diff --git a/docs/papers/covid_19/preprocess/mixing_matrix/contacts_extrapolation.pdf b/docs/papers/covid_19/preprocess/mixing_matrix/contacts_extrapolation.pdf
deleted file mode 100644
index e91c7bda89..0000000000
Binary files a/docs/papers/covid_19/preprocess/mixing_matrix/contacts_extrapolation.pdf and /dev/null differ
diff --git a/docs/papers/covid_19/preprocess/mixing_matrix/contacts_extrapolation_vic.tex b/docs/papers/covid_19/preprocess/mixing_matrix/contacts_extrapolation_vic.tex
new file mode 100644
index 0000000000..c692389d95
--- /dev/null
+++ b/docs/papers/covid_19/preprocess/mixing_matrix/contacts_extrapolation_vic.tex
@@ -0,0 +1,9 @@
+\subsection{Contact matrices construction}
+For each location $L$ (home, school, work, other locations) the age-specific contact matrix $\mathbf{C^L} = (c_{i,j}^L) \in \mathbb{R}_{+}^{16 \times 16}$ is defined such that $c_{i,j}^L$ is the average number of contacts that a typical individual aged $i$ has with individuals aged $j$. As there is no contact survey available for Australia that is complete across all age groups, the matrices $\mathbf{C^L}$ were obtained by extrapolating contact matrices from the United Kingdom, being a country included in the POLYMOD study in 2005 \cite{RN141}. The original matrices from the United Kingdom are denoted $\mathbf{Q^L} = (q_{i,j}^L) \in \mathbb{R}_{+}^{16 \times 16}$, where $q_{i,j}^L$ is defined using the same convention as for $c_{i,j}^L$. The matrices $\mathbf{Q^L}$ were extracted using the R package ``socialmixr'' (v 0.1.8) and then adjusted to account for age distribution differences between Victoria and the United Kingdom.
+
+Let $\pi_j$ denote the proportion of people aged $j$ in Victoria, and $\rho_j$ the proportion of people aged $j$ in the United Kingdom. The contact matrices $\mathbf{C^L}$ were obtained from:
+$$
+c_{i,j}^L = q_{i,j}^L \times \frac{\pi_j}{\rho_j} . 
+$$
+
+In sensitivity analysis, we replaced the United Kingdom with Belgium as the proxy country from which survey data were used for constructing Victoria's matrices. Belgium was chosen as the POLYMOD country with the closest similarity in age distribution of Australia and Belgium in 2005.
diff --git a/docs/papers/tex/covid_library.bib b/docs/papers/tex/covid_library.bib
index b4b0964ea0..cc6d1ff197 100644
--- a/docs/papers/tex/covid_library.bib
+++ b/docs/papers/tex/covid_library.bib
@@ -449,6 +449,19 @@ @article{RN66
    type = {Journal Article}
 }
 
+@article{RN141,
+   author = {Mossong, Joël and Hens, Niel and Jit, Mark and Beutels, Philippe and Auranen, Kari and Mikolajczyk, Rafael and Massari, Marco and Salmaso, Stefania and Tomba, Gianpaolo Scalia and Wallinga, Jacco and Heijne, Janneke and Sadkowska-Todys, Malgorzata and Rosinska, Magdalena and Edmunds, W. John},
+   title = {Social Contacts and Mixing Patterns Relevant to the Spread of Infectious Diseases},
+   journal = {PLOS Medicine},
+   volume = {5},
+   number = {3},
+   pages = {e74},
+   DOI = {10.1371/journal.pmed.0050074},
+   url = {https://doi.org/10.1371/journal.pmed.0050074},
+   year = {2008},
+   type = {Journal Article}
+}
+
 @misc{RN58,
    volume = {2021},
    number = {15 June},
diff --git a/docs/papers/tex/vic_supplement.tex b/docs/papers/tex/vic_supplement.tex
index e6cdf38ee5..efb2224f3f 100644
--- a/docs/papers/tex/vic_supplement.tex
+++ b/docs/papers/tex/vic_supplement.tex
@@ -7,6 +7,8 @@
 \usepackage{soul}  % For highlighting text
 \usepackage{floatrow}  % For changing default font in tables
 \usepackage[font={sf, it}]{caption}
+\usepackage{amsmath}
+\usepackage{amssymb}
 
 \floatsetup[table]{font=sf}
 
@@ -33,6 +35,7 @@
 % Methods
 \input{../covid_19/model_description.tex}
 \input{../covid_19/stratifications/agegroup.tex}
+\input{../covid_19/preprocess/mixing_matrix/contacts_extrapolation_vic.tex}
 \input{../covid_19/stratifications/clinical.tex}
 \input{../covid_19/preprocess/case_detection.tex}
 \input{../covid_19/stratifications/tracing.tex}