Study protocol: The Improving Care of Acute Lung Injury Patients (ICAP) study

To be eligible for enrollment in the ICAP study, participants must be mechanically ventilated and meet criteria for the diagnosis of ALI/ARDS, as defined by the American-European Consensus Conference [13].

Patients are excluded if they meet any of the following eight criteria:

1. More than 96 hours between ALI/ARDS diagnosis and enrollment.

2. More than five days of mechanical ventilation during the present hospitalization before enrollment.

3. Pre-existing ALI/ARDS when transferred to a study ICU.

4. Pre-existing illness with a life expectancy of less than six months.

5. Any limitation of care at the time of enrollment (for example no cardiopulmonary resuscitation).

6. Previous lung resection.

7. Inability to speak or understand English.

8. No fixed address.

These criteria exclude patients with substantial exposure to critical care before enrollment (exclusion criteria 1, 2, and 3), high short-term mortality risk unrelated to ALI/ARDS (exclusion criteria 4 and 5), or significant barriers to prescribed outcome evaluations (exclusion criteria 6, 7, and 8).

Exposures assessed in the ICAP study fall into three major categories: patient-based, clinical management, and ICU organizational exposures (Table 1) [14]. Patient-based exposures assessed at the time of study enrollment include demographics, comorbidities, admission diagnosis, severity of illness, and medications taken before hospitalization. Pilot testing indicates that medical chart abstraction for one-time measurement of these exposures takes about 30 minutes.

Patient-based and clinical management exposures measured while the participant is in the ICU vary over time, so the exposure measurements must account for this time-dependent course of critical care after ALI/ARDS diagnosis. Of particular importance, given the study objectives, are the participants' mechanical ventilator settings (for example. tidal volume and pressures) and associated arterial blood gas values. These important time-varying exposures are measured with the greatest frequency (twice daily). Other patient-based and clinical management exposures are measured once daily. On the basis of pilot testing, the time required for daily collection of these exposures is about 45 minutes per patient-day (Table 1).

ICU organizational exposures measured in the ICAP study include staff:patient ratios, ICU occupancy, and use of protocols for specific ICU therapies (for example. LTVV and intensive insulin therapy [15]). Although infrequently considered in previous studies [16], recent research has demonstrated that ICU organizational exposures have an important impact on short-term patient outcomes [17‐19]. Thus, prospective measurement of these variables and their association with long-term patient outcomes has the potential to inform clinical practice. Because some of these ICU organizational exposures vary less frequently than other exposures, they are measured twice weekly.

When available, existing validated measurement instruments are used for exposure assessment. The Acute Physiology and Chronic Health Evaluation (APACHE) II system, Sequential Organ Failure Assessment (SOFA) score, and Charlson comorbidity index are validated predictors of short-term mortality and other outcomes in ICU patients [20‐22]. Similarly, validity and reliability have been demonstrated for the Richmond Agitation–Sedation Scale (RASS) [23, 24] and the Confusion Assessment Method for the ICU (CAM-ICU) [25] as measures of sedation and delirium, respectively, among ICU patients. Delirium, measured with the CAM-ICU, is an important independent predictor of mortality in mechanically ventilated patients [26].

Poor long-term outcomes in ALI/ARDS survivors may reflect either pre-existing morbidity or persistent effects of critical illness and/or ICU treatments. For example, three studies that retrospectively measured baseline QOL in critically ill patients found significant global decrements in QOL when compared with population norms [27‐29]. Thus, long-term QOL decrements in ICU survivors (in comparison with population norms) probably reflect both poor baseline status and potential adverse effects of critical illness. To understand the relative contributions of these factors, patients' baseline status before hospitalization must be assessed and compared with their status at follow-up.

Because ALI/ARDS patients are admitted to the ICU under emergent conditions, their ideal baseline (for example, pre-admission) status is difficult, if not impossible, to obtain. Consequently, most studies of ICU survivors do not control for baseline status when measuring outcomes during follow-up [16]. To help in addressing this methodological issue, the ICAP study will use standardized surveys for the retrospective estimation of baseline QOL, physical functional status, and hearing handicap in patients surviving their ICU stay (Table 2). Pilot tests indicate that administration of these retrospective surveys takes about 60 minutes per participant (Table 2), more than double the time required for survey administration during follow-up. This additional time results from fatigue, inattention, and interruptions related to medical care while administering these surveys on the hospital ward.

