BACKGROUND: Expenditures on specialty medications for autoimmune conditions (SpRx-AIC) have increased considerably in recent years, raising affordability concerns for employers and other plan sponsors and resulting in greater patient cost-sharing. Among those commercially insured, prior studies have shown differential patterns of health care utilization in association with wage, though no data are available for SpRx-AIC. Notably, out-of-pocket costs associated with SpRx-AIC have been shown to impact medication adherence, particularly for low-income households.

OBJECTIVE: To assess the association of wage status on SpRx-AIC and health care services use and cost among employees with employer-sponsored health insurance.

METHODS: Employee health care claims and wage data were obtained from the IBM Watson MarketScan database for calendar year 2018. Midyear employee wage data were used as a basis for allocating employees into annual income quartiles: $47,000 or less, $47,001-$71,000, $71,001-$106,000, and $106,001 or more. The lowest quartile was further divided into 2 groups ($35,000 or less and $35,001-$47,000) to better evaluate subgroup differences at lower wage levels. Outcomes included monthly days supply of SpRx-AIC, medication discontinuation rates (medication cessation for ≥ 90 days), proportion of days covered (PDC), medical services utilization rates per 1,000, and allowed payment amounts. Generalized linear regressions were used to assess differences while adjusting for patient characteristics, including age, gender, plan type, region, median household income, deductible amount, comorbidity index, and psychiatric diagnostic scores.

RESULTS: From a sample of more than 2 million enrollees, 148,761 (7.2%) were identified as having an autoimmune disorder of interest. Of those, 17,096 (11.5%) had filled at least one SpRx-AIC prescription. Following adjustment, SpRx-AIC use was significantly less among the lowest wage group compared with the highest wage group (10.1% vs 11.7%; P < 0.0001). Days supply was significantly lower in the lowest wage group (244.4 vs 258.0; P < 0.001), as was PDC (0.74 vs 0.76; P < 0.001). In the lowest wage group, medical services utilization was significantly higher for inpatient admissions (0.08 vs 0.05; P = 0.002) and emergency department visits (0.52 vs 0.16; P < 0.0001). There were no significant differences among wage groups in SpRx-AIC discontinuation, outpatient services use, or health care costs.

CONCLUSIONS: Low-wage employees with autoimmune conditions are significantly less likely to use an SpRx-AIC and have a lower monthly supply and PDC when SpRx-AIC was used. They are more likely to be admitted to the hospital and have more emergency department visits. These findings raise concerns about employer benefit design inequities for SpRx-AIC access and the resulting potential adverse impact on health care costs and employee functional status.

DISCLOSURES: National Pharmaceutical Council, Genentech, and TrialCard provided funding support for this study, with funding administered by the National Alliance of Healthcare Purchaser Coalitions. Genentech and TrialCard provided comments regarding the final manuscript draft; National Pharmaceutical Council employees were actively engaged in study design, analysis and interpretation of results, and manuscript preparation.

Dr Sherman is a consultant to National Alliance of Healthcare Purchaser Coalitions. Mr Sils and Ms Westrich were employees of the National Pharmaceutical Council at time of study. Ms Kamen is an employee of IBM Watson Health.

Plain language summary

Some drugs used to treat conditions such as asthma and arthritis can be expensive. Our research shows that low-wage employees are less likely to be taking one of these drugs than employees who earn more. These results highlight inequalities in health care services that need attention. Employers can use the results of this study to improve health benefits design. Medical staff can use these results to make sure all patients receive high-quality care.

Implications for managed care pharmacy

Our study findings highlight the existence of significant wage-based differences in specialty pharmaceutical (SpRx) utilization and adherence. Additional research can help to identify more explicitly the causes of reduced SpRx use among low-wage workers, including treatment bias that may exist in clinician prescribing practices, as well as reasons for patient abandonment and medication discontinuation. With improved understanding, appropriate benefits design and patient support can help to improve benefits equity in relation to SpRx use.

Since their emergence as a pharmacotherapy option in the 1980s, the availability and indications for use of biologic specialty pharmaceuticals (SpRx) have markedly expanded. More than 600 are currently US Food and Drug Administration approved for use in humans,1 and the pace of SpRx development continues to accelerate.2 For individuals with autoimmune disorders (including rheumatoid arthritis, psoriasis, psoriatic arthritis, multiple sclerosis, inflammatory bowel disease, and asthma), these treatments offer the potential for modifying disease progression and in nearly all settings represent a substantial improvement over earlier symptomatic treatment options.

The role and value of SpRx in treatment of autoimmune conditions is well established.3,4 Their ability to mitigate disease progression is associated with preservation of functional status, reduced need for surgical intervention, and reduced likelihood of employment-related disability.3,4 Improved treatment outcomes have allowed affected individuals to remain functional and productive members of their respective communities.

