Dataset Information

A Disease Identification Algorithm for Medical Crowdfunding Campaigns: Validation Study.

ABSTRACT:

Background

Web-based crowdfunding has become a popular method to raise money for medical expenses, and there is growing research interest in this topic. However, crowdfunding data are largely composed of unstructured text, thereby posing many challenges for researchers hoping to answer questions about specific medical conditions. Previous studies have used methods that either failed to address major challenges or were poorly scalable to large sample sizes. To enable further research on this emerging funding mechanism in health care, better methods are needed.

Objective

We sought to validate an algorithm for identifying 11 disease categories in web-based medical crowdfunding campaigns. We hypothesized that a disease identification algorithm combining a named entity recognition (NER) model and word search approach could identify disease categories with high precision and accuracy. Such an algorithm would facilitate further research using these data.

Methods

Web scraping was used to collect data on medical crowdfunding campaigns from GoFundMe (GoFundMe Inc). Using pretrained NER and entity resolution models from Spark NLP for Healthcare in combination with targeted keyword searches, we constructed an algorithm to identify conditions in the campaign descriptions, translate conditions to International Classification of Diseases, 10th Revision, Clinical Modification (ICD-10-CM) codes, and predict the presence or absence of 11 disease categories in the campaigns. The classification performance of the algorithm was evaluated against 400 manually labeled campaigns.

Results

We collected data on 89,645 crowdfunding campaigns through web scraping. The interrater reliability for detecting the presence of broad disease categories in the campaign descriptions was high (Cohen κ: range 0.69-0.96). The NER and entity resolution models identified 6594 unique (276,020 total) ICD-10-CM codes among all of the crowdfunding campaigns in our sample. Through our word search, we identified 3261 additional campaigns for which a medical condition was not otherwise detected with the NER model. When averaged across all disease categories and weighted by the number of campaigns that mentioned each disease category, the algorithm demonstrated an overall precision of 0.83 (range 0.48-0.97), a recall of 0.77 (range 0.42-0.98), an F₁ score of 0.78 (range 0.56-0.96), and an accuracy of 95% (range 90%-98%).

Conclusions

A disease identification algorithm combining pretrained natural language processing models and ICD-10-CM code-based disease categorization was able to detect 11 disease categories in medical crowdfunding campaigns with high precision and accuracy.

SUBMITTER: Doerstling SS

PROVIDER: S-EPMC9257615 | biostudies-literature |

REPOSITORIES: biostudies-literature

ACCESS DATA

Similar Datasets

Project description:There has been growing recognition of the popularity of medical crowdfunding and research documenting how crowdfunding arises from, and contributes to, social and health inequities. While many researchers have surmised that racism could well play a role in medical crowdfunding campaign outcomes, research on these dynamics has been limited. No research to date has examined these dynamics among the most successful medical crowdfunding campaigns, focusing instead on average users' experiences or specific patient subpopulations. This paper analyzes key characteristics and demographics of the 827 most successful medical crowdfunding campaigns captured at a point in time in 2020 on the popular site GoFundMe, creating the first demographic archetype of "viral" or highly successful campaigns. We hypothesized that this sample would skew towards whiter, younger populations, more heavily represent men, and reflect critical illnesses and accidents affecting these populations, in addition to having visually appealing, well-crafted storytelling. Analysis supported these hypotheses, showing significant levels of racial and gender disparities among campaigners. While white men had the greatest representation, Black and Asian users, and black women in particular, were highly underrepresented. Like other studies, we find evidence that racial and gender disparities persist in terms of campaign outcomes as well. Alongside this quantitative analysis, a targeted discourse analysis revealed campaign narratives and comments reinforced racist and sexist tropes of selective deservingness. These findings add to growing calls for more health research into the ways that social media technologies shape health inequities for historically marginalized and disenfranchised populations. In particular, we underscore how successful crowdfunding campaigns, as a both a means of raising funds for health and a broader site of public engagement, may deepen and normalize gendered and racialized inequities. In this way, crowdfunding can be seen as a significant technological amplifier of the fundamental social causes of health disparities.

