Hepatitis C treatment by non-specialist providers in the direct-acting antiviral era.

Abstract

Background:

Hepatitis C Virus (HCV) remains under-treated in the United States and treatment by non-specialist providers can expand access. We compare HCV treatment provision and treatment completion between non-specialist and specialist providers.

Methods:

This retrospective study used claims data from the Healthcare Cost Institute from 2013–2017. We identified providers who prescribed HCV therapy between 2013 and 2017, and patients enrolled in private insurance or Medicare Advantage who had pharmacy claims for HCV treatment. We measured HCV treatment completion, determined based on prescription fills for the minimum expected duration of the antiviral regimen. Using propensity-score weighted regression, we compared the likelihood of early treatment discontinuation by the type of treating provider.

Results:

The number of providers prescribing HCV treatment peaked in 2015 and then declined. The majority were gastroenterologists, although the proportion of general medicine providers increased to 17% by 2017. Among the 23,463 patients analyzed, 1008 (4%) discontinued before the expected minimum duration. In the propensity-score weighted analysis, patients treated by general medicine physicians had similar odds of treatment discontinuation compared to those treated by gastroenterologists (OR 1.00, CI: 0.99–1.01, p=0.45). Results were similar when comparing gastroenterologists to non-physician providers (OR 1.00, CI: 0.99–1.01, p=0.53) and infectious diseases specialists (OR 1.00 CI: 0.99–1.01, p=0.71)

Conclusions:

HCV treatment providers remain primarily gastroenterologists, even in the current simplified treatment era. Patients receiving treatment from general medicine or non-physician providers had a similar likelihood of treatment completion, suggesting that removing barriers to the scale-up of treatment by non-specialists may help close treatment gaps for hepatitis C.

Introduction:

Direct acting antiviral agents (DAAs) for the treatment of hepatitis C virus infection (HCV) have simplified treatment for this highly prevalent disease. Whereas treatment once required specialty training and intensive monitoring, the revolutionary simplification of antiviral regimens since 2014 has opened the door for new providers, especially non-specialists, to offer HCV treatment.^1,2 Expanding HCV treatment access is an important component of national HCV elimination plans, but recent estimates suggest that 30 to 50% of HCV-infected individuals with commercial insurance or Medicare have not received treatment.^3–5

Low availability of HCV treatment providers may be a barrier to treatment in some areas.^6,7 Even where providers are available, patients may face logistical barriers accessing specialty care.^8,9 We previously showed that the number of HCV treatment providers increased substantially in the first year of DAA approval, but that the majority of this expansion was specialists.¹⁰

Insurance coverage restrictions are an important barrier to provision of HCV treatment by primary care providers. In 2017, 73% of Medicaid programs required a specialist prescriber for coverage of HCV treatment.¹¹ Several studies have demonstrated the effectiveness of non-specialist treatment in the DAA era, but most reflect the experience in specific settings including some that included training or mentorship for non-specialists.^12–15

We used administrative data to describe changes in the specialty composition of HCV treatment providers, and to test whether treatment prescribed by a non-specialist was associated with changes in the rate of treatment completion. We hypothesized that there would not be an association between provider specialty and treatment completion.

Methods:

We conducted a retrospective study to characterize providers associated with prescription claims for HCV treatment. We then used a propensity score-weighted regression to test whether early treatment discontinuation was associated with provider specialty.

Data Source:

We used 2013–2017 data from the Healthcare Cost Institute (HCCI), which contains claims from over 50 million insured individuals each year from 3 large insurers across all 50 United States and Washington DC, including employment-based, health insurance marketplace, and Medicare Advantage plans.¹⁶

Provider-level analysis:

We identified unique providers associated with at least one pharmacy claim for an HCV antiviral medication. Provider specialty was determined based on information associated with provider claims. Providers labelled as “gastroenterologists” included both gastroenterologists and hepatologists; “general medicine” included internists, family physicians, and preventative medicine physicians; “non-physician providers” included registered nurses, nurse practitioners, and physicians’ assistants. All other providers except infectious diseases specialists were labelled “other or unknown.” Provider ZIP codes were linked to the Area Health Resource File.¹⁷ Area-level variables included a measure of urbanicity using the Rural Urban Continuum codes, in which counties were categorized as either being metro counties, metro-adjacent counties, or non-metro and non-adjacent.¹⁸ We plotted the number of HCV treatment providers from 2013 to 2017 and described trends in the distribution of providers’ specialty and urbanicity.

