Features you want to see in the package about hospitalnetwork

[tjibbed]

Hi Pascal,

Shouldn't generating the edgelist and/or matrix from the raw database be part of the basic analyses? A rather important functionality of the script, I would say :)

I would suggest some functionality of introducing errors into the simulated data, in particular to test if each step of the data management side of the script identifies and handles the errors correctly. I would suggest the following errors:

• Incorrect date format (e.g. some records are dmy, while others are mdy)
• Overlapping admissions (simple, duplicate a multi-day stay and shift it one day, or introduce a one-day stay on day two of stay >3 days.
• Duplicate records
• Missing values , or spaces introduced.

[ClementMassonnaud]

Hi, I started writing some code to generate errors in a fake dataset as part of writing tests to check that the Data management functions work correctly. For now I generate errors sporadically within the tests but I can easily isolate the code to make it a separate function. Which would be more appropriate. Doing that I also got some errors running the adujst_overlap() function. I must look into it a bit more. Best, Clément Le ven. 16 août 2019 à 21:24, Tjibbe Donker <[email protected]> a écrit :

…

Hi Pascal, Shouldn't generating the edgelist and/or matrix from the raw database be part of the basic analyses? A rather important functionality of the script, I would say :) I would suggest some functionality of introducing errors into the simulated data, in particular to test if each step of the data management side of the script identifies and handles the errors correctly. I would suggest the following errors: - Incorrect date format (e.g. some records are dmy, while others are mdy) - Overlapping admissions (simple, duplicate a multi-day stay and shift it one day, or introduce a one-day stay on day two of stay >3 days. - Duplicate records - Missing values , or spaces introduced. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#6?email_source=notifications&email_token=AG5UAT4MKDXF4VH7OVWQY7TQE22BBA5CNFSM4ILBA2NKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD4OYDRA#issuecomment-522027460>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AG5UAT56XGR7M47VX2F4WODQE22BBANCNFSM4ILBA2NA> .

[tjibbed]

Brilliant, thanks Clément. I fixed the adjust_overlapping_stays() function, it couldn't handle the column naming I worked through in the other functions. It didn't pop up on my radar because the data I was working on didn't have any overlapping admissions. Should be fine now, can you double-check? Cheers, Tjibbe Op vr 16 aug. 2019 om 17:12 schreef Clement Massonnaud < [email protected]>:

…

Hi, I started writing some code to generate errors in a fake dataset as part of writing tests to check that the Data management functions work correctly. For now I generate errors sporadically within the tests but I can easily isolate the code to make it a separate function. Which would be more appropriate. Doing that I also got some errors running the adujst_overlap() function. I must look into it a bit more. Best, Clément Le ven. 16 août 2019 à 21:24, Tjibbe Donker ***@***.***> a écrit : > Hi Pascal, > > Shouldn't generating the edgelist and/or matrix from the raw database be > part of the basic analyses? A rather important functionality of the script, > I would say :) > > I would suggest some functionality of introducing errors into the > simulated data, in particular to test if each step of the data management > side of the script identifies and handles the errors correctly. I would > suggest the following errors: > > - Incorrect date format (e.g. some records are dmy, while others are > mdy) > - Overlapping admissions (simple, duplicate a multi-day stay and shift > it one day, or introduce a one-day stay on day two of stay >3 days. > - Duplicate records > - Missing values , or spaces introduced. > > — > You are receiving this because you are subscribed to this thread. > Reply to this email directly, view it on GitHub > < #6?email_source=notifications&email_token=AG5UAT4MKDXF4VH7OVWQY7TQE22BBA5CNFSM4ILBA2NKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD4OYDRA#issuecomment-522027460 >, > or mute the thread > < https://github.com/notifications/unsubscribe-auth/AG5UAT56XGR7M47VX2F4WODQE22BBANCNFSM4ILBA2NA > > . > — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#6?email_source=notifications&email_token=AMUT655KYKPCXN26MCFHFXLQE27WDA5CNFSM4ILBA2NKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD4O4I2Y#issuecomment-522044523>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AMUT652WLXTEJ37RSN4V563QE27WDANCNFSM4ILBA2NA> .

[ClementMassonnaud]

Hi, Great, thank you. I'll do that ASAP Best, Clément Le lun. 19 août 2019 à 16:31, Tjibbe Donker <[email protected]> a écrit :

…

Brilliant, thanks Clément. I fixed the adjust_overlapping_stays() function, it couldn't handle the column naming I worked through in the other functions. It didn't pop up on my radar because the data I was working on didn't have any overlapping admissions. Should be fine now, can you double-check? Cheers, Tjibbe Op vr 16 aug. 2019 om 17:12 schreef Clement Massonnaud < ***@***.***>: > Hi, > I started writing some code to generate errors in a fake dataset as part of > writing tests to check that the Data management functions work correctly. > > For now I generate errors sporadically within the tests but I can easily > isolate the code to make it a separate function. Which would be more > appropriate. > > Doing that I also got some errors running the adujst_overlap() function. I > must look into it a bit more. > > Best, > Clément > > Le ven. 16 août 2019 à 21:24, Tjibbe Donker ***@***.***> a > écrit : > > > Hi Pascal, > > > > Shouldn't generating the edgelist and/or matrix from the raw database be > > part of the basic analyses? A rather important functionality of the > script, > > I would say :) > > > > I would suggest some functionality of introducing errors into the > > simulated data, in particular to test if each step of the data management > > side of the script identifies and handles the errors correctly. I would > > suggest the following errors: > > > > - Incorrect date format (e.g. some records are dmy, while others are > > mdy) > > - Overlapping admissions (simple, duplicate a multi-day stay and shift > > it one day, or introduce a one-day stay on day two of stay >3 days. > > - Duplicate records > > - Missing values , or spaces introduced. > > > > — > > You are receiving this because you are subscribed to this thread. > > Reply to this email directly, view it on GitHub > > < > #6?email_source=notifications&email_token=AG5UAT4MKDXF4VH7OVWQY7TQE22BBA5CNFSM4ILBA2NKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD4OYDRA#issuecomment-522027460 > >, > > or mute the thread > > < > https://github.com/notifications/unsubscribe-auth/AG5UAT56XGR7M47VX2F4WODQE22BBANCNFSM4ILBA2NA > > > > . > > > > — > You are receiving this because you commented. > Reply to this email directly, view it on GitHub > < #6?email_source=notifications&email_token=AMUT655KYKPCXN26MCFHFXLQE27WDA5CNFSM4ILBA2NKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD4O4I2Y#issuecomment-522044523 >, > or mute the thread > < https://github.com/notifications/unsubscribe-auth/AMUT652WLXTEJ37RSN4V563QE27WDANCNFSM4ILBA2NA > > . > — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#6?email_source=notifications&email_token=AG5UAT6XZN33Q33VWCBUVKTQFJR6HA5CNFSM4ILBA2NKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD4SJF3Q#issuecomment-522490606>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AG5UAT3TX2WBHK3EBMKQD7LQFJR6HANCNFSM4ILBA2NA> .

