Physical activity trajectories of hospitalized patients: a latent class analysis

Abstract

ABSTRACT Background: Sedentary behavior (SB) and a lack of physical activity (PA) increase the risk of functional decline and medical complications in hospitalized patients. Longitudinal methods are necessary to study activity trends during hospitalization and to examine whether there are patients with specific activity patterns who would be at risk for low PA and functional decline. Aim: To examine patients’ daily time spent on SB and PA throughout hospitalization and to identify activity subgroups and patient-related factors associated with the distinct activity trajectories. Methods: In this observational mono-center study data of 512 adults hospitalized in 14 hospital wards were longitudinally analyzed. Patients’ SB and PA measured with an accelerometer were utilized for statistical subgrouping. Subgroups were identified using latent class mixed modeling, characteristics were compared through variance, proportion, mean and median testing. Factors associated with subgroup placement were identified using multinomial logistic regression. Results: Three subgroups were identified: a low active group (n=77) with a mean daily PA of 33 minutes, a moderate active group (n=260) with a mean PA of 80 minutes, and an active group (n=175) with a mean PA of 174 minutes. Factors associated with placement into the low active group were: higher BMI [odd ratio OR 1.054 (95% CI: 1.002-1.108)] [OR 1.097 (95% CI: 1.037-1.161)], lower handgrip strength score [OR 0.968 (95% CI: 0.942-0.994)] [OR 0.957 (95% CI: 0.930-0.984)] and longer hospital length of stay (HLOS) [OR 1.050 (95% CI: 1.016-1.085)] [OR 1.065 (95% CI: 1.022-1.110)] when compared to the moderate and active group. Higher ADL-dependency [OR 0.370 (95% CI: 0.140-0.980)] was associated with placement into the low active group when compared to the active group. Conclusion and key findings: Hospitalized patients can have different activity trajectories throughout admission in which three distinct subgroups could be identified. BMI, handgrip strength, HLOS, and ADL-dependency were all factors mildly associated with subgroup placement. Patients’ activity levels might cohere with latent variables and constructs. Impact statement Predicting patients’ activity trajectories is intricate. Physical therapists should monitor patients' activity and target interventions towards subgroups with low PA rather than generically target interventions towards predefined groups based on predefined clinical factors. Word count: 349 Keywords: Sedentary behavior, physical activity, hospitalization, latent class analysis.