Integrating public health policies in the epidemiological modeling of hepatitis C with LEHC tool: application in Austria

Hepatitis C is a silent disease that afflicts people all over the globe. This virus affects people according to unknown timetables, which imposes an important role on health services in terms of human resources, technology and capital. Its evolution is well-documented in the natural history of the disease [1].

Globally, the hepatitis C virus (HCV) is the seventh leading cause of hepatic mortality [2], often related to hepatocellular carcinoma. Transmission networks of the disease are often associated with significant costs, either being liver transplants or indirect costs.

One of the HCV characteristics is its concentration of higher prevalence rates among given population segments: people who inject drugs (PWID) [3‐5], people who have been incarcerated [6‐8], individuals who have received blood products [9, 10].

To address the public health issue of HCV, the WHO has defined an elimination goal for the disease by 2030, aiming at reducing disease incidence by 90% and mortality by 65% [11]. This target has set the discussion about targeting population groups that are more prone to be infected, with the concept being characterized as micro-elimination. There are currently over ten groups that meet this criterion [12]: individuals with Aboriginal and indigenous background, people born in specific periods, children whose mother was infected with HCV during pregnancy, hemodialysis patients, individuals coinfected with HCV/HIV, PWID, migrants from high HCV prevalence countries, people who have received blood products, prisoners, transplant recipients, war veterans [13], baby boomers [14], among others.

Projections for the HCV prevalence have been considered to be between 0.5% and 3.5% in Europe [15, 16] and 7–8% [17] in other continents. The presented values have been subject to updates in epidemiological revisions [18], with a decrease in HCV prevalence rates being verified to be 0.2–1.5% in Europe [19]. In Austria, it has been estimated that the prevalence of HCV infections was 0.46%, and the prevalence of viremic prevalence around 0.34% in 2008 [20]. For this reason, including Austria in the first modeling phase of the LEHC project was seen as essential.

The LEHC tool aims to evaluate and simulate diverse adaptations of public health policies according to the modeled epidemiological characteristics of the disease and the needs of a country to achieve the HCV elimination goal by 2030. By modeling the impact of the dynamics between both referred elements, it is possible to generate a landscape on HCV epidemiology for every year until 2030. Therefore, the model is also aligned with the sustainable development goals timetable. The study of the 24 health policies impact and their different levels of implementation resulted in different estimates on HCV epidemiology for different population groups: PWID, blood products, vertical transmission¹, remnant population. The natural history of HCV is reflected in the indicators that result from the epidemiological dynamics on model projections.

This project is being promoted by the Public Health Unit of the Health Sciences Institute of the Catholic University of Portugal. Along with Austria, other countries were included in the project’s initial modeling phase (Romania, Bulgaria, Portugal, Spain) while others are also being currently worked.

The LEHC model has the potential to forecast and gamify² different applications of public health policies while calculating its impacts on HCV epidemiology. It is possible to assess the modeling results for six population groups: remnant population, prisoners, PWID, blood products, vertical transmission and total population. The remnant population was calculated according to the formula: remnant population = total population − (blood products population + prisoners and ex-prisoners + PWID and ex-PWID). Due to its low number, the vertical transmission population was included in the remnant population.

The LEHC epidemiologic model integrated a total of 1100 active Markov chains³ branches among the identified population groups, which resulted in data that characterize the HCV epidemiological evolution. Health states and respective transition probabilities are also included in the branches system, being defined according to the natural history of HCV as identified by Salomon et al. [1] in 2003. The Markov chains component is applied for the period between 1950 until the present time, featuring data that correspond to indicators, such as incidence/prevalence rates as well as diagnosis, retain in care, treatment and sustained virologic responses (SVR) rates. The availability of data regarding epidemiological indicators for HCV is linked with research in the mid-1990s, driven by the ground-breaking hepatitis C virus discovery in 1989. As a result, when no data were available or the existing information did not meet the full criteria for the statistical series, proxies were produced to atone the modeling components. Proxies were contemplated as assumptions, being assessed and validated by the country’s National Advisory Board (NAB) and were subject to the mathematical evaluation by comparing the model results with empirical data.

It is assumed that the statistical data used for the LEHC model are correct, considering that they are either collected from peer-reviewed journals or have their source in official data of countries or regions. In the case where it was not possible to retrieve data from these sources, proxies had to be produced, with the assumption that the leading experts in each country of the LEHC project are able to provide an accurate overview of the epidemiological reality.

However, it is known that even data that were subject to peer review have often existed solely in micro-populations while being related to the methodology used in data collection. It must be noted that even official data from countries has some frailties, such as the difference of the total number of inhabitants by over one million individuals, the source or country on the origin of the data, the opinion of the most acknowledgeable experts, which is always a personal view based on vast experience.

Around 100 key opinion leaders (KOL) from each NAB have also participated in surveys based on an adaptive conjoint analysis (ACA) approach, taking a global average of 2h for completion. The resulting data were used to develop the modeling component related to the 24 public health policies (additional details in htttp://www.​letsendhepc.​com/​)⁴ in each country. These policies have a distinctive scale of implementation being contemplated on the surveys, in which every KOL attributes a different weight to different policy levels. This process allows the differentiation of forecast results by having distinctive impacts on the Markov chains component, according to a cure cascade⁵ perspective. Modeling in Austria was supported by the participation of seven KOLs in completing the ACA surveys.