Yet these benefits come at an appreciable and growing cost. Over the past 10 years, SpRx expenditures for autoimmune disorders have risen from $3.60 to $24.60 per member per month5,6 owing to new medications, expanded indications, and updated treatment guidelines,7 with current trends anticipated to moderate as biosimilars become increasingly available.8 In 2020, SpRx spending accounted for 53% of total pharmaceutical spending in the United States, with SpRx treatment of autoimmune disorders representing 14.3% of total pharmaceutical expenditures.8 Concerns regarding ongoing affordability of these SpRx and other clinical services have prompted employers and other plan sponsors to increase cost-sharing through plan design, with use of consumer-directed health plans, higher deductible thresholds, copay accumulator adjustment programs, or other means.9

For many employed individuals, and particularly those in lower-income categories, prior studies have illustrated differential patterns of health care utilization in association with wage,10,11 at least some of which are likely driven by affordability concerns. Additionally, out-of-pocket costs associated with SpRx use for autoimmune disorders have also been shown to impact medication adherence, particularly for low-income households.12,13 As an expansion of previous research to characterize broad-based health care utilization and costs by employee wage category,10 the goal of this study was to characterize utilization and cost patterns of SpRx for treatment of autoimmune disorders among employees in different wage categories, with the expectation that results can inform policy and benefit design considerations regarding SpRx access.

DATA AND SAMPLE

This study examined the specialty drug use patterns among 2,071,980 active employees working at least 30 hours per week who were enrolled in employer-sponsored health insurance coverage for the entire 12 months of 2018 through a self-insured health plan, and for whom wage data were available. Our research focus was limited to conditions in which specialty drugs were used on a chronic basis to permit evaluation of adherence and discontinuation rates. Doing so resulted in exclusion of conditions such as cancer, hepatitis, and other disorders for which specialty medications are often used, but for shorter and/or variable periods. As a result, the study population was limited to individuals with one of the following conditions commonly treated with a specialty medication: rheumatoid arthritis, atopic dermatitis, multiple sclerosis, psoriasis (plaque psoriasis and psoriatic arthritis), Crohn’s disease, and asthma. International Classification of Diseases, Tenth Revision, Clinical Modification principal diagnosis codes (Supplementary Table 1, available in online article) from the medical claims data were used to identify these populations. A total population defined as having at least one of these conditions was also included.

The research dataset was derived from the IBM Watson Health MarketScan Database, which consists of deidentified outpatient, inpatient, and pharmaceutical claims of approximately 40-50 million privately insured patients each year. These claims originate from more than 150 large employers offering self-insured health insurance coverage for eligible enrollees, with coverage in all 50 states. The database includes patient demographics, wage data, benefits eligibility, enrollee-specific health benefits design detail, and medical and pharmacy claims data. Claims data include actual (reimbursed) payment amounts and health plan and employee spending. Of note, third-party support for patient out-of-pocket health care expenses, including copay assistance, is included as an employee payment and is not separately differentiated. These detailed data permit analysis and reporting of specialty medication use through either the pharmacy or medical benefits.

Medications used to treat any of the identified autoimmune conditions were identified from the IBM Watson Micromedix medication list and then split into 2 categories for the purpose of this analysis. First, we defined “All Medications” to include any that could be used in the treatment of any of the autoimmune conditions of interest. Second, we defined the subset of “Specialty Medications” to include all biologic specialty medicines for these conditions as defined in the US Food and Drug Administration Purple Book.1 Full lists of specialty and nonspecialty medications included in the analysis are shown in Supplementary Table 2.

Additionally, in order to understand the significance of specialty medications available through the pharmacy plan relative to the medical plan, we also evaluated the comparative use of these 2 delivery channels among the study population.

WAGE COHORT AND CLINICAL DATA DEFINITIONS

Midyear employee wage data were used as a basis for allocating employees into annual income quartiles: $47,000 or less, $47,001-$71,000, $71,001-$106,000, and $106,001 or more. The lowest quartile was further divided into 2 groups ($35,000 or less and $35,001-$47,000) to enable a better understanding of differences in health care use at lower wage levels. We examined the association of wage category and use of medications for treatment in aggregate for all autoimmune conditions (either all medications or specialty drugs only). Outcomes included the likelihood of treatment with a specialty medication, number of days supplied per patient, adherence to specialty medications (using a proportion of days covered [PDC] metric), specialty medication discontinuation rate (defined as a period of 90 or more days without specialty medication), and the utilization of medical care services (emergency department [ED] use, outpatient use, and inpatient admission rates). Costs associated with overall medication and specialty medication use for autoimmune conditions, as well as medical care services use, were also included in the analysis, in aggregate for all autoimmune conditions and for each specific autoimmune disorder (full definitions in Supplementary Table 3).

STATISTICAL ANALYSIS

As described in a prior analysis,10 we regressed each utilization measure using wage group indicators and covariates: age and sex categories, ZIP code–based median household income,14 geographic census region, health plan contract type (individual, individual plus spouse, individual plus child, individual plus children, or family), net deductible as a percentage of annual wages, comorbid condition prevalence (Charlson Comorbidity Index15 and Psychiatric Diagnostic Groupings16), an indicator for being a salaried employee, an indicator for being part of a union, and an indicator for living in a rural area (full definitions in Supplementary Table 4). Notably, with significant variation in out-of-pocket costs within specific plan design types (for example, a $6,000 deductible consumer-directed health plan vs a $1,500 deductible consumer-directed health plan), net deductible as a percentage of annual wages was used instead of health plan type to more explicitly control for employee-level health care affordability.

Continuous outcomes were modeled using generalized linear models with a Gaussian family, and binary outcomes were modeled using generalized linear models with a logit link. In the descriptive and multivariate analyses, we compare each of the 4 lower wage bands with the greater than $106,000 wage band, in which access issues to SpRx were least likely to be evident. All wage comparisons were based on α = 0.0125 (the Bonferroni correction for multiple comparisons, a conservative approach).

