Effect of Teledermatology on Access to Dermatology Care Among Medicaid Enrollees

Importance  Access to specialists such as dermatologists is often limited for Medicaid enrollees. Teledermatology has been promoted as a potential solution; however, its effect on access to care at the population level has rarely been assessed.

Objectives  To evaluate the effect of teledermatology on the number of Medicaid enrollees who received dermatology care and to describe which patients were most likely to be referred to teledermatology.

Design, Setting, and Participants  Claims data from a large California Medicaid managed care plan that began offering teledermatology as a covered service in April 2012 were analyzed. The plan enrolled 382 801 patients in California’s Central Valley, including 108 480 newly enrolled patients who obtained coverage after the implementation of the Affordable Care Act. Rates of dermatology visits by patients affiliated with primary care practices that referred patients to teledermatology and those that did not were compared. Data were collected from April 1, 2012, through December 31, 2014, and assessed from March 1 to October 15, 2015.

Main Outcomes and Measures  The percentage of patients with at least 1 visit to a dermatologist (including in-person and teledermatology visits) and total visits with dermatologists (including in-person and teledermatology visits) per 1000 patients.

Results  Of the 382 801 patients enrolled for at least 1 day from 2012 to 2014, 8614 (2.2%) had 1 or more visits with a dermatologist. Of all patients who visited a dermatologist, 48.5% received care via teledermatology. Among the patients newly enrolled in Medicaid, 75.7% (1474 of 1947) of those who visited a dermatologist received care via teledermatology. Primary care practices that engaged in teledermatology had a 63.8% increase in the fraction of patients visiting a dermatologist (vs 20.5% in other practices; P < .01). Compared with in-person dermatology, teledermatology served more patients younger vs older than 17 years (2600 of 4427 [58.7%] vs 1404 of 4187 [33.5%]), male patients (1849 of 4427 [41.8%] vs 1526 of 4187 [36.4%]), nonwhite patients (2779 of 4188 [66.4%] vs 1844 of 3478 [53.0%]), and individuals without comorbid conditions (1795 of 2464 [72.8%] vs 1978 of 3024 [65.4%]) (P < .001 for all comparisons). Conditions managed across settings varied; teledermatology physicians were more likely to care for viral skin lesions and acne (3405 of 7287 visits [46.7%]), whereas in-person dermatologists were more likely to care for psoriasis and skin neoplasms (10 062 of 27 347 visits [36.8%]).

Conclusions and Relevance  The offering of teledermatology appeared to improve access to dermatology care among Medicaid enrollees and played an especially important role for the newly enrolled.

Access to specialists such as dermatologists is limited in the United States. In dermatology, this is due to a shortage of dermatologists and geographic misdistribution.^1,2 Access is particularly limited among patients with Medicaid coverage and those treated by safety net primary care professionals.¹ Many dermatologists do not accept new Medicaid patients,³ and Medicaid patients make up a significantly smaller fraction of dermatologists’ patient population than would be expected based on the number of Medicaid enrollees in the United States. Those with Medicaid insurance also have to wait longer than privately insured patients to obtain appointments.⁴

Teledermatology, defined as the delivery of dermatology care at a distance using telecommunications technologies, may improve access. Dermatology is well suited to telehealth technology because the visual presentation of a disease is critical for the diagnosis and management of skin conditions. Teledermatology is classified into several types including (1) store and forward teledermatology, in which digital images and clinical information are captured and transmitted to dermatologists for asynchronous review, and (2) synchronous (interactive) video visits. Furthermore, teledermatology can use a consultative model (health care professional to health care professional) or a direct care model (health care professional to patient).² Consultative store and forward teledermatology, in which a primary care physician obtains a consultation from a remotely located dermatologist, is one of the most common forms of telemedicine.^2,5 In theory, teledermatology can improve access by bringing dermatologists to underserved communities and decreasing travel time for patients. Prior literature on teledermatology has demonstrated high rates of diagnostic accuracy, diagnostic concordance, and patient and health care professional satisfaction.^2,6 However, the effect of teledermatology on access to care at the population level has rarely been assessed.

