A cross-sectional study of social inequities in medical crowdfunding campaigns in the United States

For most of American history, if you couldn't pay for medical care, your options were charity, debt, or doing without. Then, a new system emerged—one that promised to democratize that charity, allowing anyone with a story and a Wi-Fi connection to reach thousands of potential donors. GoFundMe became the emblem of that promise. Need a kidney transplant? Start a campaign. Diagnosed with cancer and drowning in bills? Tell your story, share your link, and let the crowd decide. The pitch was access. The pitch was fairness. Nora Kenworthy and colleagues went and measured whether that was true. Medical crowdfunding, where individuals use online platforms to solicit donations from their social networks for health needs, has grown into a mainstream financial strategy in the United States. Kenworthy and colleagues note that experts had already raised a fundamental warning: if the crowd is the mechanism by which people gain access to care, what happens when the crowd's decisions mirror and magnify existing social biases? To answer that, the team built a randomized sample of six hundred thirty-seven US medical GoFundMe campaigns and analyzed them for race, gender, age, and fundraising outcomes. Building that sample was harder than it sounds. The researchers wrote a program that queried GoFundMe's medically categorized campaigns across every US ZIP code, generating a known population of one hundred sixty-five thousand nine hundred twenty-five campaigns. From there, they drew eight hundred twenty-two at random, filtered out campaigns that didn't meet inclusion criteria or had been taken down, and arrived at their final six hundred thirty-seven. Because GoFundMe collects no demographic data, the team inferred race, gender, and age from campaign text, names, and photos. Three coders of different racial backgrounds rated every campaign, achieving an inter-rater agreement score of 0.819 on a zero-to-one scale. They tracked outcome measures including total donations, average donation size, and number of donors, while also cataloguing campaign features: photos, videos, updates, and the campaigner's Facebook friend count. They were careful to note the ethical limits of this approach and the sampling biases introduced by platform search algorithms. Those caveats matter, but so do the results. The first thing Kenworthy and colleagues found is that the demographics of medical crowdfunding don't reflect the demographics of medical need. The sample skews white—disproportionately so—while people of color, and Black women in particular, are underrepresented. That's the cruel inversion at the heart of this story: African Americans, as the paper states, are disproportionately sicker, less insured, and more medically indebted. They need this safety net the most. They show up in it the least. Children, meanwhile, appear less often than adults in the sample, though their campaigns perform differently once they're there—more on that in a moment. Then come the outcomes, and this is where the findings get concrete and uncomfortable. The average campaign in this sample received forty-three donations. The mean average donation was about eighty-four dollars. Only nine point two percent of campaigns met their stated financial goal. This is not a system reliably rescuing people. It's a lottery with bad odds—and the odds are not evenly distributed. In linear regression models of average donation size, being coded as Black was associated with receiving twenty-two dollars and twenty-two cents less per donation compared with white recipients, at a p-value of 0.030. The count of donations told a similar story. In Poisson regression—a model designed for count data—Black recipients had a significantly lower rate of donations, and non-Black people of color fared even worse, with a larger negative coefficient. Genderqueer recipients showed the most severe disadvantage, with a Poisson estimate of negative 1.51. These disparities held up even after accounting for social network size. Child campaigns were a partial exception. They attracted more donations in number—a Poisson estimate of positive 0.28—but smaller average gifts, about eighteen dollars less per donation than adult campaigns. So children mobilize more donors but smaller checks. That's a different disadvantage, not an absence of one. Now, here's what Kenworthy and colleagues tested next, and it matters. GoFundMe tells its users: share your campaign, engage your network, post updates, add photos. The implicit message is that effort and presentation determine success. The data say otherwise. Campaign features under users' control—photos, videos, updates—showed very low correlations with monetary outcomes in the Spearman analysis. Meanwhile, the size of a campaigner's Facebook network, which is not something you can simply manufacture, had a small but highly significant positive association with the number of donations. And who has larger networks? In this sample, men and white campaigners. Women's Facebook friend counts were significantly lower—the coefficient in that model was negative 0.126, with a p-value below 0.001. Non-Black people of color also showed smaller networks. So race and gender shape your network, your network shapes your donations, and your choice of photo barely registers. The racial disparities in donations persisted even when network size was included as a covariate. That means network disadvantage explains part of the racial gap—but not all of it. The crowd appears to be making independent judgments, and those judgments track race. This brings us to what Kenworthy and colleagues call digital care labor, and it's one of the most striking findings in the paper. Women are doing the organizing. Among self-fundraisers, sixty-seven percent were women. Among those fundraising on behalf of someone else—a parent, a sibling, a friend—eighty-two percent were women. Both of those gender imbalances are statistically overwhelming, with chi-square values in the hundreds. The authors describe the labor involved: sustaining relationships, managing information flows, keeping friends and family updated, handling the emotional weight of public medical appeals, capturing and posting images. This is an extension of the unpaid care work that already falls disproportionately on women, now migrated onto digital platforms. And despite shouldering most of that organizing burden, women do not come out ahead. In the Poisson model without network size included, being female was associated with slightly fewer donations—a small but significant negative effect. When Facebook friend count was added, that gender effect disappeared, suggesting network size mediates some of the difference. The picture that emerges is still damning: women work harder to run these campaigns, and the structural disadvantages encoded in their smaller networks swallow that effort. Kenworthy and colleagues describe crowdfunding as a system that makes existing social hierarchies legible and consequential. Women's digital care labor is required to participate in a marketplace that then undervalues them. So what do we do with this? The research limits are real: GoFundMe doesn't share its data publicly, so the team lacked information on page views, socioeconomic status, and medical severity—all factors that might add texture to these findings. Without richer platform data, researchers are forced to rely on methods like facial recognition to infer race, which carries its own serious ethical problems. The call for transparency from crowdfunding companies is therefore not just procedural. It's about whether the public can actually audit a system that now functions as an informal health financing market. On policy, the authors stake out two positions. First, platform accountability—data access, transparency, the ability for researchers and policymakers to examine what's actually happening on these sites. Second, and more fundamental: crowdfunding is, in their words, "wholly at odds with, and will never be a replacement for, a rights-based system of care which enables all people, regardless of identity, to access necessary healthcare." That framing is deliberate. The problem isn't that GoFundMe has a bug. The problem is the system that made GoFundMe necessary. The promise was democratization. What Kenworthy and colleagues measured was a biased marketplace—one where the color of your skin, the size of your social network, and your gender determine whether the crowd comes through for you. The old inequities didn't disappear when healthcare financing moved online. They were reformatted, accelerated, and handed back to us as a feature. This lecture was created by ennepō. Go to https://ennepo.ai to Discover, Create and Follow the latest research in your field. Read when you can. Listen when you want to.