Dijkhof, M. (Mike) (2021) Uncovering the preoperative features that underlie a postoperative complication in the onco-geriatric population using Partial Least Squares Discriminant Analysis. thesis, Human Movement Sciences.
Full text available on request.Abstract
Background: Cancer incidence increases especially in the older population and more patients with increased complication risks will be administered to the hospital. Using a Partial Least Squares Discriminant Analysis (PLS-DA), the perioperative features that differ between patients developing a complication at home and those that recover uncomplicated were analysed. Methods: Using data from an observational cohort study within the onco-geriatric population a PLS-DA classifier was built. Patients aged 65 years and older scheduled for oncologic surgery were included. Exclusion criteria were severe mobility, hearing, vision, and cognitive impairments. Physical activity (PA) was monitored using a Fitbit from study intake until three months postoperative. At study intake and hospital discharge the Timed Up and Go (TUG), handgrip strength, Hospital Anxiety and Depression Scale (HADS), (instrumented) Activities of Daily Living ((i)ADL) and Groningen Frailty Index (GFI) were administered. Based on the ICF-model domains, features were selected based on these data combined with demographic and clinical data. Model outcomes were evaluated using AUC, accuracy, sensitivity, specificity, precision, recall and F1-score. Feature importance was assessed using the VIP-scores. Results: The final dataset consisted of 32 patients used for PLS-DA with 7 latent variables. The PLS-DA accuracy was 93.8% with a specificity of 95.5%. A precision of 90% and sensitivity of 90% resulted in a F1-score of 90%. The most important preoperative features were alcohol use, daily average step count and MVPA, TUG-score, handgrip strength, scores on the HADS depression domain, iADL, and GFI-score. Discussion: The PLS-DA performance was high and only misclassified two patients. However, due to the small and unbalanced dataset, the outcomes of this study should be interpreted with caution and should serve as direction for further research rather than a definitive outcome. Nonetheless, these results indicate that PA measures as well as other measures in the function and disability component of the ICF-model are valuable additions for complication risk assessment in the onco-geriatric population.
| Item Type: | Thesis (UNSPECIFIED) |
|---|---|
| Supervisor name: | Lamoth, prof. dr. C.J.C. and Lahr, dr. M.M.H. and Leeuwen, prof. dr. B.L. van |
| Faculty: | Medical Sciences |
| Date Deposited: | 13 May 2022 09:34 |
| Last Modified: | 13 May 2022 09:34 |
| URI: | https://umcg.studenttheses.ub.rug.nl/id/eprint/3291 |
Actions (login required)
 |
View Item |