DOES UNCONSCIOUS SOCIOECONOMIC BIAS INFLUENCE TELE-EVALUATION OF OBSTRUCTIVE SLEEP APNEA? A TELE-EQUITY EXPLORATORY ANALYSIS

ABSTRACT

Introduction: Telemedicine, once of limited scope, has become common and widespread due to the present and ongoing SARSCoV- 2 pandemic. Center to home delivery, the most common model, allows for convenient and efficient care. Concurrent with this groundshift, there is increasing attention to disparities in medical services, and how these disparities may impact patient outcomes. Telemedicine could be used to help bridge barriers to timely quality care, however, patient access and longstanding institutional biases may limit the potential. Healthcare providers must actively develop systems to ensure that telemedicine is optimized for people across the income spectrum. This exploratory analysis examined how economic disparities in patients being evaluated for obstructive sleep apnea may be associated with providers clinical impressions. The objective was to study the inter-method reliability of pre-test probability of obstructive sleep apnea assessed via telemedicine and in-person evaluations, and to compare that reliability between income classes. Methods: This is a secondary analysis of a pre-pandemic interrater reliability study, conducted between March 2017 and January 2019. Our researchers completed a randomized, blinded study comparing the pre-test probability of obstructive sleep apnea between an in-person physician and a separate physician seeing the same patient via televideo conferencing. Patients referred to the University of Rochester (UR) Sleep Center were eligible for the study. Women and men 30-70 years old were invited to participate. The patients were not necessarily referred to the center for evaluation of sleep disordered breathing. Patients with dementia, hearing or visual loss, severe psychiatric or developmental illness, or not fluent in English were excluded. Patients had adequate computer literacy, access to high speed internet, and a computing device with appropriate video camera and microphone.The primary objective of the original study was to assess the interrater reliability between the in-person and telemedicine raters for pre-test probability of sleep apnea (high, moderate, or low). Providers used clinical judgement from the history and examination to determine pre-test probabilityFor this present analysis, we assessed the inter-method reliability separately for strata defined by reported annual income level low income (< $50,000), middle income ($50,000-$100,000), and high income (> $100,000). Reliability was quantified for each stratum using weighted kappa statistics given the ordinal nature of the outcome variable, pre-test probability of obstructive sleep apnea (high, moderate, or low). Weighted kappa statistics were compared between the income strata (high vs. middle, high vs. low, middle vs. low). The operant statistic assumed an approximate standard normal distribution under the null hypothesis of equal kappa values in the two income strata. The Bonferroni method was used to adjust the p-values for the three pairwise comparisons performed among the three income strata. Results: Data from 53 patients were avaiable for this analysis. 11 of these patients were in the low income group, 22 in the middle income, and 16 were in the high income group. 9 patients did not include their income bracket, and were not included in the analysis. Inter-method reliabilities, assessed using weighted kappa, were 0.83 (low income), 0.24 (middle income), and 0.66 (high income). When comparing between the strata, the kappa statistics were significantly different (p=0.005) between the low and moderate income groups. There was a trend between the high and moderate income groups that did not meet statistical significance (p=0.07). Conclusion: The intermethod reliability was substantial in the low income stratum, moderate in the high income stratum, and slight in the middle income group based on the kappa statistic. There was a significant difference in the reliability values of telemedicine versus in-person assessments between the low and middle income brackets, and there was a trend between the high and mo er te groups. Since the raters were unaware of the patients income levels, this association might suggest possible unconscious bias in evaluating for OSA. It may also suggest that beyond access to telemedicine technology, the quality of the care may also be influenced by socioeconomic factors. With telemedicine in its early stages, it is important to develop this technology that will minimize biases that could result from patients economic fortunes.