Recent Research

# Recent Research
### Sam Abbott (👩‍🔬; @seabbs)
### University of Bristol
### 2019/11/18

---

# Outline

## 1. Who am I?

## 3. Research

## 4. Reproducibility

## 5. Software

## 6. Interactive tools

---
class: center, middle, inverse

# Who am I?
---

## Slides: [bit.ly/seabbs-recent-research](http://bit.ly/seabbs-recent-research)

## Slide code:  [bit.ly/seabbs-recent-research-code](http://bit.ly/seabbs-recent-research-code)

## Code: [github.com/seabbs](https://www.github.com/seabbs)

## Personal site + blog: [samabbott.co.uk](https://www.samabbott.co.uk)

## Twitter: [@seabbs](https://twitter.com/seabbs)

---

# Background

* Postdoctoral researcher in infectious disease dynamics at University of Bristol (since September).1

* PhD: Modelling tuberculosis and the optimal usage of the BCG vaccination.2

* Previously: Mathematics, theoretical physics, applied mathematics, and data science in finance.

.footnote[
[1] More detail here: [samabbott.co.uk](https://www.samabbott.co.uk/)
[1] Thesis (in bookdown): [bit.ly/seabbs-thesis](http://bit.ly/seabbs-thesis)
]

---
# Interests

* Infectious disease dynamics.

* Improving infectious disease modelling practice and the scientific workflow more generally.

* Informing policy and enabling others to do so in a more robust manner.

* Engaging the wider community with science and actively involving them where possible.

---
class: center, middle, inverse

# Research

---

# Publications1

---

# Data analysis

#### Evidence for associations between BCG vaccination and TB outcomes.1

* Some evidence in the literature for non-specific effects of the BCG vaccine.
 
* Selected several that are most commonly reported and then used logistic regression to explore the evidence for associations.
 
#### Missing data in the Enhanced Surviellance (ETS) system2

* Structure of missing data.
    
* Logistic regression used to explore the relationships between demographic variables and missing data for key outcomes.

[1] Paper: [doi.org/10.1016/j.vaccine.2019.06.056](https://doi.org/10.1016/j.vaccine.2019.06.056); Code: [github.com/seabbs/ExploreBCGOnOutcomes](https://github.com/seabbs/ExploreBCGOnOutcomes)

[2] Draft: [samabbott.co.uk/ETSMissing/articles/paper.html](https://www.samabbott.co.uk/ETSMissing/articles/paper.html); Code: [github.com/seabbs/ETSMissing](https://github.com/seabbs/ETSMissing)
]
 
---

# Statistical modelling

#### Estimating the impact of the 2005 change in BCG vaccination policy1

* Regression analysis looking at TB incidence rates pre and post the change in BCG policy in the targeted populations.

* Developed multiple model structures and confounders using expert knowledge.

* Ranked models based on fit to the available data (WAIC and LOO).

#### Modelling zoonotic tuberculosis burden (*in progress*)

* Data sourced from multiple sources, cleaned, combined, and mapped
    
 * Modelled using GAMs in a bayesian framework so that error in outcome and error in variables can be accounted for. 
 
 * Presented as an R package to be reused as data improves.

[1] Preprint: [doi.org/10.1101/567511](https://doi.org/10.1101/567511); Code: [github.com/seabbs/DirectEffBCGPolicyChange](https://github.com/seabbs/DirectEffBCGPolicyChange)

---

# Dynamic modelling

#### Developed a model of BCG vaccination and TB transmission in the UK.

* Semi-stochastic dynamic transmission model with a high -> low risk latency structure.

* Age-stratified with realistic demographics.

* Externally forced based on non-UK born incidence.

#### Model fitting

* Model fitting pipeline developed in R including SMC-SMC (using `LibBi` and `rbi`) model comparison (AIC), parameter sensitivity (PRCC), and scenario comparison (mixing structure).

* Final model fit (at PhD hand in) was extremely poor even with post-hoc calibration.

* Conclusion: Model overly complex for approach and needed simplification.

[1] [samabbott.co.uk/thesis/8-model-development.html](https://www.samabbott.co.uk/thesis/8-model-development.html)

---

# Other

#### Dynamic model of BCG vaccination and TB transmission globally

* Aim: What is the incidence rate thresold at which an effective TB vaccine should be given to adults rather than neonates (assuming limited duration).

* Model implemented using `pomp`.

* Project dropped in favour of previous modelling project.

#### Contributing to improving cost effectiveness tooling1

* BCEA package review and overhaul.
    
* Profiling Markov modelling code + speeding up using a functional approach and C code.

[1] https://github.com/giabaio/BCEA/issues

---
class: center, middle, inverse

#Software
---

# [`{getTBinR}`](https://bit.ly/getTBinR)1, 2, 3

.footnote[
[1] Docs: [bit.ly/getTBinR](https://bit.ly/getTBinR), 
[2] Paper: [bit.ly/getTBinR-paper](http://bit.ly/getTBinR-paper)
[3] Blog: [github.com/seabbs/hpru-blog-post-getTBinR](https://github.com/seabbs/hpru-blog-post-getTBinR)
]
---

# [`{idmodelr}`](https://bit.ly/idmodelr)1

.footnote[
[1] Docs: [bit.ly/idmodelr](https://bit.ly/idmodelr), [2] Presentation: [bit.ly/idmodelr-satrday-cardiff](https://bit.ly/idmodelr-satrday-cardiff)
]

---
class: center, middle, inverse

# Interactive tools
---

# Explore infectious disease models

```r
#install.packages("shiny")
shiny::runGitHub("exploreidmodels", "seabbs")
```

.footnote[
[1] More interactive tools here: [samabbott.co.uk/#projects](https://www.samabbott.co.uk/#projects)
]
---
class: inverse, center, middle

# Thanks for listening!

## Slides: [bit.ly/seabbs-recent-research](http://bit.ly/seabbs-recent-research)

## Slide code:  [bit.ly/seabbs-recent-research-code](http://bit.ly/seabbs-recent-research-code)

## Code: [github.com/seabbs](https://www.github.com/seabbs)

## Personal site + blog: [samabbott.co.uk](https://www.samabbott.co.uk)

## Twitter: [@seabbs](https://twitter.com/seabbs)