EMPLOYEE CHARACTERISTICS

From an initial sample of more than 2 million employees, 148,761 (7.2%) were identified as having an autoimmune disorder of interest. Table 1 shows that employees within lower wage groups were significantly more likely to have a diagnosed autoimmune condition than those in the highest wage group (P < 0.0001).

Table

TABLE 1 Proportion of Employees With Autoimmune Diagnosis and Use of Specialty Medication by Annualized Wage Category

TABLE 1 Proportion of Employees With Autoimmune Diagnosis and Use of Specialty Medication by Annualized Wage Category

Annualized wage band $35,000 and under $35,001-$47,000 $47,001-$71,000 $71,001-$106,000 $106,001 and over Total
Employees, N 246,611 257,016 531,919 518,215 518,219 2,071,980
With an autoimmune diagnosis
  N 17,504 18,677 39,149 37,770 35,661 148,761
  Unadjusted mean, % 7.1 7.3 7.4 7.3 6.9 7.2
  Predicted mean, % (95% CI) 6.7 (6.5-6.8) 6.7 (6.6-6.8) 6.6 (6.5-6.7) 6.6 (6.5-6.7) 6.2 (6.2-6.3)
  P value < 0.0001a < 0.0001a < 0.0001a < 0.0001a
With a specialty drug (of those with an autoimmune diagnosis)
  N 1,550 1,880 4,378 4,711 4,577 17,096
  Unadjusted mean, % 8.9 10.1 11.2 12.5 12.8 11.5
  Predicted mean, % (95% CI) 10.1 (9.5-10.7) 11.0 (10.5-11.6) 11.8 (11.4-12.2) 12.3 (12.0-12.7) 11.7 (11.2-12.1)
  P value < 0.0001a 0.06 0.74 0.007a

Authors’ analysis of 2018 health care administrative data from the IBM Watson Health MarketScan database. Predicted values adjusting for all covariates.

aDenotes a statistically significant difference between a wage category and the category of $106,001 and over (P < 0.0125).

Of those employees who had an autoimmune condition, 17,096 (11.5%) had filled at least one prescription for an SpRx through the pharmacy benefit, and 2,202 (1.5%) had filled at least one prescription for an SpRx via the medical benefit, with 616 (0.3%) receiving at least one prescription for an SpRx via both the medical and pharmacy benefit. As there were no significant differences found in SpRx use among wage groups through the medical benefit and the relative number of employees using an SpRx through the medical benefit was low compared with the pharmacy benefit, subsequent analyses address only specialty medications filled through the pharmacy benefit.

The likelihood of having a specialty medication filled through the pharmacy benefit was significantly less for employees in the lowest wage group compared with those in the highest wage group (10.1 vs 11.7%; P < 0.0001), despite an increased proportion of employees having a relevant autoimmune condition.

Detailed demographics information regarding employees with autoimmune conditions taking SpRx is shown in Table 2. Employees in the lower wage categories were younger, more likely to be female, and more likely to be hourly workers in comparison with employees in higher wage groups. Employees with lower wages were more than twice as likely to be enrolled in an employee-only benefit plan compared with employees in higher wage groups and were nearly 4 times more likely to pay deductibles as a greater proportion of wages (1.6% of wages or greater) than employees in the highest wage group. In addition, employees in the lowest wage category had significantly higher scores on both the Charlson Comorbidity Index and Psychiatric Diagnostic Groups than employees in the highest wage group, indicating a higher disease burden of both chronic physical and psychiatric illnesses.

