Necessity of Interrupted Time Series Analysis in Evaluating the Impact of PHS Risk Identification and Introduction of Direct-Acting Antiviral Therapy and Share 35 Implementation
We think more necessity of interrupted time series (ITS) analysis as an alternative trend modeling method that can estimate the changes in rates over time while taking preexisting trends into account in uneven observation time.2 For example, in Figure 1, PHS-identified risk recovered donors, the ITS can estimate average annual percent changes of PHS-identified risk recovered donors using 3 time stages: before modification (2005-2012), during the introduction of modification (2013), and after the full implementation of modification (2014-2015). Overdispersion is commonly observed in healthcare utilization.2 If overdispersion is observed in preliminary results, log-link function and maximum likelihood function can be added to adjust final ITS regression model.1 Therefore, the ITS can better characterize the impact of PHS 2013 modification on recovered donor trends.
According to Flemming et al's3 population-based cohort study, there were remarkable declines (approximately 30%) in listing for hepatitis C virus (HCV)-related decompensated liver disease in liver transplant (LT) waitlist in the era of direct-acting antivirals (DAAs) therapy compared to patients with other etiologies (hepatitis B, nonalcoholic steatohepatitis) as conjunction to a significant overall rise in waitlist among patients with HCV. However, the actual proportion of DAA users among HCV patients awaiting LT is largely unknown. Moreover, 1-year post-LT survival rate improved4 since the Share 355 policy implementation in June 2013 that prioritized patients with model for model for end-stage liver disease score of 35 or higher within the donor’s Organ Procurement and Transplantation Network region before any local candidates with Model for End-Stage Liver Disease less than 35 as response to the increasing medical acuity of patients with end-stage liver disease and persistent organ donor shortage. In short, DAA therapy and Share 35 policy occurred simultaneously when PHS modification was implemented in 2013.
Establishing causality is not plausible in observation studies even analysis from nationwide database, and it is difficult to adjust confounding factor, such as socioeconomic status by Kaplan-Meier analysis that were used in Table 3, 1-year posttransplant outcomes during 2012 to 2014 from 19- to 34-year-old HCV donors by Pruett et al.1 One of the major strength of ITS studies is that the model partially accounts for other external factors, such as DAA therapy introduction and Share 35 policy implementation influencing the outcomes.2 In addition, the ITS studies might easily detect overdispersion effect that the variance occasionally tends to be greater which could lead to incorrect estimation of the standard errrors.2 Therefore, the ITS, as one of time series analysis, can provide more precise information how “PHS 2013 modification” and potential confounding factors, “DAA therapy” and “Share 35” influences PHS-identified risk donor recovery and 1-year posttransplant outcomes, especially patients with HCV and awaiting LT.