Key Points

Question听 How is stratifying patients referred to the dermatology department for a new lesion into simple and interpretable risk categories associated with patient outcomes?

Findings听 In this single-institution quality improvement cohort study of 37鈥�478 patients referred to a dermatology department, a risk stratification pathway that was based on categorical referral reasons and risk factors provided by primary care physicians identified and prioritized patients at a higher risk of skin cancer (overall and melanoma specifically).

Meaning听 The study findings suggest that stratifying patients based on this outlined pathway might expedite care for patients who require a biopsy and subsequent treatment, ensuring that individuals at higher risk for skin cancer receive timely treatment.

Importance听 Access to timely dermatologic care remains a challenge, especially for patients with new skin lesions. Assessing the efficiency of new triage pathways may assist in better resource allocation and shorter time to care.

Objective听 To evaluate whether a rule-based triage system was associated with better skin cancer risk stratification of patients and reduced wait times.

Design, Setting, and Participants听 This retrospective quality improvement cohort study of patients referred to Stanford University dermatology clinics was conducted between November 2017 and January 2023. A rules-based triage system based on a priori鈥揹etermined high-risk lesion characteristics was implemented.

Exposures听 Referral reasons and risk factors of patients provided by their primary care physicians.

Main Outcomes and Measures听 Biopsy results of patients (diagnosis of any skin cancer and melanoma) at their visit or within 6 months after the visit. Regression models were used to assess the association between risk factors at referral and (1) biopsy outcomes and (2) time to first visit, adjusting for sociodemographic factors.

Results听 Among 37鈥�478 patients (mean [SD] age, 54 (18) years; 21鈥�292 women [57%]), the rates of aggregate biopsy, malignant biopsy specimens, and melanoma were comparable across patients seen after (n鈥�=鈥�12鈥�302) and before (n鈥�=鈥�25鈥�176) the implementation of the new triage pathway. Patients seen through the lesion pathway had a higher risk of having malignant biopsy results (adjusted risk ratio [aRR], 1.6; 95% CI, 1.4-1.9) and melanoma (aRR, 2.0; 95% CI, 1.2-3.2) than those not seen through the pathway. Lesions that were concerning to referring clinicians for skin cancer were associated with an increased risk of skin cancer (all skin cancer: aRR, 2.8; 95% CI, 2.2-3.5; melanoma: aRR, 2.02; 95% CI, 1.1-3.7). Patients in the 3 high-risk lesion groups were seen faster in the new triage pathway (mean reduction, 26 days; 95% CI, 18-34 days).

Conclusions and Relevance听 In this study, a new automated, rules-based referral pathway was implemented that expedited care for patients with high-risk skin cancer. This reform may have contributed to improving patient stratification, reducing the time from referral to first encounter, and maintaining accuracy in identifying malignant lesions. The findings highlight the potential to optimize clinical resource allocation by better risk stratification of referred patients.

There is a well-described undersupply and maldistribution of dermatologists in the US^1,2 that has been associated with excessive wait times averaging 30 to 100 days for appointments.^2,3 For patients with new skin lesions, long wait times can be associated with advanced presentations at diagnosis and in turn more aggressive treatments, extensive resource use, and poor outcomes.⁴

The current referral process is frequently followed by delays and errors.⁵ Specialty referrals are often handwritten or free-text descriptors that frequently require clinician input to appropriately triage and schedule patients, in which is associated with increased physician administrative burden. In practices without formal triage workflows, patients are often defaulted to the next available appointment, unless specific 鈥渞ed-flag鈥� complaints are identified by front-line staff. This lack of consistency leaves much room for improvement. Prior work has demonstrated that lesion-directed screening in the general population revealed skin cancer at rates similar to full-body skin examinations but was more than 5 times faster.⁶ This finding suggests a time-efficient and cost-effective approach for skin cancer detection through single-lesion risk stratification.

In 2021, we instituted a quality improvement (QI) intervention (Figure) to our referral and triage system reform to prioritize patients at higher risk for skin cancer. We outlined a strategy to evaluate the feasibility and efficiency of matching risk-stratified referrals with the timeliness of visits, allowing for greater practice efficiency by predicting the necessity of in-person clinic procedures and subsequent skin cancer treatment. We present this rules-based, single-lesion referral pathway as a specific model for this broad approach.

