Observation Status Stays With Low Resource Use Within Children’s Hospitals

We conducted a retrospective cohort study of observation hospitalizations of children aged 0 to 18 years from January 1, 2019, to December 31, 2019, using the Pediatric Health Information System (PHIS) (Children’s Hospital Association, Lenexa, KS). PHIS is an administrative database containing patient data from 48 tertiary pediatric hospitals. For each encounter, the PHIS database contains demographic characteristics and up to 41 International Classification of Diseases, 10th Revision, Clinical Modification diagnosis and procedure codes. Daily itemized billing, along with charge information, is also available in PHIS. Transfers in and out of facilities and neonatal and surgical All Patient Refined Diagnosis Related Groups (APR-DRGs) were excluded. Additionally, we excluded 6 California hospitals in which observation is not recognized and consequently not reported in PHIS, leaving 42 hospitals in the study cohort.

In PHIS, charges are converted to estimated costs by using 29 department-specific cost-to-charge ratios collected from each hospital each year. Costs were categorized as either room or use. Room costs encompass overhead and staffing costs including nursing and other general facility costs (eg, building maintenance fees, electricity, administration) necessary to care for patients. Use costs include those costs incurred when clinical services are rendered to the patient and include pharmacy, laboratory, radiology, clinical, and supply costs. These provider-modifiable costs vary from stay to stay and reflect the consumption of hospital resources. After costs were determined and categorized, the use ratio (ie, the proportion of total costs that were nonroom) was calculated by dividing the use cost by the total costs for the encounter. Observation encounters were then divided into quartiles on the basis of these use ratios. The lowest quartile, reflecting the smallest ratio of use costs to total costs, was defined as the LRU quartile.

We summarized categorical variables with frequencies and percentages and used medians with interquartile ranges (IQRs) for continuous variables. We compared clinical and demographic characteristics including age, sex, race or ethnicity, payer, route of admission (direct versus ED), hospital-level case mix index using the Hospitalization Resource Intensity Score for Kids method,¹⁶  and presence of complex chronic conditions (CCCs)¹⁷  across quartiles using χ² or Wilcoxon rank tests as appropriate. We evaluated the proportion of LRU observation stays across PHIS hospitals using a χ² test. Next, we modeled associations between demographic and clinical characteristics and LRU using a generalized linear mixed-effects model with a hospital random effect to account for clustering of patients within hospitals. Lastly, a classification and regression tree (CART) analysis was performed to identify combinations of characteristics associated with LRU stays. To evaluate diagnoses associated with LRU stay, we aggregated APR-DRGs by body system using PHIS service lines (ie, respiratory, gastrointestinal, etc). CART models are ideal for identifying distinct combinations of factors that have a different likelihood of being an LRU observation stay. We used cost-complexity pruning¹⁸  to prevent the tree from overfitting the data. All statistical analyses were performed by using SAS version 9.4 (SAS Institute, Inc, Cary, NC), and P values <.05 were considered statistically significant. Because we used deidentified data, the Children’s Mercy Hospital Institutional Review Board deemed the study exempt.

We identified 174 315 observation status encounters meeting inclusion criteria at PHIS participating hospitals between January 1, 2019, and December 31, 2019 (Table 1). We found that children classified under observation status tended to be young (52.0% <5 years old), male (53.8%), and non-Hispanic white (49.7%). The majority (55.2%) of stays were billed to a public payer and were admitted via the ED (74.1%).

After calculating use ratios, we divided stays into quartiles. The lowest quartile of use ratio (ie, the LRU quartile) had use costs <31.8% of total costs and was represented by 44 422 hospitalizations. When compared with all observation status stays, we found that those in the LRU quartile had a greater proportion of infants <1 year of age (28.7%), had fewer CCCs (82.6% with no CCCs), were longer in duration (30 hours [IQR 20–45]), and were more likely to be admitted via direct admission (40.7%) than other quartiles (all P < .001).

Although LRU observation status hospitalizations occurred at all included PHIS hospitals, there was marked variation in their frequency. LRU observation hospitalizations varied from 1.3% to 57.6% with a median of 21.2% (IQR 13.0%–37.9%) (Fig 1).

Using multivariable regression analysis, we observed that younger children, especially those <1 year of age (odds ratio [OR] 3.3; 95% confidence interval [CI] 3.1–3.4), those without a CCC (OR 3.6; 95% CI 3.2–4.0), and those directly admitted to the hospital (OR 4.2; 95% CI 4.1–4.4) were more likely to be associated with an LRU observation stay (Table 2, Supplemental Table 3). Those with previous hospitalization within 7 days (OR 1.4; 95% CI 1.3–1.5) and those with public insurance (OR 1.1; 95% CI 1.1–1.2) were also associated with small increased odds of LRU stay.

We used CART analysis to describe the combination of characteristics most associated with experiencing an LRU observation stay (Fig 2). We observed that 40.0% of children who were directly admitted, 50.1% of children who were directly admitted with no CCCs, and 69.5% of children who were directly admitted without medical complexity and a diagnosis of a respiratory illnesses experienced LRU observation stays. In contrast, LRU observation stays occurred for only 20.4% of children admitted through the ED.

With this study, we provide new information about the clinical and demographic characterization of observation status stays with LRU among pediatric patients. Young patients, those who were previously healthy, and those directly admitted to the hospital were more likely to experience an LRU observation stay. We describe that the combination of characteristics most likely to place a patient at risk for an LRU stay includes direct admission, lack of a CCC, and respiratory illness. Additionally, we found notable variation in the frequency of LRU observation stays across children’s hospitals. The description of clinical and demographic features associated with LRU observation stays identifies priority populations for further study with the eventual goal of developing targeted strategies to decrease unnecessary hospitalizations, thereby increasing the delivery of cost-effective and efficient care.

