Variation in organ donation after brain death (DBD) per million population varies markedly between countries, within country regions, between and within intensive care units (ICU). These circumstances also apply to end-of-life decisions in the ICU.

The investigators studied all ICU deaths in Sweden between 2014-2017 in ICUs that, as routine, registered treatment plan (no treatment limitation and/or treatment limitation) and DBD.

The investigators hypothesized that ICUs with high proportion of treatment limitation (withholding or withdrawing life sustaining treatment) also had less proportion of DBD.

The Swedish Intensive Care Registry (SIR) collects a comprehensive data set of patient characteristics, intensive care procedures, severity of illness scores, workload and outcomes using detailed guidelines.

Continuous data are collected as raw data, validated locally and transferred electronically to the registry for central validation (confirmed to be within prespecified limits and inconsistencies and illogical entries identified). If necessary, data are returned for correction and revalidation before being accepted and added to the master database. No data are changed by SIR. At SIR's homepage there are numerous open reports, updated as soon as new incoming data are processed by SIR.

SIR has an annual 2-day course for all ICU's in registration techniques and problems and daily telephone support possibilities concerning registration questions, problems and guidance. SIR has, however not yet a systematic on-site auditing program.

Missing data regarding treatment strategy are registered and are accounted for in the study. There are no missing data regarding organ donation.

The investigators calculate the proportion of treatment limitations per ICU as the number of deaths with any treatment limitation divided by all deaths in the same ICU.

Continuous variables are expressed as mean (standard deviations) or median (interquartile range).

Differences in proportions are analysed using the χ2-test. Changes over time are analysed using the non-parametric trend test. Survival are examined using the Kaplan-Meier estimate and differences in survival are analysed with the log-rank test.

Logistic regression is used to examine associations between patient and ICU characteristics and DBD as dependent variable. Mixed-effects logistic regression clustered per ICU is used to assess associations between treatment limitation and DBD after adjusting for patient age, sex, comorbidities, the Simplified Acute Physiology Score III (SAPS3) probability of death, principal diagnostic category, presence of treatment limitation and type of hospital.

Significance was assumed if P < 0.05.

Inclusion Criteria:

All ICU deaths in Sweden between 1/1 2014 and 31/12 2017 from

1. general and neurological intensive care units,
2. which recorded treatment strategy according to national guidelines and
3. send data according to SIR's protocol for the follow-up of all deceased intensive care patients (DBD).

Exclusion Criteria:

1. Special units like paediatric (N=4), burn (N=2) and thoracic (N=8) intensive care units, due to low coverage ratio of documented treatment plan decisions in combination with a low ICU mortality rate.
2. General (N=3) ICUs that do not register documented treatment plan.

The study population consisted of all ICU deaths (12,072) between 2014-2017 in general and neurological intensive care units which all recorded treatment strategy in their daily routine work.

Special units like paediatric, burn and thoracic intensive care units were excluded