Table

TABLE 2 Employee Demographic Characteristics by Annualized Wage Category

TABLE 2 Employee Demographic Characteristics by Annualized Wage Category

Annualized wage band $35,000 and under $35,001-$47,000 $47,001-$71,000 $71,001-$106,000 $106,001 and over
Employees, N 1,550 1,880 4,378 4,711 4,577
Age group, mean
  Males, age, %
    18-30 4.26 5.43 4.13 3.61 1.03
    31-40 7.74 8.94 10.67 9.30 10.18
    41-50 9.29 10.37 11.47 15.03 18.11
    51-64 12.00 12.02 17.63 24.64 32.05
  Females, age, %
    18-30 7.55 7.39 4.91 3.23 0.68
    31-40 13.16 12.87 9.96 8.36 6.12
    41-50 17.87 15.80 16.10 13.48 12.87
    51-64 28.13 27.18 25.13 22.35 18.96
  Chi-square P value < 0.0001a < 0.0001a < 0.0001a < 0.0001a
Salary status, mean
  Salaried, % 23.35 17.45 32.96 51.73 79.33
  Hourly, % 61.16 69.47 53.56 35.60 10.40
  Unknown, % 15.48 13.09 13.48 12.67 10.27
  Chi-square P value < 0.0001a < 0.0001a < 0.0001a < 0.0001a
Coverage tier, mean
  Employee only, % 56.19 49.04 38.72 36.38 26.17
  Employee + spouse, % 11.61 11.97 13.89 15.60 16.12
  Employee + child, % 3.87 5.48 3.86 2.61 3.36
  Employee + children, % 6.90 7.45 5.53 6.43 7.52
  Family, % 15.42 20.59 26.56 32.84 42.98
  Chi-square P value < 0.0001a < 0.0001a < 0.0001a < 0.0001a
Region, mean
  South, % 47.29 42.45 40.32 35.09 47.29
  Northcentral, % 29.16 34.84 34.76 24.69 20.08
  Northeast, % 9.87 13.14 14.53 22.18 26.17
  West, % 13.68 9.57 10.39 18.04 24.86
  Chi-square P value < 0.0001a < 0.0001a < 0.0001a < 0.0001a
Union status, mean
  Union, % 15.87 19.47 30.17 28.30 9.53
  Not union, % 68.84 70.59 61.01 62.28 74.61
  Unknown, % 15.29 9.95 8.82 9.42 15.86
  Chi-square P value < 0.0001a < 0.0001a < 0.0001a < 0.0001a
Area median income, mean
  $68,761 and under, % 36.39 35.80 32.28 21.78 13.68
  $68,762-$80,700, % 25.35 30.16 29.31 23.94 20.19
  $80,701-$96,330, % 18.71 23.88 23.94 27.91 25.41
  Over $96,331, % 19.55 10.16 14.48 26.36 40.73
  Chi-square P value < 0.0001a < 0.0001a < 0.0001a < 0.0001a
Deductible as percentage of wage, mean
  0%-0.7% 0.71 7.71 17.34 32.22 43.74
  0.7%-1.6% 13.29 22.18 19.12 12.38 15.91
  1.6%-3.2% 58.06 40.16 27.62 20.31 15.75
  Over 3.2% 4.90 0.00 0.00 0.00 0.00
  Missing, % 23.03 29.95 35.93 35.09 24.60
  Chi-square P value < 0.0001a < 0.0001a < 0.0001a < 0.0001a
Charlson Comorbidity Index, mean 0.78 0.73 0.70 0.69 0.61
  P value < 0.0001a < 0.0001a < 0.0001a < 0.0001a
Psychiatric diagnostic groups, mean 0.10 0.07 0.07 0.07 0.05
  P value < 0.0001a 0.16 0.004a 0.037a
Rural indicator, mean, % 90.77 87.39 90.77 93.42 96.68
  Chi-square P value < 0.0001a < 0.0001a < 0.0001a < 0.0001a

Authors’ analysis of 2018 health care administrative data from the IBM Watson Health MarketScan database.

aDenotes a statistically significant difference between a wage category and the category of $106,001 and over (P<0.0125).

PATTERNS OF MEDICATION USE

Table 3 details the patterns of specialty medication use following adjustment for demographic and health variables. The PDC was significantly lower for employees in the 2 bottom wage groups (0.74) compared with those in the highest wage group (0.76; P ≤ 0.01). Days supply of specialty medication was also significantly less (244.4) for employees in the lowest wage group compared with those in the highest wage group (258.0; P = 0.001). Although the 90-day discontinuation rate trended higher among the lowest wage group (0.32) than the highest wage group (0.29), it was not statistically significant.

Table

TABLE 3 Medication Usage by Annualized Wage Category

TABLE 3 Medication Usage by Annualized Wage Category

Annualized wage band $35,000 and under $35,001-$47,000 $47,001-$71,000 $71,001-$106,000 $106,001 and over
Proportion of days covered specialty prescriptions
  N 1,550 1,880 4,378 4,711 4,577
  Unadjusted mean 0.73 0.74 0.75 0.76 0.77
  Predicted mean (95% CI) 0.74 (0.72-0.75) 0.74 (0.73-0.76) 0.75 (0.74-0.76) 0.75 (0.75-0.76) 0.76 (0.76-0.77)
  P value (relative to over $106,000) 0.0007a 0.01a 0.02 0.12
Medication discontinuation of specialty prescriptions
  N 1,550 1,880 4,378 4,711 4,577
  Unadjusted mean 0.34 0.33 0.31 0.30 0.27
  Predicted mean (95% CI) 0.32 (0.29-0.35) 0.31 (0.29-0.33) 0.30 (0.29-0.32) 0.30 (0.29-0.31) 0.29 (0.27-0.30)
  P value (relative to over $106,000) 0.04 0.12 0.13 0.14
All prescriptions days supply
  N 1,550 1,880 4,378 4,711 4,577
  Unadjusted mean 1,369.5 1,328.5 1,343.9 1,310.0 1,204.3
  Predicted mean (95% CI) 1,337.6 (1,287.7-1,387.4) 1,334.0 (1,289.8-1,378.2) 1,331.8 (1,303.6-1,359.9) 1,308.7 (1,282.4-1,335.1) 1,225.8 (1,194.9-1,256.6)
  P value (relative to over $106,000) 0.0005a 0.0003a < 0.0001a < 0.0001a
Specialty prescriptions days supply
  N 1,550 1,880 4,378 4,711 4,577
  Unadjusted mean 237.6 244.2 250.0 257.0 264.1
  Predicted mean (95% CI) 244.42 (237.91-250.94) 250.77 (245.00-256.54) 252.38 (248.70-256.05) 255.89 (252.44-259.33) 257.96 (253.92-261.99)
  P value (relative to over $106,000) 0.001a 0.06 0.06 0.43

Authors’ analysis of 2018 health care administrative data from the IBM Watson Health MarketScan database. Predicted values adjusting for all covariates.

aDenotes a statistically significant difference between a wage category and the category of $106,001 and over (P < 0.0125).

