Overview of Epidemiology Model
A schematic overview of the study epidemiologic model is shown in Figure S1. The model was built in Microsoft Excel. Mid-year population estimates for each country were taken from the US Census Bureau [
24]. The model was created using the following steps, which were conducted or calculated for each country. First, the annual number of Fontan procedural codes for the most recent 5 years available were extracted from hospital or claims databases and averaged to give a single estimate for the 2010–2020 period. Second, the frequency of surgery conducted in last 5 years by country was applied to mid-year populations since 1972. And third, an uptake curve for the Fontan procedure (overall) and subtype utilisation curves were modelled and used to adjust the annual numbers of AP, LT-TCPC and EC-TCPC Fontan surgeries since 1972. A literature search was conducted to determine survival rates and the median age at Fontan surgery for each Fontan procedure type. Long-term (30-year) survival curves for each procedure type were generated and used to estimate the number of patients alive in 2020. The median age at which Fontan procedure types occurred was then applied to stratify this prevalence estimate by age cohort. The prevalence estimates were then extrapolated through to 2030.
1.
Estimation of the number of Fontan procedures conducted annually between 1972 and 2020
1.1
Extraction of Fontan procedural codes from databases for the 2010–2020 period
The number of Fontan procedures conducted annually was estimated by extracting the most recent 5 years of available data (between 2010 and 2020) of Fontan procedure codes from a claims or administrative database for each of the countries except Australia and New Zealand. For Australia and New Zealand, the number of procedures were taken from the binational registry [
14]. Data for these two countries were grouped together throughout the above-outlined steps.
Procedure codes for all three types of Fontan surgeries (AP, LT-TCPC and EC-TCPC) were extracted from the available claims or hospital databases, with hospital databases being the preferred data source. Table S1 summarises the databases and the procedure codes used. For example, a US claims database, the COMPILE database, was used to extract procedure codes for the US. Data for the most recent 5 years available (over the 2010–2020 period) were extracted and entered into the model (Table S1). Wherever possible, the procedure counts were checked against another hospital or claims database. For example, procedure counts from the US COMPILE database were cross-checked against reports from the National Inpatient Sample (NIS) by Akintoye et al. [
25] (Table S1).
1.2
Application of the uptake curve to estimate the number of Fontan procedures occurring prior to 2010
The extracted procedure counts (Sect.
1.1) were converted to procedure rates (defined as number of procedures per 100,000 inhabitants per year). These procedure rates were applied to their respective national populations between 1972 and 2020. To account for (1) gradual uptake or implementation of Fontan surgery into surgical procedure catalogue of national healthcare institutions and (2) utilisation of different procedure types between 1972 and 2020; a Fontan uptake curve was created and applied (Figure S2). This uptake curve was modelled based predominantly on data from The Australia and New Zealand Fontan Registry, and it assumed a slow initial uptake of the procedure as surgery teams developed more experience with the procedure [
26]. The uptake curve plateaued by 2001 and therefore the average rate of procedures after this point was assumed to represent 100% uptake for Fontan surgeries. In other words, there was no capacity constraint to conduct Fontan surgeries in any country. Using this assumption, the curve was applied to adjust and estimate the average procedure rates in each country between 1972 and 2010.
1.3
Applying subtype utilisation curves to obtain the average annual number of AP, LT-TCPC and EC-TCPC procedures
Subtype utilisation curves were developed to model the uptake of each of the procedure types, within this overall rate of uptake. These curves were modelled based on data from The Australia and New Zealand Fontan Registry and the systematic literature review by Kverneland et al. [
26] (Figure S3). They were applied to the annual number of overall conducted Fontan procedures by year and country since 1972 to yield annual procedure numbers for AP, LT-TCPC and EC-TCPC by year and country.
A comprehensive literature search was conducted to identify published estimates of the following parameters (all split by procedure type): median age at Fontan procedure, 30-day mortality rates and long-term survival rates (10, 15, 20, 25, 30 and 35-years post-Fontan).
PubMed was searched in December 2020 for relevant articles published at any time, with a focus on those from the last 5 years. The following search terms for the title or abstract were used: ‘mortality’ or ‘survival’ and ‘fontan’ or ‘TCPC’ or ‘total cavopulmonary anastomosis’ and ‘long term’ or ‘long-term’ or ‘intermediate’ and ‘outcome’ or ‘result’ or ‘prognosis’. The literature search results were filtered to exclude non-human studies, articles not written in the English language and those without abstract or full-text available. Duplicate records were also removed. The remaining articles were reviewed and excluded if: (1) data were not based on any of the 11 countries included in this study, (2) the study sample size was low (n < 100 participants), (3) methodology was unclear and unable to be reproduced, (4) the study was specific to a narrow subset of patients which may not be representative (e.g. a focus on patients with specific aetiology or only adult patients) (5) if the reported data had a wide variance, (6) if data were outdated (prior to 2000), or (7) if the study author indicated a high degree of uncertainty.
For each of the parameters, an average was calculated from all published estimates. These averages were assumed to apply to all 11 countries, as healthcare systems are similarly developed across these countries and country-specific data were scarce in the literature.
3.
Calculation of Fontan prevalence overall and by age and subtype categories, using data derived from literature search
Survival curves for the procedure subtypes were created (based on literature-derived survival rates) and used to calculate prevalence of persons living with Fontan. The survival curves (Figure S4) were created using 30-day mortality after the last Fontan surgery as the input for timepoint 0 (i.e. for a procedure type with a 30-day mortality of 10%, the survival curve began at 90% at timepoint 0) (Figure S4). Between timepoint 0, and the first year of input from the literature (10 years post-Fontan), a linear decrease in survival was assumed. This assumption was also applicable between the other years of literature input (for example, between 10 and 15 years post-Fontan). Beyond the last year of input from the literature, a year-on-year decrease in survival of 1.25% was assumed. This decrease is similar to the year-on-year mortality for a general population, and is also in line with reports of a constant 1.3% mortality rate per year beyond 20 years post-Fontan from Dabal et al. [
27]. These survival curves (Figure S4) were applied to the number of AP, LT-TCPC and EC-TCPC patients between 1972 and 2020 to yield prevalence estimates. The average age of patients at time of surgery (by procedure type) was then applied to estimate the prevalence split by age group.
4.
Verification of the model’s outputs
Wherever possible, the model outputs were cross-checked with a second or multiple other sources from that country. Published real-world data on Fontan prevalence were available from two studies from Australia and New Zealand (Schilling et al. [
14], Iyengar et al. [
28]) and from one study from Sweden (Dahlqvist et al. [
29]). These results were compared with the estimates from our model to determine whether our data were consistent with the available literature.
The epidemiology model was used to extrapolate the prevalence of Fontan through to 2030. In order to project this information, US census bureau projections of population size were used and stable procedure rates between 2020 and 2030 were assumed. In addition, we assumed that factors such as access to care and proportion of pregnancy terminations remained stable.