Evaluation of prognostic indices in elderly hospitalized patients

Prognosis informs the physician's decision-making process, especially for frail older adults. So far, any non-disease-specific index has proven full evidence for routine use in clinical practice. Here, we aimed at assessing, prospectively, the calibration and discriminating accuracy of validated prognostic indices in a cohort of elderly hospitalized patients.

This was a prospective observational study that enrolled elderly patients (n = 100). The patients' assessment included clinical variables, as well as the following five prognostic indices of mortality: (i) Levine index (2007); (ii) Walter index (2001); (iii) CARING (C, primary diagnosis of cancer; A, ≥ 2 admissions to the hospital for a chronic illness within the last year; R, resident in a nursing home; I, intensive care unit admission with multiorgan failure, NG, noncancer hospice guidelines [meeting ≥ 2 of the National Hospice and Palliative Care Organization's guidelines]) criteria of Fischer (2006-2011); (iv) Silver Code of Di Bari (2010); and (v) Burden of Illness Score for Elderly Persons of Inouye (2003).

Patients' clinical characteristics: 70% women (age 86.20 ± 0.69 years), 30% men (age 85.40 ± 1.07 years), Comorbidity Illness rating scale (CIRS) 4.3 ± 0.61 and Barthel Index 28 ± 0.54. Walter and Burden of Illness Score for Elderly Persons scores showed similar prediction rates when compared with the expected validated values (ancova: F = 14.00, P < 0.008). Burden of Illness Score for Elderly Persons was the most calibrated and accurate index (receiver operating characteristic curve 0.72; P < 0.02).

None of the assessed prognostic indices, in a "real world" scenario, afforded the optimal predictive accuracy (receiver operating characteristic curve 0.90); all these indices are still far from a robust answer to the prognosis in older age, reflecting a poor ability to encompass the spectrum of frailty. Effort should be made to tailor the prognostication in geriatrics, moving from a disease-centered model to a precision model, tailored to the frail phenotype. Geriatr Gerontol Int 2017; 17: 1015-1021.