Correlation Between Recent Ratings and Past Ratings, Controlled for Sampling Errors^a.

Platforms and physician sample	Physicians, No.	Reviews, No.	Reviews per physician, mean No.	Estimated correlation of recent and past ratings, r
All platforms
All physicians	310 849	7 262 161	23.36	0.55
Physicians with ≥10 reviews	190 135	6 577 268	34.60	0.56
Physicians with ≥20 reviews	113 301	5 422 074	47.86	0.58
Physicians with ≥50 reviews	30 197	2 800 649	92.75	0.61
HealthGrades	168 130	1 748 293	10.40	0.54
Vitals	209 710	4 143 315	19.76	0.44

^a The sample is composed of physicians who have reviews in both the recent 3 years (August 1, 2018, to August 1, 2021) and prior (before August 1, 2018). To further examine the correlation between recent and past ratings, we used the linear mixed random effect mode to control for error from small sample size. A linear random mixed effect model is used to estimate the correlation between the physician random effects in recent ratings and past ratings. A small value of estimated correlation indicates the large discrepancy between the recent and past reviews. Because patients might submit reports to more than 1 platform, which could bias the correlation downward, we reexamined whether documented effects across all platforms we repeat the above analysis separately on the 2 major platforms, HealthGrades and Vitals.