Intensive Care Units (ICU) are an important, but troubled, part of modern health care systems. While it seems likely that both the technical and structural elements of ICU care are important determinants of relevant ICU outcomes, little is known about how the structure of ICU care affects outcomes. One element of potential importance is the way that ICU physicians (intensivists) organize themselves to provide ICU care, particularly at night. The dominant, historical ("standard") model of intensivist staffing involves an intensivist who is present during daytime hours, but then takes "call" at night from home. But, in recent years there has been widespread concern about whether patients experience adverse events or worse outcomes related to a lesser level of expertise and care readily available at night in hospitals. Only two studies, both from single ICUs, and both using simple before vs. after study designs, have conducted interventional studies directly comparing a "standard" intensivist staffing model with a "24-7" model of nighttime intensivist coverage via shift work, i.e. with the daytime intensivist giving way at the late afternoon to a nightshift intensivist who remained in the hospital and covered the ICU until morning. Those two studies found contradictory effects of the intervention. But despite the absence of clear data indicating a benefit to ICU patients associated with having intensivists remain in the hospital overnight, there has been a major movement around the world towards ICU staffing models utilizing shift work to ensure such coverage. The potential impact of such a change in staffing paradigm is large, with possible effects on all the other major stakeholders involved in ICU care: families, nurses, and house officers. Both benefits and detriments are possible.

On the one hand, moving to a shift work model from a model in which a single intensivist becomes overworked and sleep-deprived as a result of being responsible for care both day and night, has the potential to reduce the sleep deprivation, job distress, and burnout prevalent among intensivists with standard staffing models. But, it would also require more intensivists, a serious challenge given the worsening intensivist manpower shortage. Also, there are many detrimental effects of shift work on humans, including negative effects on motor function, cognition, sleep, job satisfaction, mood, errors, and cardiovascular health. Shift work is the most common reason that Emergency Medicine physicians give for leaving that field. The physical availability of an intensivist around-the-clock might also influence the problems mentioned of family dissatisfaction with communication in ICUs, and poor communication/ teamwork with physicians often perceived by ICU nurses. In ICUs of teaching hospitals, where relatively inexperienced house officers typically remain in the ICU overnight, the nighttime presence of an attending physician might influence residents' perceptions of domains such as teaching, and clinical autonomy.

This purpose of this study is to rigorously compare the effects of two different intensivist staffing models, specifically the current standard model, and a 24-7 staffing model enabled via shift work. This study will be conducted in two ICUs, one academic with house officers who remain in ICU overnight (the Medical ICU at Health Sciences Center), and one in a community hospital which currently lacks overnight, in-ICU physicians (the Victoria General Hospital).

This study is designed to improve upon both prior studies. To obviate the problems with using historical controls inherent in those before-vs-after study designs, our study will alternate the two staffing models (e.g. A-B-A-B). Also, the investigators will rigorously assess the effect of 24 hour intensivist presence on all major stakeholders, i.e. patients, families, intensivists, nurses, and house officers.

Design & Procedures - General Procedures

The investigators will directly compare 2 distinct models of intensivist coverage in the 2 participating ICUs in Winnipeg:

A) The standard model: A single intensivist staffs an ICU for 7 days. He/she will is present during daytime hours, and takes call from home afterwards. This is the current staffing paradigm in both participating ICUs.

B) The 24-7 in-house coverage model: 24-7 in-hospital coverage by an intensivist is enabled by splitting each 24 hour period into two shifts. There will, as in the standard model, be a single intensivist covering the ICU during the day shifts for one week. The day shift will run 8 am to 5:30 pm on weekdays, and 8 am to 3 pm on Saturday and Sunday. The night shift intensivist will arrive and take over at 5:30 pm on weekdays, and 3 pm on weekends and remain in the hospital until 8 am. Call rooms will be provided to allow the night shift intensivist to sleep, if the workload permits.

The interventional part of the study will last 32 weeks, comprising 4 blocks of 8 weeks each. It will run from October 2008 until June 2009. During each 8 week block, each ICU will be staffed under model A or B. Thus we will alternate the models: A-B-A-B in one ICU, and B-A-B-A in the other. This alternating design obviates the problem with historical controls; any nonrandom difference between the two staffing models should be seen to come and go in this design.

