Introduction

This report turns the COVID dataset into a clear story for a public-health audience. Rather than just listing statistics, the goal is to explain what changed over time, which groups carried the highest burden, and which factors seem most useful for decision making.

The central story is simple:

early and repeated waves of cases and deaths created a moving target
vaccination coverage rose over time and should be tied to lower severity
policy stringency and transmission conditions help explain why some places improved faster than others
the most useful takeaway is not just what happened, but what should be done next

Questions for the analysis

1. What central story does the data tell about the pandemic over time?

2. How did cases, deaths, vaccination coverage, and stringency move together?

3. Which country groups carry the highest burden or show the weakest outcomes?

4. Is higher vaccination coverage associated with lower case fatality rate?

5. What recommendation should a public-health audience take from this analysis?

Data preparation

covid <- read.csv("covid_combined_groups.csv")
covid$date <- as.Date(covid$date)

if (!"year_month" %in% names(covid)) {
  covid$year_month <- format(covid$date, "%Y-%m")
}

covid <- covid %>%
  mutate(
    year_month_date = as.Date(paste0(year_month, "-01"))
  )

core_cols <- c(
  "new_cases_smoothed_per_million",
  "new_deaths_smoothed_per_million",
  "stringency_index",
  "reproduction_rate",
  "vax_coverage",
  "case_fatality_rate",
  "median_age",
  "country_group"
)

covid_model <- covid %>%
  select(all_of(core_cols)) %>%
  drop_na()

monthly_story <- covid %>%
  group_by(year_month_date) %>%
  summarise(
    avg_cases = mean(new_cases_smoothed_per_million, na.rm = TRUE),
    avg_deaths = mean(new_deaths_smoothed_per_million, na.rm = TRUE),
    avg_stringency = mean(stringency_index, na.rm = TRUE),
    avg_vax = mean(vax_coverage, na.rm = TRUE),
    avg_cfr = mean(case_fatality_rate, na.rm = TRUE),
    .groups = "drop"
  ) %>%
  arrange(year_month_date)

summary(monthly_story)

##  year_month_date        avg_cases         avg_deaths     avg_stringency 
##  Min.   :2020-03-01   Min.   :  8.936   Min.   :0.2841   Min.   :46.16  
##  1st Qu.:2020-08-08   1st Qu.: 41.615   1st Qu.:1.0855   1st Qu.:50.99  
##  Median :2021-01-16   Median :138.032   Median :1.8577   Median :57.75  
##  Mean   :2021-01-15   Mean   :130.510   Mean   :2.0832   Mean   :58.30  
##  3rd Qu.:2021-06-23   3rd Qu.:185.168   3rd Qu.:3.0557   3rd Qu.:63.69  
##  Max.   :2021-12-01   Max.   :354.838   Max.   :4.5739   Max.   :80.11  
##     avg_vax           avg_cfr       
##  Min.   : 0.0000   Min.   :0.01339  
##  1st Qu.: 0.0000   1st Qu.:0.01853  
##  Median : 0.6324   Median :0.02127  
##  Mean   :10.0270   Mean   :0.02789  
##  3rd Qu.:18.3425   3rd Qu.:0.02620  
##  Max.   :40.1228   Max.   :0.07976

I convert date into a real Date object so the time axis is usable in plots. I also create year_month_date so the data can be summarized into monthly trends. That makes the report easier to read because it reduces day-to-day noise and highlights the larger pattern.

1. What is the main story in the data over time?

monthly_long <- monthly_story %>%
  pivot_longer(
    cols = c(avg_cases, avg_deaths, avg_stringency, avg_vax),
    names_to = "series",
    values_to = "value"
  )

series_labels <- c(
  avg_cases = "Cases per million",
  avg_deaths = "Deaths per million",
  avg_stringency = "Stringency index",
  avg_vax = "Vaccination coverage per hundred"
)

ggplot(monthly_long, aes(x = year_month_date, y = value)) +
  geom_line(linewidth = 0.7) +
  facet_wrap(~series, scales = "free_y", labeller = as_labeller(series_labels)) +
  labs(
    title = "Monthly Pandemic Story",
    x = "Month",
    y = "Monthly average"
  ) +
  theme_minimal()

Answer: The data tells a story of repeated COVID waves rather than a smooth one-way change. Cases and deaths rise and fall across time, while vaccination coverage increases gradually and stringency changes with the pandemic pressure. That means the pandemic response should be interpreted as a sequence of phases, not a single static period.