Our goals were to address this gap in knowledge and assess the effect of a single teledermatology initiative to improve access to care among Medicaid enrollees. This initiative focused on San Joaquin County in California’s Central Valley, which has fewer dermatologists per capita than the national mean (1.2 vs 3.6 per 100 000 population).^7,8 To improve access to dermatologists, Health Plan of San Joaquin (HPSJ), a Medicaid managed care plan, began covering teledermatology services in April 2012. The study aims were to assess whether the introduction of teledermatology increased the number of Medicaid enrollees who received dermatology care and the characteristics of patients who were referred to teledermatology vs in-person dermatology care.

Box Section Ref ID

Health Plan of San Joaquin is a Medicaid managed care plan with 289 356 enrollees as of December 31, 2014. This study describes HPSJ’s experience with teledermatology from April 1, 2012, through December 31, 2014. During the study period, HPSJ experienced considerable growth (126%) in enrollees. This growth was fueled by geographic expansion into neighboring Stanislaus County and Medicaid expansion after the implementation of the Affordable Care Act.⁵ This study was approved by the institutional review board of the RAND Corporation, who waived the need for informed consent for this study of deidentified HPSJ health claims data.

In April 2012, HPSJ began covering teledermatology services delivered by Direct Dermatology. The teledermatology model incorporated consultative and direct care elements. Some of the primary care physicians who referred patients to teledermatology provided background on and obtained digital images of the patient’s skin problems and then uploaded the necessary information to a secure website. They received a consultation report within 2 days. However, over time, most primary care practices that referred patients to teledermatology used a more direct care model. Patients with skin problems were referred by their physician to 2 brick and mortar clinics to have photographs taken and to be interviewed by a teledermatology nurse. The remote dermatologist then reviewed the history and images, and the diagnosis and treatment was relayed to the patient and referring primary care physician. Teledermatology nurses also called the patient to review the diagnosis and treatment plan. If necessary, an in-person follow-up visit was scheduled at the clinic, where a teledermatologist was regularly onsite. Most of these follow-up appointments were to perform a biopsy or to address more complex cases. As such, the introduction of the teledermatology intervention in this population increased virtual and in-person access to dermatologists. When describing visit patterns, both types of visits provided by the teledermatology vendor were labeled as teledermatology, because they were delivered as part of a single, cohesive intervention.

Health Plan of San Joaquin provided deidentified health plan claims data and enrollment information for all enrollees with at least 1 day of enrollment from July 1, 2011, through December 31, 2014. Data provided included enrollee sex, age, zip code of residence, assigned primary care physician or practice, enrollment period, race/ethnicity, and use of health care services (eg, site of care, date of service, and diagnoses).

The 9-month period from July 1, 2011, to March 31, 2012, constituted the preteledermatology period. When describing visit patterns after teledermatology was offered, all enrollees in HPSJ from April 1, 2012, through December 31, 2014 (n = 382 801), were included. In subanalyses, the sample was restricted to those continuously enrolled from July 1, 2011, through December 31, 2014 (continuous enrollment defined as least 5 of 6 months in 2011 and 11 of 12 months in 2012, 2013, and 2014; 57 113 patients [14.9% of the entire population]) and those newly enrolled in 2014 (enrolled for ≥1 day in 2014 with no prior enrollment; 108 480 patients [28.3%]). All enrollees were included in our primary analysis because the continuously enrolled population (14.9%) is less likely to be generalizable owing to significant instability in Medicaid plans; however, for the primary analysis we conducted a sensitivity analysis limited to the continuously enrolled.⁹

We found a large amount of variation across practices in the use of teledermatology. Anecdotally the variation was driven by the primary care physician’s comfort and knowledge of the model as well as availability of in-person dermatology care (ie, within their multispecialty group or geographically close by).

The following 2 categories of enrollees were developed: those affiliated with practices that used teledermatology (user practices) and those affiliated with practices that did not (nonuser practices). User practices were defined as those with patients who had more than 20 teledermatology visits during the intervention period. These practices represented 12.3% of all in-network primary care practices who were the primary health care providers for 79.1% of all enrollees.

