Hepatology partnership model of care.

The hepatology partnership aimed to deliver expert assessment of HCV and specialised care using a hub and spoke model. The hub was based at the Sunshine Coast University Hospital, with HCV treatment being delivered to patients in the community by GPs following assessment at community clinics local to patients. During treatment, patients and GPs were supported by the maintenance of close links with the specialist team who led the intervention from secondary care. Referrals to the service were received from GPs, following triage, patients were assessed in the community and then discussed at weekly multidisciplinary team (MDT) meetings, which included consultants and nurses involved in the intervention. Initial appointments were scheduled and undertaken by nurses, with assessment of liver fibrosis using the mobile Fibroscan technology undertaken in the community setting. Patient education was also nurse-led and undertaken in the community setting. Patients requiring treatment for their infection had a drug regimen recommended via the specialist-based hub which was prescribed by the General Practitioner (GP), therefore patients were not required to visit a hospital for treatment unless fibrosis assessment revealed cirrhosis in which case review in secondary care was organised. If treatment failed, the process was repeated, with the exception of the fibrosis assessment, which was only completed once. The service was designed to reduce flow into the hospital setting and provide patients with timely and appropriate care, close to their community. Telehealth and MDT case conferencing was available to ensure a connection between primary and secondary care, and was utilised on an ad-hoc basis.

Usual care (historical comparator).

Before the implementation of the intervention, there was a significant reliance on secondary care for the provision of hepatology services. Referrals to the service at the Sunshine Coast hepatology department were received from primary care, where an initial appointment was booked for a specialist-led, hospital outpatient clinic. Following consultation with a hepatology specialist the patient had a fibrosis assessment using the Fibroscan technology, by a hospital-based nurse who also delivered patient education at this appointment. A follow-up appointment was scheduled, requiring another hospital outpatient visit and after treatment had been prescribed and completed, a final hospital outpatient clinic visit was required to check whether cure had been achieved. If treatment failed, the process was repeated, with the exception of the fibrosis assessment, which was only completed once.

Transition probabilities.

Where available, we used data collected from the study to estimate transition probabilities for the model. For transitions that could not be estimated from primary data, such as death from all causes, we used estimates that were published in the literature. For this type of evaluation, it is not uncommon to use a variety of sources to estimate transition probabilities [10]. We used prospectively collected data to estimate the probability of receiving treatment, the probability of a patient achieving cure and the probability of treatment failing, for both groups. To inform estimates for the usual care group, data was collected for 12 months prior to the new model of care. For estimates associated with the intervention, data was collected over a 20-month study period. To estimate the rare probability of re-infection after virus has been cleared, the same rate that was used in a previously published study was used for this model [7]. All-cause mortality was calculated using age and gender specific estimates from the Australian Bureau of Statistics that represent the modelled cohort [11]. Due to the low percentage of patients in our study being cirrhotic, the background risk of hepatocellular carcinoma in this cohort is very low. As such, due to the unlikeliness of death due to HCV, we assumed no difference in all-cause mortality and HCV-related mortality. The model’s cycle length was weekly, meaning that all probabilities were transformed using a standardised formula [6]. The probability of remaining in a health state was simply calculated as 1 minus the sum of the other probabilities associated with leaving that health state [12]. Sources of evidence and values used in the model are summarised in Table 2.

Quality of life.

Due to a large body of work in the area of HCV quality of life already existing, health utility estimates for use in the model were extracted from the published literature and were not prospectively collected as part of the study. The new drug regimen available for patients living with HCV are widely reported to be safe, with very few adverse events reported in studies from around the world [15,16]. This is supported by the data collected in this project, where there were no serious adverse events reported by patients undergoing treatment, either pre- or post-implementation of the partnership. The literature shows that the new drug regimens prescribed to patients living with HCV are well tolerated and that compliance is generally high, even in vulnerable patient populations, so we have assumed that patient experience of treatment, including patient reported quality of life is largely homogenous [17,18]. Several papers stratify health-related quality of life associated with having HCV according to severity categories, and as such, in our base case we assumed that estimates for ‘moderate’ disease would represent the utility of being ‘infected’ in our model. We also assumed that if treatment failed, and patients restarted their treatment, that health utility would be the same as a first-time infection. In sensitivity analysis, we explored changes to the results if the health utility for the infected health states were varied according to the reported ranges in previously published literature. In comparison to the two infected health states, having the virus cleared was associated with an increased health utility.

