ATLAS Tutorial: Data Sources - Data Density

Summary

This video tutorial explores the data density report feature in Atlas, a tool within the OHDSI framework. It explains how to visualize data accumulation across different domains like conditions, drugs, and procedures over time. The report offers insights into the total number of records, records per person, and distinct concepts per person, using graphs and box plots to illustrate data distribution and density. This analysis helps users understand the growth and stability of data within their domains.

Takeaways

• 🗺️ The script introduces the data sources capability of Atlas, focusing on the data density report.
• 🔍 Users can select a data source and explore various reports, starting with the data density report.
• 📊 The first graph in the data density report shows the total number of data rows by time and data domain.
• 📅 The x-axis of the graph represents calendar months, and the y-axis shows the number of records.
• 🔑 Each data domain (conditions, drugs, observations, procedures, and visits) is represented by a separate line series.
• 👁️ Hovering over a line in the graph reveals the specific calendar month and the number of records for a domain.
• 📈 The second graph normalizes the data to show records per person, helping to understand data growth relative to the number of persons.
• 🧐 By hovering, one can see the average number of records per person in a specific domain for a given month, like 2.45 records in the condition occurrence domain for October 2015.
• 📊 The third graph provides a box plot for concepts per person across selected data domains, showing the distribution of distinct information.
• 📊 The box plot includes minimum, maximum, median, interquartile range, and the 10th and 90th percentiles for the distribution of distinct concepts.
• 🔎 Hovering over the box plots allows for further exploration of the data distribution.
• 🔗 For more information on Atlas and additional data sources reports, the script directs viewers to Odyssey.org.

Q & A

What is the main feature of the data sources capability in Atlas?

The main feature of the data sources capability in Atlas is the ability to select a data source and explore it through a series of reports, such as the data density report.

What does the data density report in Atlas show?

The data density report in Atlas shows a series of graphs that represent the total number of rows of data by time and by data domain, records per person, and concepts per person across selected data domains.

How is the x-axis represented in the first graph of the data density report?

In the first graph of the data density report, the x-axis represents calendar months.

What does the y-axis represent in the first graph of the data density report?

In the first graph, the y-axis represents the number of records in the data source.

What information can be obtained by hovering over a specific line in the first graph of the data density report?

By hovering over a specific line, one can see the series represented, the calendar month shown, and the number of records in the data source for that domain in that calendar month.

What does the second graph in the data density report show?

The second graph shows records per person, normalizing the counts to be records per person by dividing the total number of records by the number of persons.

How does the third graph in the data density report differ from the first two?

The third graph focuses on concepts per person, showing the number of distinct concepts per person across different data domains, represented by box plots that display the distribution of distinct concepts.

What does the minimum value in the condition occurrence domain's box plot represent in the third graph?

The minimum value represents that every individual in the condition occurrence table has at least one distinct concept.

What does the maximum value in the condition occurrence domain's box plot indicate?

The maximum value indicates that there is an individual with as many as 800 distinct concepts in the condition occurrence domain.

What additional resources can be found on Odyssey.org for more information about Atlas and its data sources reports?

Odyssey.org provides additional details about Atlas, its capabilities, and other data sources reports that can enhance the understanding of the platform.

Outlines

00:00

📊 Data Density Report Overview

This paragraph introduces the Data Density Report feature within Atlas's data sources capability. Users can select a data source and explore various reports, with a focus on the data density report in this context. The report is visualized through several graphs that provide insights into data accumulation over time and across different data domains. The first graph displays the total number of data rows by time and data domain, with each domain represented by a line on the graph. Hovering over a line reveals specific data counts for a given month and domain. The domains covered include conditions, drugs, observations, observation periods, procedures, and visits. The second graph shifts the focus to records per person, normalizing the data to understand data density relative to the number of persons. The third graph presents concepts per person across selected data domains, using box plots to show the distribution of distinct concepts, including minimum, maximum, median, and interquartile ranges. This comprehensive approach allows users to gauge data growth and stability, and to understand the distinct information available per person.

Keywords

💡Data Sources

Data sources refer to the origins of data that are used within the Atlas system to generate reports. These sources provide the raw information that is analyzed and visualized in various reports, such as the data density report. In the video, selecting a data source is the first step to explore and analyze the available data.

💡Data Density Report

The data density report is a specific type of report within Atlas that visualizes the volume and distribution of data over time and across different data domains. This report includes several graphs that help users understand the accumulation and density of data. It is central to the video's demonstration of how to use the Atlas system.

💡OMOP Common Data Model

The OMOP (Observational Medical Outcomes Partnership) Common Data Model standardizes data from disparate sources to facilitate large-scale analytics. The video references this model to explain how data is organized and analyzed in the Atlas system, ensuring consistency and comparability across different datasets.

💡Graph

Graphs are visual representations of data used in the data density report to show trends and distributions. The video discusses various types of graphs, such as line graphs and box plots, which help in interpreting data related to different domains like conditions, drugs, and procedures.

