Repeated encounters for actinic keratoses in Medicare patients: a retrospective cohort study

Dear Editor,

Actinic keratoses (AKs) are common premalignant skin lesions (Gupta et al., 2002, Landis et al., 2014, Salasche, 2000). Per year, approximately 2 in 5 men and 1 in 5 women in Medicare have an AK encounter (Navsaria et al., 2022). Most prior studies on AKs (Yeung et al., 2018) cannot address the burden of AKs in individual patients, however, anecdotally, many patients with AKs return with new AKs at future visits, but evidence on whether AKs are chronic and intermittent is lacking.

For this retrospective cohort study, we used a random, de-identified sample of Medicare beneficiaries (2009–2018) aged ≥65 years without Medicare Advantage coverage (fee-for-service only) with continuous Parts A, B and D coverage. We included beneficiaries with at least one outpatient encounter with an AK diagnosis code (ICD-9 702.0, ICD-10 L57.0) who had at least one year of Medicare enrollment without an AK prior to their initial AK diagnosis encounter in our dataset (Cohen et al., 2020). We excluded beneficiaries with HIV or organ transplant (Table S1 and S2). This study was IRB approved (MD Anderson Cancer Center 2019–0966).

We evaluated demographics and clinical characteristics of the included beneficiaries and evaluated AK encounters, including treatments. Treatments included in-office destruction (identified by CPT codes) (Table S3), topical treatment with fluorouracil, imiquimod, ingenol mebutate, or diclofenac 3% (identified by National Drug Codes) (Table S4), photodynamic therapy (identified by CPT codes) (Table S5).

We evaluated the risks and timing of future AK encounters to determine the chronicity of AKs. Date of initial AK encounter was the start of follow-up and end of follow-up was last encounter in data set or death. First, we calculated the mean cumulative function, the average number of cumulative AK encounters at time points after initial follow-up (after initial AK). Second, we used the direct adjustment of cumulative incidence functions to calculate the risk over time of a 2^nd AK encounter after the 1^st, a 3^rd after the 2^nd, a 4^th after the 3^rd, a 5^th after a 4^th, and a 6^th after a 5^th. We adjusted for age, gender, race, and skin cancer history. Third, we used a Prentice-Williams-Peterson (PWP) gap time model (Prentice et al., 1981) with robust sandwich variance to evaluate risk factors for developing future AKs after an initial AK, adjusted for age, gender, race, skin cancer history, type of AK treatment, and Charlson comorbidity index. The PWP gap time model is similar to Cox proportional hazard regressions but incorporates all the repeated AK events over time and establishes new baseline risks for each timeframe at risk.

Of the 4,999,999 beneficiaries, we identified 850,108 beneficiaries aged ≥65 years with one year of Medicare enrollment prior to an initial AK diagnosis encounter (Table 1). Included beneficiaries had on average 4.4 (SD 2.7) years of follow-up, a mean of 3.7 (SD 4.1) AK encounters, and 711,400 (83.7%) beneficiaries had at least one dermatology encounter.

Table 1.

Demographics of patients with at least 1 AK diagnosis

Patient Characteristics (850,108 patients)	N (%)
Sex
Female	502137 (59.1)
Male	347971 (40.9)
Age at initial AK diagnosis
65–69	268707 (31.6)
70–74	191290 (22.5)
75–79	156811 (18.5)
80–84	116936 (13.8)
85–89	74929 (8.8)
90+	41435 (4.9)
Race and ethnicitya
White	809740 (95.3)
African American	2417 (0.3)
Asian	6875 (0.8)
Hispanic	16305 (1.9)
American Indian/Alaska Native	1542 (0.2)
Other	3527 (0.4)
Unknown	9702 (1.1)
Census region
Midwest	205443 (24.2)
South	349012 (41.1)
Northeast	136139(16.0)
West	158722 (18.7)
Other/unknown	792 (0.1)
Skin cancer historyb
No	615023(72.4)
Yes	235085 (27.7)
Keratinocyte carcinoma	162266(19.1)
Squamous cell carcinoma	55190 (6.5)
Basal cell carcinoma	55455 (6.5)
Melanoma	28920 (3.4)
AK Encounter Characteristics (3,114,909 encounters)	N (%)
Encounters with a dermatologist	2248866 (72.2)
Treatments
Destructionc
1 AK treated	742076 (23.8)
2–14 AKs treated	1408114 (45.2)
15+ AKs treated	210450 (6.8)
Topical medicationd
Fluorouracil	69760 (2.2)
Imiquimod	14021 (0.5)
Ingenol mebutate	1924 (0.1)
Diclofenac	1966 (0.1)
Photodynamic therapye	39073 (1.3)
No treatment identified	665146 (21.4)

AK, actinic keratosis

^aRace and ethnicity defined by CMS RTI race code – categories include Unknown, Non-Hispanic White, Black (or African American), Other, Asian/Pacific Islander, Hispanic, and American Indian/Alaska Native. (RTI, 2022)

^bSee Supplement Table S8 for definition of skin cancer history and code list. Numbers of patients with each skin cancer type sums to fewer than the total number of patients with skin cancer history as some skin cancer history codes are not specific to a cancer type.