Project description:Americans are increasingly relying on crowdfunding to pay for the costs of healthcare. In medical crowdfunding (MCF), online platforms allow individuals to appeal to social networks to request donations for health and medical needs. Users are often told that success depends on how they organize and share their campaigns to increase social network engagement. However, experts have cautioned that MCF could exacerbate health and social disparities by amplifying the choices (and biases) of the crowd and leveraging these to determine who has access to financial support for healthcare. To date, research on potential axes of disparity in MCF, and their impacts on fundraising outcomes, has been limited. To answer these questions, this paper presents an exploratory cross-sectional study of a randomized sample of 637 MCF campaigns on the popular platform GoFundMe, for which the race, gender, age, and relationships of campaigners and campaign recipients were categorized alongside campaign characteristics and outcomes. Using both descriptive and inferential statistics, the analysis examines race, gender, and age disparities in MCF use, and tests how these are associated with differential campaign outcomes. The results show systemic disparities in MCF use and outcomes: people of color (and black women in particular) are under-represented; there is significant evidence of an additional digital care labor burden on women organizers of campaigns; and marginalized race and gender groups are associated with poorer fundraising outcomes. Outcomes are only minimally associated with campaign characteristics under users' control, such as photos, videos, and updates. These results corroborate widespread concerns with how technology fuels health inequities, and how crowdfunding may be creating an unequal and biased marketplace for those seeking financial support to access healthcare. Further research and better data access are needed to explore these dynamics more deeply and inform policy for this largely unregulated industry.

Project description:ObjectivesMedical crowdfunding is a rapidly growing practice where individuals leverage social networks to raise money for health-related needs. This practice has allowed many to access healthcare and avoid medical debt but has also raised a number of ethical concerns. A dominant criticism of this practice is that it is likely to increase inequities in access to healthcare if persons from relatively wealthy backgrounds, media connections, tech-savvy and educational attainments are best positioned to use and succeed with crowdfunding. However, limited data has been published to support this claim. Our objective in this paper is to assess this concern using socioeconomic data and information from crowdfunding campaigns.SettingTo assess this concern, we present an exploratory spatial analysis of a new dataset of crowdfunding campaigns for cancer-related care by Canadian residents.ParticipantsFour datasets were used: (1) a medical crowdfunding dataset that included cancer-related campaigns posted by Canadians, (2) 2016 Census Profile for aggregate dissemination areas, (3) aggregate dissemination area boundaries and (4) forward sortation area boundaries.ResultsOur exploratory spatial analysis demonstrates that use of crowdfunding for cancer-related needs in Canada corresponds with high income, home ownership and high educational attainment. Campaigns were also commonly located near city centres.ConclusionsThese findings support concerns that those in positions of relative socioeconomic privilege disproportionately use crowdfunding to address health-related needs. This study was not able to determine whether other socioeconomic dimensions such as race, gender, ethnicity, nationality and linguistic fluency are also correlated with use of medical crowdfunding. Thus, we call for further research to explore the relationship between socioeconomic variables and medical crowdfunding campaigning to explore these other socioeconomic variables and campaigns for needs unrelated to cancer.

Project description:BACKGROUND:Genetic sequencing is critically important to diagnostic health care efforts in the United States today, yet it is still inaccessible to many. Meanwhile, the internet and social networking have made crowdfunding a realistic avenue for individuals and groups hoping to fund medical and research causes, including patients in need of whole exome genetic sequencing (WES). OBJECTIVE:Amplify Hope is an educational program designed to investigate what factors affect the success of medical crowdfunding campaigns. We conducted a needs assessment, a series of 25 interviews concerning crowdfunding, and provided training on best practices identified through our assessment for 11 individuals hoping to run their medical crowdfunding campaigns to raise money for patients to access trio WES to identify the mutated proteins that caused their apparent inherited disease. METHODS:The crowdfunding education was given in a 30-day training period with resources such as webinars, fact sheets and a crowdfunding training guide emailed to each participant. All campaigns were launched on the same date and were given 30 days to raise the same goal amount of US $5000. Reviewing the 4 crowdfunding campaigns that raised the goal amount within the 30-day period, we sought to identify features that made the 4 crowdfunding campaigns successful. In addition, we sought to assess which factors the resulting 75 donors report as influencing their decision to donate to a campaign. Finally, we investigated whether crowdfunding campaigns for exome sequencing had an impact on increasing applicant's and donors' knowledge of genomics. RESULTS:Of the 86 study inquiries, 11 participants submitted the required forms and launched their crowdfunding campaigns. A total of 4 of the 11 campaigns raised their goal amounts within 30 days. CONCLUSIONS:We found that social media played an important role in all campaigns. Specifically, a strong social media network, an active outreach process to networks, as well as engagement within the study all correlated with a higher success rate. Amplify Hope donors were more likely to support projects that were near their fundraising goals, and they found video far more effective for learning about genomics than any other medium.

