ATLAS ERROR: Reviewing Atlas

[mizq7]

@shossain-mizzou @vasanthi014 @mosaa While exploring Atlas, I see two issues that probably need attention and modification.

Issue 1 Upon choosing MU CDM and 'Dashboard' as Data Sources in Atlas, it becomes apparent that the X-axis, depicting patients' age on the second graph titled "Age at First Observation," spans from 0 to 2000. Consequently, the graph appears distorted. To rectify this, kindly adjust the X-axis range from 0 to 100. Please find the graph below for reference -

Issue 2 In the third graph on the same Altas interface, "Cumulative Observation" you will see 56 percent of population are identified with negative years. For example, if we cite the graph below as an example, it will mean that 57% of the patient population in this database are observed in the data for negative 3.6 years or more. I am not sure why so many patients are identified with negative years.

[mizq7]

@shossain-mizzou In addition to the two graphs above, it looks like the fourth graph "Persons With Continuous Observation By Month" also warrant some fixes. In this graph, the X-axis depicts the calendar month, while the Y-axis represents the number of individuals/people. Upon hovering over any specific calendar month, we observe the count of individuals continuously monitored within that timeframe. Looking into the graph below, see that the temporal coverage for this dataset spanning from January 1800 to January 2020 covering a total of 2,640 months which seems unrealistic. Given this long timeframe, the graph does not look having appropriate representation of the dataset while the main purpose of this graph is to demonstrate that this database enables exploration of monthly cohorts, revealing the population count for each month and illustrating population stability across time in the database. Ideally, for this graph, the temporal starting point could be the year 2000 because there is a huge jump of patient population after 1997. See below the graph.

[vasanthi014]

@shossain-mizzou In addition to the two graphs above, it looks like the fourth graph "Persons With Continuous Observation By Month" also warrant some fixes. In this graph, the X-axis depicts the calendar month, while the Y-axis represents the number of individuals/people. Upon hovering over any specific calendar month, we observe the count of individuals continuously monitored within that timeframe. Looking into the graph below, see that the temporal coverage for this dataset spanning from January 1800 to January 2020 covering a total of 2,640 months which seems unrealistic. Given this long timeframe, the graph does not look having appropriate representation of the dataset while the main purpose of this graph is to demonstrate that this database enables exploration of monthly cohorts, revealing the population count for each month and illustrating population stability across time in the database. Ideally, for this graph, the temporal starting point could be the year 2000 because there is a huge jump of patient population after 1997. See below the graph.

We have looked into this issue and this is happening because there are about 9000 patients with null or 0 date of birth in our CDM datalake.

[mizq7]

@Yaswitha-MU @vasanthi014 @shossain-mizzou : I see a discrepancy in data output for ICD-10 code G30.9. When I search for this code in Altas, it shows some 6,716 records count but No 'Person Count'. See below the picture:

But when I search the same code in i2b2, I see a total of 3,255 patients with a diagnosis of ICD-10 code G30.9. Please see the picture below:

What could possibly be the explanation for this?

[Yaswitha-MU]

Reference

This isn't really a discrepancy but how atlas works. We might not have cache for that table to pull the data in search. Also, this isn't the way to compare patient counts. You must create a cohort to do so. Ontologies are different and you will have N number of rows for same dx code.

[mosaa]

I think Saber shall look into this to fix the PC column.

[mizq7]

@Yaswitha-MU @vasanthi014 @shossain-mizzou I would like to bring to your notice to other items I encountered today. While trying to create new concept set and cohort definition in the Atlas platform, I see two different outputs for the same ICD-10 code related to 'Herpes Simplex' virus coded as A60. In Atlas with ICD-10 code specified, it does not show any records or patient count for A60 while in the i2b2, it generates a total of 2736 patients. See below -

Is it a mapping issue? How to resolve this?

[mosaa]

@vasanthi014: How are you tracking these? Are you creating separate ticker for each of them to replicate and resolve?

[vasanthi014]

Yes. I will track them as separate tickets in OMOP repo. Some issues here are related to the tickets already created by Yaswitha. I will create the rest and refer them here.

[shossain-mizzou]

Dashboards reports are more accurate

[shossain-mizzou]

@Yaswitha-MU @vasanthi014 @shossain-mizzou I would like to bring to your notice to other items I encountered today. While trying to create new concept set and cohort definition in the Atlas platform, I see two different outputs for the same ICD-10 code related to 'Herpes Simplex' virus coded as A60. In Atlas with ICD-10 code specified, it does not show any records or patient count for A60 while in the i2b2, it generates a total of 2736 patients. See below - Is it a mapping issue? How to resolve this?

@mizq7, You were looking at wrong codes. You should be searcing with vocabulary type not class type

[shossain-mizzou]

@vasanthi014, we need to identify why patient count is showing 0 where we have multiple records. Probably this is a mapping issue, creating inconsisency across different functionality of the app.

[shossain-mizzou]