The study was conducted from November 2017 to January 2023. The QI intervention implemented a customized electronic medical record referral order that allows referring clinicians to select prespecified referral reasons. We then designed a rule-based system that automatically categorized the patients into high risk or average risk based on referral reasons (Figure). In cases of multiple reasons, we assigned patients to the highest-priority category.

Biopsy results, scheduling data, and demographic characteristics were extracted from the enterprise data warehouse using targeted queries and then analyzed using natural language processing tools in Python.⁷ To evaluate the accuracy of the inferred biopsied results, a board-certified dermatologist examined all cases of malignant biopsy specimens and 500 randomly chosen cases of nonmalignant biopsy specimens. The QI process was exempt from institutional review board approval, and we received institutional review board approval from Stanford University to publish the outcome of this study, who waived informed consent.

We evaluated 3 biopsy-related outcomes: (1) biopsy rates (proportion of referred patients who underwent biopsy); (2) rates of malignant biopsy specimens (proportion of patients with confirmed skin cancers in any biopsied lesions); and (3) biopsied melanoma specimen rates (proportion of patients with confirmed melanoma among any biopsied lesions). For each referral reason, we defined the control group as all pre鈥搕riage reform patients and post鈥搕riage reform patients not assigned with that reason. The adjusted relative risk (aRR) for (1) malignant biopsy specimens and (2) melanoma biopsy specimens were evaluated using Poisson regressions, adjusting for age, sex, race and ethnicity, and median household income. Differences in the referral to first encounter time for each reason were assessed using linear regressions. We used R software (version 4.1.1; R Foundation) for statistical analyses (with a robust variance estimator accounting for unequal variance across patients) and conducted a sensitivity analysis by excluding self-referred patients.

Our study included 37鈥�478 patients (mean [SD] age, 54 [18.6] years; 21鈥�292 women [57%]; 1408 [4%] African American, 7740 [21%] Asian, 3844 [10%] Hispanic, and 21鈥�148 [56%] White individuals) (eTable 1 in Supplement 1). Comparing patients before and after the new triage pathway, we saw broadly similar biopsy rates (9.7% and 10.1%, respectively), malignant biopsy specimen rates (28.5% and 25.3%, respectively) and melanoma biopsy specimen rates (2.5% and 3.7%, respectively) (Table 1). Baseline demographic characteristics associated with biopsy outcomes included age (15.0% of patients who were older than 80 years underwent biopsy, with 56.8% positive for cancer) and race (823 White patients underwent a biopsy, of which 34.6% yielded malignant results; eTable 2 in Supplement 1).

We found that the single-lesion cohort had a higher biopsy rate (15.1% vs 8.3%; P鈥�<鈥�.001) despite a relatively similar malignant biopsy specimen rate (28.2% vs 22.8%; P鈥�=鈥�.04), suggesting that although this cohort underwent more biopsies, diagnostic accuracy was maintained. Within this cohort, specific risk factors associated with higher biopsy rates and malignant biopsy specimen rates included lesions that were bleeding and undergoing rapid evolution (24.2% and 32.1%, respectively), or lesions in which prior clinicians had concerns for melanoma (23.4% and 25.0%, respectively). Regarding new melanoma diagnoses, referrals with concern from a referring clinician for melanoma had the highest rate of malignant biopsy specimens at 9.4%.

Adjusting for demographic risk factors, patients with specific lesions of concern had a higher risk of having malignant biopsy results (aRR, 1.6; 95% CI, 1.4-1.9) and melanoma results (aRR, 2.0; 95% CI, 1.2-3.2) than those without (Table 2). Patients with lesions that were concerning to a clinician for (1) general skin cancer and (2) melanoma also had higher risks of malignant biopsy results (aRR, 2.8; 95% CI, 2.2-3.5; aRR, 2.02; 95% CI, 1.1-3.7) and higher risks of melanoma (aRR, 2.8; 95% CI, 1.3-6.0; aRR, 7.7; 95% CI, 2.5-23.8).