Modeling hepatitis C in Austria was based on gathered data from multiple sources, such as national entities (Statistics Austria, Suchthilfe Wien, Ministry of Justice, Red Cross Austria, etc) [21‐23, 32]. and published studies [20, 24, 25, 31]. When no data were found to complete statistical series elements such as specific years, stratification by age or by gender, proxies were produced as detailed in the methodology. Estimates were based on both national and international studies regarding the populations covered by the LEHC project: Remnant population and vertical transmission [20, 24‐26], people who inject drugs [28, 29, 31], prisoner populations [32], blood products population.

Empirical evidence resulting from the LEHC project corroborates the view that the path for eliminating hepatitis C until 2030 requires a higher focus on finding and maintaining patients within the grasp of the healthcare system. Given the high prevalence and incidence rates among population groups, such as PWID and prisoners, it is important that measures focus on the micro-elimination concept.

It is also important to consider the risk of overlapping populations, especially between PWID and prisoner populations, and to a much lesser extent between PWID and blood products. It was not possible to find published epidemiological data that could separate these populations in the remaining LEHC countries. As a result, in the modeling process, a proxy was created whereby it is assumed that about 60% of the prison population would have been HCV-infected in the past by i.v. drug use, subtracting this number from the PWID population. A similar situation was applied to the relationship between blood products and PWID, seeking to correct double interpretation situations.

Considering the modeling results, there are several situations in which the number of patients with compensated cirrhosis is higher than the total number of HCV infected individuals. This situation is verified as people who are successfully treated for HCV will still have pre-existing conditions such as the cirrhotic state. Considering the increasing number of people undergoing treatment and thus cured, it is natural that even with new HCV infection cases, in a given moment there will be less HCV infected individuals than individuals with compensated cirrhosis.

In addition, diagnosis and linkage to care measures are paramount to meet the HCV elimination target, being aligned with the sustainable development goals. In Austria, it has been confirmed that 15 out of the 24 public health policies considered for the study are currently fully implemented.

Despite the fact that other countries are also included in the LEHC project, it was considered that introducing external readings of comparison between country policies and epidemiological indicators could be an inducement of bad practices for health authorities. In the LEHC project’s developed work in 12 countries, the dynamics of public health policies were verified to be very independent, even in common issues.

The LEHC project accomplished the modeling of hepatitis C in terms of public health policy impact on the disease epidemiology for the remnant population and other population groups in Austria. The dynamics of this tool allowed the gamification of public health policies, which provided an insight into how different extents of specific measures have an impact on the HCV epidemiology.

This model has the ability to forecast the extent to which a specific public health policy might exert influence on a particular population group, in a pool of 24 public health policies focused on HCV considered in the LEHC project. This is essential in identifying patterns that have the potential to not only allow countries to address given policies to achieve the HCV elimination goal by segmenting the population but also to know the why, who, what, where and when. The tool allows rehearsing public health policy concepts before implementing them in the field, thus anticipating the outcomes.

In the light of the cure cascade system, modeling results appoint awareness and prevention as the dimensions in which public health policies might have the broadest impact in lessening HCV prevalence values in Austria. Therefore, given these areas of influence, there is a clear indication that achieving the HCV elimination target will require additional efforts besides medication development. Diagnosing HCV in individuals who are unaware of their condition and linking them to healthcare services will determine the success of achieving the HCV elimination goal.

According to the modeling results, the elimination target will not be met by 2030 in Austria. Nonetheless, in the case that all the 24 public health policies are fully implemented, it will not only be possible to meet the elimination target, but it might happen as soon as 2026. As a result, it is seen that policies are fundamental in providing insights into combating HCV in order to achieve its elimination.

The quality of the modeling always depends on data quality used as input, on the quality of the analysis and on the knowledge about the public health policies’ impact in the field. In LEHC (or any other model), it is paramount to deal with the issue of quality as well with data availability.

Since there are no confidence intervals in modeling, the fitting test was performed by overlapping the results obtained by the LEHC modeling tool, on the known epidemiological curve or admitted into a given geographic region, in this case, a country. In the LEHC tool, only countries where this curve was at least 95% similar in the previous years were selected for publishing data; however, it must be noted that even national data are often limited in their reading due to being either very regional (in the case of Austria, most of the information comes from Tyrol), from samples sometimes incident on populations that do not represent the country (studies with prisoners or PWID), or epidemiological studies that do not cover the entire population. This dependence on adjusted fitting based on known epidemiological evidence is a limitation of the tool, but endogenous to the modeling process. No country which worked in the LEHC project showed strong robustness of epidemiological data.

Considering that the LEHC model was based on the best available data from Austria, it is also important to state that obtaining national data was challenging, which imposed the need of producing proxies from Austrian regions, to extrapolate and interpolate national data, and to create statistical series having proxies as a basis composed by other secondary data series. Therefore, the Austrian LEHC model is the best possible to achieve, considering the data available. A call for action on the research and publication of epidemiological data regarding hepatitis C in Austria is of the utmost importance in filling in data gaps. This project is fully available to analyze and integrate all data that the scientific community provides for this project.

In perspective, the dynamic vision for this project should contemplate an annual review for integrating new official data and other epidemiological data that might have been produced in the meantime. Similarly, the authors comprise the modeling results of LEHC for 31 December 2018, in which it is based on statistical data available up to the referred data and this being designated as “LEHC: An integrated solution for HCV—application in Austria”.