[tjibbed]

Reading through the minutes of the NeWIS meeting, I noticed we've forgotten about the general summary statistics. We discussed that for each database / country we should extract summary stats such as number of patients, number of hospitals, mean length of stay, etc.

I'll start writing a function to extract these.

[tjibbed]

Going through the workflow of the package, isn't it an idea to include the summary statistics of the entire database into the HospiNet class? We can call the functions extracting these stats from the database within the hospinet_from_patient_database() function.

If the HospiNet is not constructed from the patient database directly, but for instance using a pre-constructed edgelist, it just remains empty (NULL). However, if it was constructed using the patient database, it would give us the opportunity to easily export the data as one file.

I think this would greatly improve the collection of data back from all participants.

[ClementMassonnaud]

Hi, I committed modifications to the creation of the fake patient database. I added the option to introduce errors in the database (wrong date format, missing values and overlapping stays) In the meantime, I created some tests to check that the functions in DataManagement.R work correctly. It seems to be ok, all the errors are handled as expected. I got no errors for overlapping stays, however I introduced rather simple overlapping errors. I can't be sure that the function will work properly in the case of more complicated and intricate overlaps. Also I could only test the checkBase() function on a small database. I don't know what time it would take on a database of several millions lines... My computer is not powerful enough to generate such a big database with the create_fakeDB() function... Maybe one of you guys can try to have an idea. In particular I'm worried about the behavior of the adjust_overlap() function on a big database due to the "while()" within. Best, Clément

…

On Mon, 19 Aug 2019 at 11:31, Tjibbe Donker ***@***.***> wrote: Brilliant, thanks Clément. I fixed the adjust_overlapping_stays() function, it couldn't handle the column naming I worked through in the other functions. It didn't pop up on my radar because the data I was working on didn't have any overlapping admissions. Should be fine now, can you double-check? Cheers, Tjibbe Op vr 16 aug. 2019 om 17:12 schreef Clement Massonnaud < ***@***.***>: > Hi, > I started writing some code to generate errors in a fake dataset as part of > writing tests to check that the Data management functions work correctly. > > For now I generate errors sporadically within the tests but I can easily > isolate the code to make it a separate function. Which would be more > appropriate. > > Doing that I also got some errors running the adujst_overlap() function. I > must look into it a bit more. > > Best, > Clément > > Le ven. 16 août 2019 à 21:24, Tjibbe Donker ***@***.***> a > écrit : > > > Hi Pascal, > > > > Shouldn't generating the edgelist and/or matrix from the raw database be > > part of the basic analyses? A rather important functionality of the > script, > > I would say :) > > > > I would suggest some functionality of introducing errors into the > > simulated data, in particular to test if each step of the data management > > side of the script identifies and handles the errors correctly. I would > > suggest the following errors: > > > > - Incorrect date format (e.g. some records are dmy, while others are > > mdy) > > - Overlapping admissions (simple, duplicate a multi-day stay and shift > > it one day, or introduce a one-day stay on day two of stay >3 days. > > - Duplicate records > > - Missing values , or spaces introduced. > > > > — > > You are receiving this because you are subscribed to this thread. > > Reply to this email directly, view it on GitHub > > < > #6?email_source=notifications&email_token=AG5UAT4MKDXF4VH7OVWQY7TQE22BBA5CNFSM4ILBA2NKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD4OYDRA#issuecomment-522027460 > >, > > or mute the thread > > < > https://github.com/notifications/unsubscribe-auth/AG5UAT56XGR7M47VX2F4WODQE22BBANCNFSM4ILBA2NA > > > > . > > > > — > You are receiving this because you commented. > Reply to this email directly, view it on GitHub > < #6?email_source=notifications&email_token=AMUT655KYKPCXN26MCFHFXLQE27WDA5CNFSM4ILBA2NKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD4O4I2Y#issuecomment-522044523 >, > or mute the thread > < https://github.com/notifications/unsubscribe-auth/AMUT652WLXTEJ37RSN4V563QE27WDANCNFSM4ILBA2NA > > . > — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#6?email_source=notifications&email_token=AG5UAT6XZN33Q33VWCBUVKTQFJR6HA5CNFSM4ILBA2NKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD4SJF3Q#issuecomment-522490606>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AG5UAT3TX2WBHK3EBMKQD7LQFJR6HANCNFSM4ILBA2NA> .