Patient-level analysis:

Patients were included if they had a prescription claim for an all-oral HCV DAA regimen, they had continuous enrollment for 12 months before and 6 months after the month of their initial DAA prescription, and they had at least one physician claim during this period. We only included the first HCV treatment episode for each patient, with a gap in treatment of over 30 days constituting a new episode. We excluded patients with missing zip codes (n=781) and incomplete information treatment regimens (n=449), such as monotherapy with a drug usually given in combinations (Table S1).

The main dependent variable was early discontinuation, determined if discontinuation of treatment occurred before the end of expected duration of a patient’s regimen, as determined from the package insert.¹⁹ (Table S1) We chose this measure because it is a process outcome that providers can influence, the high efficacy of direct acting antivirals makes it likely that patients who complete treatment are cured, and it is easily measured in administrative data.^19,20

The main independent variable was provider specialty, as described above, using provider information from the first HCV treatment prescription received by each patient, to avoid capturing refills placed by covering providers. Patient age was reported categorically in 10-year increments, and race/ethnicity information was not available. Comorbidities were identified using ICD-9/10 codes occurring in the 12 months prior to the start of HCV treatment. (Table S2)^21,22 Possible opioid use disorder (OUD) was defined if the patient had either a diagnosis of opioid use disorder, a prescription for buprenorphine for OUD, or a diagnosis of a condition related to injection drug use.²¹ The Charlson-Deyo comorbidity score was calculated for each patient using diagnosis codes in physician claims.²² While we could not adjust for individual treatment regimens, we included an indicator variable for whether the regimen contained ribavirin, as this agent has a number of side effects that may lead to earlier discontinuation.²³ Area-level variables were determined using the Area Health Resource File as described above. We also included binary variables for whether the county is designated a high-poverty county (≥20% of population is poor) or a low-employment county (<65% of population age 18–64 is employed) by the US Department of Agriculture Economic Research Service.¹⁷ Finally, we accounted for medication cost by calculating the average daily out-of-pocket cost during the treatment episode, as reported in the dataset.

We used inverse-probability of treatment weighting to estimate the association between specialty and early treatment discontinuation, to account for confounding by indication, in which patients with more complex medical illness may be more inclined to visit a specialist, whereas patients with greater socioeconomic deprivation may be less able to access a specialist. In either case, these variables may also be associated with the likelihood of treatment completion.

We followed the method described by McCaffery and colleagues to generate propensity scores in the setting of multiple exposure groups^24–26 (Table S5). We then used a logistic regression model to estimate the probability of treatment completion, adjusting for individual covariates as well as accounting for propensity score weights. This approach, known as “doubly robust estimation,” helps provide valid inferences if either the propensity score model or the outcome model is correctly specified.²⁷ All analyses were conducted using R version 4.0.0, using the “twang” and “svydesign” packages for regression, and the “ggplot” package to generate plots.^28–31

Results:

Trends in HCV treatment providers over time:

The number of providers prescribing HCV treatment in each quarter increased after the approval of all-oral DAA regimens in 2014, peaked in the 2nd quarter of 2015, and then decreased thereafter. This change was driven by gastroenterology specialists, who were the majority of providers in each year. The number and proportion of general medicine physicians increased from 319 (14%) in 2013 to 1073 (17%) in 2017. (Figure 1, Table S3)

Figure 1:

Trend in the number of HCV treatment providers by specialty

Trends in characteristics of HCV patients treated by provider specialty

A total of 24,463 patients were included in the analysis. Of these, 62% were male, 89% aged between 45 and 75 years, and 91% lived in urban counties. About a third (32%) had cirrhosis, 11% had opioid use disorder, and 4% had HIV. Gastroenterologists treated the most patients. Patients treated by gastroenterologists were less likely to have HIV/AIDS diagnosis, opioid use disorder, or to live in non-metro, high poverty or low employment counties (Table 1). Table S4 shows changes in types of insurance coverage, an increasing proportion of patients with opioid and alcohol use disorders, and a decreasing proportion of cirrhosis from 2014–7.

Table 1:

