Evaluation of Otolaryngology Provider Bias in Portal Messaging

Patient messaging portals are widely used in clinical practice and are linked to improved patient outcomes, but they are also associated with provider burnout, a phenomenon that is understood to exacerbate implicit bias. This study evaluates provider bias in responses to patient portal messages within an academic otolaryngology practice.

A total of 167,866 patient messages and provider responses were extracted from May 2018 to April 2023 within the Otolaryngology department, along with patient demographic and surgical data. Provider response length and response time were calculated as surrogates for implicit bias and compared across patient populations by demographics and economic class (estimated by zip code and using insurance type), with and without adjusting for surgical and message factors.

Of the total messages, 32,447 initial patient messages that received a provider reply were included in the analysis. In unadjusted comparisons, response time and length did not differ significantly by gender, ethnicity, or income. Medicare patients received 3% shorter replies (p = 0.009). White patients had a faster response time than those with unknown race (420.0±5.8 vs 469.0±37.7 minutes, p = 0.01). In adjusted models, provider responses varied significantly by message category and patient status. Compared with postoperative questions (reference), paperwork/insurance messages had 41.5% shorter response delays, imaging 26.9% shorter, and appointment/scheduling 21.2% shorter (all p<0.001). Nonsurgical patients had 12.7% longer adjusted response delays (p<0.001).

When controlling for relevant patient and message factors, surrogates of provider bias in portal messages showed limited associations and no relationship with patient demographics.