Participants' retrospective report of their baseline status may be influenced by their current status at the time of this assessment, leading to recall bias in the baseline measurement. Thus, the ICAP study will explore patient proxies as an alternative method of estimating baseline status. In a subgroup of consecutive consenting participants, close contacts will be identified and asked to complete the same baseline surveys that were completed by patients. Proxy responses will then be compared with those of the participants to assess the overall level of agreement and any systematic differences between the two response groups. Understanding differences between participant and proxy-based assessments will enable investigators to adjust for any systematic differences in proxy responses obtained during follow-up when participants are not available for direct assessment (for example, owing to hospitalization, impaired physical or mental condition, or incarceration).

Outcomes assessed in the ICAP study include mortality and medical outcomes, organ impairment, functional status, and quality of life (Table 3). Outcomes are assessed at 3, 6, 12, and 24 months after ALI/ARDS diagnosis in a research clinic staffed by trained research assistants, nutritionists, and physical therapists. When patients are unable to attend their clinic appointment, visits to their home or rehabilitation facility or telephone interviews will be conducted. Each follow-up visit requires about four hours (Table 3).

Outcomes are assessed with standardized instruments, most of which are used widely in critical care research [30]. Some instruments, such as the Medical Outcomes Study Short Form 36-Item Health Survey (SF-36) [31], have been specifically validated in critically ill populations, whereas others (for example, the EuroQOL (EQ-5D) [32], Activities of Daily Living (ADL) [33], and Instrumental Activities of Daily Living (IADL) [34] surveys) have been validated only in more general patient populations. Because certain issues being explored within the ICAP study have not been widely investigated in critically ill patients, these assessments will require instruments validated predominantly in non-ICU patient populations. For example, the ICAP study is the first large study to measure long-term swallowing impairment and hearing handicap among ICU survivors. Corresponding instruments (for example, Sidney Swallowing Questionnaire and Hearing Handicap Inventory for Adults – Screener) were selected on the basis of validation studies completed in other patient populations [35, 36].

The ICAP study collects a large amount of data regarding patient exposures and outcomes. To reduce measurement error, a comprehensive data quality assurance program will be employed. As outlined in Table 4, this program includes several components: first, a comprehensive training, certification and ongoing quality assurance review of study coordinators; second, duplicate data entry with extensive validity checks conducted by the data entry staff and database software; and third, regular review of the entire database, by means of customized descriptive statistics algorithms, to assess potential outlier, illogical and missing data values. This multi-faceted approach will help to ensure the accuracy of data in the ICAP study.

While in the ICU, most eligible patients cannot provide informed consent to participate in the ICAP study because of sedation, delirium, or physical or emotional distress. Furthermore, pilot testing of the ICAP study protocol revealed that patient proxies are frequently not available on a timely basis. Delays in obtaining consent impede the accurate assessment of patient exposures that cannot be obtained retrospectively (for example, sedation and delirium status). Thus, to facilitate prompt participant enrollment without biasing the study sample against patients without readily available proxies, a waiver of consent was requested from the Institutional Review Board of each participating site. Because the study protocol poses minimal risk to patients, this waiver request was approved, thus allowing observational data to be collected, without consent, on eligible patients during their hospitalization. Patients who survive their ICU stay are then asked for consent to participate in the long-term follow-up portion of the study.

An important challenge facing longitudinal research in ICU survivors is the threat of selection bias from loss to follow-up. Participants who are lost to follow-up may not be representative of the original study sample as a result of increased morbidity (hindering appointment attendance) or improved outcomes (increasing mobility and capacity to move to distant locations) [14]. Thus, participant retention strategies will have a vital role in the success of the ICAP study (Table 5). Retention efforts will begin immediately once eligible patients have consented and will continue for the duration of the study, using frequent telephone, written, and in-person communication with each participant. For purposes of scheduling follow-up appointments, participants will be actively tracked with the following sequence of events: three telephone calls to the primary contact number; telephone calls to alternative numbers; signature-required/registered letters; telephone calls and letters to alternate contacts; local and online telephone directory searches; unscheduled home visits; and confirmation of vital status from government databases. The proportion of patients contacted and the number of visits scheduled and completed will be reviewed on a weekly basis to track the implementation and performance of the retention strategies.

