Characterizing the lack of diversity in musculoskeletal urgent care website content

Study design and setting

To investigate the diversity in website content for MUCCs, we performed a comprehensive search for all MUCCs in the United States. In June 2021, our study population was determined and included all MUCCs in the United States located using Google Maps (Mountain View, CA, USA). We used the phrases “XX musculoskeletal urgent care”, “XX orthopedic urgent care”, and “XX MSK urgent care”, where XX was replaced with the two-letter state postal abbreviation. All search results were investigated on Google Maps to determine if they met the classification for a MUCC. Once a MUCC was located on Google Maps, Google (Mountain View, CA) was used to find the MUCC website. In addition, the list of MUCCs we developed was compared to a prior list of MUCCs created by Yousman et al. [14] to ensure all possible MUCCs were included.

Participants/study subjects

The classification of a MUCC was a clinic that had same-day appointments or walk-in appointments. Orthopedic surgical centers that offered to schedule appointments the next day or the day after were excluded prior to collecting website graphic data. We excluded general urgent care centers and orthopedic clinic offices prior to collecting website graphic data. We found a total of 619 MUCCs in 49 states, excluding Delaware which had no open MUCCs. To collect information from MUCC websites, we examined content on the webpage that did not require scrolling (above the fold). Given that prior research has shown that website users spend more time looking at information above the fold [15, 16], we elected to only include website content that was visible without scrolling on the main page of the website. All website pages were viewed on a 13” laptop screen to standardize the size of the website page. We did not click on additional website tabs to look for other graphics. Videos were excluded. No information was collected on website graphics that did not feature a model or only included body parts such as a hand. If there was not a prominent website graphic, the MUCC was omitted from analysis. A total of 235 prominent website graphics were discovered. As some MUCC websites had multiple locations listed and patients would be able to access the MUCC website through different location specific search terms, we included all MUCC locations in our analyses. Any MUCCs for which we were unable to collect data were excluded from the analysis.

Variables, outcome measures, data sources, and bias

We collected the following yes/no variables for each prominently featured website graphic (above the fold and visible without scrolling) that included human figures (either full models or partial models with identifiable characteristics) on the website to portray a patient (model)(s): only white model(s), multiracial presence, presence of women, presence of obese/overweight model, white feature model, and female feature model. We selected race/ethnicity, gender, and body habitus as our primary variables given that prior research in orthopedic surgery has demonstrated that patients from racial or ethnic minority backgrounds, women, and obese patients may have difficulty accessing orthopedic care and may have worse outcomes following orthopedic surgery [17–19]. To minimize bias, two authors reviewed each graphic to collect the above variables and a third author was included if any disagreement regarding the variables arose, which occurred infrequently. A study by Tirrell et al. utilized an author as a reviewer to assess race and gender of models included in social media posts by various plastic surgery groups [20]. Rankin et al. used a similar methodology including the use of two author reviewers for direct-to-consumer advertising in total joint arthroplasty [21]. Prior research has demonstrated that APIs used to classify race/ethnicity and gender of graphics can have low accuracy which is why we utilized multiple reviewers instead of an API [22, 23]. Multiracial presence was defined as when a website graphic portrayed individuals from multiple racial groups (including White, Hispanic, Black, and Asian) at the same level of perspective. A graphic would not qualify for multiracial presence if a white model was the focus with models with different races in the background; all individuals had to be featured equally. A feature model was defined as the model that was most prominent in the graphic either by being the largest graphic or being in the foreground of the graphic. If a graphic included models from multiple racial groups with a white model in the foreground, then this graphic was classified as having a white feature model. Rankin et al. used a similar classification scheme with non-white focused advertisements which are equivalent to multiracial graphics in this study and white-focused advertisements which are equivalent to white feature model graphics in this study [21]. We chose to use this classification scheme as we felt that graphics with white models in the foreground with multiple racial groups in the background does not portray true equity in representation. As in Rankin et al. [21], the presence of overweight/obese models was determined as when models had an estimated body mass index of greater than 30. While this is a more subjective measure, each graphic was reviewed by two of the authors and confirmed by a physician.

Individual MUCCs were classified according to their affiliation as either nonaffiliated (without a connection to a hospital or practice), extension (a MUCC associated with a private practice or nonacademic hospital), or academic (associated with a teaching hospital). They were also classified according to their geographic region (Northeast, South, Midwest, West) according to the U.S. Census definitions of region [24]. Cash payment prices (the required payment for an appointment for a patient without insurance) for individual MUCCs were collected in a previous study by Yousman et al. [14].

Primary and secondary study outcomes

Our primary study goal was to characterize the racial, gender, and body type diversity present in website content for MUCCs across the United States. To achieve this, we collected website graphic information as explained above and identified the race, gender, and body type of the model(s) portrayed in each graphic.