Project description:BACKGROUND:There are a range of perceived gaps and shortcomings in the publicly funded Canadian health system. These include wait times for care, lack of public insurance coverage for dental care and pharmaceuticals, and difficulties accessing specialist care. Medical crowdfunding is a response to these gaps where individuals raise funds from their social networks to address health-related needs. OBJECTIVE:This study aimed to investigate the potential of crowdfunding data to better understand what health-related needs individuals are using crowdfunding for, how these needs compare with the existing commentary on health system deficiencies, and the advantages and limitations of using crowdfunding campaigns to enhance or augment our understanding of perceived health system deficiencies. METHODS:Crowdfunding campaigns were scraped from the GoFundMe website. These campaigns were then limited to those originating in the metropolitan Vancouver region of two health authorities during 2018. These campaigns were then further limited to those raising funds to allow the treatment of a medical problem or related to needs arising from ill health. These campaigns were then reviewed to identify the underlying health issue and motivation for pursuing crowdfunding. RESULTS:We identified 423 campaigns for health-related needs. These campaigns requested CAD $8,715,806 (US $6,088,078) in funding and were pledged CAD $3,477,384 (US $2,428,987) from 27,773 donors. The most common underlying medical condition for campaign recipients was cancer, followed by traumatic injuries from collisions and brain injury and stroke. By far, the most common factor of motivation for crowdfunding was seeking financial support for wages lost because of illness (232/684, 33.9%). Some campaigns (65/684, 9.5%) sought help with purchasing medical equipment and supplies; 8.2% (56/684) sought to fund complementary, alternative, or unproven treatments including experimental interventions; 7.2% (49/684) sought financial support to cover travel-related costs, including in-province and out-of-province (49/684, 7.2%) travel; and 6.3% (43/684) campaigns sought help to pay for medication. CONCLUSIONS:This analysis demonstrates the potential of crowdfunding data to present timely and context-specific user-created insights into the perceived health-related financial needs of some Canadians. Although the literature on perceived limitations of the Canadian health system focuses on wait times for care and limited access to specialist services, among other issues, these campaigners were much more motivated by gaps in the wider social system such as costs related to unpaid time off work and travel to access care. Our findings demonstrate spatial differences in the underlying medical problems, motivations for crowdfunding, and success using crowdfunding that warrants additional attention. These differences may support established concerns that medical crowdfunding is most commonly used by individuals from relatively privileged socioeconomic backgrounds. We encourage the development of new resources to harness the power of crowdfunding data as a supplementary source of information for Canadian health system stakeholders.

Project description:BackgroundCommon disease-specific outcomes are vital for ensuring comparability of clinical trial data and enabling meta analyses and interstudy comparisons. Traditionally, the process of deciding which outcomes should be recommended as common for a particular disease relied on assembling and surveying panels of subject-matter experts. This is usually a time-consuming and laborious process.ObjectiveThe objectives of this work were to develop and evaluate a generalized pipeline that can automatically identify common outcomes specific to any given disease by finding, downloading, and analyzing data of previous clinical trials relevant to that disease.MethodsAn automated pipeline to interface with ClinicalTrials.gov's application programming interface and download the relevant trials for the input condition was designed. The primary and secondary outcomes of those trials were parsed and grouped based on text similarity and ranked based on frequency. The quality and usefulness of the pipeline's output were assessed by comparing the top outcomes identified by it for chronic obstructive pulmonary disease (COPD) to a list of 80 outcomes manually abstracted from the most frequently cited and comprehensive reviews delineating clinical outcomes for COPD.ResultsThe common disease-specific outcome pipeline successfully downloaded and processed 3876 studies related to COPD. Manual verification indicated that the pipeline was downloading and processing the same number of trials as were obtained from the self-service ClinicalTrials.gov portal. Evaluating the automatically identified outcomes against the manually abstracted ones showed that the pipeline achieved a recall of 92% and precision of 79%. The precision number indicated that the pipeline was identifying many outcomes that were not covered in the literature reviews. Assessment of those outcomes indicated that they are relevant to COPD and could be considered in future research.ConclusionsAn automated evidence-based pipeline can identify common clinical trial outcomes of comparable breadth and quality as the outcomes identified in comprehensive literature reviews. Moreover, such an approach can highlight relevant outcomes for further consideration.