MEDICAL CARE SERVICES UTILIZATION

Table 4 illustrates all-cause health care services use for common service categories, including ED visits, inpatient admissions, and outpatient services. The all-cause use of ED services was highest among employees in the lowest wage group and was significantly higher than those in the highest wage group (238 visits per 1,000 patients vs 134 visits per 1,000 patients; P < 0.0001). ED service usage was also significantly higher among employees in all other lower wage categories than those in the highest wage group (P ≤ 0.001).

Table

TABLE 4 Medical Services Utilization Patterns by Annualized Wage Category

TABLE 4 Medical Services Utilization Patterns by Annualized Wage Category

Annualized wage band $35,000 and under $35,001-$47,000 $47,001-$71,000 $71,001-$106,000 $106,001 and over
Inpatient admissions per 1,000 patients
  N 1,550 1,880 4,378 4,711 4,577
  Unadjusted mean 134 112 112 89 75
  Predicted mean (95% CI) 78 (65-94) 75 (63-88) 71 (63-79) 60 (54-67) 53 (46-60)
  P value (relative to over $106,000) 0.002a 0.003a 0.002a 0.12
Outpatient services per 1,000 patients
  N 1,550 1,880 4,378 4,711 4,577
  Unadjusted mean 67,979 63,216 63,041 62,000 59,539
  Predicted mean (95% CI) 65,411 (62,453-68,368) 62,648 (60,028-65,268) 62,655 (60,986-64,324) 61,753 (60,188-63,318) 61,265 (59,434-63,096)
  P value (relative to over $106,000) 0.03 0.43 0.30 0.68
Emergency department visits per 1,000 patients
  N 1,550 1,880 4,378 4,711 4,577
  Unadjusted mean 520 377 293 228 160
  Predicted mean (95% CI) 238 (215-262) 212 (193-232) 178 (166-190) 160 (150-171) 134 (122-146)
  P value (relative to over $106,000) < 0.0001a < 0.0001a < 0.0001a 0.001a

Authors’ analysis of 2018 health care administrative data from the IBM Watson Health MarketScan database. Predicted values adjusting for all covariates.

aDenotes a statistically significant difference between a wage category and the category of $106,001 and over (P < 0.0125).

Inpatient admission rates were also highest among employees in the lowest wage group (78 admissions per 1,000 patients) and significantly higher than those in the highest wage group (53 admissions per 1,000 patients; P = 0.002). Inpatient admission rates were also significantly higher among employees in the $35,001-$47,000 and $47,001-$71,000 wage categories compared with those in the highest wage group (P ≤ 0.003). Analysis of autoimmune condition-specific admission rates by wage category could not be performed owing to insufficient sample size.

We observed a trend toward more outpatient services use by employees in the lowest wage group (65,411 per 1,000 patients) compared with those in the highest wage group (61,265 per 1,000 patients), but findings were nonsignificant (P = 0.03).

HEALTH CARE SPENDING PATTERNS

Spending on overall medical services, prescriptions, and total health care did not reveal any statistically significant relationships (Table 5). However, employees in the lowest wage group trended toward reduced prescription spending ($47,918) and increased medical spending ($14,398) compared with those in the highest wage group ($49,635 and $12,242, respectively).

Table

TABLE 5 Health Care Cost Data for Employees by Annualized Wage Category

TABLE 5 Health Care Cost Data for Employees by Annualized Wage Category

Annualized wage band $35,000 and under $35,001-$47,000 $47,001-$71,000 $71,001-$106,000 $106,001 and over
Total allowed amount per patient
  N 1,550 1,880 4,378 4,711 4,577
  Unadjusted mean, $ 61,326 60,699 61,320 62,038 63,049
  Predicted mean (95% CI), $ 62,316 (59,837-64,795) 62,052 (59,856-64,247) 62,064 (60,665-63,462) 61,621 (60,309-62,932) 61,877 (60,343-63,411)
  P value 0.75 0.89 0.85 0.77
Medical allowed amount per patient
  N 1,550 1,880 4,378 4,711 4,577
  Unadjusted mean, $ 15,011 12,865 12,458 12,532 11,858
  Predicted mean (95% CI), $ 14,398 (12,704-16,093) 12,663 (11,162-14,165) 12,463 (11,507-13,419) 12,437 (11,540-13,334) 12,242 (11,193-13,291)
  P value 0.02 0.63 0.74 0.75
Prescription allowed amount per patient
  N 1,550 1,880 4,378 4,711 4,577
  Unadjusted mean, $ 46,316 47,834 48,862 49,506 51,191
  Predicted mean (95% CI), $ 47,918 (46,075-49,761) 49,389 (47,757-51,022) 49,601 (48,561-50,641) 49,184 (48,209-50,159) 49,635 (48,494-50,776)
  P value 0.10 0.80 0.96 0.49

Authors’ analysis of 2018 health care administrative data from the IBM Watson Health MarketScan database. Predicted values adjusting for all covariates.

Employer health benefit strategies have had a largely unintended and significant adverse impact on employees with lower wages. Traditionally, benefit designs have included a set of health plan options available to eligible employees irrespective of wage, without consideration of employee cost burden. In 2020, only 7% of employers offered a financial subsidy for employees with lower wages,9 despite the fact that current health benefit offerings can reach 21% or more of employee earnings.10 Paradoxically, this subpopulation generally has the greatest prevalence of unhealthy lifestyle behaviors and chronic conditions,17 resulting in greater need for and use of health care services and prescription medications.