In terms of wait times, we saw a slight increase over time (eFigure 2 in Supplement 1); patients in the a priori鈥揹etermined high-risk categories, who were assigned an expedited lesion slot through a rules-based approach, were seen a mean of 66 days after referral, which was 26 days earlier (95% CI, 18-36 days; Table 2) than the rest of patients. We did not observe a difference in wait times between White patients and African American, Asian, and Hispanic patients (2 days; 95% CI, 鈭�8 to 11 days), although skin cancer was more common in former (eTable 2 in Supplement 1). The sensitivity analysis that excluded self-referral patients yielded similar results for both biopsy outcomes and wait times (eTable 3 in Supplement 1).

This cohort study provides a granular model for how known skin cancer risk factors, when applied at the health system level, may better identify higher-risk patients requiring timely in-clinic care. The single lesion referral pathway was associated with similar biopsy frequency and accuracy, better risk stratification, and shorter wait times for patients with higher risks for skin cancer. Future QI efforts are focused on a matching capacity to further reduce waits for these and other identified higher-risk groups, including patients undergoing skin cancer screening. The UK鈥檚 National Institute for Health and Care Excellence recommended a specialist appointment within 28 days of referral for concerning lesions.⁸ In addition, multiple studies identified anxiety and stress that long wait times cause for patients with possible skin cancer diagnoses,⁹ underscoring the benefits of earlier diagnosis.

We found similar biopsy and cancer rates to those of previous studies. Sherban et al¹⁰ found that 27% of biopsied lesions via full-body skin examinations were malignant, similar to our findings of 27.7% for patients referred for skin cancer screening. We found that patients with a family or personal history of skin cancer were more likely to have a malignant biopsy specimen diagnosis, which is well established in prior work.^11,12

This study鈥檚 findings have limitations: given the observational nature, we cannot establish definitive causality, and the risk stratification and reduced wait times could be subject to uncontrolled confounding. Moreover, triage priorities can vary across practices. For example, this algorithm prioritized skin cancer risk over referrals for symptomatic lesions, although the latter can be associated with decreased quality of life. Such nonprioritized groups can experience relatively longer waits, highlighting the importance of understanding the postdeployment effect of triage decisions. Lastly, this cohort was from a single academic center, limiting the generalizability even after adjusting for patient demographic characteristics. In future work, we will explore a machine learning鈥揳ugmented triage pathway integrating inputs, including lesion images.

In this cohort study, we instituted a new referral pathway to our academic center鈥檚 dermatology department. This was associated with improved patient stratification for skin cancer diagnoses and shortened referral to first encounter time by up to 30 days for higher-risk groups. Our findings might improve in-clinic resource allocation and reduce patient wait time by prioritizing those more likely to require biopsies.

Accepted for Publication: April 19, 2024.

Published Online: June 26, 2024. doi:10.1001/jamadermatol.2024.1832

Corresponding Author: Albert S. Chiou, MD, MBA, Department of Dermatology, School of Medicine, Stanford University, Stanford, CA 94305 (achiou@stanford.edu).

Author Contributions: Drs Chiou and Chen had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Dr Chen and Ms Gui were co鈥搇ead authors.

Concept and design: Chen, Gui, Ko, Daneshjou, Chiou.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Chen, Gui, Adu-Brimpong, Daneshjou, Chiou.

Critical review of the manuscript for important intellectual content: All authors.

Statistical analysis: Chen, Yao, Adu-Brimpong, Chiou.

Administrative, technical, or material support: Javitz, Golovko, Ko, Chiou.

Supervision: Chen, Golovko, Ko, Daneshjou, Chiou.

Conflict of Interest Disclosures: Dr Daneshjou reported personal fees from L'oreal, Frazier Healthcare Partners, Pfizer, DWA, and VisualDx and stock options with MDAlgorithms and Revea outside the submitted work. No other disclosures were reported.

Funding/Support: Research reported in this article was partially supported by Stanford MedScholars (Drs Gui and Adu-Brimpong) and Stanford Data Science Postdoctoral Fellowship (Dr Chen).

Role of the Funder/Sponsor: The funding organizations had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Data Sharing Statement: See Supplement 2.

Additional Contributions: We acknowledge the Value-Based Care program at Stanford Healthcare for analyst support.