Patient characteristics by specialty of treating provider

	Gastroenterology	General Medicine	Non-Physician Providers	Infectious Diseases	Other Providers
Total number of patients	14935	2475	1524	3237	1292
Gender
Male	9043 (61%)	1602 (65%)	2001 (62%)	1026 (67%)	845 (65%)
Female	5892 (39%)	873 (35%)	1236 (38%)	498 (33%)	447 (35%)
Age Category
18–24	169 (1%)	28 (1%)	31 (1%)	19 (1%)	11 (1%)
25–34	345 (2%)	55 (2%)	72 (2%)	49 (3%)	25 (2%)
35–44	544 (4%)	102 (4%)	130 (4%)	79 (5%)	34 (3%)
45–54	2433 (16%)	405 (16%)	563 (17%)	279 (18%)	222 (17%)
55–64	7414 (50%)	1216 (49%)	1636 (51%)	730 (48%)	668 (52%)
65–74	3425 (23%)	576 (23%)	698 (22%)	319 (21%)	292 (22%)
75+	605 (4%)	93 (4%)	1107 (3%)	49 (3%)	40 (3%)
Insurance
Private	7995 (54%)	1285 (52%)	1754 (54%)	708 (46%)	712 (55%)
Medicare Advantage	6940 (46%)	1190 (48%)	1483 (46%)	816 (53%)	580 (45%)
Opioid Use Disorder
No	13400 (90%)	2153 (87%)	2875 (89%)	1327 (87%)	1135 (88%)
Yes	1535 (10%)	322 (13%)	362 (11%)	197 (13%)	157 (12%)
Alcohol Use Disorder
No	13708 (92%)	2217 (90%)	2920 (90%)	1366 (90%)	1154 (89%)
Yes	1227 (8%)	258 (10%)	317 (10%)	158 (10%)	138 (11%)
Cirrhosis
No	10333 (69%)	1674 (68%)	2016 (62%)	1163 (76%)	810 (63%)
Yes	4602 (31%)	801 (32%)	1221 (38%)	361 (24%)	482 (31%)
HIV/AIDS
No	14792 (99%)	2267 (91%)	3155 (97%)	1187 (78%)	1229 (95%)
Yes	143 (1%)	208 (79%)	82 (3%)	337 (22%)	63 (5%)
Median Charlson Score (interquartile range)	2 (1–4)	2 (1–5)	2 (1–4)	3 (1–7)	2 (1–5)
Ribavirin prescription
No	12675 (85%)	2115 (85%)	2747 (85%)	1338 (88%)	1095 (85%)
Yes	2260 (15%)	360 (15%)	490 (15%)	186 (12%)	197 (15%)
Urbanicity
Metro	13794 (92%)	2231 (90%)	2884 (89%)	1369 (90%)	1185 (92%)
Metro-Adjacent	838 (6%)	169 (7%)	251 (8%)	113 (7%)	73 (6%)
Nonmetro	303 (2%)	75 (3%)	102 (3%)	42 (3%)	34 (2%)
High Poverty County
No	13309 (89%)	2057 (83%)	2912 (90%)	1341 (88%)	1142 (89%)
Yes	1626 (11%)	418 (17%)	325 (10%)	183 (12%)	150 (12%)
Low Employment County
No	14167 (95%)	2197 (89%)	3076 (95%)	1444 (95%)	1219 (94%)
Yes	768 (5%)	278 (11%)	161 (5%)	80 (5%)	73 (6%)

Treatment completion and provider specialty status

Overall, 1,008 (4%) patients discontinued before the expected duration. The proportion with early discontinuation was similar among specialties: 4% for gastroenterology and infectious diseases, 5% for general medicine and non-physicians, and 6% for providers with other or unknown specialties. Using an unweighted, bivariate logistic regression, treatment by general medicine physicians was associated with a greater odds of early discontinuation compared to gastroenterologists (OR 1.24, 95% CI 1.01–1.51, p = 0.03). In a multivariate logistic regression without propensity-score weighting (Table 2), the odds of early discontinuation were no longer significantly different between specialties except for providers with Other or Unknown specialty. Other variables significantly associated with early discontinuation included age under 25, private insurance, opioid use disorder, higher Charlson score, a ribavirin-containing regimen, and increased out of pocket cost.

Table 2:

Unweighted logistic regression on outcome of treatment discontinuation

	OR	95% CI	p-value
Specialty (ref Gastroenterology)
General Medicine	1.17	0.95–1.44	0.13
Non-Physician Provider	1.13	0.93–1.35	0.22
Infectious Diseases	0.93	0.69–1.23	0.62
Other or Unknown	1.42	1.10–1.81	0.006
Gender (ref Female)
Male	0.90	0.79–1.03	0.14
Age Category (ref 55–64)
18–24	1.98	1.28–2.96	0.001
25–44	0.93	0.59–1.38	0.73
35–44	1.03	0.74–1.41	0.85
45–54	0.91	0.76–1.09	0.32
65–74	0.91	0.76–1.10	0.35
75+	1.11	0.76–1.58	0.56
Insurance (ref private)
Medicare Advantage	0.47	0.4–0.55	<0.001
Opioid Use Disorder	1.79	1.48–2.15	<0.001
Cirrhosis	0.91	0.78–1.05	0.23
Alcohol Use Disorder	1.07	0.86–1.33	0.78
HIV/AIDS	1.01	0.95–1.07	<0.001
Charlson score	1.09	1.07–1.12	<0.001
Regimen (ref No ribavirin)
Ribavirin containing	2.45	2.09–2.86	<0.001
Average daily out of pocket medication cost ($)	1.01	1.00–1.01	<0.001
Year (2014 ref)
2015	0.7	0.58–0.83	0.02
2016	0.81	0.67–1.29	0.36
2017	1.03	1.00–1.01	0.002

