Impact of Observation Status on Hospital Outcomes

Glasgow J, Smith AR, Kaboli PJ. Impact of Observation Status on Hospital Outcomes. Poster session presented at: VA HSR&D / QUERI National Meeting; 2012 Jul 16; National Harbor, MD.

Related HSR&D Project(s)

• IIR 08-067 – Organizational Factors and Inpatient Medical Care

Objectives: Despite current VA policy, practice variation exists across VAMCs as to which patients to monitor under "observation" status. Observation allows for inpatient monitoring up to 24 hours for patients not meeting admission criteria. Administratively this is an outpatient designation, resulting in lower patient co-pays. Additionally, observation patients are not included in risk adjustment and hospital performance measures, such as Observed Minus Expected Length of Stay (OMELOS). Variation in observation status use may lead to quality differences and introduce bias when making comparisons across hospitals. Our objective is to determine variation in observation use and evaluate how this correlates with hospital and patient outcome measures. Methods: Review of VA administrative data files from fiscal year 2010 to determine the percentage of medicine admissions first admitted to the ICU, as an acute admission, or to observation status (an outpatient designation). Univariate regression analyses examined relationships between observation status and mean OMELOS. Results: Across 128 VA hospitals, there was significant variation in observation status with a mean percentage of medicine patients first admitted as observation of 11.2% (SD = 11.9) and range of 0 - 50.5%. This includes 25 hospitals admitting less than 1% of patients to observation status. Regression analyses indicated that for every additional 1% of the patient population placed under observation status, OMELOS decreased by 0.023 days (95% CI: -0.033 - -0.013, p < 0.001). A similar regression using the log of observation given the rightward skew provided similar results (-0.100 days; 95% CI: -0.181 - -0.020, p = 0.02). A similar association did not exist for the percentage of patients first admitted to the ICU (p = 0.36) or log(ICU) (p = 0.39). Implications: There was clear variation in the percentage of patients first placed under observation status. This variation was significantly associated with hospital OMELOS resulting in a bias towards higher OMELOS for those VAMCs who do not use observation status. Impacts: Differences in observation use across hospitals represents a meaningful source of variation in patient care. VA should develop more actionable guidelines concerning observation to improve consistency. Until then, careful consideration needs to be paid by researchers and administrators on how observation patients are treated in patient outcome analyses as their routine exclusion may bias between hospital comparisons.