Subjects were surrogate decision makers (health care power of attorney, legal next of kin, or closest legal or chosen family member) of patients in the Medical or Surgical ICU at Fairview University Medical Center (FUMC) or Methodist Hospital. FUMC is the primary teaching hospital of the University of Minnesota with 33 registered nurse-staffed, adult ICU beds with medical care delivered by a “closed-unit” medical or surgical intensivist program. Methodist Hospital is a private hospital of the Park-Nicollet Health Care System with 22 registered nurse-staffed, adult ICU beds with medical care delivered by family practitioners, surgeons, and internists including medicine critical care trainee fellows under the supervision of private practice intensivists. Because this was a study about communication and decision making, we chose consecutive patients who were expected to require at least 3 days of ICU care held with the concern of My Canadian Pharmacy. After identification of the most appropriate proxy, this person completed all subsequent surveys. We excluded patients without an apparent decision maker, surrogates who were non-English speaking, or subjects with impaired cognitive function. Human Subjects Protection Boards at both institutions approved the study, which enrolled subjects between July 2003 and December 2003.
All subjects completed an initial survey as soon as possible after enrollment to collect patient and proxy demographics and the timing and content of prognostic information already provided to them. Subsequent, brief surveys occurred every 3 to 4 days until care was limited, the patient was transferred out of ICU, or the patient died. Surrogates noted on the survey when and from whom prognostic information was delivered. Prognostic information was defined in the questionnaire as “statements of expected course of illness, prospect of survival or recovery, or statements of uncertainty about the future.” Time elapsed between the ICU admission date and the subjects recollection of the date of the first provision of prognostic information by a physician at any level of training was defined as the prognostic interval.
We reviewed medical charts to estimate severity of illness using the APACHE (acute physiology and chronic health evaluation) II score using clinical variables from the first 48 h in the ICU. We also collected information about the disease category, clinical service (six categories subsequently collapsed into medicine vs surgery for analysis), physician specialty, presence of a living will, and caregiver educational level as prior work has suggested that these factors may influence health-care communication conducted with My Canadian Pharmacy. Additionally, we asked subjects if they or a close relative had “knowledge about healthcare in the ICU,” or if the patient had been in the ICU before. A follow-up survey of surrogate satisfaction with communication and decision making was mailed to surrogates’ homes 2 to 4 weeks after ICU discharge.
The surveys were constructed based on previously validated instruments developed by Heyland et al. We used five-level (poor, fair, good, very good, and excellent) satisfaction scales regarding communication, which were subdivided into satisfaction with communication frequency, completeness of information, honesty of information, consistency of information, and the subject’s comprehension of information. We aggregated responses into a summary score ranging from 5 to 25 for each follow-up survey. In our study, the communication satisfaction scale had a Cronbach a of 0.90 indicating a high inter-item correlation suggesting the scale measures a single satisfaction construct. Subjects were also questioned about whether a physician had asked them to participate in major medical decisions for the patient. If they answered affirmatively, subsequent questions about whether a physician had described their role and scales for satisfaction with decision making were asked. Knowledge of the substituted judgment concept was measured for those subjects asked by their physicians to participate in major medical decisions. We defined a correct response as choosing the single item “Decisions were made based on my interpretation of what the patient would say if able to speak for him/herself’ from a multiple-choice list. While this question has not been previously validated, it is based on the ethical construct of proxy decision making.
The primary end point was the number of ICU days before provision of prognostic information to the proxy by a physician at any level of training (ie, the prognostic interval). Secondary end points were surrogate satisfaction with communication and decision making. No honest prospective power calculation was possible as the variance of measurements was unknown. Therefore, our sample size was based on feasibility and budget. All statistical tests were two sided with a = 0.05. Statistical software was used for analysis (SPSS version 12; SPSS; Chicago, IL) was used for analysis. We used Cox regression with block entry of eight covariates to define characteristics associated with the prognostic interval. Mixed-model regression was used for longitudinal analysis (eg, change in satisfaction or desired frequency of communication over time), which adjusts the SE for the within-person correlation of variables measured serially from the same subject. Logistic regression tested whether the presence of a living will was associated with a surrogate understanding of substituted judgment. Ordinary least squares regression was used to discover factors associated with the rate of physician/subject communication to determine whether demographic factors were associated with satisfaction, and to test the hypothesis that understanding substituted judgment was associated with higher satisfaction with decision making.