Participants were recruited from a convenience sample of patients cared for at four primary care clinics located at three teaching hospitals in the greater Boston area. These comprised an urban geriatric clinic, a suburban geriatric clinic, an urban adult primary care clinic, and a suburban adult primary care clinic. Recruitment occurred between 1 September 2007 and 30 May 2008. Clinic staff gave all scheduled English speaking patients aged 65 or over a leaflet outlining the study after patients registered for their clinic visit, which was scheduled as part of their usual care. At the end of the visit, clinic staff asked patients if they were interested in participating in the study. If patients indicated interest, the research team initially interviewed them for eligibility. Eligibility criteria included ability to communicate in English, ability to provide informed consent, and absence of moderate or severe cognitive impairment based on a short portable mental status questionnaire (SPMSQ) score of ≥7 (scores <7 indicate moderate or severe cognitive impairment).25
Study design and randomisation
After we obtained informed consent, all patients who met the eligibility criteria were randomised into one of two groups: listening to a verbal narrative describing advanced dementia (control group) or listening to the same verbal narrative followed by watching a two minute video depicting a patient with advanced dementia (intervention group). We used simple randomisation based on a computer generated scheme. Individual assignments were concealed in numbered envelopes, half of which were made available to each interviewer. One randomisation list was generated for all four clinics. At the end of the trial, the randomisation order of participants was checked against the computer generated list. A trained member of the research team followed a structured script to collect data in a quiet room in the clinic area.
For both groups, the interviewer read aloud the verbal narrative describing advanced dementia (see appendix on bmj.com). This description was based on the functional assessment staging (FAST) stage 7a.26
The FAST criteria include seven stages of dementia, with the later stages depicting more advanced disease. Stage 7 is further broken down into six substages (7a-7f). Stage 7a is generally considered the threshold for advanced dementia. The narrative states that advanced dementia is an incurable illness of the brain caused by many years of Alzheimer’s disease or a series of strokes. Its salient features are the inability to communicate understandably with others, inability to walk without assistance, and inability to feed oneself.
Participants randomised to the intervention group viewed the video decision support tool on a portable computer after listening to the same verbal narrative. The two minute video depicts the principal features of advanced dementia as described in the narrative. The video presents an 80 year old female patient with advanced dementia together with her two daughters in the nursing home setting (www.bmj.com/video/care_preferences_dementia.dtl
; also available at www.ACPdecisions.com
). The patient fails to respond to their attempts at conversation (inability to communicate). The patient is next shown being pushed in a wheelchair (inability to ambulate). Lastly, the patient is fed pureed food (inability to feed oneself). Before filming we obtained consent from her designated healthcare proxy to film the patient and to use the video for research purposes.
The development of the video followed a systematic approach,27
starting with a review of the literature on dementia and advance care planning. We then used a panel of physicians with an iterative process of comments to review the design, content, and structure of the video intervention. This panel included five geriatricians and five neurologists, all of whom specialise in the care of patients with dementia.
The video was filmed without the use of prompts or stage directions to convey a candid realism.28
The principal investigator (AEV) did all filming and editing, following previously published filming criteria.29
The video is accompanied by the same narration that was used in the verbal description arm of the study.
Data collection and other variables
At all four study sites, two members of the research team (AEV and AEJ), who were not blinded to the randomisation group, used structured questionnaires to interview participants before and after they listened to the verbal narrative alone or listened to the narrative and watched the video. At the baseline structured interview (15 minutes) we collected demographic data and data on health status and knowledge about advanced dementia. Sociodemographic data included age, race (self reported), sex, educational status, and marital status. Health status was self rated on a Likert scale as excellent, very good, good, fair, or poor. Participants were also asked if they had had a diagnosis of dementia and whether they had known a person with advanced dementia. We assessed knowledge of advanced dementia with five true/false questions that asked whether advanced dementia is curable and if patients with advanced dementia are able to communicate with others, recognise family members, ambulate, and feed themselves. Knowledge scores therefore ranged from 0-5, with higher scores indicating better knowledge.