^c1 AK treated – CPT 1700; 2–14 AKs treated – CPT 17003; 15+ AKs treated – CPT 17004

^dIdentified by National Drug Codes on the same date as an AK diagnosis code or within the following 14 days

^ePhotodynamic therapy identified by CPT codes on the same date as an AK diagnosis code (see Supplement for code list)

Figure 1a shows the cumulative mean number of AK encounters (excluding initial AK encounter) was 0.69 (95% CI 0.69–0.69) at 1 year after initial AK diagnosis, 1.77 (95% CI 1.77–1.78) at 3 years, and 2.91 (95% CI 2.90–2.92) at 5 years. When evaluated by region, the cumulative mean number of AK encounters ranged from 0.68 (Midwest and South) to 0.75 (Northeast) at 1 year after initial AK diagnosis and from 2.75 (Midwest) to 3.07 (West) at 5 years (Table S6).

Figure 1: Mean cumulative function and Adjusted cumulative incidence curves.

a. shows the mean cumulative function, the average number of cumulative AK encounters experienced by patients in the dataset at time points after the start of follow-up (after initial AK)

b. The incidence of subsequent AK encounters for the first five AK intervals in our study (the incidence of the 2^nd AK encounter after having a 1^st, the incidence of the 3^rd after a 2^nd, of a 4^th after a 3^rd, and a 5^th after a 4^th). These are adjusted for age, gender, race, and skin cancer history.

Figure 1b shows adjusted cumulative incidence curves, which plot the incidence of subsequent AK encounters for the first five AK intervals in our study (the incidence of the 2^nd AK encounter after having a 1^st, the incidence of the 3^rd after a 2^nd, of a 4^th after a 3^rd, and a 5^th after a 4^th). Using the PWP gap time model, we observed statistically significant decreased risk of future AK encounters with increasing age (HR 0.928 per 10-year increase in age, 95% CI 0.925–0.932) and worsening Charlson comorbidity index (HR 0.997 per 1-unit increase, 95% CI 0.995–0.999), and increased risk with male gender (HR 1.36, 95% CI 1.35–1.37), White race (HR 1.15, 95% CI 1.14–1.17), history of skin cancer at initial AK diagnosis (HR 1.18, 95% CI 1.17–1.19), and receiving topical AK treatment (topical medication or photodynamic therapy; HR 2.30, 95% CI 2.27–2.33), adjusted for geographic region (Table S7).

We report a cohort of more than 850,000 Medicare beneficiaries. Patients aged ≥65 years who were seen for AKs would be seen nearly twice more for AKs in the following three years, for approximately one AK encounter per year. This provides evidence for the chronic, intermittent nature of AKs in individuals. We observed the time to next AK encounter decreased with each subsequent AK encounter. Additionally, our results showed differences in rates of AK encounters between regions in multivariable analysis. Patients with an increased risk of future AK encounters were more likely to be younger, healthier, male, white, and have history of skin cancer.

These findings are noteworthy, as there are currently no standardized guidelines on the length of follow-up for patients with AKs (Work et al., 2018). While interventions such as regular dermatology follow-up and sun protection education (Naylor et al., 1995, Thompson et al., 1993), should be regularly discussed with all AK patients, knowledge of risk factors and information on the likelihood of needing continued care for AKs over time can be a source of additional motivation for long-term behavioral changes.

Our study has limitations. This sample of fee-for-service Medicare beneficiaries aged ≥65 years may not be nationally representative, especially of patients ≤65. Claims data relies on the accuracy of diagnostic codes and, though we used a validated approach (Cohen et al., 2020), it is possible that not all of the AK encounters represent new incident AKs. Additionally, claims data do not include all patient characteristics.

AKs affect a significant portion of US adults aged ≥65 years and are a considerable healthcare burden. Understanding the patterns of AK encounters and course of disease can help guide clinical care for patients with AKs and can support future AK research.

Data Availability Statement:

Datasets related to this article are available for request under a data use agreement with CMS and can be found at resdac.org, hosted at the Research Data Assistance Center on behalf of the Centers for Medicare and Medicaid Services. More information on ‘Data Disclosures and Data Use Agreements’ for CMS data can be found at https://www.cms.gov/Research-Statistics-Data-and-Systems/Files-for-Order/Data-Disclosures-Data-Agreements/Overview.