Health Plan of San Joaquin provided each enrollee’s chosen or designated primary care practice by month. Enrollees were assigned to a practice (and by extension to user or nonuser practice categories) using the following steps. (1) If any single primary care practice accounted for most of the total enrolled days, the enrollee was assigned to that practice (74.3% of all enrollees). (2) If no single practice accounted for most of the days, but most days were affiliated with a set of user practices or nonuser practices, the enrollee was assigned to the user or nonuser practice cohort, respectively (8.8% of all enrollees). (3) Remaining enrollees (16.9%) remained unassigned and were excluded from the analyses comparing the use of health care services between user and nonuser practices.

All visits to in-person dermatologists (primary or secondary specialty of dermatology) and any type of visit to the teledermatology provider (which were flagged by HPSJ in claims) were identified. The teledermatology provider also had some in-person visits that were identified based on billing codes.

We compared the leading dermatologic conditions treated by teledermatology, in-person dermatologists, and primary care practices (primary specialty of family medicine, family practice, general practice, internal medicine, or pediatrics). All dermatology visits were categorized by the 3-digit code from the International Classification of Diseases, Ninth Revision; as such, 1 visit could contribute more than 1 diagnosis. The number of comorbid conditions was calculated using the Charlson Comorbidity Index.¹⁰

Data were analyzed from March 1 to October 15, 2015. Demographic, socioeconomic, comorbidity, and use of health care service variables were compared across enrollees who received care from teledermatology and in-person dermatologists using χ² tests and unpaired t tests. Comorbidities were identified using data from July 1, 2011, to December 31, 2012; thus, our analyses using comorbidities were limited to a subset of the enrollee population with data from this earlier period. The main outcomes were (1) the percentage of enrollees with at least 1 visit to an in-person dermatologist or teledermatology and (2) total number of visits with in-person dermatologists or teledermatology per 1000 enrollees.

The periods before teledermatology (July 1, 2011, to March 31, 2012) and after teledermatology (July 1, 2013, to March 31, 2014) were defined to assess intervention effects. These periods were chosen to compare use of health care services in a similar 9-month period because seasonal differences in dermatology use could exist.

We calculated the percentage of change from before to after the entry of teledermatology for the primary outcomes by comparing enrollees affiliated with user vs nonuser practices and comparing the continuously and the newly enrolled patients. For the first outcome (percentage of enrollees with ≥1 visit to a dermatologist), an adjusted figure is presented that controls for patient demographics (age, sex, and race/ethnicity) and estimates the probability of at least 1 visit in a period, if the individual was enrolled for the full 9 months of the period before teledermatology or the period after teledermatology. We needed to account for time enrolled because in any given 9-month period, most of the enrollees had noncontinuous enrollment that affects the probability that they will visit a dermatologist. For this adjustment, the log odds of at least 1 visit, controlling for time enrolled and demographic characteristics, was modeled. The model also included an indicator for teledermatology user practices, an indicator equal to 1 for the period after the teledermatology coverage change, and their interaction. The marginal estimated probability of at least 1 visit from the mean observed demographic distributions in the full sample for each analysis, while assuming that each enrollee had the full 9 months of enrollment, is presented. For the second outcome (visit rate), we modeled the number of visits in each period for user and nonuser practices, controlling for enrollee demographics and length of time enrolled. The estimated marginal mean number of visits, scaled to visits per 1000 individuals during the 9-month study period, was reported. To accommodate the large number of zeros and greater-than-Poisson variances observed in the data, a zero-inflated negative binomial model was used with cluster-adjusted standard errors to account the potential for 2 observations (1 in the period before teledermatology and 1 in the period after) per enrollee. Statistical significance of the treatment effect was assessed by comparing the marginal means of the periods before and after teledermatology in the user vs nonuser practices.

We also compared leading diagnoses (defined as diagnoses that account for >5% of all skin condition visits to ≥1 of the settings) managed by teledermatology, in-person visits to a dermatologist’s office, and visits to primary care physician offices using χ² tests. In this analysis, the health care visit was the unit of analysis, and therefore a single enrollee could contribute more than 1 visit, and individual visits could contribute more than 1 diagnosis.

