Hospital‐free days: A novel measure to study outcomes for emergency department care

Abstract

Traditionally, researchers have evaluated the quality of emergency care using subsequent adverse events, like mortality or return hospital visits. The selection of these well-defined, single events as outcomes for epidemiological studies has strengths, including ease of interpretation. Yet any particular outcome only captures part of the story of how patients fare after receiving care in the emergency department (ED).

For example, mortality is a clinically significant outcome. It fails, however, to capture quality of life decrements that may be important to patients and families.^{1, 2} Furthermore, its rarity necessitates prohibitively large sample sizes. By contrast, composite outcome measures such as major adverse cardiac events require less statistical power but assume that all component parts (e.g., death and hospitalization) are valued equally by patients.

Hospital-free days (HFD) offer a potential solution, naturally combining morbidity and mortality to create an easily calculated, patient-centered measure sensitive to high-quality emergency care. Essentially synonymous with days alive and out of hospital (DAOH), and similar in concept to other metrics including healthy days at home (HDAH), HFD is gaining acceptance in health services, epidemiological, and comparative effectiveness research.^{3, 4} However, these metrics have only recently been used in emergency medicine research.⁵ In this commentary, we outline the HFD approach, discuss unique considerations to using HFD to study emergency medicine outcomes, and propose development of standardized approaches for emergency medicine research.

Calculating HFDs

Unlike event counts such as mortality rates and revisit rates, the HFD approach aims to measure a conceptual construct of health health rather than illness. Like some longstanding measures, such as disability-adjusted life year, HFD begins with an idealized amount of “full health,” a quota from which illness and death are subtracted.

To calculate HFD, a time period after an anchor event (e.g., presentation to the ED or discharge from the ED or hospital) must be selected. For ED encounters, 30 days has often been used to measure mortality or readmissions after ED visits and may also represent an appropriate window to calculate HFD.^{6, 7} For outcomes expected to be highly responsive to ED treatment, a 9-day time window may be most appropriate.⁸ Depending on the specific research question and context, time frames of 60 or 90 days might also be appropriate.

Next, the number of days in which patients are neither alive nor hospitalized are then subtracted from the total days in the chosen time period. Importantly, there is no single definition for which types of services should be counted, highlighting the importance of transparent reporting of how HFD, HDAH, or DAOH are defined for any individual study as well as an opportunity to incorporate patient perspectives into these definitions.

Nearly all definitions subtract days for which patients have expired or spend in an acute care hospital, including long-term acute care. The HFD measure typically does not subtract days spent in post–acute care, such as skilled nursing facilities or inpatient rehabilitation, while HDAH does subtract those days.^{3, 4} Neither measure typically subtracts time spent in residential nursing homes or hospice. Studies vary with regard to classification of subsequent ED visits that do not result in hospital admission, either ignoring these encounters entirely or considering each ED visit as a half or a full day.^{9, 10} Finally, some definitions of HDAH subtract days that patients receive postacute home health visits while others do not.¹¹ Measures including these contacts with the health care system will be more sensitive to morbidity relative to mortality.

Benefits of HFDs as an outcome measure

HFD captures a more nuanced, and potentially more accurate, picture of patients' overall health than alternatives. For example, a study evaluating a clinical pathway for patients discharged from the ED with undifferentiated abdominal pain may report that 20% of patients were admitted over the subsequent 30 days. These readmissions might represent a mix of brief observation stays, patients who experience prolonged hospitalization due to a missed serious diagnosis, and patients who expire during their readmission. Focusing on readmission rates alone does not consider illness severity, duration of hospitalization, or number of readmissions, while focusing on mortality may only consider patients with very severe illness. HFD offers one solution, therefore, that incorporates magnitude into an otherwise binary outcome. Depending on the context, HFD may better account for the range of outcomes which patients may experience and may correlate with patient-centered outcomes such as including functional status.¹²

Another advantage of the HFD approach is the feasibility of data collection. HFD is well suited to pragmatic trials and observational studies that measure outcomes using electronic health record or claims data, since the only data requirements are hospitalization dates (dates of admission and discharge), ED presentation dates, and dates of death. These reduced data requirements offer a pragmatic approach to patient follow-up with considerably lower costs and missing data than survey-based approaches. However, traditional clinical trials or survey-based studies may also easily incorporate HFD as an outcome. To implement this, a clinical study that surveys patients for 30-day outcomes could collect data on not just readmissions but also how many days were spent in different health care settings during that interval or conduct electronic follow-up using regional health exchange and state death record data.

