Obstacles to successful treatment of hepatitis C in uninsured patients from a minority population

216 chronic HCV mono-infected patients were treated at Jackson Memorial Hospital since 2014. All patients who were eligible for testing for SVR 12 weeks post treatment completion before August 1, 2017 were included in this retrospective electronic health record (EHR) review. To be eligible for treatment, patients needed to be referred by a primary care provider and assessed for financial assistance eligibility. Individuals needed to be illicit drug and alcohol free for 6 months in non-cirrhotic patient or 3 months for cirrhotic patients. Treatment was selected and initiated based on American Association for the Study of Liver Diseases (AASLD) Guidelines [8], taking into account genotype, cirrhosis status and pre-treatment viral load. Patients who ultimately took their first dose of medication (as documented in the EHR) were included. This study was approved by the Institutional Review Board at the University of Miami and Jackson Health System’s Office of Research and Grants.

The primary variables collected were insurance status and the results of a post treatment viral load test 12 or more weeks after cessation of therapy. SVR was achieved if the patient was negative for detectable virus at this time. Additionally, age at treatment initiation, sex, race, ethnicity, genotype, pre-treatment viral load and cirrhosis status were collected. We also assessed whether the patient was compliant with appointments and whether complications were reported. All of these covariates were analyzed as distinct clinical variables.

Age at first treatment with a DAA regimen was documented at treatment initiation and recorded in years. Sex, race, and ethnicity were collected and defined per criteria of the EHR. All patients categorized as Asian, other, or other/white were combined into a single Asian/other race category. Patients who did not list an ethnicity category (n = 4) were classified as non-Hispanic. An additional variable combining race and ethnicity was created in which the non-Hispanic patients were assigned to their designated race categories, such as non-Hispanic black and non-Hispanic white. Non-compliance was defined as missing multiple clinic appointments or laboratory assessments, or being documented in the EHR as a “non-compliant patient”. All other patients were designated as compliant. Complications were defined as any treatment related complication that the patient reported and was recorded in the patient EHR, including headache, fatigue and difficulty understanding instructions. Treatment completion was assessed by patient report and documented in the EHR. The presence of cirrhosis was defined as the patient being documented as having the diagnosis of cirrhosis based on clinical or radiologic criteria.

All statistical analyses were conducted using SAS 9.4. Chi square tests were used for categorical variables. For variables in which expected cell frequencies were less than 5, Fisher Exact Tests were conducted. Continuous variables that were not well described by a normal distribution were tested using Wilcoxon rank-sum tests. Univariable and multivariable logistic regression utilizing all predictor variables was performed to determine associations between potential predictor variables and insurance status and between potential predictors and being tested for SVR. 95% confidence intervals for odds ratios were calculated. Significance was determined at a two-sided alpha level of 0.05 for all tests.

Initially, we analyzed our data for differences in clinical covariates and treatment outcomes between insured and uninsured patients (Table 1). 216 patients were included in this study and their corresponding data was used for analysis. The mean age of treated patients was approximately 56 years; the youngest was 25 and the oldest was 77. The median age was 58 and the interquartile range was 52–62 years. Males and females were almost equal in number (109 and 107, respectively). The distribution of race, as recorded in the EHR, was 153 whites, 59 blacks, 2 Asians, 1 other and 1 other/white. The documented ethnicity distribution was 123 Hispanics and 89 Non-Hispanics, with four patients lacking an ethnicity designation (of which two were insured and two were uninsured, p-value 0.6990). Most patients were compliant with appointments and laboratory assessments and approximately half experienced some type of complication or side effect of treatment. Pre-treatment viral load ranged from 840 to 166,211,511 IU/mL, with a mean of 4,247,796 and median of 1,655,329 (interquartile range was 434,953–4,676,597). The overall HCV Genotype (GT) distribution was 81% GT1, 9% Type GT2, 7% Type GT3, six patients with GT4, and one patient with a mixed GT1 and GT2 infection. The majority of patients were non-cirrhotic, completed treatment and attained SVR.

71% of patients in this cohort (n = 154) had no health insurance (Table 1). Insured patients were older (p-value < 0.0001) with a median age of 60 (interquartile range was 56–65) compared to 57 (interquartile range was 50–61) in the uninsured. Sex, race and ethnicity were not significantly associated with insurance status. In addition, compliance was not associated with insurance status. The development of complications or side effects was also not associated with insurance status. The average pre-treatment viral load in the insured was 5,973,784 IU/mL with a median of 1,806,502 (interquartile range 535,360–3,793,470) and it was not statistically different from the uninsured that had a mean of 3,552,918 and median of 1,527,253 (interquartile range 388,600–4,762,484). HCV Genotype distribution was also not associated with insurance status. Similar proportions of insured patients completed treatment as uninsured patients (90% vs 86%, respectively, p-value 0.4261). However, cirrhosis was associated with insurance status with significantly more cirrhotics being insured (40% vs 22%, respectively, p-value 0.0049). Univariable logistic regression revealed similar results as described above, however in multivariable logistic regression, only age and cirrhosis remained significantly associated with insurance status (Table 2). Overall, there were relatively few differences between our insured and uninsured patients.

