Health 4.0: Data Science for Global Public Health

class: center, middle, inverse, title-slide

# Health 4.0: Data Science for Global Public Health
## <div class="line-block">Harvard T.H. Chan School of Public Health<br />
NIOSH Education and Research Center Environmental and Occupational Medicine and Epidemiology Program</div>
### <div class="line-block">Thierry Warin, PhD<br />
Professor, HEC Montreal</div>

---

class: center

# Consent and Disclosures

This presentation is being recorded; I consent to this recording.

I have no relevant financial interests to disclose.

Thierry Warin, PhD

---

.pull-left[

### Privacy Notice (for participants)

In order to protect your privacy we have:

- Muted your microphones

- Disabled your cameras

Please enter your questions in the chat box.

### Session Evaluation (all participants complete)

Please scan the QR code using your phone camera or the QR app on your phone.

]

.pull-right[

If you prefer a survey link to enter into your browser, email Ann (<mailto:abackus@hsph.harvard.edu>), or click [[here]](https://harvard.az1.qualtrics.com/jfe/form/SV_elKbdJcrhSV2MUS
), or wait for the last slide (The QR code and link will appear on the last slide at the end of the presentation).

Please complete and submit the evaluation, using the QR code or link.

Questions:  <mailto:abackus@hsph.harvard.edu>

Thank you.

]

---
class: inverse

> Learning objectives

Participants will be able to:

1) describe the transformative role of data science for public health;

2) give examples of data science in public health policies;

3) differentiate data science from artificial intelligence applications in public health;

4) discuss public health policies in light of data science.

> This (and more) Material: <https://warin.ca/posts/conference-health4/>

---

## 3 Lessons

1. Data science and AI are about code, and are also mostly about protocols

2. Governments should consider data science as a top priority for public health policies = **Augmented Intelligence**

3. Public health as a **platform**: the new **business model** of the health sector?

---
class: inverse, center, middle

# Introduction

---

## Introduction

### IR 4.0 and its impact on the global economy

### Architectural revolution

- architectural innovation for companies

- architectural innovation for governments (Government 4.0)

---

## From the digital economy to the platform economy

---
class: inverse

## The A.I. revolution

---
class: inverse

## The A.I. revolution

---
class: inverse

## The A.I. revolution

### ... to Society 5.0 (from Japan)

---
class: inverse, center, middle

# Outline

---

## Outline

1. Transformative Role of Data Science for Public Health

2. Examples of AI in public health policies

3. Difference between data science and artificial intelligence

4. Data Science in public health policies: A case study

---
class: inverse, center, middle

# 1. Transformative Role of Data Science for Public Health

---

## 1. Transformative Role of Data Science for Public Health

### What is medicine and what is public health?

> To answer this question, we will talk about statistics a lot. When I say a lot, I mean a lot.

---

## 1. Transformative Role of Data Science for Public Health

.panelset[
.panel[.panel-name[Not this kind of AI]

]

.panel[.panel-name[Not this kind of DS]

]

.panel[.panel-name[This kind of AI]

]

.panel[.panel-name[This kind of DS]

]
]

---

## 1. Transformative Role of Data Science for Public Health

First dimension: medicine and public health

- The medical field has also always been statistics heavy.

- Public Health has always been statistics heavy.

Second dimension: prevention and treatment

- The practice of medicine in our societies is both about **prevention** and **treatment**.

- In both dimensions, statistics are present.

- Besides statistics and its use in order to help diagnosis, **treatment** involves a biological perspective.

---

## 1. Transformative Role of Data Science for Public Health

> A public policy is about determining attainable goals for a society.

Once the goals are defined, for instance **a higher life expectancy** for the overall society, it is a matter of:

- **allocating the resources** (funding and human resources) and

- **organizing an efficient system**.

---

## 1. Transformative Role of Data Science for Public Health

The question we are interested in here is the architectural innovation brought by AI and data science to improve the efficiency of the system.

Why does it matter?

-	It is **massive**

-	It is indeed a **revolution**

-	It is **scalable**

---

## 1. Transformative Role of Data Science for Public Health

The question we are interested in here is the **architectural innovation** brought by **AI** and **data science** to improve the efficiency of the system.

Why does it matter?

-	It is massive

-	It is indeed a revolution

-	It is scalable

> It matters because it will change public health policies.

