H.-M. Chen, Y.-L. Chen, C. C.-H. Yang
Background Length of stay (LOS) in acute psychiatric inpatient settings is clinically and operationally important, but practical early markers for identifying patients at risk for prolonged hospitalization remain limited. Objective To evaluate whether sleep instability measured during the early inpatient observation period was associated with prolonged post-enrollment hospitalization in acute psychiatric inpatients. Methods This observational study included 94 consecutive patients admitted to an acute psychiatric inpatient unit. Monitoring began after study enrollment, and participants contributed 5 to 7 available nightly observations from wrist actigraphy and routine nursing observation; observations were not required to occur on consecutive calendar days. The primary exposure was intraindividual sleep variability (IISV), defined as the within-subject standard deviation of nightly total sleep time across available post-enrollment observations. The primary outcome was prolonged post-enrollment length of stay, defined using the 75th percentile of the cohort distribution. Age- and sex-adjusted logistic regression models evaluated associations, discrimination, calibration, and risk stratification. Total length of stay from admission was examined as a temporally distinct sensitivity estimand. Results Participants were enrolled a median of 4 days after admission (IQR 2-7) and contributed 5 to 7 available nightly observations, which accumulated over a median of 8 calendar days (IQR 7-10); 39 of 94 observation series (41.5%) fell on consecutive calendar days. Prolonged post-enrollment hospitalization was defined as more than 24.0 days, yielding 23 events among 94 participants. Greater actigraphy-derived IISV was associated with prolonged post-enrollment hospitalization (adjusted odds ratio [OR], 2.30 per 1-hour increase; 95% CI, 1.08-4.93; P=.03), with an area under the receiver operating characteristic curve (AUC) of 0.756 (95% CI, 0.626-0.852; optimism-corrected, 0.725) compared with 0.692 for a model containing age and sex alone. Nursing-derived IISV showed an association of similar direction (adjusted OR, 3.10; 95% CI, 1.15-8.39; P=.03; AUC, 0.723; 95% CI, 0.571-0.837; optimism-corrected, 0.690). A combined model containing both measures did not improve discrimination over actigraphy alone (AUC, 0.751). Conclusions Sleep instability measured during the early inpatient observation period, derived independently from wearable actigraphy and from routine nursing sleep documentation, was associated with prolonged post-enrollment hospitalization. Because the outcome interval overlaps the period during which the sleep observations were accumulated, these findings describe an association across the early inpatient course rather than a prediction made at enrollment. The results support further evaluation of repeated sleep variability as a longitudinal inpatient measure; the variable observation schedule, single-center sample, and absence of external validation mean that clinical implementation remains premature.