Outcomes and cascade of care definitions

The primary endpoint evaluated was sustained virologic response (SVR) at least 12 weeks after treatment completion. Secondary endpoints included achievement of each individual stage in the CoC as well as time to treatment approval after DAA prescription. For the purposes of this study, the CoC represented the progression from referral to the VUMC ID clinic through HCV evaluation, prescription, initiation, and completion of treatment, and achievement of a SVR at least 12 weeks after completing treatment. Definitions of each step within the CoC are found in Table 1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Cascade of care definitions.

https://doi.org/10.1371/journal.pone.0199174.t001

When individual patients did not proceed through the CoC, data was collected from the EMR on the reason(s) for lack of advancement. A patient was considered lost to follow-up if ≥ 5 attempts to contact the patient were made by phone as well as a letter sent to the patient’s most recent address with no response over at least three months. If a patient missed an initial appointment in the ID Clinic, one outbound call was made by pharmacist and a voicemail was left if there was an option to do so.

Data collection

Dates of clinic visits were obtained from the Epic scheduling system while all other outcomes were collected from the EMR. Patient characteristics assessed in the EMR prior to an evaluation included patient age, gender, ethnicity, insurer, and home zip code. When available in referring paperwork or the EMR, additional data points were collected including cirrhosis, active illicit substance use, HIV co-infection, insurance status, psychiatric history, and gender. Additional characteristics were confirmed at the time of evaluation, including: HCV genotype, HCV treatment history, fibrosis stage, HIV co-infection, history of and/or ongoing illicit substance or injection drug use (IDU), history of and/or ongoing alcohol abuse, and psychiatric disorder.

Medicaid recipients were identified as having Medicaid as a primary insurer for prescription coverage at the time of HCV prescription. Patients were considered co-infected with HIV if they were labelled with ICD10 code B20. Cirrhosis was defined as meeting any of the following criteria: anatomic ultrasound showing changes consistent with cirrhosis; ultrasound with acoustic radiation force impulse predicting F3-F4 or F4 fibrosis; FIB-4 score ≥3.25; Fibrosure of ≥0.72; or a liver biopsy with Metavir score F4. Diagnosed psychiatric disorder included patients labelled with an ICD10 including F01-F69 and F80-F99. Ongoing alcohol abuse was defined as >5 drinks on most days of the week as reported by the patient. Ongoing illicit substance or injection drug use (IDU) was defined as use within 3 months of evaluation as reported by the patient. The “baby boomer” age cohort was defined as any patient born between January 1, 1945 and December 31, 1965.

All patients were assessed for HCV treatment based on AASLD/IDSA Guidelines for the Treatment of Hepatitis C from the same group of providers working within the VUMC ID Clinic.

Statistical analysis

Categorical variables were described using number of patients and percentage, while continuous variables were described using, mean, median, standard deviation, and interquartile range. Demographic characteristics selected a priori included gender, ethnicity, insurance type, HIV co-infection, cirrhosis, psychiatric disorder, ongoing illicit substance use, and baby boomer age cohort. Outcomes were binary variables for retained status (indicating movement from one step in the cascade to the next), medication approval status, as well as the time to medication approval (measured in days).

To further investigate associations between patient demographic characteristics and outcome variables, analyses were performed using logistic regression for binary outcomes, a cox proportional hazard model for time to event (i.e., medication approval) outcomes, and multiple linear regression model for continuous outcomes initially fit without adjusting for covariates. This was followed by multivariable models that controlled for gender, insurance type, HIV co-infection, cirrhosis, psychiatric disorder, and illicit substance use. For the analysis of time to medication approval using a multiple linear regression model, logarithmic transformation was performed to reduce skewedness. To avoid case-wise deletion of records with missing covariates, we employed a multiple imputation method with 10 imputation samples using predictive mean matching. All statistical analyses were performed using the programming language R version 3.3.0.

Linkage to care

As seen in historical and modern HCV CoC studies, linkage to care after referral was a significant barrier to CoC completion.[7, 8, 10, 11] Overall, 36% of patients with a scheduled appointment did not complete medical evaluation. Patients with Medicaid were 79% less likely to complete an evaluation as those without Medicaid, even when controlling for other factors. Previous studies have highlighted the disparity in this population obtaining access to treatment; however, our study demonstrates that this population may have difficulties accessing and completing other elements of medical care related to HCV.[6, 8, 9] Women were three times less likely to complete an evaluation in our cohort. While the reason for this gender disparity is likely complex, efforts should be focused on engaging women at the time of screening and referral, particularly in light of recent increased rates of HCV diagnosis in women of childbearing age and the subsequent risk of vertical transmission.[12]

Strategies to improve linkage to care should be further implemented and evaluated in the DAA era. These include engaging mental health and social work as well as coordinating substance abuse services for certain patients.[13] A number of interventions to improve linkage and access employed by the US Department of Veterans Affairs are translatable to those outside a single payer model and should be explored, including utilizing telemedicine and electronic technologies to reach rural and underserved areas, utilizing nonphysician advanced practice providers, and establishing services that address substance abuse.[14] As HCV treatment becomes more streamlined for most patients, shifting HCV treatment from specialty clinics to the primary care setting is an opportunity to improve effective linkage and completion of medical evaluation.[14, 15]