We identified that the hospitalizations with the highest odds of LRU observation stay were those after direct admission (OR 4.2; 95% CI 4.1–4.4). Direct admissions are intended to be a means of initiation of hospitalization without a patient having received care in that facility’s ED. These admissions, which represent ∼25% of all unscheduled hospitalizations in the United States, were initially intended to offload the high patient volume seen in tertiary hospitals’ EDs.¹⁹  Policies and procedures surrounding direct admissions are variable among hospitals, and in 2016, a national survey of 108 hospitals revealed that only one-third of respondents had formal criteria in place to assess the appropriateness of direct admissions.²⁰  Our findings linking direct admission with LRU hospitalization suggest that analysis of direct admission policies may be necessary to further evaluate the value of these stays.

For hospitals, the costs associated with caring for children are substantial and growing. Previous work indicates that the cost for hospitals to care for observation stays is similar to the cost to care for inpatient hospitalizations of similar duration.¹³  Despite similar costs, observation status stays are billed at an hourly rate as outpatient encounters¹⁰  and reimbursements are likely reduced, resulting in net financial losses for hospitals. This financial liability may become increasingly important with the rising use of observation stays in recent years.^14,15,21  Hospital systems focused on their financial health may benefit from reducing unnecessary observation encounters, and the LRU quartile may include patients who could have been successfully treated outside of the hospital.

Although reduction in the number of observation stays would undoubtably decrease health care costs related to hospitalization, strategies aimed at length of stay reduction, such as dedicated observation units, may help improve throughput and efficiency while decreasing expenditures. Dedicated observation units are uncommon in freestanding children’s hospitals, and when they are present, they are heterogeneous in their structure and function.^22,23  In this study, we identified specific patient typologies that were most likely to result in an LRU hospitalization. If patients with characteristics most associated with LRU observation stays were proactively identified and preferentially assigned to an observation unit, strategies to streamline care delivery and prioritize early discharges may result. Thus, observation units may promote high-value care while simultaneously decreasing costs.

In addition to risking the financial health of children’s hospitals, admitting children to the hospital when care could be safely delivered outside of the hospital has a profound effect on patients and their families. The financial ramifications for families of children who are hospitalized come in the form of increased transportation and food costs as well as lost wages resulting from missed work.^24–26  Furthermore, hospitalizations remove children from their social network and expose them to repeated stressful situations.²⁷  Classifying observation stays on the basis of use ratio may help to identify groups of stays that do not provide sufficient value to the patient and the family to justify hospitalization.

The decision to admit a patient to the hospital reflects a real-time cost/benefit analysis. Although challenging for patients and their families, hospitalization allows for more timely access to diagnostic and treatment modalities that would be either unavailable or notably delayed if administered in the outpatient setting. The totality of services rendered, which we quantified and stratified on the basis of use ratios, may provide insight into some of the value that a hospitalization provides to the patient. Importantly, an additional benefit of hospitalization is the close and careful monitoring of the patient by health care professionals, which was not evaluated in our study. Ultimately, those stays that provided little tangible use benefits to the patient should be evaluated further to determine if the benefit provided by close monitoring and observation justifies the hardships created by hospitalization or if continued follow-up in the clinic setting is reasonable.

Our study should be interpreted in the context of several limitations. First, as stated above, our use of the use ratio to describe a population of children who received minimal nonroom services during their hospitalization is a use metric only and does not address whether a patient clinically warranted hospitalization (under observation or inpatient status). Recently, the move to “safely do less” has encouraged clinicians to decrease unnecessary testing and limit interventions to those necessary to achieve the desired patient outcomes.^28–32  For example, foregoing invasive and costly testing in favor of thorough physical examination and close observation in the diagnosis and treatment of bronchiolitis may represent high-value care while also falling into the LRU quartile. Ultimately, decreased use may be a marker of appropriately following these guidelines, and hospitalizations falling into the LRU category should not be viewed as appropriate or inappropriate solely on the basis of their use ratio.

A second important limitation is that the PHIS database, although robust, does not represent community hospitals. Consequently, the composition of clinical and demographic characteristics as well as the strategies to reduce unnecessary hospitalizations may differ for these institutions compared with tertiary care children’s hospitals. For example, when evaluating the appropriateness of admission, direct admission policies, and use ratios, the type and size of the hospital may be important considerations.

Finally, in our study, we used data collected before the coronavirus disease 2019 pandemic, which has impacted patient care and administrative processes throughout the hospital.³³  Increased use of personal protective equipment, changes in staffing, and implementation of cohorting policies have likely increased the total cost of hospitalization without directly impacting use costs. Future studies are needed to describe the impact that coronavirus disease 2019 has had on hospitalization costs.

Use ratios allow for identification of hospitalizations with minimal nonroom resource use. LRU stays may contain a cohort of patients who could have safely received treatment outside of the hospital setting. LRU observation stays are variable across children’s hospitals and occur most frequently in generally healthy young children and those directly admitted. Future evaluation of these populations may reveal ways to reduce avoidable hospitalizations.

APR-DRG

All Patients Refined Diagnosis Related Groups

CART

Classification and Regression Tree Analysis

CCC

complex chronic condition

CI

confidence interval

ED

emergency department

IQR

interquartile range

LRU

low resource use

OR

odds ratio

PHIS

Pediatric Health Information System