Our secondary study goal was to analyze what factors increased the likelihood of diverse website content such as MUCC affiliation, MUCC region, and cash payment cost. We achieved this by performing analysis to investigate which factors were most strongly associated with increased website content diversity.

Statistical analysis

Chi-squared and univariate logistic regression analyses were performed to analyze trends in MUCC website content and to investigate factors that were associated with increased diversity in website graphics. All statistical analysis was performed using Stata Version 16.1 (StataCorp, College Station, TX).

Overview

Overall, 14% (32/235) of MUCCs had graphics on their websites that included individuals from multiple racial groups at equal levels of perspective, 80% (188/235) of graphics featured only white models, 57% (135/235) of graphics included women, 2% (5/235) included obese or overweight models, 86% (203/235) of graphics had a white feature model, and 34% (81/235) of graphics had a female feature model (Table 1).

Differences in website graphics by region

Website graphics with a white feature model differed by region (p < 0.01) with the South having the highest proportion of website graphics with white feature models and the Northeast having the lowest proportion. Website graphics with a female feature model differed by region (p < 0.01) with the Northeast having the highest proportion of website graphics that had a female model as the most predominant figure in the graphic and the South having the lowest proportion. There were no differences in the proportion of website graphics that included body type diversity between geographic regions in the United States (p > 0.05) (Table 1).

Differences in graphics by Medicaid acceptance and cash payment price

For graphics at MUCCs that accepted Medicaid, 76% (120/158) included only white models, 18% (28/158) had multiracial presence, 60% (95/158) included women, 3% (4/158) included obese/overweight models, 84% (133/158) included white feature models, and 34% (53/158) included female feature models. For graphics at MUCCs that did not accept Medicaid, 88% (68/77) included only white models, 5% (4/77) had multiracial presence, 52% (40/77) included women, 1% (1/77) included obese/overweight models, 91% (70/77) included white feature models, and 36% (28/77) had female feature models. The average cash payment price at the 205 MUCCs for which we had data was $251 (standard deviation (SD): 95) for MUCCs with graphics that included only white models, $233 (SD: 85) with graphics that had multiracial presence, $245 (SD: 98) with graphics that included women, $282 (SD: 103) with graphics that included obese/overweight models, $252 (SD: 93) for graphics with white feature models, and $245 (SD: 98) for graphics with female feature models (Table 1).

Factors associated with multiracial presence in website graphics

MUCCs that accepted Medicaid were more likely to have a multiracial presence in their website graphics (OR = 1.11, 95% CI: 1.00–1.22, p = 0.04). Additionally, the presence of women in website graphics was associated with an increased likelihood of multiracial presence (OR = 1.18, 95% CI: 1.08–1.29, p < 0.001). Region, MUCC affiliation, and cash payment price were not associated with multiracial presence in website content (Table 2).

Factors associated with women in website graphics

Websites that had a multiracial presence in their graphics were also more likely to feature women (OR 1.44, 95% CI: 1.18–1.76, p < 0.001). Region, MUCC affiliation, cash payment price, and MUCC Medicaid acceptance were not associated with the presence of women in website content (Table 3).

Limitations

Our study has several limitations. First, as no centralized database for MUCCs exists, our search strategy may have missed some MUCC locations given that Google Maps is limited to registered buildings and is partially crowd sourced. However, we feel that our search was comprehensive and encompassed all MUCCs that could be found on the internet. In addition, we were able to compare our list with a previous list produced by Yousman et al. [14] which provided us with a way to further search for MUCCs. Second, it is possible that we misclassified the perceived race/ethnicity or body type of models featured in the website graphics. We minimized this risk by having multiple authors review the website graphics and come to a consensus when a disagreement arose, which occurred infrequently. Misclassification of perceived race/ethnicity or gender could be more likely if the graphic did not include a full human figure but instead included only a partial human figure. However, a very small number of graphics did not include full human figures and we only included graphics with partial figures if they had identifiable characteristics. We excluded graphics that only had specific body parts such as a hand as the risk of misclassification would be higher. Third, we only analyzed graphics that were above the fold. It is possible that websites included graphics on other parts of the website. However, given that most people focus on content above the fold [15, 16], we felt this was a reasonable criterion for data collection. In addition, all data was collected on 13” laptops to try and standardize the website content that was visible without scrolling.

Ethics approval and consent to participate

This research was classified as “Not Human Subjects Research” by the Yale University Institutional Review Board so further ethics approval was not required and no consent to participate was needed. All methods were carried out in accordance with relevant guidelines and regulations.

Consent for publication

Not applicable.

Competing Interests

The authors have no conflicts of interest related to this manuscript.