Preoperative Noninvasive Cardiac Testing in Older Adults With Hip Fracture

A Multi-Site Study

Liron Sinvani, MD; Daniel A. Mendelson, MD; Ankita Sharma, DO; Christian N. Nouryan, MA; Joanna S. Fishbein, MPH; Michael G. Qiu, PhD, MD; Roman Zeltser, MD; Amgad N. Makaryus, MD; Gisele P. Wolf-Klein, MD


J Am Geriatr Soc. 2020;68(8):1690-1697. 

In This Article



A 2-year retrospective chart review of patients aged 65 years and older admitted with acute hip fracture was conducted across seven hospitals (three tertiary and four community hospitals) from April 1, 2014, to December 31, 2015. In all, 1,079 charts were identified and analyzed.

The TTS explanatory variables included preoperative TTE performed (yes/no); day of the week for admission (weekday: Monday-Friday; weekend: Saturday or Sunday); age category (65-74, 75–84, or ≥85 years); sex; race/ethnicity; hospital type (tertiary, community); services category (medical, surgical); marital status (married/civil union, single/separated/divorced, widowed); and Charlson Comorbidity Index (CCI).


Descriptive statistics (mean and standard deviation [SD] for continuous variables, frequencies and proportions for categorical variables) were calculated. To assess the association between potential explanatory factors and each of the three patient outcomes (TTS, LOS, and in-hospital mortality), correlation of patients within hospitals was first assessed (specifically, the intraclass correlation coefficient [ICC] was computed). Because the ICC was negligible (≤.07), for all three outcomes, hierarchical modeling was not necessary for analysis.

For in-hospital mortality, logistic regression modeling was used. For both TTS and LOS outcomes, Kaplan-Meier product-limit curves were constructed for categorical factors with each outcome separately along with Cox regression for continuous factors of interest. Given the variability in when TTE was performed across subjects, Cox regression extended for time-dependent covariates was performed for each endpoint.

The frequency and proportion of patients who underwent a preoperative TTE were computed overall and by hospital, along with the exact 95% binomial confidence interval (CI) for the overall rate. Associations between patient demographics and clinical characteristics with preoperative TTE status were assessed using the chi-square test for categorical data and the Wilcoxon rank sum test for continuous data, as appropriate.

For TTS, an extended Cox regression model using preoperative TTE as a time-varying effect was modeled (whereby a patient's status on this variable can change during the course of their admission).

For LOS, the time to event (where the event is having been discharged alive) was measured in days between surgery and discharge. In cases where the patient died before discharge, their status was censored at date of death. Covariate adjustment for time between admission and surgery was included along with a binary variable (Yes/No) for preoperative TTE status; postoperative TTE status was included as a time-dependent covariate in the LOS analyses.

For all outcomes, a univariate screen was first performed to assess candidate predictor variables; any factors yielding a result of P < .1 was considered for inclusion in the respective final multivariable model for that outcome. Final multivariable models were selected using backward elimination and α = .05 significance level, and in some cases, preoperative TTE was kept in the model due to the primary interest in this factor for this study. All analyses were conducted using SAS, v.9.4 (SAS Institute, Cary, NC).