nrodemund
commented
4 months ago Hi ad 1) This is an error in raw data, unfortunately I won't be able to address to it. It seems to happen if patient, who was planned for IMC, never reaches IMC (in most cases triage for a more urgent patient). As TimeOfStay is capped to LastSignalTime. This happened in a minor amount of cases (45). We may create another rule of deselection to filter out these cases.
For studies I recommend setting a minimal icu time, i.e. 6 hours, as this will filter out most cases of no interest.
2) from a rigourous perspective yes, I also did that. Note though, that this may not reduce the "date errors" of all automated data (
- If someone dies out of hospital: TimeOfStay is in seconds/minutes due to highly granular data, OffsetOfDeath is in days as it is retrieved from official data. Theoretically up to 23h59min error can happen.
- If someone dies in hospital: The "official" time of death may not represent the actual. This is, because the death in system may be documented on station release, which is up to 4 hours after actual death. There is no way to determine this in austria 3) admitting physician's clinical impression, it is a SAPS3 field. That is why the "AdmissionForm" is added, to show that it may have observer bias 4) The OffsetOfDeath field is combined and at least 180 days. The EstimatedObservationTime indicates when it is a year. (simple answer: all patients with admissionyear <2021 have one year) 5) It is difficult to find out what happened here (most likely human error), I will look into this
Thanks for your input! Next update will contain Hospital discharge days and loinc codes!
best regards, Niklas
Am Do., 27. Juni 2024 um 05:16 Uhr schrieb Yi-Le Ning < @.***>:
Dear Administrators of the SICdb,
I have successfully deployed SICdb V1.0.6, a great critical care database, on my Linux server. Recently, I have been exploring the SICdb V1.0.6 database, and I would like to confirm a few points regarding it:
-
TimeOfStay Field: Does this field refer to the ICU stay duration that includes the surgery time, while the ICUOffset is the actual time of ICU admission (which can also be understood as the time of transfer to the ICU after surgery)? Therefore, the value of the TimeOfStay field should be greater than the value of the ICUOffset field. The difference between the TimeOfStay and ICUOffset fields represents the actual ICU stay duration. However, I have found several cases where the values of the TimeOfStay and ICUOffset fields are the same (CaseID 423817, 995058, 320834, 600783, 624033, 719288, 965353, 489596, 904148, 321468, 466580, 151811, 647403, 888798, 784743, 105176, 891163, 234068, 609003, 975159, 209179).
OffsetOfDeath Field: The OffsetOfDeath represents the survival duration (if not null). From a rigorous perspective, should we set the initial observation time as the moment of ICU admission? Thus, the difference between the OffsetOfDeath field (if not null) and the ICUOffset field (similar to TimeOfStay Field) is the actual survival duration ?
AdmissionFormHasSepsis Field: How is the sepsis diagnosis labeled in this field? Is it based on the admitting physician's clinical impression or according to the objective Sepsis 3.0 criteria?
Outcome: Are the available outcomes for all patients confirmed to include ICU mortality, 28-day, 90-day, and 180-day mortality? Since the V1.0.6 data lacks the HospitalDischargeDay field data (using ICU admission as the observation starting point), we can currently calculate in-hospital outcomes but cannot determine the in-hospital survival duration (from ICU admission as the observation starting point).
Data Inconsistencies: I have found a few logical inconsistencies in the data, such as individuals with an empty OffsetOfDeath field but an in-hospital outcome of death (34 cases), and an ICU outcome of death (36 cases). Another significant logical error regarding death is patients with a non-empty OffsetOfDeath field and an ICU outcome of death but an in-hospital outcome of survival (CaseID 103553, 881546, 587891, 411376, 452151, 545370, 974105, 453844, 991006, 522927, 549662, 725039). In light of the above issues, may I propose some logical corrections? For instance, for individuals with an empty OffsetOfDeath field and an ICU outcome of death, I could use the actual ICU stay duration (the value of the TimeOfStay field minus the ICUOffset field) as their OffsetOfDeath value (time of death). When the OffsetOfDeath field is non-empty and the ICU outcome is death, the in-hospital outcome should be corrected to death.
Thank you for your attention to these matters.
— Reply to this email directly, view it on GitHub https://github.com/nrodemund/sicdb/issues/5, or unsubscribe https://github.com/notifications/unsubscribe-auth/A335S4OEWWPEAOPCVT3GMALZJN7YRAVCNFSM6AAAAABJ7BJPEKVHI2DSMVQWIX3LMV43ASLTON2WKOZSGM3TMNZZGA4DOMQ . You are receiving this because you are subscribed to this thread.Message ID: @.***>
Dear Administrators of the SICdb,
I have successfully deployed SICdb V1.0.6, a great critical care database, on my Linux server. Recently, I have been exploring the SICdb V1.0.6 database, and I would like to confirm a few points regarding it:
TimeOfStay Field: Does this field refer to the ICU stay duration that includes the surgery time, while the ICUOffset is the actual time of ICU admission (which can also be understood as the time of transfer to the ICU after surgery)? Therefore, the value of the TimeOfStay field should be greater than the value of the ICUOffset field. The difference between the TimeOfStay and ICUOffset fields represents the actual ICU stay duration. However, I have found several cases where the values of the TimeOfStay and ICUOffset fields are the same (CaseID 423817, 995058, 320834, 600783, 624033, 719288, 965353, 489596, 904148, 321468, 466580, 151811, 647403, 888798, 784743, 105176, 891163, 234068, 609003, 975159, 209179).OffsetOfDeath Field: TheOffsetOfDeathrepresents the survival duration (if not null). From a rigorous perspective, should we set the initial observation time as the moment of ICU admission? Thus, the difference between the OffsetOfDeath field (if not null) and the ICUOffset field (similar toTimeOfStay Field) is the actual survival duration ?AdmissionFormHasSepsis Field: How is the sepsis diagnosis labeled in this field? Is it based on the admitting physician's clinical impression or according to the objective Sepsis 3.0 criteria?Outcome: Are the available outcomes for all patients confirmed to include ICU mortality, 28-day, 90-day, and 180-day mortality? Since the V1.0.6 data lacks the HospitalDischargeDay field data (using ICU admission as the observation starting point), we can currently calculate in-hospital outcomes but cannot determine the in-hospital survival duration (from ICU admission as the observation starting point).Data Inconsistencies: I have found a few logical inconsistencies in the data, such as individuals with an empty OffsetOfDeath field but an in-hospital outcome of death (34 cases), and an ICU outcome of death (36 cases). Another significant logical error regarding death is patients with a non-empty OffsetOfDeath field and an ICU outcome of death but an in-hospital outcome of survival (CaseID 103553, 881546, 587891, 411376, 452151, 545370, 974105, 453844, 991006, 522927, 549662, 725039). In light of the above issues, may I propose some logical corrections? For instance, for individuals with an empty OffsetOfDeath field and an ICU outcome of death, I could use the actual ICU stay duration (the value of the TimeOfStay field minus the ICUOffset field) as their OffsetOfDeath value (time of death). When the OffsetOfDeath field is non-empty and the ICU outcome is death, the in-hospital outcome should be corrected to death.Thank you for your attention to these matters.