Measuring men's preferences for involvement in medical care: getting the question right

Now that active involvement by patients in their health care is widely endorsed, valid and reliable methods for determining preferences for involvement in treatment decision making are essential. Relatively little methodological work has been conducted to compare and contrast their reliability and validity. Available single-item measures exist to determine preferences, ranging from 'menu-based' questions to simpler Likert-type scales.

Within a larger community survey of 514 men aged 50-70 years in Sydney, Australia, we compared two measures to assess their preferences for involvement in medical decision making. Using the 'menu-based' Control Preference Scale (CPS), men were classified as preferring to be either 'passive' or 'active' during decision making or to share ('shared') with their doctors on an equal basis. Men also were classified as preferring to be either 'passive' or 'active' according to a Likert-scale measure.

Agreement between the two measures was 'poor' (kappa=0.19). While 24.9% of participants were classified as preferring a 'passive' role in treatment decision making according to the CPS, almost half (47.9%) were so classified according to Arora and McHorney's measure. In the absence of a 'shared' response option on the Arora and McHorney measure, 45.3% of men classified as preferring a 'shared role' on the CPS were instead categorized as 'passive' using Arora and McHorney's measure. Predictors of preferring a 'passive' role also differed, depending on the measure employed. Only occupational skill level predicted men's preferences for a 'passive' role when measured by the CPS [odds ratio (OR)=1.67; 95% CI 1.09-2.55] (P=0.02). For the Arora and McHorney's measure of preferences for involvement, men were significantly more likely to prefer a 'passive' role if they were older [adjusted odds ratio (AOR)=1.06, 95% CI 1.02-1.09] (P=0.001), currently smoking (AOR=1.86, 95% CI 1.09-3.17) (P=0.02) and had higher chance health locus of control scores (AOR=1.26; 95% CI 1.01-1.56) (P=0.04). Having been employed or previously employed in an occupation of a lower skill level was also significantly and independently predictive of a passive role (AOR=2.35, 95% CI 1.57-3.50) (P<0.001).

Single-item measures of decisional preferences have poor convergent validity. Characteristics associated with preference classifications also differ, depending upon the measures used. These results suggest that research efforts should be directed towards developing psychometrically robust measures to determine decisional preferences.