💡Calendar Month

Calendar month refers to the time intervals used on the x-axis of the graphs in the data density report. It helps in tracking the accumulation and density of data over time. The video uses calendar months to show how data records grow and change from month to month.

💡Data Domain

Data domains are categories or types of data within the OMOP Common Data Model, such as conditions, drugs, observations, procedures, and visits. The video explains how each domain is represented in the data density report and how users can analyze data specific to each domain.

💡Records per Person

Records per person is a measure used in the data density report to normalize data by dividing the total number of records by the number of persons. This metric helps in understanding the data density relative to the population size. The video demonstrates how to interpret this measure to assess data accumulation per individual.

💡Distinct Concepts per Person

Distinct concepts per person refer to the unique pieces of information, such as distinct conditions or procedures, associated with each person in the dataset. The video explains how this metric is visualized using box plots to show the distribution of unique concepts across different data domains.

💡Box Plot

A box plot is a type of graph used in the data density report to show the distribution of distinct concepts per person. It illustrates the median, interquartile range, and outer ranges of the data. The video describes how to interpret box plots to understand the spread and central tendency of the data.

💡Odyssey.org

Odyssey.org is the website mentioned at the end of the video for more information about the Odyssey initiative, which includes Atlas and other data analysis tools. The video directs viewers to this site for additional details and resources related to the data sources and reports discussed.

Highlights

Introduction to Atlas' data sources reporting capability

Accessing data density report through the data sources menu

Visualization of total number of data rows by time and data domain

X-axis represents calendar months, Y-axis shows record counts

Line series for each data domain supported by OMA command data model

Data domains include conditions, drugs, observations, procedures, and visits

Interactivity feature to view specific domain's data for a given month

Example of 33 million records in condition occurrence domain in March 2015

Understanding data accumulation and density across domains over time

Second graph shows records per person, normalizing data counts

Y-axis represents records per person, considering total number of persons

Example of average 2.45 records per person in October 2015 for condition occurrence

Analyzing data density growth in relation to total data or data per person

Third graph displays concepts per person across selected data domains

Box plot representation of the distribution of distinct concepts per person

Example of a maximum of 800 distinct concepts for an individual in the condition occurrence domain

Box plot details including minimum, maximum, median, interquartile, and outer ranges

Explorability of box plots for a comprehensive understanding of data distribution

Conclusion of the data density report summary and invitation to Odyssey.org for more information

Transcripts

[Music]

today I'm going to walk you through some

of the reports within the data sources

capability of Atlas if you click on data

sources you can select the data source

that you are interested in exploring and

select one of a series of reports today

we'll look at the data density report

when the data density report loads

you'll see a series of graphs on the

screen the first graph represents the

total number of rows of data by time and

by data domain in the Oh mop command

data model so here you can see on your

x-axis calendar months and on your

y-axis you see the number of records in

your data source and there's a align

series for every data domain in that we

support in the OMA command data model

that includes conditions drugs

observations observation periods

procedures and visits if you hover over

any one particular line you will see the

series that you are represented the

calendar month that's shown and the

number of records in your data source

for that domain in that calendar month

so here for example we can see that in

this data source in the condition

occurrence domain there are 33 million

records in March of 2015 this graph

allows you to see how data are

accumulating across your data domains

over time and to understand which data

domains have the greatest density of

information the next graph shows you

records per person similar to the first

graph this graph is showing you calendar

month on the x-axis and showing you a

line series for each of the data domains

however now the y-axis has normalized

the the counts to be records per person

taking the total number of records as

represented in the first graph but

dividing it by a common denominator

represent

by the number of persons so here if I

hover over any particular graph I can

get an understanding of the number of

Records per person in a given data

domain for a given calendar month

so here I've highlighted that this

source in October of 2015 in the

condition occurrence domain has an

average of two point four four or five

records per person this graph allows you

to understand whether or not the data

density seems to be growing as a

function of the total data or as a

function of the amount of data per

person and the relative stability that's

included there in the third graph that's

shown on the data density report is

concepts per person and here this is

giving you a sense of how much distinct

information exists per person across

selected data domains on the x-axis we

show the concept type represented by the

different data domains including

conditions drugs observations procedures

and on the y-axis we're showing you the

number of distinct concepts per person

and each graph represents a box plot

showing you the distribution of distinct

concepts so here on the condition

occurrence domain we can see that the

distribution of distinct concepts per

person has a minimum of one for

everybody that's in the condition

occurrence table and a maximum of 800

meaning that there is an individual who

has 800 distinct concepts in the

condition occurrence domain this

provides you the full distribution

including the median here representing

that that 50% of patients have 12 or

more distinct concepts the interquartile

range represented by a P 25 and a P 75

as well as an outer range represented by

the 10% and the 90th percent percentile

each of these box plots can be explored

by hovering over them this concludes our

summary of the data density report for

more information about odyssey including

additional details of Atlas and

additional data sources reports check

at Odyssey org