A variety of outcomes will be compared between the two intensivist staffing models. The primary outcome, which is the one for which the sample size analysis was performed, is the intensivists' burnout scores.

In order to avoid contamination between the staffing models, patients (and family participants) whose ICU stay is long enough to include time under both models will be excluded from analysis. Also excluded will be patients who are directly transferred from one ICU in Winnipeg to another. For patients who are admitted more than once to a participating ICU during the study, only the initial ICU admission will be included in the analysis.

The intensivists rotate on a weekly basis, and they will be provided questionnaires at the end of each week of ICU service. ICU nurses will be surveyed at the end of each 8 week block of a given staffing model. House officers in the ICU at HSC rotate on a 4 week basis, and they will be provided questionnaires at the end of their final week of their ICU rotations.

Even though the study design makes it likely (apart from the possibility of seasonal differences over the 32 weeks of the study) that participants' characteristics will be balanced between the two staffing models, this is not a randomized study and our primary analysis will use regression modeling to adjust for baseline characteristics of subjects. The investigators will use ordinary least squares or quantile regression for continuous variables, and logistic regression for binary variables. This approach permits assessment for differential effects between the two sites by including: (i) an indicator variable representing the separate sites, and (ii) interaction terms between that indicator variable and other covariates. A relevant aspect of the analysis is that outcomes for the intensivists, nurses and house officers derive from questionnaires, and that a single individual could complete the questionnaire more than once. Regression using General Estimating Equation methodology will therefore be used to account for correlation of responses within subjects.

Design & Procedures - Specific Aims Specific Aim#1: To compare clinical outcomes for ICU patients cared for under the two intensivist staffing models.

• The Department of Medicine has long maintained a comprehensive clinical database of all patients admitted to Winnipeg ICUs. It contains information about patients' demographics, illness type and severity, comorbid conditions, and ICU and hospital course, including the outcomes for this study. The investigators will obtain a deidentified data file containing this information for patients admitted to the participating ICUs during the study period.

Specific Aim#2: To compare family satisfaction with ICU care under the two intensivist staffing models.

• On each day for each of the participating ICUs, the investigators will attempt to recruit the next-of-kin of one newly admitted patient. The investigators will use a 2 stage consent process for these participants. The investigators will use the Family Satisfaction Survey - ICU 24 for assessing family satisfaction with care. This well-validated tool has two subscales, and has been extensively used in Canadian ICUs.

Specific Aim#3: To compare the work-related personal/emotional burden for intensivists working under the two intensivist staffing models.

• The intensivists will be asked to complete multiple questionnaires. Before starting, and after concluding the intervention, all Winnipeg intensivists will be surveyed. After each one week rotation, the intensivists who worked that week in the participating ICUs will be surveyed. A single consent provided before beginning the study intervention will ask the intensivists to consent for all the questionnaire that they will receive. The initial survey establishes background information, and asks for their opinions and preferences relating to the two different intensivist staffing models. The concluding survey repeats those questions. The weekly questionnaires contain the items for the 5 scales that will be analyzed; four of those scales are validated. The fifth scale, Role Conflict, aims to assess an interaction that only occurs in the shift work staffing model, i.e. the interface between 2 intensivists doing shift-work. The investigators were unable to find any scale that addresses this sort of issue, and thus created the items ourselves.

Specific Aim#4: To compare ICU nurses' perceptions of working alongside the two intensivist staffing models.

• At the end of each 8 week block of the study, we will ask all nurses in the participating ICUs to complete a questionnaire. The questionnaire asks for background information, about the mix of shifts they worked over the prior 8 weeks, and includes items from 5 validated scales.

Specific Aim#5: To compare ICU house officers' perceptions of working under the two intensivist staffing models.

• At the end of each 4 week rotation of ICU house officers in the Medical ICU at HSC, we will ask them to complete a questionnaire. The questionnaire asks for background information, the number of previous ICU rotations they completed, and items from 4 scales. The Autonomy and Role Conflict scales are validated. The Clinical comfort and Education/learning scales have been previously used, but have not been formally validated.

Inclusion Criteria:

• ICU admission contained entirely within the period of a single intensivist staffing format during the study period.

Exclusion Criteria:

• Repeat ICU admissions during the study period
• ICU admissions that overlapped more than one intensivist staffing format during the study period

All patients admitted to the participating Intensive Care Units during the study period