Costs.

Costs associated with each health state were measured and valued in 2018 Australian dollars. We used a previously published method for costing the intervention, which included costs associated with staff, space and the resources required to undertake the intervention [19]. Costs were associated with the ‘infected’ and ‘treatment fails’ health states as it is these health states where patients have contact with clinical care and staff, space and consumable costs are accrued. Staff costs such as those associated with triage, booked appointments and patient education were measured and included for both the new model and usual care using a monthly costing survey. Equipment costs, such as the mobile Fibroscanner, travel costs and other costs such as stationery, maintenance contracts and clinic space hire were also measured using routinely collected finance department data, and valued for both the new model and usual care. There are no costs associated with the ‘virus cleared’, ‘treatment’ or ‘dead’ health states. We did not include the cost of drug regimens for treatment of HCV for two reasons–first, the cost of the regimen is heavily subsidised by the Federal government and is not a cost that is of interest to local decision-makers, as they do not foot the bill for treatment. Secondly, because the purpose of the intervention is to increase uptake of HCV treatment, including drug treatment costs would unfairly disadvantage the intervention in comparison to a less effective model of care, where less people are receiving treatment. We did not measure patient out of pocket costs, and the model does not include personal lost productivity or other societal costs. As such, our costing perspective is best described as a restricted healthcare perspective [20].

Model outputs.

The main purpose of this evaluation is to estimate the expected value for money of the new model of care for patients with HCV, compared to usual care. This is indicated by the incremental cost-effectiveness ratio (ICER), where the mean change to costs associated with the new model of care is divided by the mean change in quality adjusted life-years (QALYs) [21]. The ICER is compared against a cost-effectiveness threshold, which is assumed to be the willingness to pay (WTP) for an additional QALY, with ICERs that fall under the threshold deemed to be ‘cost-effective’. The threshold for this study is $28,000 (AUD), which is based on the most recently published estimates for the Australian healthcare setting [22].

Handling uncertainty.

To quantify the impact of uncertainty relating to the model’s inputs on the model outputs, probabilistic sensitivity analysis was undertaken. Ten thousand iterations of the model were undertaken, using Monte Carlo simulations, where each simulation takes a random draw from each parameter’s distribution [10]. Transition probabilities and health utilities were assigned beta distributions, while costs were assigned gamma distributions, which reflects the skew that is usually associated with this type of data [10]. The change to costs and the change to QALYs was recorded for each model simulation, producing 10,000 pairs of incremental costs and effects. However, given that the ratio of two numbers has awkward statistical properties, practical issues with using the ICER for decision-making exist [23]. To simplify this ratio information into a single number, the net monetary benefit (NMB) framework is used. Results are converted from the ICER to a NMB value, through the linear rearrangement of the ICER equation, as follows:

NMB = (WTP threshold x Change in Effects)–Change in Costs

The interpretation of cost-effectiveness becomes particularly simple using the NMB: a positive NMB indicates that a strategy is cost-effective and a negative NMB indicates that a strategy is not cost-effective [24]. Using the net benefit framework gives decision-makers a clear framework for choosing to adopt, or not adopt the intervention being evaluated, without needing an in-depth understanding of the ratio of two numbers.

We also changed key parameter values in the model to reflect plausible scenarios, allowing us to explore aspects of uncertainty in the evaluation and to test the robustness of the model. This also increases the extent of information that is available to decision-makers. We considered six alternate scenarios: (1) where the utility value for the infected health state was lower than in the baseline model; (2) where the probability of cure is the same for both the usual care and intervention group; (3) where the probability of dying from infection in the intervention group is higher than in the baseline model; (4) where the decision-maker’s willingness to pay for health benefits is $0; (5) where the probability of being treated is the same for both the usual care and intervention group; and (6) where the time horizon for the model was extended from one year, to ten years.