We next assessed the transition of our cohort though the HCV treatment cascade [22] based on insurance status. The HCV care cascade describes the progression of patients from diagnosis of infection to cure [22]. However, the HCV treatment cascade does not include screening, diagnosis, and linkage to care. In order to elucidate clinically relevant transition points in the HCV treatment cascade, we only focused on steps involved specifically in HCV treatment and two steps were added: treatment completion and 12-week post-treatment viral load testing. Figure 1 presents a schematic of the patient numbers as they move through the treatment cascade and it includes data from our complete cohort of 216 patients that initiated HCV treatment. As found in the previous analysis, there were no significant differences between the insured and uninsured patients as they progressed through the HCV treatment cascade. As expected, patient attrition occurred as individuals moved through the treatment cascade.

Of the 216 patients who initiated treatment, 87.5% of patients (n = 189) completed treatment (90% of the insured and 86% of the uninsured, p-value 0.4261). Within these 189 patients, 177 (94%) obtained a viral load 12 weeks after treatment completion (93% of the insured and 94% of the uninsured, p-value 0.7516). Within the population that obtained a viral load at 12 weeks post-treatment (177 patients), 172 patients (97%) had undetectable HCV RNA levels achieving SVR (98% of the insured and 97% of the uninsured, p-value 1.0000) (Fig. 2). Interestingly, we found a large unexpected occurrence of patient dropout at the stages of completing treatment and being tested for SVR 12 of more weeks after treatment completion (Fig. 1). Neither of these parameters are typically described as phases of the HCV care cascade [22].

Interestingly, in those tested for SVR, cirrhotics attained similar rates of cure as non-cirrhotics (95 and 98% respectively, p-value 0.3630). Likewise, cirrhotics and non-cirrhotics had comparable rates of treatment completion (85% vs 89%, respectively, p-value 0.3942) and comparable rates of obtaining a 12-week post treatment viral load to assess the attainment SVR (97% vs 92%, respectively, p-value 0.2169) (Additional file 1: Figure S1).

The largest drop-off of patients along the HCV treatment cascade occurred between treatment initiation and treatment completion. 27 patients (12.5% of the cohort who began treatment) failed to reach the next step of the HCV treatment cascade. These patients were younger (p-value 0.0081), with a median age of 54 years (interquartile range 41–59) compared to a median age of 58 for those who completed treatment (interquartile range 53–62). A greater proportion of females completed treatment than males (93% vs 82%, respectively, p-value 0.0087). Lastly, patients who completed treatment were more compliant (p-value < 0.0001) (Table 3). Race, ethnicity, complications, insurance status, pre-treatment viral load, genotype and cirrhosis were not associated with treatment completion in univariable or multivariable analysis (Additional file 2: Table S1).

Since treatment completion has been studied in other populations [23] and is influenced by a myriad of clinical factors not necessarily related to insurance coverage, we next analyzed covariates that correlate with adherence for SVR testing (Table 4). Although 189 patients completed treatment, only 177 (82% of the cohort, 94% of those who completed treatment) were assessed for SVR attainment 12 weeks after treatment completion. There were no statistically significant differences in age between those tested and those who were not (p-value 0.8764). Neither sex, race, ethnicity nor insurance status were associated with being tested for SVR. However, as expected, compliant patients were significantly more likely to be tested for SVR than non-compliant patients (p-value < 0.0001). Furthermore, every compliant patient who completed treatment was ultimately tested for SVR, while approximately one-third of non-compliant patients who completed treatment failed to obtain a 12-week post treatment viral load. Importantly, pre-treatment viral load did not differ between patients tested and untested for SVR: median 1,675,814 (interquartile range 330,073–4,736,831) vs 1,172,716 (interquartile range 470,264–4,408,714). Tested and untested patients had similar complication rates, genotype distribution and cirrhosis rates. Univariable logistic regression yielded similar results; however, the compliance odds ratio and corresponding p-value were not estimable because, as stated previously, there were no compliant patients who completed treatment but did not obtain a 12-week post treatment viral load. Analogous results were obtained in multivariable logistic regression (Additional file 3: Table S2).

Of the 216 patients who started treatment, 88% completed treatment per guidelines, based on patient self-report (90% of the insured and 86% of the uninsured, p-value 0.4261). Of those who completed treatment, 94% were tested for SVR, or 93% of the insured and 94% of the uninsured. The SVR rate in those tested was 97% (98% in the insured and 97% in the uninsured) (Figs. 1 and 2). Overall, in our study cohort, uninsured patients have similar treatment outcomes and high cure rates as insured patients. However, patients were lost to follow-up as they progressed through the HCV treatment cascade and this was not correlated with insurance status.