Of the 382 801 patients enrolled for at least 1 day from April 1, 2012, to December 31, 2014, 8614 (2.2%) had 1 or more visits with a dermatologist. During the study period, 7287 teledermatology visits occurred across 4515 patients (mean of 1.6 visits per patient) and 27 287 in-person dermatology visits across 4522 patients (mean of 5.4 visits per patient). Teledermatology accounted for 21.0% of all visits with dermatologists and served 48.5% of patients who interacted with a dermatologist. Among teledermatology visits, 80.2% were store and forward visits, and 19.8% were in-person follow-up visits with teledermatologists.

The number of dermatology visits per 1000 enrollees (visit rate) increased from 0.50 to 1.24 from the third quarter of 2011 to the first quarter of 2014, after which it fell to 0.91 in the fourth quarter of 2014 (Figure). In 2014, a rapid influx of new enrollees occurred owing to the Affordable Care Act. Compared with in-person dermatology visits, teledermatology served more patients younger than 17 years (2600 of 4427 [58.7%] vs 1404 of 4187 [33.5%]; P < .001), male patients (1849 of 4427 [41.8%] vs 1526 of 4187 [36.4%]; P < .001), nonwhite patients (2779 of 4188 [66.4%] vs 1844 of 3478 [53.0%]; P < .001), and individuals without comorbid conditions (1795 of 2464 [72.8%] vs 1978 of 3024 [65.4%]; P < .001) (Table 1).

In unadjusted models, the percentage of enrollees with 1 or more visits to a dermatologist (including in-person and teledermatology visits) increased from 0.9% to 1.3% among user practices from the period before teledermatology to the period after. The number of visits per 1000 enrollees increased from 31.29 to 42.73.

When adjusting for age, sex, race/ethnicity, and time enrolled from the period before teledermatology to the period after, the percentage of enrollees with 1 or more visits with a dermatologist (including in-person and teledermatology visits) increased from 1.2% to 1.9% among user practices and from 1.2% to 1.5% among nonuser practices. The difference in the growth rate between user and nonuser practices (63.8% vs 20.5%; P < .001) was statistically significant. In addition, the number of visits per 1000 enrollees increased from 40.55 to 66.72 among user practices and from 37.09 to 59.38 among nonuser practices. The difference in the growth rate comparing user and nonuser practices (64.6% vs 60.1%; P < .001) was smaller but also statistically significant (Table 2).

We compared the use of dermatology in 2014 among the newly enrolled and continuously enrolled patients for the entire study period. Fewer of the newly enrolled patients had 1 or more visits with a dermatologist compared with the continuously enrolled patients (2.2% vs 3.0%). They also had a lower visit rate, with 45.50 vs 112.83 visits per 1000 enrollees. Among the newly enrolled patients, a larger proportion of overall visits with dermatologists were via teledermatology (43.0% vs 17.5%) compared with the continuously enrolled patients (Table 3).

The leading 2 categories of diagnoses from teledermatology visits were viral skin lesions and acne, which together accounted for 3405 of 7287 visits (46.7%). In contrast, the leading 2 categories of diagnoses for in-person dermatology visits were acne and neoplasms of uncertain behavior, which together accounted for 10 062 of 27 347 visits (36.8%) (Table 4).

Our sensitivity analysis limited the before-and-after comparison of the use of dermatology services to the continuously enrolled patients who were assigned to a primary care practice (45 113 [11.8%]). In adjusted models, user practices experienced greater growth than nonuser practices in the percentage of enrollees with 1 or more visits with dermatologists (93.2% growth rate vs 59.9%; P = .07). However, nonuser practices had a slightly higher growth rate in visits per 1000 enrollees (109.3% vs 103.1%; P = .09), which was fueled by greater use of in-person dermatologists by nonuser practices (eAppendix in the Supplement).

Results indicate that the introduction of teledermatology achieved the aim of increasing access to dermatologists in a large Medicaid managed care plan. Patients among primary care practices that engaged in teledermatology were more likely to have 1 or more visits with dermatologists, and the rate of visits with dermatologists also increased. After it was introduced, teledermatology served half of all enrollees who received any dermatology care. Among new enrollees, teledermatology served three-fourths of patients with any dermatology care in 2014.