In our study of commercially insured adults, we found evidence that these health benefit strategies impacted both specialty medication access and usage as well as health care services utilization for employees in lower wage categories. Despite having a greater prevalence of autoimmune conditions, employees in the lowest wage category were nearly 14% less likely to have a specialty medicine to treat those conditions than employees in the highest wage group. When employees in low-wage categories did use specialty medicines, they had lower days supply and lower PDC. Compared with their higher-earning colleagues, employees in low-wage categories also had significantly higher all-cause inpatient admissions and ED visits. These observed income-associated differences in SpRx use may have a detrimental impact on employee functional status and health care utilization patterns.

There are likely numerous reasons for the observed disparity in SpRx access and usage, with affordability of SpRx a likely significant contributor. In the setting of employer-sponsored insurance, coverage policies for specific medications vary by pharmacy benefit plan and may well be driven by pharmacy benefit manager/manufacturer agreements and rebates. For employers, the usual model of pharmacy benefit design includes at least 1 if not 2 specialty medication tiers, with preferred vs nonpreferred tiering often based on pharmacy benefit manager/manufacturer contracting and generally subject to coinsurance, which generally results in greater cost-sharing and out-of-pocket expense. Low-wage employees enrolled in high-deductible health plans may opt to forgo more costly treatment options in favor of less expensive alternatives or other, more immediate social needs, such as housing or food.11 Low income has been identified as a distinct risk factor for suboptimal adherence and greater SpRx discontinuation rates.13,18 Additionally, cost-related abandonment, or the failure of a patient to purchase and receive the initial SpRx prescription because of excessive out-of-pocket cost,8 is a possibility that cannot be evaluated in our claims-based analysis and may well have contributed to the observed reduction in SpRx utilization rates among employees with low wages.

In addition to the observed differences in SpRx access and utilization, health care services utilization patterns also varied by wage category. The increased ED use and hospitalization rates noted among employees with lower wages were consistent with our prior findings of employees with commercial insurance,10 suggesting that the tendency toward more reactive health care use may be common both in broad populations and among individuals with autoimmune conditions.

Of note, presenting autoimmune disease severity may also be worse among individuals in lower socioeconomic groups,19-21 suggesting one plausible explanation for the increased inpatient admissions and ED visits by employees in lower wage groups observed in our results. This finding may be attributed to delayed diagnosis, health care affordability and/or access concerns, suboptimal patient compliance with recommended care, access barriers to care, cultural beliefs, health literacy, disease-related impairment in earnings potential, or other reasons.19-21 This literature suggests a potential justification for disproportionately greater SpRx use among workers with lower wages, yet the opposite was seen in our study population.

Interestingly, although the use of ambulatory care services trended to be greater among employees with lower wages, it did not reach statistical significance (P = 0.03). This observation is in contrast to our previous research findings of reduced ambulatory care services use among workers with lower wages in comparison with their higher-earning counterparts,10 suggesting that perhaps suboptimal symptom control of autoimmune conditions may drive greater ambulatory services use.

Although there were no significant differences in spending patterns between wage groups, there was a trend in the lowest wage group toward greater medical spending and reduced prescription spending relative to the highest wage group. This trend suggests the possibility that greater use of specialty medication in the lowest wage group could potentially reduce the increased hospital admissions and ED use seen in this group. Although total health care spending might remain unchanged, the potential for improved disease control could potentially result in a reallocation of resources, resulting in less use of hospital-based care in favor of more ambulatory services use.

CLINICAL IMPLICATIONS

Clinician bias may influence SpRx prescribing patterns for patients with autoimmune conditions. In a study of rheumatology clinic patients, low socioeconomic status was associated with worse patient functional status at presentation, which declined more rapidly over time despite patient access to care.21 Reduced SpRx prescribing has also been reported in Medicaid enrollees relative to individuals enrolled in commercial insurance.22 Medicare Advantage patients with low personal income or living in low ZIP code-based household income areas had lower prescribing rates of disease-modifying antirheumatic drugs.23

Some of these findings may be a consequence of structural racism, which may also contribute to health inequities.24 Race-related differences in SpRx use for autoimmune disorders have been summarized in recent reviews, including socioeconomic status, access and health care delivery system-level factors for rheumatoid arthritis,25 multiple sclerosis,26 and asthma,27 with an additional single-center research study among patients with psoriasis.28 Additional analyses can help to identify and effectively address the root causes of these observed race-related health inequities.

For health care practitioners, a better understanding of sources of treatment bias in relation to prescribing patterns may be helpful. Equipped with this additional information, both clinicians and policymakers can more readily identify where practice or policy changes may be most beneficial in enhancing the equitable delivery of health care for patients with autoimmune disorders.

RESEARCH IMPLICATIONS

Study results point to the need for additional research approaches to address remaining knowledge gaps and to overcome the limits of a commercial claims-based analysis. More data are needed on the causes of the reduced supply of SpRx seen in this study, including evaluation of racial bias that may exist in clinician prescribing practices, as well as reasons for patient abandonment and medication discontinuation. Additional outcomes data regarding the impact of reduced SpRx use on employee work performance and other work-related outcomes are also lacking.

