Surviving ARDS: The Influence of Quality of Care and Individual Patient Characteristics on Quality of Life

Design Patients will be included in the study at the beginning of ICU stay in a participating hospital (t0) and survivors will be followed over a period of 12 months after discharge from ICU. If patients have been treated before in a referring hospital, a retrospective assessment of their health status during their stay in this hospital and during inter-hospital transport will be performed. Four follow-up measurements (at discharge from ICU as well as 3, 6 and 12 months after discharge) will be performed on individual patient level.

Sample \& Recruitment Hospitals of the "ARDS Network Germany" and other hospitals which provide care for patients with ARDS were invited to participate in the study.

It is envisaged to screen 2600 patients for the presence of ARDS and to include 2400 patients as members of the source population. An in-hospital mortality rate of 30% is assumed; therefore, we expect to analyze a study population of 1500 patients at t1 (discharge from ICU). For each follow-up-assessment (t2-t4: 3 months, 6 months, 12 months after discharge), loss-to-follow-up rates due to death, withdrawal of consent or other reasons are estimated to be about 10%, resulting in slightly more than 1000 patients at 12-months-follow-up.

Measurements The main predictor is quality of care provided in the ICUs of the participating hospitals. The following main indicators were chosen: qualification of physicians, implementation of routine daily multiprofessional ward rounds with documentation of daily therapy goals, number of ventilated patients per, and membership of the hospital in the "ARDS Network Germany". All quality indicators will be assessed by questionnaire at the institutional level.

Main outcome measures of the DACAPO study are HRQoL and return to work among survivors of ARDS (see study outcomes).

Moderating variables are gender, socio-economic status, prevalent and incident psychopathological symptoms (PHQ-D, PTSS-14) and the availability of social support (F-SozU K-14).

Additional variables comprise general medical ICU parameters (e.g. comorbidity, prognostic scores (SAPS (Simplified Acute Physiology Score) II, SAPS III ) and an organ dysfunction score (SOFA (Sequential Organ Failure Assessment score)) and parameters relating to ARDS (e.g. cause and severity) and its treatment (e.g. ventilation parameters, use of supportive care measures, critical events (hypoglycaemia, hypoxia)) as well as socio-demographic and psycho-social characteristics.

Additionally, costs will be assessed as direct costs in terms of treatment costs at the ICU.

Data Sources and Data Collection Patients's general medical characteristics and medical data relating to ARDS and its treatment will be assessed during ICU stay and gathered through electronic case report forms (eCRFs). Data on the direct costs of treating patients with ARDS in the ICU will be gathered from hospital records. A 12-months mortality follow-up will be performed. Local municipal population registries will be contacted to obtain data on mortality at the patient level.

At baseline, caregivers will provide proxy report of patients' socio-demographic data, at follow-ups (discharge from ICU, 3 months, 6 months, 12 months), patients themselves will complete self-report questionnaires on socio-demographic and psycho-social characteristics as well as on the main outcome measures.

Data on indicators of quality of care will be assessed through questionnaires administered to the directors of the hospitals/ICUs.

Data Management The open source software OpenClinica, version 3.1, will be used for electronic data collection and data management. eCRFs will be used for data entry. A continuous assessment of data quality will be performed by exporting data sets to statistical packages to run further error and plausibility checks. Data from paper-pencil-questionnaires completed by patients will be entered into the database twice in order to ensure high data quality.

Statistical Analysis A linear multilevel regression model will be fitted both for the physical and the mental component summary score of the SF-12. A non-parametric multilevel regression model will be used for the outcome "return to work". The modelling procedure will be applied to each of the operationalizations of the predictor "quality of care".

All analyses will be carried out using SAS 9.4 software and STATA version 12 software.