Examining the context and helpfulness of family companion contributions to older adults’ primary care visits

TL;DR

In 30 older adult–family companion primary care dyads, companions mostly facilitated information exchange, and their verbal activity and autonomy-related behaviors varied with patients’ approach to managing health. Patients who preferred family involvement in decisions rated companions as more helpful, but a pre-visit agenda-setting intervention did not significantly change companion behaviors.

Key findings

• Companion communication was predominantly autonomy enhancing: such behaviors averaged 35 times per visit and represented 81% of all companion behavior, with clarifying or expanding patient information the dominant behavior (55% of behaviors; average 25 times per visit). Autonomy detracting behaviors averaged 8 times per visit (19%), most commonly correcting or blaming the patient (average 2 times per visit; 5%).
• Patients who delegated health management were less verbally active than co-managers or self-managers (16.8% vs. 42.6% and 40.3% of statements; p<0.001), while their companions were more verbally active (45.2% vs. 14.5% and 13.8%; p<0.001). Companions of delegating patients also performed more autonomy enhancing behaviors (73.4 vs. 37.4 and 25.2 per visit; p<0.05) and more autonomy detracting behaviors (26.0 vs. 6.8 and 5.2; p<0.01).
• Patients who preferred to leave medical decisions to family/close friends rated companions as significantly more helpful than patients with shared or independent decision-making preferences (mean 6.7, 4.6, and 2.5, respectively; p=0.05). Visits were shorter for patients who preferred to leave decisions to family/close friends (20.5 minutes) than for those preferring shared decisions (38.8 minutes) or independent decisions (31.6 minutes; p<0.05).
• The pre-visit agenda-setting intervention did not significantly affect companion behaviors: intervention companions contributed more autonomy enhancing behaviors than controls (47.2 vs. 43.5 per visit), most notably facilitating doctor understanding (37.1 vs. 30.9), but differences were not statistically significant.

Methods

• Analysis of audio-taped primary care visits and linked patient surveys from a pilot patient-companion communication intervention conducted in an academic geriatrics clinic between April and August 2012; final sample was 30 older adult–companion dyads after excluding a joint visit of two patient participants accompanied by an adult child from 32 dyads.
• Patients were age 65 or older, English speaking, able to communicate by telephone and read large print, and typically accompanied by an unpaid family member or friend; significant cognitive impairment was excluded. Dyads were randomized to a pre-visit agenda-setting checklist intervention or usual care, visits were audio-recorded, and patients and companions completed surveys immediately after and by telephone 2 weeks later.
• Visit dialogue was coded with the Roter Interaction Analysis System for verbal activity; transcripts were coded in Nvivo 10 using Clayman’s autonomy enhancing/autonomy detracting companion behavior categories. Analyses used SAS 9.4 PROC GLM for unbalanced analysis of variance, with treatment group included as a dichotomous measure; small sample size precluded adjustment for many patient factors.

Limitations

• Stated: this was an exploratory study using a convenience sample of older patients and companions from a single academic geriatric clinic; the small sample precluded multivariate regression models or stratification by patient/companion characteristics.
• Stated: older adults with significant cognitive impairment were excluded, so findings cannot address older adults with dementia; patient preferences for involving family in decision-making were assessed with a single question, could not definitively be linked to the specific companion present, and could not distinguish preferences across different decisions or health issues.
• Evident: subgroup cells were very small (e.g., n=3 delegating management; n=3 preferring to leave decisions to family/close friends), and the analysis did not adjust for confounders, so the reported associations are preliminary and cannot establish causality.