Alternative research approaches are also needed because of the likely underrepresentation of employees with lower wages in our dataset. Employees in lower wage groups likely have inconsistent insurance coverage through employers if they have any at all, work more part-time positions without insurance, have a higher proportion of those who are uninsured or underinsured, and may seek more care through the public safety net outside of the typical insurance structure. As this analysis focused on employees with low wages with employer-sponsored health insurance, the magnitude of the disparity in access to care may be significantly larger when accounting for the larger pool of workers with lower wages.

POLICY IMPLICATIONS

From a policy perspective, the observed findings of lower SpRx use among low-wage workers raises further questions about the ability of the US health care system to achieve the aspirational goal of equitable access to health care. Although these data do not provide a definitive explanation for the observed findings, they highlight an apparent inequity and the need for a better understanding of the primary contributors to lower SpRx use among low-income individuals.

Employers who are committed to addressing systemic workplace disparities should consider how their benefit design can be a used as tool for improving access and providing equitable care to their employees. Considerations include providing multiple health plan options, wage-based benefit design with premium and/or Health Savings Account subsidies, predeductible coverage of medications for chronic conditions, and eliminating copay accumulator adjustment programs, in addition to fostering employee benefits literacy to promote selection of optimal plan design.

LIMITATIONS

This study has several limitations. First, in relation to the study population, only self-insured employers that provided IBM Watson Health with health care claims data and employee wage information were included. Although the study size was large, the results may not be generalizable to all employers.

Second, the study did not adjust for differences within disease characteristics between the different wage groups, given the limitations of the claims dataset. Additionally, patterns of condition-specific health care services use could not be statistically evaluated owing to comparatively small sample sizes of some individual conditions.

Third, individual wages are not a necessarily reflective of combined household income, which may have confounded the observed results. As in previous research,10 we controlled for ZIP code–level median household income using imputed data from the 2018 American Community Survey to mitigate this concern. We also adjusted for health plan enrollment category (single, employee plus spouse, and employee plus family) to further account for differences in household income.

Fourth, wage may also be a reflection of other individual characteristics such as education, literacy, and social vulnerability. Additionally, the interrelationship between wage and race/ethnicity is complex, making it possible that racial and ethnic bias may confound the observed findings. Therefore, our conclusions may be driven by factors other than wage. However, wage is a readily accessible measure available for analysis. Future research will help to clarify the relative contributions of these other factors.

Fifth, the observed results may not be generalizable to other conditions not addressed in this analysis. The unique therapeutic benefit and timing of clinical impact for each specialty drug may have resulted in different adherence patterns.

Sixth, third-party copay assistance, such as pharmaceutical manufacturer copay cards, was not included in the research dataset and may have impacted our interpretation of the results. However, given the widespread use of copay support, we do not believe the results were meaningfully impacted by the lack of these data. Additional research is underway to clarify this consideration.

Our use of claims data for this analysis precluded us from determining whether the observed patterns in SpRx use and discontinuation were a function of patient abandonment or discontinuation of provided prescriptions, a function of variable provider prescribing based on patient wage category, or a consequence of shared decision-making between patients and clinicians regarding SpRx use. It is possible that race, ethnicity, or socioeconomic bias may have also influenced physician prescribing behaviors and biased some clinicians against use of SpRx based on their perception of patient sociodemographic details. Insight into the relative significance of these contributors will necessitate use of other data sources, which were not available for this analysis.

Reduced specialty medication access and usage among employees with low wages appears to represent a significant inequity in employee-sponsored health benefits. Employees with low wages also use more high-intensity health care services, perhaps indicating poorer condition control and worse functional status than higher-earning employees. More equitable health benefit design may help to mitigate the observed findings, which may also be helpful in efforts to identify and address inequities in the delivery of health care services.

ACKNOWLEDGMENTS

The authors would like to acknowledge the support of Kimberly Jinnett, PhD, Principal, Health Services Organization & Policy, E4A, USMA at Genentech, who provided helpful suggestions for the manuscript.