Project description:BackgroundMedical crowdfunding has emerged as a growing field for fundraising opportunities. Some environmental trends have driven the emergence of campaigns to raise funds for medical care. These trends include lack of medical insurance, economic backlash following the 2008 financial collapse, and shortcomings of health care regulations.ObjectiveResearch regarding crowdfunding campaign use, reasons, and effects on the provision of medical care and individual relationships in health systems is limited. This study aimed to explore the nature and dimensions of the phenomenon of medical crowdfunding using a visual analytics approach and data crawled from the GoFundMe crowdfunding platform in 2019. We aimed to explore and identify the factors that contribute to a successful campaign.MethodsThis data-driven study used a visual analytics approach. It focused on descriptive analytics to obtain a panoramic insight into medical projects funded through the GoFundMe crowdfunding platform.ResultsThis study highlighted the relevance of positioning the campaign for fundraising. In terms of motivating donors, it appears that people are typically more generous in contributing to campaigns for children rather than those for adults. The results emphasized the differing dynamics that a picture posted in the campaign brings to the potential for medical crowdfunding. In terms of donor's motivation, the results show that a picture depicting the pediatric patient by himself or herself is the most effective. In addition, a picture depicting the current medical condition of the patient as severe is more effective than one depicting relative normalcy in the condition. This study also drew attention to the optimum length of the title. Finally, an interesting trend in the trajectory of donations is that the average amount of a donation decreases with an increase in the number of donors. This indicates that the first donors tend to be the most generous.ConclusionsThis study examines the relationship between social media, the characteristics of a campaign, and the potential for fundraising. Its analysis of medical crowdfunding campaigns across the states offers a window into the status of the country's health care affordability. This study shows the nurturing role that social media can play in the domain of medical crowdfunding. In addition, it discusses the drivers of a successful fundraising campaign with respect to the GoFundMe platform.

Project description:BackgroundThe electronic health record (EHR) contains a wealth of medical information. An organized EHR can greatly help doctors treat patients. In some cases, only limited patient information is collected to help doctors make treatment decisions. Because EHRs can serve as a reference for this limited information, doctors' treatment capabilities can be enhanced. Natural language processing and deep learning methods can help organize and translate EHR information into medical knowledge and experience.ObjectiveIn this study, we aimed to create a model to extract concept embeddings from EHRs for disease pattern retrieval and further classification tasks.MethodsWe collected 1,040,989 emergency department visits from the National Taiwan University Hospital Integrated Medical Database and 305,897 samples from the National Hospital and Ambulatory Medical Care Survey Emergency Department data. After data cleansing and preprocessing, the data sets were divided into training, validation, and test sets. We proposed a Transformer-based model to embed EHRs and used Bidirectional Encoder Representations from Transformers (BERT) to extract features from free text and concatenate features with structural data as input to our proposed model. Then, Deep InfoMax (DIM) and Simple Contrastive Learning of Visual Representations (SimCLR) were used for the unsupervised embedding of the disease concept. The pretrained disease concept-embedding model, named EDisease, was further finetuned to adapt to the critical care outcome prediction task. We evaluated the performance of embedding using t-distributed stochastic neighbor embedding (t-SNE) to perform dimension reduction for visualization. The performance of the finetuned predictive model was evaluated against published models using the area under the receiver operating characteristic (AUROC).ResultsThe performance of our model on the outcome prediction had the highest AUROC of 0.876. In the ablation study, the use of a smaller data set or fewer unsupervised methods for pretraining deteriorated the prediction performance. The AUROCs were 0.857, 0.870, and 0.868 for the model without pretraining, the model pretrained by only SimCLR, and the model pretrained by only DIM, respectively. On the smaller finetuning set, the AUROC was 0.815 for the proposed model.ConclusionsThrough contrastive learning methods, disease concepts can be embedded meaningfully. Moreover, these methods can be used for disease retrieval tasks to enhance clinical practice capabilities. The disease concept model is also suitable as a pretrained model for subsequent prediction tasks.