HFD may capture a more patient-centered outcome than other metrics. One study specific to older adults suggests this may be the case.¹³ Ultimately, patients care most about health, which this approach measures as time spent outside of health care facilities. In the example above, patients may be willing to return to the hospital to have their abdominal pain reevaluated if it means that they avoid severe illness. However, only preliminary work has been done to confirm patient perspectives on HFD overall as well as incorporate feedback on methodological choices when using this approach.^{14, 15}

Challenges of HFDs

One challenge is the interpretation of HFD, both for clinicians and for patients. A difference in 5 days alive and out of the hospital over a 30-day period is likely to represent a considerable benefit, but interpreting the clinical significance of a 0.5-day improvement in HFD, for example, requires more context. A recent study reporting the minimum clinically important difference (MCID) for 180 days following ICU discharge offers an approach that could be used to determine an MCID for ED patients at shorter time frames more relevant to post–ED discharge.¹⁵

Another challenge for studies that seek to use HFD is obtaining data outside of individual health systems. Out-of-hospital death can be available through linkage to the National Death Index or through state vital registration systems. To obtain out-of-hospital ED visitation and hospitalization dates, regional health exchanges can provide data for external health systems, including dates of ED presentation and inpatient admission, enhancing the accuracy of HFD. Administrative claims data sets, such as Medicare, Medicaid, or commercial insurance claims, or state all payer claims databases, also contain sufficient information to calculate HFD. Survey-based approaches can augment or replace these routinely captured data if not available.

There are key statistical issues in the use of HFD which remain unresolved. Distribution of the outcome is important to consider depending on the study population and must be considered when selecting statistical tests. Importantly, HFD was initially developed for studying populations with considerable expected health care needs, such as Medicare beneficiaries and survivors of serious illness.^{3, 4} Healthier patients with low-acuity illness may be expected to have data skewed toward full health and maximal HFD, while patients with severe illness may be skewed in the direction of poor health and relatively few HFD. Analyses of populations with expected high rates of HFD accrual, such as healthy pediatric populations, require additional validation before utilizing this metric.

While HFD incorporates quality of life more than existing alternatives, a limitation of HFD is that it cannot fully capture quality of life.³ Two patients may have very different levels of symptoms and function at home but still be considered to have equivalent HFD if those symptoms do not require hospital care.

The application of HFD in studies that include patients who are hospitalized following an ED encounter also presents challenges. HFD accounts for hospital and post–acute care length of stay, which make this measure relevant for time-sensitive interventions delivered in the ED that seek to mitigate the severity and length of illness. However, many additional factors may drive HFD that occur following the emergency phase of care. For admitted patients, a clear causal pathway should be established between ED decisions, diagnostics, and management and long-term health outcomes for HFD to be relevant in this context. For these reasons, the choice whether to initiate the follow-up period at the day of ED visitation or the day of hospital discharge may also depend on the goals of the study (Table 1).

TABLE 1. Challenges and proposed solutions to incorporate HFD into emergency medicine research.