The primary statistical analysis for the ICAP study will assess the casual effect of LTVV on patient mortality. Exposure to LTVV will be evaluated with a compliance algorithm from a related randomized trial [4]. This analysis is complex because within the observational study design, the primary exposure (for example, LTVV) is 'time-dependent' and 'dynamic', meaning that the exposure varies over time and depends on a patient's 'at risk status' (for example, being alive, being in hospital, and receiving mechanical ventilation), which also changes during the study period. Appropriate statistical methods for the causal analysis of such time-varying, dynamic treatment regimes have been developed recently [37]. These methods, known as structural models, demonstrate that the analysis of such data with standard statistical techniques (for example, a Cox proportional hazards model with LTVV and risk factors modeled as time-varying covariates) can lead to biased results. More specifically, a bias can result when two conditions occur: first, there is a measured time-dependent risk factor for survival that also predicts subsequent exposure to LTVV, and second, past LTVV exposure affects the subsequent level of the risk factor. Within the ICAP study, these conditions may be met for important time-varying risk factors, such as organ failure and lung compliance, which may be associated with both death and compliance with LTVV while also being associated with future levels of these exposures. To avoid such bias in the statistical analysis, we plan to use structural models to estimate the causal effect of LTVV on mortality.

Additional analyses of LTVV will include assessing its casual effect on functional status and quality of life. Because these outcomes can be assessed only in survivors, the analysis of functional status and quality of life, at a specified time point after enrollment, is complicated by the high mortality rates of critically ill patients [38]. An analysis that restricts the study population to observed survivors can suggest a treatment effect simply because the type of patients who survive differ between treatment regimes. Recently developed statistical methods that address this survivor bias [38‐40] will be extended to accommodate dynamic, time-varying treatments.

Secondary analyses in the ICAP study will examine the impact of ICU exposures, other than LTVV, on long-term outcomes including mortality, organ impairment, functional status, and quality of life. Of particular interest are exposures previously shown to affect short-term outcomes, including delirium [26], sedative use [41], ICU organizational characteristics [18, 42], and tight glucose control with intravenous insulin therapy [43]. In addition, there are many other unanswered clinical questions that the ICAP study may inform, such as the association between the dose and duration of use of steroids, and functional status. The relationships between these exposures and long-term outcomes in ALI/ARDS survivors currently remain unclear, as do these exposures' relevant mechanisms of action. To inform this latter question, analyses will explore the effect of characterizing the exposure 'dose' as cumulative (for example, total exposure during ICU stay), maximum/minimum (highest or lowest value attained during ICU stay), or proportional (percentage of ICU stay greater than a threshold). Ultimately, these secondary analyses seek to extend the results of previous studies and generate new hypotheses about how existing clinical practices affect long-term patient outcomes.

On the basis of the actual incidence of ALI/ARDS observed at the participating hospitals during a previous randomized trial [4], the ICAP study seeks to enroll 520 patients during its 2-year enrollment period. With this sample size, the ICAP study would be the largest existing prospective cohort study of long-term outcomes in ALI/ARDS survivors [16] – significantly larger than the largest previous study, which enrolled 107 survivors [44].

Because there are no available methods for computing power or sample size for the statistical methods described above, we estimated statistical power for the ICAP study with conventional methods. On the basis of preliminary data regarding the short-term mortality rates of ALI/ARDS patients in the ICAP study and long-term mortality rates observed in previous studies [8, 45], we expect that 50% of patients receiving more than the median level of LTVV will die within two years of ALI/ARDS diagnosis. Under these assumptions, the ICAP study has 86% power to detect a relative hazard of 0.7 for 2-year mortality, comparing equal-sized groups of patients receiving a higher versus a lower proportion of LTVV (two-tailed α = 0.05). However, this calculation does not consider that the treatment groups may be unbalanced with regard to confounding factors for patient mortality. To address this deficiency, we conducted a simulation to assess whether 520 patients, using an analysis that accounts for confounding, would be adequate to detect a 30% relative reduction in mortality at two years. The majority of the simulation scenarios yielded a power of more than 70%.

The benefit of the ICAP study's proposed sample size, in comparison with previous studies of ALI/ARDS survivors, is most apparent when analyzing other patient outcomes. For example, one important secondary outcome is the physical functioning domain of the SF-36 quality of life survey. Physical functioning has the greatest precision (for example, it reflects the greatest number of response categories) of the eight SF-36 domains and clinically represents an important domain for understanding the impact of ALI/ARDS [8]. Recent expert panels have suggested that the minimum clinically important difference in the SF-36 physical functioning domain for patients with chronic pulmonary conditions is 10 points [46, 47]. The power of the ICAP study to detect a 10-point difference at 2-year follow-up, comparing patients receiving a higher proportion of LTVV with patients receiving a lower proportion, is much greater than in previous studies of ALI/ARDS survivors that had sample sizes of less than 84 survivors [8, 45, 48‐50] (Figure 2). Thus, the ICAP study will have sufficient power to detect associations of clinical relevance that could not be adequately investigated in earlier studies with smaller sample sizes.