5-YEAR SURVIVAL-DATA IN SURGICAL DECISION-MAKING - WHAT ASPECTS OF GRAPHICAL DATA INFLUENCE PATIENTS PREFERENCES

Five-year survival data have formed the crux of surgical decision maki ng regarding the question: what treatment is best for my patient? Yet little research has been done to assess how patients' expressed prefer ences regarding treatment are influenced by 5-year survival data displ ayed in a graphical format. This study was designed to assess (a) how patients' attention to features of survival curves in the short, mediu m, and long term influences their choices of treatments, and (b) wheth er patients attend to sharp declines in mortality vs. gradual declines in mortality in their decision making. The study was designed as cros s-sectional structured interviews of patients and was set up in a univ ersity-based Department of Veterans Affairs Medical Center; 202 patien ts (mean age = 65.7 years, SD = 9.9, range = 29-88) were seen consecut ively in a general medicine continuity clinic. Five survival graphs we re presented to patients. Each graph contained a pair of survival curv es. Each curve pair represented the survival curves for two alternativ e unidentified treatments for an unidentified medical condition. The L T curve - with better long-term, worse short-term survival (LT) - rema ined fixed in each survival curve pair, while the second curve - with better short-term, worse long-term survival (ST)varied short-term surv ival. The ST curve across all survival graphs consistently yielded no immediate risk at the time of treatment (year 0) and a 22% 5-year surv ival, which is 15% worse than the 5-year survival of the LT curve. We varied survival in years 1-4 of the ST curve over the five curve pairs . Patients were sequentially presented with survival curve pairs in in creasing ST survival (order A) or in decreasing ST survival (order B). For each survival curve pair, patients were asked to express a prefer ence for treatment and to identify the single feature of the survival curve pair that most influenced their choice. Patients receiving the o rder-B presentation were significantly less likely to choose the ST tr eatment option on the curve pair reflecting actual medical data than p atients receiving the order-A presentation (chi square = 6.06, df = 1, P = 0.014). Further, order-B patients reported attending to more of t he medium-term features of the curves. Patients receiving the order-A presentation tended to focus on the end-point results (short-term or l ong-term, chi square = 11.2, df = 1, P = 0.0008). This study demonstra ted that the presentation of an ordered array of curve pairs with sequ ential variation of the features of the curves influences both patient s' expressed preferences for treatments as well as the features of the curves that they report as most influencing their choices. These resu lts indicate that patient attention and preferences may be highly infl uenced with an intervention comprised of a decision-making exercise.