---
class: inverse, center, middle

# 2. Examples of AI in public health policies

---

## 2. Examples of AI in public health policies

.panelset[
.panel[.panel-name[Concepts]

Personal data have a new dimension:

Before: vital signs

Now: life style: structured and unstructured data

Results: thanks to machine learning, we can build **predictive models** with massive data, combining vital signs with life-style data in a personalized way.

]

.panel[.panel-name[Today]

]

.panel[.panel-name[Tomorrow]

]

.panel[.panel-name[Augmented Intelligence]

]
]

---

.panelset[
.panel[.panel-name[Predicting strokes]

Features Information of 'healthcare-dataset-stroke-data.csv' stored here: <https://www.warin.ca/datalake/health_data/healthcare-dataset-stroke-data.csv>.

Code here: <https://warin.ca/shiny/conferences/harvard-health/>

1) id: unique identifier

2) gender: "Male", "Female" or "Other"

3) age: age of the patient

4) hypertension: 0 if the patient doesn't have hypertension, 1 if the patient has hypertension

5) heart_disease: 0 if the patient doesn't have any heart diseases, 1 if the patient has a heart disease

]

.panel[.panel-name[Predicting strokes]

6) ever_married: "No" or "Yes"

7) work_type: "children", "Govt_jov", "Never_worked", "Private" or "Self-employed"

8) Residence_type: "Rural" or "Urban"

9) avg_glucose_level: average glucose level in blood

10) bmi: body mass index

11) smoking_status: "formerly smoked", "never smoked", "smokes" or "Unknown"*

12) stroke: 1 if the patient had a stroke or 0 if not

]

.panel[.panel-name[Code]

```r
# Load data
data_stroke = readr::read_csv("https://www.warin.ca/datalake/health_data/healthcare-dataset-stroke-data.csv")
```

```r
# knn with k-fold cross-validation
modelknnStroke <- caret::train(form = stroke ~ age + ever_married + smoking_status + work_type  , data = trainStroke,
                method = "knn",
                tuneGrid = expand.grid(k = seq(1, 101, by = 2)),
                preProcess = c("center", "scale") ,
                trControl = trainControl(method = "cv", number = 3))
```

]

.panel[.panel-name[Results]

![](index_files/figure-html/unnamed-chunk-18-1.png)

]

.panel[.panel-name[Code]

```r
# Knn predict on the test data set
knnPredictStroke <- predict(modelknnStroke, newdata = testStroke)
caret::confusionMatrix(knnPredictStroke, as.factor(testStroke$stroke))
```

]

.panel[.panel-name[Results]

```
## Confusion Matrix and Statistics
## 
##           Reference
## Prediction   0   1
##          0 972  49
##          1   0   0
##                                          
##                Accuracy : 0.952          
##                  95% CI : (0.937, 0.9643)
##     No Information Rate : 0.952          
##     P-Value [Acc > NIR] : 0.5379         
##                                          
##                   Kappa : 0              
##                                          
##  Mcnemar's Test P-Value : 7.025e-12      
##                                          
##             Sensitivity : 1.000          
##             Specificity : 0.000          
##          Pos Pred Value : 0.952          
##          Neg Pred Value :   NaN          
##              Prevalence : 0.952          
##          Detection Rate : 0.952          
##    Detection Prevalence : 1.000          
##       Balanced Accuracy : 0.500          
##                                          
##        'Positive' Class : 0              
## 
```

]

---
class: inverse, center, middle

# 3. Difference between data science and artificial intelligence

---

## 3. Difference between data science and artificial intelligence

AI (and ML) is very useful to do what we just defined.

Now data science, **the science and the art of extracting information from structured and unstructured data**, is incredibly useful for public policies.

Data science can bring a more **macro*economic role** perspective. **AI is about the mechanics, Data Science is AI applied to a domain**, here, health.