Several types of patients were more likely to be referred to teledermatology. Teledermatology generally served a healthier, younger population with more targeted, potentially less severe skin conditions. Although some common diagnoses occurred at in-person and telederamatology visits (eg, acne), in-person dermatology visits had higher rates of use for neoplasms and psoriasis and greater variation in diagnoses. These patterns suggest that in-person dermatologists served many of the patients in greatest need, but teledermatology expanded access to a new group of enrollees with different demographics and different needs. Teledermatology also played an important role for the newly insured and may be a particularly attractive option for those having difficulty establishing a relationship with an in-person dermatologist. Many newly insured patients have unmet health care needs that existing health care professionals in a community could have difficulty meeting in the near term. The dermatology visit rate among the newly insured was less than half that of the continuously enrolled. This difference is a bit surprising given that other work has noted that sicker patients (who have deferred use of services on enrollment) typically enroll first. If no barriers were associated with obtaining dermatologic care as a new patient, one would expect to see higher rates of use among newly insured patients.¹¹

The health plan’s primary goal in covering teledermatology was to increase use of dermatologic care. In the future, the aim of teledermatology interventions may be to substitute for in-person visits, thereby reducing costs (eg, travel time and costs and lost productivity for patients) from a societal perspective.^2,12

While teledermatology was being implemented in HPSJ, a substantial increase occurred in the use of in-person dermatologists. Sensitivity analyses focused exclusively on the continuously enrolled patients revealed that nonuser practices had a greater increase in dermatology visits per capita than user practices. Why this is the case is unclear. The continuously enrolled patients in this particular analysis represented 11.8% of all enrollees, and they are very different than the general Medicaid population. In general, nonuser practices may not have referred their patients to teledermatology because they observed fewer barriers to dermatologic care and/or because of new dermatologists in their system, and this possibility is reflected in the observed increase of in-person dermatology visits.

Several limitations deserve consideration. First, assignments (eg, user practices, assignment to a primary care practice) not directly observable had to be made to categorize and analyze the data. Second, this study only describes the experiences of HPSJ with teledermatology services in California, and thus how generalizable our results are to other teledermatology interventions in other regions with other health plans remains unclear. Third, HPSJ experienced a great deal of growth during the study period and as a result, enrollees had variable exposure to teledermatology, and enrollees who joined owing to the Affordable Care Act’s Medicaid expansion are likely different from a demographic perspective. Fourth, HPSJ could only provide 9 months of preteledermatology data to describe baseline trends. Fifth, the study lacks an ideal control group because all practices were eligible to use teledermatology, but only certain practices took advantage of the service. Finally, although teledermatology appears to have increased access to dermatologists, whether the use of these services observed after the introduction of teledermatology suggests adequate access is unclear because no national standard exists on appropriate dermatology use.

The clinical model implemented by this teledermatology provider may have important lessons for the field of telehealth. The model evolved from a purely virtual model to one that incorporates brick-and-mortar clinics to accommodate in-person evaluations. This novel model is more resource intensive than an exclusively virtual model, but it was successful in expanding access. Future research should assess the quality of services provided and the satisfaction of patients and referring health care professionals with this model of teledermatology compared with others. The results show that this particular model, and teledermatology in general, have promise for other Medicaid plans that struggle with meeting the specialty care needs of traditional as well as new enrollees.

The offering of teledermatology appears to improve access to care among Medicaid enrollees and played an especially important role for newly enrolled patients. This particular model, and teledermatology in general, offer promise for other Medicaid plans that struggle with meeting the specialty care needs of traditional as well as new enrollees.

Corresponding Author: Lori Uscher-Pines, PhD, RAND Corporation, 1200 S Hayes St, Arlington, VA 22202 (luscherp@rand.org).

Accepted for Publication: March 9, 2016.

Published Online: May 4, 2016. doi:10.1001/jamadermatol.2016.0938.

Author Contributions: Dr Uscher-Pines had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Uscher-Pines, Mehrotra, Mulcahy, Mehrotra.

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

Drafting of the manuscript: Uscher-Pines.

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

Statistical analysis: All authors.

Obtained funding: Uscher-Pines, Mehrotra, Mehrotra.

Administrative, technical, or material support: Uscher-Pines.

Study supervision: Uscher-Pines, Mehrotra.

Conflict of Interest Disclosures: None reported.

Funding/Support: This study was supported by the California Healthcare Foundation.

Role of the Funder/Sponsor: The funding source 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.