Open ningyile opened 2 years ago
shaou77
commented
2 years ago Hi, I am also a Chinese ICU doctor. A word about something unrelated to your question. I read an article using 28-day mortality before MIMIC IV 2.0 was released, isn't that nonsense?
ningyile
commented
2 years ago Hi, I am also a Chinese ICU doctor. A word about something unrelated to your question. I read an article using 28-day mortality before MIMIC IV 2.0 was released, isn't that nonsense?
You may have gotten an idea of the mortality of MIMIC-IV from an issue proposed previously in the community. https://github.com/MIT-LCP/mimic-code/issues/1199. In fact, MIMIC-IV V1.0 only provided in-hospital death records, so any study with MIMIC-IV V1.0(or previous version) involving 28-day, 30-day, or 60-day mortality had the wrong clinical outcome. Unfortunately, some journal editors and reviewers ignored the problem. The good news is that the dod field in the patients table in MIMIC-IV V2.0 contains 'state records’ of out-of-hospital death which was purchased by the administrator from the CDC for over $70,000, with a maximum follow-up of 1 year after their last hospital discharge. This is meant to provide a basis for calculating the 28-day mortality, etc. As out-of-hospital death data comes from state records, not like the MIMIC III from Social Security Death Index. I am concerned that there may be a subset of non-local patients (compared to Massachusetts patients) that may not be covered. So I'm here to ask the official administrator whether 28-day mortality as the clinical outcome can be used in MIMIC-IV V2.0, which aimed to ensure the accuracy of our current study.
@shaou77
Figure 1 is from the MIMIC official website.
Figure 2 is from the Regarding the availability of mortality in MIMIC-IV 1.0 data.
Prerequisites
Description
MIMIC-IV V2.0 was released on June 12, 2022. The date of death is available within the 'dod' column of the 'patients' table since V2.0, and the maximum follow-up time for each patient is exactly one year after their last hospital discharge. With out-of-hospital follow-up mortality(stop getting death from Social Security Death Index, but purchasing the data from the CDC), MIMIC-IV V2.0 provides a basis for calculating 28-day mortality(same as 30-day, 60-day, 90-day, and even up to 1-year mortality, etc.). Some scholars have used 28-day mortality as clinical outcome in their studies with MIMIC-IV V2.0.
Zhu, Z., Zhou, M., Wei, Y. et al. Time-varying intensity of oxygen exposure is associated with mortality in critically ill patients with mechanical ventilation. Crit Care 26, 239 (2022). https://doi.org/10.1186/s13054-022-04114-w
However, I noticed that the source of out-of-hospital mortality was from 'state records’, and I‘m confused with the data source(I'm an ICU physician in China mainland, I am not familiar with the US government institutions), I mean whether the data was from the state CDC(Massachusetts) or the national CDC？
If the survival data come from the CDC of Massachusetts, for after discharge of Non-Massachusetts patients(who come from other states and even from other countries), their follow-up information can be tracked and recorded by the CDC of Massachusetts? I know the vast majority of patients in MIMIC-IV should be local residents, and only a very small percentage of patients may fall into the category I mentioned above. If out-of-hospital survival data for non-Massachusetts patients could not be tracked, could I exclude this part of non-Massachusetts patients based on certain fields(for example, insurance from admissions or other fields)?
As it's important for the clinical outcome of our project, I am interested in whether metrics such as 28-day mortality can be used in MIMIC-IV V2.0. In other words, I would like to ask whether the state(Massachusetts?) CDC data cover the vast majority of patients for out-of-hospital survival information enough. If there are only a few are not covered, the impact of those few can be ignored?
In summary, 28-day mortality as the clinical outcome can be used in MIMIC-IV V2.0?
@alistairewj