Treatment prescription and initiation

Our results diverge with historical CoC models once patients completed evaluation, showing a high rate of retention in care with 89% of patients completing DAA treatment. Minimal loss between an evaluation and prescription of treatment occurred. Access through prior authorization (PA) approval to these costly medications has been identified as a barrier to DAA initiation.[6, 8, 9] We found a lower non-start rate than those previously reported in larger cohorts, with only 5% of our patients who were prescribed treatment not subsequently initiated.[5, 7, 16–18] Despite a large proportion of the evaluated cohort having Medicaid (32%), only five patients of 60 were not ultimately approved for treatment. Conversely, the TRIO Network recently reported a non-start rate of 48% in this population.[19] While Medicaid restrictions vary by state, Tennessee Medicaid restricts HCV treatment during the study period to patients with moderate to advanced fibrosis (F2 and above) with proof of alcohol and drug rehabilitation for any previous abuse as well as 6 months of sobriety.[20] The discrepancy in access to treatment by payer has been highlighted in multiple analyses and is perpetuated in this real world cohort.[5, 6, 18, 19, 21]

The median time to approval for all patients prescribed treatment was five days (IQR 3–14); lower than in several previously reported cohorts.[4, 5, 8, 19, 21] Patients without Medicaid (including those with no insurance) had a median time to approval of 4 days (IQR 2–8.5), while those with Medicaid who were approved had a median time of 30 days (IQR 10–46). Our results substantiate individual findings that for patients with Medicaid, delays in approval and initiation were common even if treatment is ultimately approved.[4, 22] Future studies should evaluate the clinical impact of treatment delays caused by difficulty accessing DAAs. Though not directly evaluated in this study, we believe our improved non-start rate and time to treatment approval are likely a result of the integrated model of our clinic with dedicated pharmacy services experienced in navigating the requirements to access therapy.

Treatment completion

Few patients were lost to follow-up between treatment initiation and confirmed completion (n = 6), and no patients stopped treatment due to adverse effects or inability to afford treatment. This step within the CoC is clearly distinguishable from historical CoC data, where completion rates were lower due to the adverse effects of interferon and ribavirin.[23] Upon univariate analyses, no differences in treatment completion rates were seen among baseline characteristics studied. When medication approval was obtained, Medicaid patients and those with ongoing illicit substance use had similar rates of treatment and CoC completion. This finding further supports the AASLD/IDSA Guidelines’ recommendation to consider treatment in patients with ongoing illicit substance use.[24]

Within this study population, 95% of patients who initiated DAA therapy completed the full prescribed course. Provider fear of non-adherence to treatment has been highlighted as a reason for not prescribing DAA treatment.[25, 26] In the modern CoC, this concern is likely driven primarily by social factors including unstable housing, mental illness, or active alcohol and substance abuse rather than severe side effects from the medication.[14, 25, 27] Given the high prescription, initiation, and completion rates seen in the population in our study, concern regarding non-adherence may not be warranted in all settings. To facilitate initiation and completion of HCV treatment, providers should place an emphasis on engaging this population with social workers, case managers, and addiction counselors.[13, 14] Within our clinic, the pharmacist played an integral role in identifying and mitigating adverse effects that may have led to treatment discontinuation and enabled patients with counseling and adherence tools to ensure treatment completion. The impact of shorter treatment durations is yet to be elucidated but is likely only to improve completion rates.

Sustained virologic response

Achievement of an SVR after treatment completion has clearly improved in the modern CoC with DAA treatment in comparison to historical standards.[28, 29] As expected, SVR rates for those that completed treatment were high (93%), with only 3 virologic failures among those with available laboratory data. Four patients were lost to follow-up after confirming treatment completion. A high intention-to-treat SVR rate was also seen in this real-world cohort, with 88% of patients who started treatment obtaining a confirmed SVR. None of the patients achieving an SVR in our evaluation were re-infected within the timeframe of the study. However, as more patients are treated and cured of HCV infection, future CoC studies should include movement beyond SVR eradication to include the frequency of patients who are re-infected.

Limitations

All subjects were adults followed at a single outpatient ID clinic. A large percentage of patients were co-infected with HIV and few were actively using illicit drugs, particularly injection drugs. Therefore, the generalizability of the population is limited. As Medicaid restrictions differ by state, findings regarding delay in and access to treatment in Medicaid patients in our cohort may not reflect those of other Medicaid programs.[20] This study does not take into account the steps in the CoC prior to a referral to the clinic, including screening and diagnosis. This study was not blinded. Due to the limited sample size, the number of variables analyzed by multivariate analysis was limited. Finally, some factors that may have impacted an evaluation completion were not available, limiting a full analysis of these factors on evaluation completion rates.