BACKGROUND: This study describes longitudinal changes in the composition and accuracy of modal life-sustaining treatment preferences as predictors of patients' treatment preferences. METHOD: Healthy outpatients age 65 and older and their surrogate decision makers recorded preferences for 4 treatments in 9 hypothetical illness scenarios 3 times over a period of 2 years. A statistical prediction model, based on patients' modal preferences, was created using initial responses and updated 2 years later. RESULTS: When reestimating the model at 2 years, 4 of 27 items in the model created using baseline responses no longer reached the threshold for inclusion, but 5 new items did meet criteria. All modal preference changes reflected a trend toward refusing treatment. Both the original and updated models were more accurate in predicting patients' preferences than were surrogates making concurrent predictions. Adding covariates (e.g., gender, age, presence of plans for future medical care) did not alter the model's predictive superiority over surrogates. CONCLUSIONS: Models using modal preferences are useful to patients, surrogates, and physicians when trying to accurately discern end-of-life treatment choices, but the models must be updated periodically.
BACKGROUND: This study describes longitudinal changes in the composition and accuracy of modal life-sustaining treatment preferences as predictors of patients' treatment preferences. METHOD: Healthy outpatients age 65 and older and their surrogate decision makers recorded preferences for 4 treatments in 9 hypothetical illness scenarios 3 times over a period of 2 years. A statistical prediction model, based on patients' modal preferences, was created using initial responses and updated 2 years later. RESULTS: When reestimating the model at 2 years, 4 of 27 items in the model created using baseline responses no longer reached the threshold for inclusion, but 5 new items did meet criteria. All modal preference changes reflected a trend toward refusing treatment. Both the original and updated models were more accurate in predicting patients' preferences than were surrogates making concurrent predictions. Adding covariates (e.g., gender, age, presence of plans for future medical care) did not alter the model's predictive superiority over surrogates. CONCLUSIONS: Models using modal preferences are useful to patients, surrogates, and physicians when trying to accurately discern end-of-life treatment choices, but the models must be updated periodically.
Authors: Catherine L Auriemma; Christina A Nguyen; Rachel Bronheim; Saida Kent; Shrivatsa Nadiger; Dustin Pardo; Scott D Halpern Journal: JAMA Intern Med Date: 2014-07 Impact factor: 21.873
Authors: Mi-Kyung Song; Sandra E Ward; Mary Beth Happ; Beth Piraino; Heidi S Donovan; Anne-Marie Shields; Mary C Connolly Journal: Res Nurs Health Date: 2009-06 Impact factor: 2.228