Materials and Methods
The study was conducted at an 800-bedded hospital providing tertiary-level care for vascular surgery patients. This was a single-centre, non-randomised descriptive between subjects, before-and-after design comparing pre-existing practice with a model of care previously shown to be effective in other surgical settings11. This comprised daily senior-led (registrar/consultant) medical liaison review provided by geriatric medicine physicians. This service was provided within normal working hours (0800-1700, Monday to Friday). Patient identification was triggered through direct liaison with the vascular surgery team and daily attendance to the vascular surgery ward to case-find.
Electronic records and patient case notes were analysed for all patients aged 65 years and older admitted for one or more nights. Notes were analysed retrospectively during two three-month periods across two consecutive years (January-March 2017 and 2018) to allow pre- and post-intervention analysis.
Data collection was conducted by a team of doctors and one medical student working within the Department of Medicine for Older People. Notes were requested from business intelligence services. Patient demographics were recorded including sex, age, admission type (emergency/elective), source of admission (home/other hospital/care home), operation type, comorbidities and frailty scores. Comorbidities were recorded using the Charlson comorbidity index12. Frailty was recorded using the Clinical Frailty Scale13. Outcome variables included LOS and for patients undergoing surgery, the number of complications suffered. Complications were recorded according to actual number and using guidance from The Clavien-Dindo system14. Other information recorded included admission to intensive care, 30-day readmission rates and inpatient mortality.
Statistical analysis was performed using SPSS Statistics 23. Between-group differences were analysed using correlation analysis, chi-square test of association, odds ratios, two-sample independent tests, and Kaplan-Meier analysis as appropriate. Multiple regression analyses were used to determine whether putative risk factors were relevant in their impact on LOS.