Methodologic issue	Proposed solution
Population baseline prevalence	HFD was developed for patients with serious illness or older patients. Populations with lower baseline prevalence of ED visits, nursing facility days, hospital days, and death than these populations will need additional research and validation. For instance, children, and younger adults experiencing acute trauma may require additional validation before utilizing a HFD metric.
Duration of follow-up period	Given the differential weighting of outcomes based on the duration of the chosen follow-up period, we recommend that metrics be reported by the assessment period in days, such as HFD-9 or HFD-30. Sensitivity analyses should explore alternative, prespecified time periods.
Initiation of follow-up period (“time zero”)	For ED-based studies of discharged patients, the timing should generally begin at the time of ED discharge. For studies that involve ED interventions among patients who are then admitted, the decision is more complex and may depend on whether the intervention seeks to improve outcomes during the subsequent hospitalization versus long-term outcomes.
Incorporation of ED visits into HFD	In some studies, HFD has not included ED treat-and-release visits in the outcome. Recent studies use versions of HFD which include ED visits by subtracting a full or half day for each ED visit. Similar to hospital revisits, ED revisits also contain important information about care quality and disease progression. While ED visits do not typically occupy an entire day, given crowding, waiting room conditions, and hallway beds; the uncertainty of a new medical diagnosis; and the fact that most of a hospitalization's medical care and diagnostics occurs in the ED, they are likely to cause a similar decrement to quality of life as a hospital day. Given this, studies of emergency care that use HFD should generally include subsequent ED visits. Further, patient-centered research is necessary to determine whether the average ED visit in a given population causes sufficient discomfort, expense, travel cost, and time to be equivalent to a full versus fractional hospital day.
Incorporation of observation stays into HFD	Observation stays, whether in the ED or on an inpatient floor under observation status or in a dedicated observation unit, should count in the same way as admission days: each day under observation should decrease the number of HFD.
ED treat-and release visits that cross a single midnight	Because even brief ED visits often cross midnight, ED visits not resulting in observation status or admission should count as a single day in the ED, on the day when the ED visit started. Where time-of-day data are available, use the number of hours in the ED to determine when a visit consists of more than 24 h in the ED.
Variable inclusion of post–acute institutional care and home health services	Reporting and subtracting time spent in post–acute care facilities or utilizing home health services from measures of HFD and/or HDAH should be tailored to the specific study question and population.
Baseline nursing home care	Depending on the context, institutionalized patients prior to the intervention being studied should be deemed “hospital-free” when they return to a baseline care facility of a similar type after their ED visit.
Elective surgeries and elective admissions	Given that many elective surgeries and admissions are responsive to the initial ED visit, we recommend including these as hospital days by subtracting them from HFD.
Binary (any days) versus continuous	Continuous measures are preferred to allow for greater statistical power and to not treat a single ED visit or hospital day as equivalent to death, undermining the strengths of the HFD approach.
Interpreting results of studies using HFD as the primary outcome	Further research is needed to evaluate differences and clinical importance of weighting that HFD implicitly places on different outcomes, such as death. HFD may also require validation for different disease conditions. When summary measures such as HFD are used, the component measures should always be reported as well to enable readers to independently assess the relative importance of each measure in determining the composite outcome.

• Abbreviation: HFD, hospital-free days.

The time-dependent nature of this outcome also must be considered, particularly when patients expire during the follow-up period. For example, a patient who dies on Day 9 of a 9-day assessment period has the same HFD as a patient who revisits the ED a single time during that period. For a 30-day assessment period, however, the patient's death on Day 9 would cause more than 20 times the reduction in HFD compared to a single ED revisit. In general, shorter periods weight nonmortality outcomes higher relative to longer periods. Longer periods may better capture the course of disease and recovery after major illness or procedures. All studies using HFD should also separately report both the rate and the average duration of the component outcomes such as hospitalization, mortality, ED visits, and nursing home days. Future studies should assess the relative tradeoffs of each assessment period to enable researchers to align their choice with the particular context and intervention they are studying.

There are few current examples of using HFD and similar approaches in the emergency medicine literature.^5-8 However, this approach is likely to gain increasing acceptance in the literature with further exploration of its advantages and disadvantages. Recognizing the potential utility as well as pitfalls of using these outcomes in studies of emergency care, we present key considerations for investigators as well as readers of the literature. We list additional key design and reporting decisions, along with proposed solutions, in Table 1.

Finally, technology and structural change in the health care system may accelerate secular trends toward more outpatient and less inpatient care.¹⁶ Ongoing monitoring of these trends, as well as longitudinal comparison to metrics that are less likely to be affected by the economics of the health care industry without an underlying change in patient health, such as functional status and mortality, will be necessary to ensure ongoing construct validity.