11 KiB
Incidence of influenza-like illness in France
Foreword
For running this analysis, you need the following software:
Emacs 25 or higher
Older versions may suffice. For Emacs versions older than 26, org-mode must be updated to version 9.x.
Python 3.6 or higher
We use the ISO 8601 date format, which has been added to Python's standard library with version 3.6.
import sys
if sys.version_info.major < 3 or sys.version_info.minor < 6:
print("Please use Python 3.6 (or higher)!")(unless (featurep 'ob-python)
(print "Please activate python in org-babel (org-babel-do-languages)!"))R 3.4
We use only basic R functionality, so a earlier version might be OK, but we did not test this.
(unless (featurep 'ob-R)
(print "Please activate R in org-babel (org-babel-do-languages)!"))Data preprocessing
The data on the incidence of influenza-like illness are available from the Web site of the Réseau Sentinelles. We download them as a file in CSV format, in which each line corresponds to a week in the observation period. Only the complete dataset, starting in 1984 and ending with a recent week, is available for download. The URL is:
http://www.sentiweb.fr/datasets/incidence-PAY-3.csv
This is the documentation of the data from the download site:
| Column name | Description |
|---|---|
week |
ISO8601 Yearweek number as numeric (year*100 + week nubmer) |
indicator |
Unique identifier of the indicator, see metadata document https://www.sentiweb.fr/meta.json |
inc |
Estimated incidence value for the time step, in the geographic level |
inc_low |
Lower bound of the estimated incidence 95% Confidence Interval |
inc_up |
Upper bound of the estimated incidence 95% Confidence Interval |
inc100 |
Estimated rate incidence per 100,000 inhabitants |
inc100_low |
Lower bound of the estimated incidence 95% Confidence Interval |
inc100_up |
Upper bound of the estimated rate incidence 95% Confidence Interval |
geo_insee |
Identifier of the geographic area, from INSEE https://www.insee.fr |
geo_name |
Geographic label of the area, corresponding to INSEE code. This label is not an id and is only provided for human reading |
The ISO-8601 format is popular in Europe, but less so in North America. This may explain why few software packages handle this format. The Python language does it since version 3.6. We therefore use Python for the pre-processing phase, which has the advantage of not requiring any additional library. (Note: we will explain in module 4 why it is desirable for reproducibility to use as few external libraries as possible.)
Download
After downloading the raw data, we extract the part we are interested in. We first split the file into lines, of which we discard the first one that contains a comment. We then split the remaining lines into columns.
from urllib.request import urlopen
data = urlopen(data_url).read()
lines = data.decode('latin-1').strip().split('\n')
data_lines = lines[1:]
table = [line.split(',') for line in data_lines]Let's have a look at what we have so far:
table[:5]Checking for missing data
Unfortunately there are many ways to indicate the absence of a data value in a dataset. Here we check for a common one: empty fields. For completeness, we should also look for non-numerical data in numerical columns. We don't do this here, but checks in later processing steps would catch such anomalies.
We make a new dataset without the lines that contain empty fields. We print those lines to preserve a trace of their contents.
valid_table = []
for row in table:
missing = any([column == '' for column in row])
if missing:
print(row)
else:
valid_table.append(row)Extraction of the required columns
There are only two columns that we will need for our analysis: the first ("week") and the third ("inc"). We check the names in the header to be sure we pick the right data. We make a new table containing just the two columns required, without the header.
week = [row[0] for row in valid_table]
assert week[0] == 'week'
del week[0]
inc = [row[2] for row in valid_table]
assert inc[0] == 'inc
del inc[0]
data = list(zip(week, inc))
Let's look at the first and last lines. We insert None to indicate to org-mode the separation between the three parts of the table: header, first lines, last lines.
[('week', 'inc'), None] + data[:5] + [None] + data[-5:]Verification
It is always prudent to verify if the data looks credible. A simple fact we can check for is that weeks are given as six-digit integers (four for the year, two for the week), and that the incidence values are positive integers.
for week, inc in data:
if len(week) != 6 or not week.isdigit():
print("Suspicious value in column 'week': ", (week, inc))
if not inc.isdigit():
print("Suspicious value in column 'inc': ", (week, inc))No problem - fine!
Date conversion
In order to facilitate the subsequent treatment, we replace the ISO week numbers by the dates of each week's Monday. This is also a good occasion to sort the lines by increasing data, and to convert the incidences from strings to integers.
import datetime
converted_data = [(datetime.datetime.strptime(year_and_week + ":1" , '%G%V:%u').date(),
int(inc))
for year_and_week, inc in data]
converted_data.sort(key = lambda record: record[0])We'll look again at the first and last lines:
str_data = [(str(date), str(inc)) for date, inc in converted_data]
[('date', 'inc'), None] + str_data[:5] + [None] + str_data[-5:]Date verification
We do one more verification: our dates must be separated by exactly one week, except around the missing data point.
dates = [date for date, _ in converted_data]
for date1, date2 in zip(dates[:-1], dates[1:]):
if date2-date1 != datetime.timedelta(weeks=1):
print(f"The difference between {date1} and {date2} is {date2-date1}")Transfer Python -> R
We switch to R for data inspection and analysis, because the code is more concise in R and requires no additional libraries.
Org-mode's data exchange mechanism requires some Python code for transforming the data to the right format.
[('date', 'inc'), None] + [(str(date), inc) for date, inc in converted_data]In R, the dataset arrives as a data frame, which is fine. But the dates arrive as strings and must be converted.
data$date <- as.Date(data$date)
summary(data)Inspection
Finally we can look at a plot of our data!
plot(data, type="l", xlab="Date", ylab="Weekly incidence")A zoom on the last few years makes the peaks in winter stand out more clearly.
plot(tail(data, 200), type="l", xlab="Date", ylab="Weekly incidence")Study of the annual incidence
Computation of the annual incidence
Since the peaks of the epidemic happen in winter, near the transition between calendar years, we define the reference period for the annual incidence from August 1st of year N to August 1st of year N+1. We label this period as year N+1 because the peak is always located in year N+1. The very low incidence in summer ensures that the arbitrariness of the choice of reference period has no impact on our conclusions.
This R function computes the annual incidence as defined above:
yearly_peak = function(year) {
debut = paste0(year-1,"-08-01")
fin = paste0(year,"-08-01")
semaines = data$date > debut & data$date <= fin
sum(data$inc[semaines], na.rm=TRUE)
}We must also be careful with the first and last years of the dataset. The data start in October 1984, meaning that we don't have all the data for the peak attributed to the year 1985. We therefore exclude it from the analysis. For the same reason, we define 2018 as the final year. We can increase this value to 2019 only when all data up to July 2019 is available.
years <- 1986:2018
We make a new data frame for the annual incidence, applying the function yearly_peak to each year:
annnual_inc = data.frame(year = years,
incidence = sapply(years, yearly_peak))
head(annnual_inc)Inspection
A plot of the annual incidence:
plot(annnual_inc, type="p", xlab="Année", ylab="Annual incidence")Identification of the strongest epidemics
A list sorted by decreasing annual incidence makes it easy to find the most important ones:
head(annnual_inc[order(-annnual_inc$incidence),])Finally, a histogram clearly shows the few very strong epidemics, which affect about 10% of the French population, but are rare: there were three of them in the course of 35 years. The typical epidemic affects only half as many people.
hist(annnual_inc$incidence, breaks=10, xlab="Annual incidence", ylab="Number of observations", main="")