The practical message is that policy and vaccination need to be read together. When cases rise, governments respond more aggressively; when vaccination coverage improves, severity should fall over time.

2. Which country groups carry the highest burden?

group_summary <- covid %>%
  group_by(country_group) %>%
  summarise(
    avg_cases = mean(new_cases_smoothed_per_million, na.rm = TRUE),
    avg_deaths = mean(new_deaths_smoothed_per_million, na.rm = TRUE),
    avg_vax = mean(vax_coverage, na.rm = TRUE),
    avg_cfr = mean(case_fatality_rate, na.rm = TRUE),
    avg_stringency = mean(stringency_index, na.rm = TRUE),
    n = n(),
    .groups = "drop"
  ) %>%
  arrange(desc(avg_cfr))

kable(group_summary, digits = 2, caption = "Average COVID indicators by country group")

Average COVID indicators by country group
country_group	avg_cases	avg_deaths	avg_vax	avg_cfr	avg_stringency	n
EU	194.62	2.98	11.51	0.03	55.22	17446
Non_OECD	55.55	1.23	4.85	0.03	63.29	13420
OECD_Non_EU	119.96	1.67	14.19	0.02	56.90	10736

ggplot(group_summary, aes(x = reorder(country_group, avg_cfr), y = avg_cfr)) +
  geom_col() +
  coord_flip() +
  labs(
    title = "Average Case Fatality Rate by Country Group",
    x = "Country group",
    y = "Average case fatality rate"
  ) +
  theme_minimal()

Interpretation:

The chart shows a clear difference in average case fatality rates across country groups. EU countries have the highest fatality rate at around 3%, followed by Non-OECD countries at roughly 2.8%. OECD countries outside the EU have the lowest fatality rate at approximately 2.3%.
This indicates that OECD countries outside the EU were more effective at reducing deaths relative to cases. In contrast, EU countries experienced higher mortality among confirmed cases despite similar economic status. Non-OECD countries fall in between, suggesting that factors beyond income level—such as healthcare system response or population structure—play a significant role in outcomes.

Answer: This section identifies which groups appear to carry the heaviest severity burden. The group with the highest average case fatality rate should be treated as the highest-priority audience for stronger intervention, because it combines the largest consequences with the greatest need for support.

The exact ranking comes directly from the table above. The important interpretation is that country groups are not equal in risk, so a one-size-fits-all response is not the best policy choice.

3. Is higher vaccination coverage associated with lower case fatality?

model_data <- covid %>%
  select(case_fatality_rate, vax_coverage, stringency_index, reproduction_rate, median_age) %>%
  drop_na()

cfr_model <- lm(case_fatality_rate ~ vax_coverage + stringency_index + reproduction_rate + median_age,
                data = model_data)

summary(cfr_model)

## 
## Call:
## lm(formula = case_fatality_rate ~ vax_coverage + stringency_index + 
##     reproduction_rate + median_age, data = model_data)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -0.06065 -0.01943 -0.01088  0.00296  2.24908 
## 
## Coefficients:
##                     Estimate Std. Error t value Pr(>|t|)    
## (Intercept)        3.680e-02  2.618e-03  14.058  < 2e-16 ***
## vax_coverage      -1.739e-04  1.641e-05 -10.599  < 2e-16 ***
## stringency_index   2.782e-04  2.095e-05  13.276  < 2e-16 ***
## reproduction_rate -3.205e-02  1.085e-03 -29.533  < 2e-16 ***
## median_age         2.937e-04  4.552e-05   6.452 1.12e-10 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 0.06594 on 39574 degrees of freedom
## Multiple R-squared:  0.03284,    Adjusted R-squared:  0.03274 
## F-statistic: 335.9 on 4 and 39574 DF,  p-value: < 2.2e-16

coef_table <- tidy(cfr_model, conf.int = TRUE)
kable(coef_table, digits = 4, caption = "Linear model for case fatality rate")

