Rothman index as a predictor for type of discharge and readmission rates in a cancer hospital: The Yale experience.

Abstract

6635 Background: There is limited data on predictors of discharge (DC) and readmission (RA) for oncology inpatients. The Rothman index (RI) is a single score calculated from 26 parameters including nursing assessments, vital signs, and laboratory tests, integrated with the hospital’s electronic medical record and automatically updated when new data are entered. Scores are inversely related to patient impairment. We evaluated the role of RI in the prediction of DC types and RA rates at the Smilow Cancer Hospital at Yale. Methods: Patients (pts) with solid tumors, length of stay longer than 24 hours, admitted between 6/1/11 and 12/31/11 were enrolled. Elective admissions were excluded. DC types were classified as home without hospice (Hm), nursing home (NH), hospice (Hp) and death (D). The RI at the time of DC was evaluated as a predictor for RA at 7 or 30 days and the type of DC. Analysis of variance (ANOVA) was used to determine whether there were statistically significant differences between RI means and DC categories, followed by pair-wise contrast using the Tukey method. Results: Among the 476 admissions from 331 pts, the median age was 63 (range 21-92, 23% age 70 or older), and 55% males. The most common tumors were lung (24%) and breast (8%). The number of DCs were 313 Hm (66%), 75 NH (16%), 55 Hp (11%) and 33 D (7%). RI at DC for H, NH, Hp and D were 74, 62, 47 and 23. All comparisons among RI at DC and type of DC were significant (p < 0.05). Among the Hm pts, the rates of RA at 7 days were significantly higher for RI < 50 (45%) and 50-60 (32%) compared to 81-90 (13%) and > 90 (7%) (Table). RA rates at 30 days were also related to RI at DC, ranging from 65% in RI < 50 to 33% in RI > 90. Conclusions: RI is a useful tool in the evaluation of inpatients by providing an objective metric at the time of DC. The high correlation between RI scores and both DC type and RA rate underscores the need for a more consistent and reproducible evaluation at the time of DC and may be used by health care providers to guide disposition planning. A prospective study is being planned at our institution. [Table: see text]