- Google trends

- Immunization discourses: We collected almost 657,000 coronavirus-related tweets between January 1 OR 2018 and November 09 OR 2020 OR using a set of predefined search terms (“vaccine” OR “vaccines” OR “vaccinate” OR “vaccination” OR “vaccineswork” OR “antivax” OR “vaccinesdontwork” OR “provax” OR “vaxwithme” OR “antivaxxers” OR “immunization”) [exact total number: 656 994 tweets]. Source: <https://warin.ca/posts/article-vaccination-anti-pro/>

- Vaccines' brand discourses: We collected vaccines’ name related tweets between January 10, 2020 and March 24, 2021 using the list of vaccines’ name currently in use as predefined search terms (“Cormirnaty” OR “CoronaVac” OR “Covaxin” OR “Covishield” OR “SputnikV”). Source: <https://warin.ca/posts/article-vaccines-name/>

---

## 3. Difference between data science and artificial intelligence

.panelset[
.panel[.panel-name[Google Trends]

.pull-left[

We can listen to the world:

]

.pull-right[

]

.panel[.panel-name[Immunization discourses]

.pull-left[

]

.pull-right[

]

.panel[.panel-name[Vaccines' brands discourses]

.pull-left[

]

.pull-right[

]

---

## 3. Difference between data science and artificial intelligence

.pull-left[
<img src="images/fig16.png" width="60%" style="display: block; margin: auto;" />

]

.pull-right[

Port-au-Prince, Spring 2018

]

---

## 3. Difference between data science and artificial intelligence

.pull-left[
<img src="images/fig17.png" width="600px" style="display: block; margin: auto;" />

]

.pull-right[

RQ: In countries embracing a booming controversial industry (fracking), is there resonance or dissonance between the public’s risk perception and the state of  academic research?</li>

- Analysing 60,000 geo-located tweets and (10,000; 600) peer-reviewed articles revealed a very high acceptability despite the major concerns of the scientific community (seismicity, health & environmental impacts, occupational health)</li>

]

---
class: inverse, center, middle

# 4. Data Science in public health policies: A case study

---

## 4. Data Science in public health policies

.panelset[
.panel[.panel-name[EpiBibR]

We can also help research with data science pipelines

-	EpibibR

This package is about global research on coronaviruses. It proposes 22 features for the 130,098 references (on March 15, 2021) and access to the data has been made as easy as possible in order to integrate efficiently in almost any researcher’s pipeline.

]

.panel[.panel-name[Global Research]

]

.panel[.panel-name[Global Health]

]

.panel[.panel-name[Global Health]

]

---
class: inverse, center, middle

# Conclusion

---

## Conclusion

The era of data is upon us. It is proliferating at an unprecedented pace, reflecting every aspect of our lives and circulating from satellites in space through the phones in our pockets.

The data revolution creates endless opportunities to confront the grand challenges of the 21st century. Yet, as the scale and scope of data grow, so must our ability to analyze and contextualize it.

Drawing genuine insights from data requires training in statistics and computer science, and subject area knowledge.

Putting insights into action requires a careful understanding of the potential ethical consequences - for both individuals and entire societies.

- Privacy issues => data privacy and AI ethics literacy

> Augmented Intelligence

---

## Conclusion

To open up the conversation, public health policies at a time of big data

- Public health as a "platform": the new "business model" of the health sector?

- The platform economy: From public health as a public good to public health as a private good

- To find other material on the topic: <https://warin.ca/posts/conference-health4/>

---
class: inverse, center, middle

# Thank you!

---

# References

- <https://warin.ca/posts/conference-health4/>

- de Marcellis-Warin N, Munoz J.M., Warin Th. ["Government 4.0 and the pandemic"](https://cmr.berkeley.edu/2020/06/government/), *California Management Review Insights*, June 24, 2020 [[DOI: 10.6084/m9.figshare.12580091.v1](https://cmr.berkeley.edu/2020/06/government/)]

- Warin T, “Global Research on Coronaviruses: An R Package”, J Med Internet Res 2020; 22(8):e19615, DOI: 10.2196/19615, PMID: 32730218, PMCID: 7423387 <https://www.jmir.org/2020/8/e19615/>

- <https://github.com/warint/EpiBibR>

- <https://www.england.nhs.uk/wp-content/uploads/2020/01/advanced-forecasting-techniques.pdf>

- Competing in the Age of AI How machine intelligence changes the rules of business by Marco Iansiti and Karim R. Lakhani (2020)

---

## Evaluation - Thierry Warin, PhD

.pull-left[

SESSION EVALUATION (all participants please complete)

Please scan the QR code using your phone’s camera or the QR app on your phone

Or enter the evaluation link into your browser <https://harvard.az1.qualtrics.com/jfe/form/SV_elKbdJcrhSV2MUS>

Or request the evaluation link from Ann <mailto:abackus@hsph.harvard.edu>

Please complete and submit the evaluation

Questions:  <mailto:abackus@hsph.harvard.edu>

Thank you.
 
]

.pull-right[

]