Growing Disparities in Patient-Provider Messaging: Trend Analysis Before and After Supportive Policy

Abstract

Background: Public policy introduced since 2011 has supported provider adoption of electronic medical records (EMRs) and patient-provider messaging, primarily through financial incentives. It is unclear how disparities in patients' use of incentivized electronic health (eHealth) tools, like patient-provider messaging, have changed over time relative to disparities in use of eHealth tools that were not directly incentivized.

Objective: This study examines trends in eHealth disparities before and after the introduction of US federal financial incentives. We compare rates of patient-provider messaging, which was directly incentivized, with rates of looking for health information on the Web, which was not directly incentivized.

Methods: We used nationally representative Health Information National Trends Survey data from 2003 to 2018 (N=37,300) to describe disparities in patient-provider messaging and looking for health information on the Web. We first reported the percentage of individuals across education and racial and ethnic groups who reported using these tools in each survey year and compared changes in unadjusted disparities during preincentive (2003-2011) and postincentive (2011-2018) periods. Using multivariable linear probability models, we then examined adjusted effects of education and race and ethnicity in 3 periods-preincentive (2003-2005), early incentive (2011-2013), and postincentive (2017-2018)-controlling for sociodemographic and health factors. In the postincentive period, an additional model tested whether internet adoption, provider access, or providers' use of EMRs explained disparities.

Results: From 2003 to 2018, overall rates of provider messaging increased from 4% to 36%. The gap in provider messaging between the highest and lowest education groups increased by 10 percentage points preincentive (P<.001) and 22 additional points postincentive (P<.001). The gap between Hispanics and non-Hispanic whites increased by 3.2 points preincentive (P=.42) and 11 additional points postincentive (P=.01). Trends for blacks resembled those for Hispanics, whereas trends for Asians resembled those for non-Hispanic whites. In contrast, education-based disparities in looking for health information on the Web (which was not directly incentivized) did not significantly change in preincentive or postincentive periods, whereas racial disparities narrowed by 15 percentage points preincentive (P=.008) and did not significantly change postincentive. After adjusting for other sociodemographic and health factors, observed associations were similar to unadjusted associations, though smaller in magnitude. Including internet adoption, provider access, and providers' use of EMRs in the postincentive model attenuated, but did not eliminate, education-based disparities in provider messaging and looking for health information on the Web. Racial and ethnic disparities were no longer statistically significant in adjusted models.

Conclusions: Disparities in provider messaging widened over time, particularly following federal financial incentives. Meanwhile, disparities in looking for health information on the Web remained stable or narrowed. Incentives may have disproportionately benefited socioeconomically advantaged groups. Future policy could address disparities by incentivizing providers treating these populations to adopt messaging capabilities and encouraging patients' use of messaging.

Keywords: communication; disparities; eHealth; inequality; policy; secure messaging; socioeconomic factors.

Prevalence of electronic health use, 2003 to 2018. The sample for provider messaging includes 32,742 total responses (average 4677 per year), and the sample for looking for health information on the Web includes 28,663 total responses (average 4090 per year). Survey weights were used to generate means reflective of the US population. Bars represent 95% CIs generated using jackknife SEs.

Electronic health use by education level, 2003 to 2018. The sample for provider messaging includes 31,672 total responses, and the sample for looking for health information on the Web includes 27,860 total responses. Survey weights were used to generate means reflective of the US population. Bars represent 95% CIs generated using jackknife SEs. Brackets represent the difference in prevalence between the highest and lowest education groups in the first and last years of the analysis. Of the overall respondents, 3% were not included in this analysis because they were missing information on education.

Electronic health use by race and ethnicity. The sample for provider messaging includes 29,484 total responses, and the sample for looking for health information on the Web includes 25,638 total responses. Survey weights were used to generate means reflective of the US population. Bars represent 95% CIs generated using SEs. Brackets represent the difference in prevalence between Hispanic and non-Hispanic white respondents in the first and last years of the analysis. Other and multiracial categories were excluded from this analysis because they did not have at least 100 observations for each sample year. Furthermore, 7% of total respondents were excluded because they did not indicate a race.