Variables significantly associated with provider specialty included: HIV/AIDS, cirrhosis, Charlson score, and living in a low employment or high poverty county. After propensity score weighting, no variable was significantly associated with provider specialty (Table S5) and propensity scores overlapped between all groups (Figure S2).

Results of the propensity score-weighted regression with a doubly robust approach indicated similar likelihood of early discontinuation between any of the non-gastroenterologist specialties and gastroenterologists (Table 3). In a sensitivity analysis, we allowed treatment discontinuation up to 7 days early; this did not change the results (Table S6).

Table 3:

Likelihood of treatment discontinuation by provider specialty after application of propensity score weights and adjustment for additional confounding

Comparison	aOR (95% CI)a	p-value
Gastroenterology vs general medicine	1.00 (0.99–1.01)	0.45
Gastroenterology vs non-physician	1.00 (0.99–1.01)	0.53
Gastroenterology vs infectious diseases	1.00 (0.99–1.01)	0.71
Gastroenterology vs other/unknown	1.01 (1.00–1.03)	0.03

^aadjusted for age, gender, insurance, opioid use disorder, cirrhosis, alcohol use disorder, HIV, Charlson score, county characteristics, ribavirin regimen, medication out of pocket cost, and year.

Discussion:

We observed a peak in the number of gastroenterologists writing prescriptions for HCV treatment shortly after DAAs were approved, followed by a decrease. These findings may reflect the successful early treatment of patients with cirrhosis, many of whom may have been waiting to receive treatment with newer regimens and were already engaged with gastroenterologists. General medicine physicians and non-physicians remained a small minority of providers prescribing HCV medications, despite an increase in numbers. Provision of treatment by non-specialists may be hindered by prior authorization requirements that explicitly require a specialist evaluation or are complex enough to deter other providers from incorporating HCV treatment into their practices.³² Our study could not directly examine these policies, as plan identifiers or benefit designs were not available in the data source. The patterns seen in our study may also reflect the provider population serving the insured population that we studied: many primary care providers might not have seen the need to incorporate treatment into their practice if there was a specialist that patients could access.

We found that patients treated by gastroenterologists were less likely to have opioid use disorder, HIV, or to live in non-urban and disadvantaged communities. Adjusting for these factors attenuated the association between specialty and discontinuation. These findings suggest that general medicine physicians were likely treating a more underserved population with a higher risk of early treatment discontinuation, although overall treatment discontinuation rates were low.

Our study adds to the literature suggesting that the primary care and non-physician workforce can provide effective HCV treatment, and supports elimination of prior authorization policies which inhibit provision of treatment by those providers. Increasing the availability of treating providers by actively engaging non-specialists may be a viable public health strategy to achieve HCV elimination. Examples of successful engagement interventions incorporate tele-mentoring, care coordination, or direct specialist support for the non-specialist provider.^33–37 Less intensive strategies may also be effective, such as directly promoting treatment to primary care providers in high-need areas, known as “public health detailing.”^38,39 These strategies can be targeted to areas where limited specialist capacity is the main barrier to treatment uptake, or to providers treating patient populations who may have difficulty accessing specialist providers.

This study is limited by use of administrative data, including the inability to determine actual medication use or intended treatment duration. We were also unable to measure sustained virologic response and other quality measures such as alcohol counseling or hepatitis B vaccination, although the outcome we focused on—treatment completion—strongly predicts virologic response. Our data include a large sample of privately insured or Medicare Advantage individuals in the US, but do not represent all HCV-infected individuals. Patients participating in multiple insurance plans would not be captured in this study. We did not have information about provider characteristics beyond specialty, especially provider training in Hep C care or collaborations between non-specialists and specialists, which may be particularly relevant for non-physician prescribers. While we attempted to control for confounding by indication using statistical methods, there may be residual confounding.

In this national retrospective study of a commercially insured cohort, we found that the provider landscape for hepatitis C treatment was still dominated by gastroenterologists but the role of non-specialists was increasing, and found a similar rate of treatment completion between specialists and non-specialist providers after adjusting for confounding. Our findings support removing barriers to the continued expansion of HCV treatment by non-specialist providers.