Participants underwent a second structured interview (15 minutes) immediately after the intervention. This included knowledge of advanced dementia, preferences for goals of care, health literacy, and, for the video group, comfort with the video decision support tool. The knowledge questions were identical to those asked in the baseline interview.
There were three options for preferences for goals of care: life prolonging care, limited care, and comfort care (see appendix on bmj.com). Researchers verbally described examples of the kinds of care implied by each goal. The first option, life prolonging care, was described as aiming to prolong life at any cost. It translates into all potentially indicated medical care that is available in a modern hospital, including cardiopulmonary resuscitation and treatment in the intensive care unit. The second option, limited care, was described as aiming to maintain physical functioning. It includes treatments such as admission to hospital, intravenous fluids, and antibiotics but not attempted cardiopulmonary resuscitation and treatment in the intensive care unit. The third option, comfort care, was described as aiming to maximise comfort and to relieve pain. Only measures that provide comfort are performed. It is compatible with oxygen and analgesics but not with intravenous treatments and admission to hospital unless necessary to provide comfort. After these explanations, participants were asked about their preferences for care if they developed advanced dementia. Participants who were unable to select a level of care were considered “uncertain.”
We assessed health literacy using the rapid estimate of adult literacy in medicine tool (REALM).30
This is a two to three minute English test of medically relevant vocabulary. It is a validated test of word pronunciation and has been shown to correlate well with tests that evaluate a range of literacy skills.30
As others have done, we defined three categories for health literacy based on the REALM scores: 6th grade and below (up to age 11; score 0-45); 7-8th grade (ages 12-13; 45-60); and 9th grade and above (age 14 and over; 61-66).31 32
For those participants randomised to the video intervention group, we used a four point Likert scale to assess the perceived value of the video by asking participants whether they had a better understanding of the disease after viewing the video, if they were comfortable watching the video, if they would recommend the video to others, and whether they thought videos would be helpful for eliciting preferences for care in other diseases like cancer.
One interviewer (AEV) contacted participants by telephone six weeks after the initial interview to determine again what their preferences would be if they had advanced dementia in exactly the same manner as the initial interview. We chose a follow-up period of six weeks to ensure that an adequate amount of time elapsed from exposure to the intervention and to assess whether the video had an enduring effect.
Our analyses were based on the decision making group to which participants were randomised. The primary outcome measure was preferences for care if they developed advanced dementia categorised as four options (life prolonging, limited, comfort, or uncertain). Additional outcomes included change in knowledge scores before and after the intervention and the stability of preferences after six weeks.
All characteristics of participants and outcomes were described by using proportions for categorical variables and means (SD) for continuous variables. We used χ2 tests to compare preferences for care (life prolonging, limited, comfort, or uncertain) between the two groups.
Two sample t tests compared change in knowledge scores before and after the intervention between the two groups. We used κ statistics to summarise the stability of preferences six weeks after the clinic interview for each group and compared the proportions who changed preferences with Pearson χ2 exact test between the two groups.
The measure for the primary outcome analysis was the unadjusted difference in proportions of participants preferring comfort care between the two study groups. We conducted secondary analyses to identify factors associated with a preference for comfort care among all participants. Bivariate analyses determined the association between individual characteristics of participants (age, sex, race, education, marital status, health status, personal history of dementia, previous relationship with a person with advanced dementia, health literacy, and randomisation group) and a preference for comfort care with Fisher’s exact test. Multivariable logistic regression analyses were used to identify factors independently associated with preferences for comfort care. Factors significant at 0.10 in the bivariate analyses were entered into a stepwise algorithm, retaining factors in the model that were significant at the 0.05 level. We used the variance inflation factor to diagnose colinearity among potential predictors.
All reported P values are two sided, with P<0.05 considered as significant. The study was designed to detect a 25% difference in the proportion of participants choosing comfort care between the two groups, assuming the rate in the verbal group was 60%. With a target of 100 patients in each group, the power of the study was estimated to be >90%. Data were analysed with SAS software, version 9.1 (SAS Institute, Cary, NC).