1. US Food & Drug Administration. Purple book database of licensed biological products. Accessed August 20, 2021. https://purplebooksearch.fda.gov/ Google Scholar
2. Research and Markets. Biologics global market opportunities and strategies to 2030: COVID-19 impact and recovery - summary. December 2020. Accessed September 8, 2021. https://www.researchandmarkets.com/reports/5232536/biologics-global-market-opportunities-and Google Scholar
3. Wolfe R, Ang D. Biologic therapies for autoimmune and connective tissue diseases. Immunol Allergy Clin North Am. 2017;37(2):283-99. doi:10.1016/j.iac.2017.01.005 MedlineGoogle Scholar
4. Rosman Z, Shoenfeld Y, Zandman-Goddard G. Biologic therapy for autoimmune diseases: an update. BMC Med. 2013;11(88). doi:10.1186/1741-7015-11-88 MedlineGoogle Scholar
5. Vora J, Gomberg J. Milliman Specialty Medical Drug 2010 Commercial Benchmark Study. Brookfield: Milliman, Inc.; November 2012. Google Scholar
6. Bunger A, Cline M, Holcomb K. Commercial Specialty Medication Research: 2019 Benchmark Projections: Milliman, Inc.; 2020. Google Scholar
7. artemetrx. 5th annual state of specialty spend and trend report. Accessed January 17, 2022. https://www.psgconsults.com/specialtyreport Google Scholar
8. The IQVIA Institute. The use of medicines in the US. The IQVIA Institute. May 27, 2021. Accessed August 20, 2021. https://www.iqvia.com/insights/the-iqvia-institute/reports/the-use-of-medicines-in-the-us Google Scholar
9. KFF. 2020 employer health benefits survey. October 8, 2020. Accessed August 16, 2021. https://www.kff.org/report-section//ehbs-2020-section-1-cost-of-health-insurance Google Scholar
10. Sherman B, Gibson T, Lynch W, Addy C. Health care use and spending patterns vary by wage level in employer-sponsored plans. Health Aff (Milwood). 2017;36(2): 250-57. doi:10.1377/hlthaff.2016.1147 MedlineGoogle Scholar
11. Pera M, Cain M, Emerick A, et al. Social determinants of health challenges are prevalent among commercially insured populations. J Prim Care Community Health. 2021;12:21501327211025162. doi:10.1177/21501327211025162 MedlineGoogle Scholar
12. Goh H, Kwan Y, Seah Y, Low L, Fong W, Thumboo J. A systematic review of the barriers affecting medication adherence in patients with rheumatic diseases. Rheumatol Int. 2017;37:1619-28. doi:10.1007/s00296-017-3763-9 MedlineGoogle Scholar
13. Solomon D, Tonner C, Lu B, et al. Predictors of stopping and starting disease-modifying antirheumatic drugs for rheumatoid arthritis. Arthritis Care Res (Hoboken). 2014;66(8):1152-58. doi:10.1002/acr.22286 Crossref, MedlineGoogle Scholar
14. United States Census Bureau. American community survey. October 28, 2021. Accessed November 11, 2021. https://www.census.gov/programs-surveys/acs Google Scholar
15. Charlson M, Pompei P, Ales K, MacKenzie C. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis. 1987;40:373-83. doi:10.1016/0021-9681(87)90171-8 Crossref, MedlineGoogle Scholar
16. Ashcraft M, Fries B, Nerenz D, et al. A psychiatric patient classification system. An alternative to diagnosis-related groups. Med Care. 1989;27(5):543-57. doi:10.1097/00005650-198905000-00009 Crossref, MedlineGoogle Scholar
17. Adler N, Boyce T, Chesney M, et al. Socioeconomic status and health. Am Psychol. 1994;49(1):15-24. doi:10.1037//0003-066x.49.1.15 MedlineGoogle Scholar
18. Li P, Blum M, Von Feldt J, Hennessy S, Doshi J. Adherence, discontinuation, and switching of biologic therapies in medicaid enrollees with rheumatoid arthritis. Value Health. 2010;13(6):805-12. doi:10.1111/j.1524-4733.2010.00764.x Crossref, MedlineGoogle Scholar
19. Groessl E, Ganiats T, Sarkin A. Sociodemographic differences in quality of life in rheumatoid arthritis. Pharmacoeconomics. 2006;24(2):109-21. doi:10.2165/00019053-200624020-00002 MedlineGoogle Scholar
20. Calixto OJ, Anaya JM. Socioeconomic status. The relationship with health and autoimmune diseases. Autoimmun Rev. 2014;13(6):641-54. doi:10.1016/j.autrev.2013.12.002 MedlineGoogle Scholar
21. Izadi Z, Li J, Evans M, et al. Socioeconomic disparities in functional status in a national sample of patients with rheumatoid arthritis. JAMA Netw Open. 2021;4(8):e2119400. doi:10.1001/jamanetworkopen.2021.19400 MedlineGoogle Scholar
22. Akenroye A, Heyward J, Keet C, Alexander G. Lower use of biologics for the treatment of asthma in publicly insured individuals. J Allergy Clin Immunol Pract. 2021;9(11):3969-76. doi:10.1016/j. jaip.2021.01.039 MedlineGoogle Scholar
23. Schmajuk G, Trivedi A, Solomon D, et al. Receipt of disease-modifying antirheumatic drugs among patients with rheumatoid arthritis in Medicare managed care plans. JAMA. 2011;305(5):480-86. doi:10.1001/jama.2011.67 Crossref, MedlineGoogle Scholar
24. Bailey Z, Krieger N, Agenor M, Graves J, Linos N, Bassett M. Structural racism and health inequities in the USA: evidence and interventions. Lancet. 2017;389(10077):1453-63. doi:10.1016/S0140-6736(17)30569-X MedlineGoogle Scholar
25. Yip K, Navarro-Millán I. Racial, ethnic, and healthcare disparities in rheumatoid arthritis. Curr Opin Rheumatol. 2021;33(2):117-21. doi:10.1097/BOR.0000000000000782 MedlineGoogle Scholar
26. Amazcua L, Rivera VM, Vazquez TC, et al. Health disparities, inequities, and social determinants of health in multiple sclerosis and related disorders in the US: a review. JAMA Neurol. 2021;78(12):1515-24. doi:10.1001/jamaneurol.2021.3416 MedlineGoogle Scholar
27. Grant T, Croce E, Matsui EC. Asthma and the social determinants of health. Ann Allergy Asthma Immunol. 2022;128(1):5-11. doi:10.1016/j.anai.2021.10.002 MedlineGoogle Scholar
28. Hodges WT, Bhat T, Raval NS, et al. Biologics utilization for psoriasis is lower in black compared with white patients. Br J Dermatol. 2021;185(1);207-09. doi:10.1111/bjd.19876 MedlineGoogle Scholar

Share

Find related content

By Author