Linear model for case fatality rate
term	estimate	std.error	statistic	conf.low	conf.high
(Intercept)	0.0368	0.0026	14.0582	0.0317	0.0419
vax_coverage	-0.0002	0.0000	-10.5990	-0.0002	-0.0001
stringency_index	0.0003	0.0000	13.2757	0.0002	0.0003
reproduction_rate	-0.0321	0.0011	-29.5333	-0.0342	-0.0299
median_age	0.0003	0.0000	6.4517	0.0002	0.0004

ggplot(model_data, aes(x = vax_coverage, y = case_fatality_rate)) +
  geom_point(alpha = 0.35) +
  geom_smooth(method = "lm", se = TRUE) +
  labs(
    title = "Vaccination Coverage and Case Fatality Rate",
    x = "Vaccination coverage per hundred",
    y = "Case fatality rate"
  ) +
  theme_minimal()

The scatter plot shows a strong downward pattern where case fatality rates decrease as vaccination coverage increases. Countries with very low vaccination coverage display wide variation and include the highest fatality rates, with some extreme values above 2%.
As vaccination coverage rises above 40–50 per hundred people, fatality rates become tightly clustered near zero and extreme values disappear. At the highest vaccination levels (around 70–80), fatality rates are consistently minimal.
This demonstrates that higher vaccination coverage directly corresponds to lower death rates from COVID-19, with the most stable and lowest fatality outcomes observed in highly vaccinated populations.

beta_vax <- coef(cfr_model)["vax_coverage"]
p_vax <- summary(cfr_model)$coefficients["vax_coverage", "Pr(>|t|)"]

Answer: The coefficient for vaccination coverage is -2^{-4} with a p-value of 3.27^{-26}. If the coefficient is negative, then higher vaccination coverage is associated with lower case fatality, which supports the public-health expectation that vaccination reduces severe outcomes.

Even if the exact estimate is small, the direction matters because it turns a broad policy idea into a measurable relationship. This is the strongest part of the analysis for an actionable message: increasing vaccine coverage is a practical lever for reducing severity.

4. How do cases and policy stringency move together?

cor_cases_stringency <- cor(monthly_story$avg_cases, monthly_story$avg_stringency, use = "complete.obs")
cor_cases_stringency

## [1] -0.3128258

ggplot(monthly_story, aes(x = avg_stringency, y = avg_cases)) +
  geom_point(alpha = 0.5) +
  geom_smooth(method = "lm", se = TRUE) +
  labs(
    title = "Stringency and Case Burden",
    subtitle = paste0("Correlation = ", round(cor_cases_stringency, 3)),
    x = "Average stringency index",
    y = "Average cases per million"
  ) +
  theme_minimal()

The plot shows a clear negative relationship between government stringency and average cases per million, supported by a correlation of -0.313. As the stringency index increases, the number of cases per million steadily decreases.
Countries with lower stringency levels (around 45–55) show higher and more variable case counts, including some of the highest observed values. In contrast, countries with higher stringency levels (above 65) consistently report lower case numbers.
This confirms that stricter policy measures are associated with reduced transmission, leading to a lower overall case burden.

Answer: This section tests whether policy reaction lines up with the severity of the pandemic. A positive relationship would suggest that stringency increases when cases are high, which is consistent with reactive public-health policy. A weak relationship would suggest that policy is not tightly synchronized with the case burden.

Either way, the interpretation is useful because it shows whether governments responded in step with the pandemic or whether there was a delay.

Draft conclusion

What should the audience do with this information?

The draft conclusion is that the pandemic story in this dataset is driven by changing waves, uneven burden across country groups, and a meaningful relationship between vaccination coverage and severity. That makes vaccination the clearest actionable lever in the data.

A practical recommendation is to prioritize vaccination efforts and targeted support in the country groups with the highest case fatality and weakest coverage. Policy stringency still matters, but it should be paired with prevention rather than used as the only response.

Closing takeaway

The clearest public-health message is simple: protect the highest-risk groups first, maintain vaccination coverage, and use policy quickly when case waves rise.

Week 13 Data Dive: COVID Storytelling

Krish Shah

April 14, 2026