Nonspecific chest pain and hospital revisits within 7 days of care: variation across emergency department, observation and inpatient visits

Patients increasingly receive care in outpatient observation services rather than being admitted to the hospital, and chest pain is the leading diagnosis among observed patients [ 1 , 2 , 3 , 4 ]. Chest pain can indicate a serious medical condition (e.g., acute coronary syndromes, including acute myocardial infarction [AMI] and unstable angina) or a relatively minor one (e.g., gastroesophageal reflux). Often the difference between a serious and minor condition is not readily apparent [ 5 , 6 ]. A relatively small percentage of patients who present to the hospital with chest pain have acute coronary syndromes or other emergent conditions requiring inpatient care [ 7 ]. However, determining the diagnosis and appropriate course of treatment for chest pain can involve clinical evaluation, testing, and monitoring. Clinicians historically had two options: hold patients with chest pain in the emergency department (ED), or admit them to the hospital for a short stay. Observation status offers a third option, allowing more time to assess the patient’s condition. The concept of hospital observation emerged in the 1970s as an alternative to ED and inpatient care [ 8 , 9 ]. For chest pain, randomized controlled trials in the 1990s and 2000s demonstrated that dedicated cardiac observation units and specific care protocols improved outcomes, reduced hospital admissions, and lowered costs [ 10 , 11 , 12 , 13 ]. The practice of hospital observation status has evolved over time, and expanded in diverse ways across different states and payers. Many public and private payers now reimburse observation services as a hospital-based, outpatient service irrespective of diagnosis, use of a specific protocol, or physical location in the hospital [ 14 , 15 ]. The Centers for Medicare & Medicaid Services defines observation services as “a well-defined set of specific, clinically appropriate services, which include ongoing short term treatment, assessment, and reassessment before a decision can be made regarding whether patients will require further treatment as hospital inpatients or if they are able to be discharged from the hospital.” [ 16 ] Despite rapid growth of observation services for chest pain [ 17 , 18 ], it is unclear whether observation services in current practice is associated with different outcomes compared with ED or inpatient care [ 19 ]. Unlike the historical randomized controlled trials previously mentioned, many hospitals do not have designated observation units or use chest pain protocols [ 18 ]. A potentially useful outcome measure following an ED, observation, or inpatient hospital visit is the percentage of patients who return to the hospital [ 20 , 21 ]. Compared with longer follow-up periods, a 7-day follow-up period emphasizes revisits that are more likely to involve acute issues related to the index visit [ 22 , 23 ]. Moreover, the first week after a hospital visit for chest pain represents a high-risk period during which atherosclerotic heart disease and AMI are leading causes of death [ 23 , 24 ]; thus, return visits for acute coronary syndromes are particularly concerning. Our primary objective was to compare the rate at which patients with nonspecific chest pain return to the hospital within 7 days after index observation visits versus after index ED and inpatient visits. Although acuity likely differs across these three types of visits, no benchmarks have been established for assessing outcomes of observation care, making ED and inpatient care useful points of comparison [ 25 ]. Secondary objectives included examining [ 1 ] the type of revisit (ED, observation, or inpatient) to determine how often patients receive a higher level of care upon return and [ 2 ] first-listed diagnoses at the revisit to assess whether patients may be returning for acute coronary syndromes, nonspecific and other causes of chest pain, or other reasons. We used 2013 and 2014 data from the Healthcare Cost and Utilization Project (HCUP) [ 26 ] State Inpatient Databases (SID), State Emergency Department Databases (SEDD), and State Ambulatory Surgery and Services Databases (SASD) [ 27 ]. We selected 10 states (Georgia, Iowa, Maryland, Nebraska, Nevada, South Carolina, South Dakota, Tennessee, Vermont, and Wisconsin) because they had complete data on observation services and encrypted patient linkage numbers that linked index visits and revisits within the same state. These states represented 13% of the U.S. population in 2014 (4% West, 23% South, 17% Midwest, and 1% Northeast). We included patients aged 18 years or older with an index visit for a first-listed (or principal) diagnosis of nonspecific chest pain. The unit of analysis was an index visit—ED visit, short-stay observation visit, or short-stay inpatient visit (defined below). Patients with multiple encounters contributed multiple index visits if the index visit met the following exclusion and inclusion criteria. We excluded index visits if the patient transferred into or out of the hospital, died in the hospital, or left against medical advice. We required that index visits have a clean period, defined as no hospitalization, observation visit, or ED visit for any reason in the past 30 days [ 28 ]. To identify events that occurred before and after the index visit, we only included index visits that occurred between February 2013 and November 2014. The primary independent variable was the type of index visit for nonspecific chest pain: 1. Index ED visit . The patient was discharged from ED care without being placed under observation or admitted. Index ED visit . The patient was discharged from ED care without being placed under observation or admitted. 2. Index observation visit . The patient was discharged from observation care within 2 nights after presentation to the hospital without being admitted. Index observation visit . The patient was discharged from observation care within 2 nights after presentation to the hospital without being admitted. 3. Index inpatient visit. The patient was discharged from short-stay inpatient care within 2 nights after presentation to the hospital. Index inpatient visit. The patient was discharged from short-stay inpatient care within 2 nights after presentation to the hospital. Determination of type of index visit was based on status at time of discharge (e.g., patients seen in the ED but then moved to observation were considered having an observation visit. We included only short-stay observation and inpatient visits (discharge within 2 nights [ 15 , 29 ] after presentation to the hospital) to increase the comparability of the three types of visits. The primary dependent variable was the presence or absence of any revisit to the hospital—meaning ED, observation, or inpatient visits—within 7 days of the discharge date from the index visit. Secondary dependent variables included type of care (ED, observation, or inpatient) and diagnosis at the revisit. For the latter, we created a categorical variable based on the following hierarchy: (1) any diagnosis of acute coronary syndromes (AMI or unstable angina), (2) any diagnosis of nonspecific chest pain and other conditions potentially associated with chest pain (e.g., other cardiac conditions, lower respiratory disease, esophageal disorders), and (3) unrelated conditions (defined in Additional file 1 Table S1) [ 5 ]. Covariates included patient sex, age, expected primary payer (Medicare, Medicaid, private, self-pay / no charge), urban/rural location, median household income quartile (derived using patient ZIP Code), as well as coexisting conditions at the index visit: prior AMI, any other ischemic heart disease, other comorbid conditions identified using Elixhauser Comorbidity software [ 30 ], designed to predict 30-day readmissions. We included 14 out of 29 comorbid conditions, which represented at least 1% of all index records in all three settings. We also included a variable representing the number of diagnoses reported on the record, because coexisting conditions may be documented more frequently for inpatient and observation than for ED visits. We estimated separate logistic regression models for the primary and secondary dependent variables including the covariates described above and hospital-level fixed effects, because care may be influenced by unknown factors that are constant across visits to a particular hospital. We used results from the models to derive risk-adjusted 7-day revisit rates for different types of index visits, applying a widely used approach for risk adjustment (described in the Additional file) [ 31 ]. We repeated these analyses stratifying by expected primary payer, because insurance is likely to influence access to ambulatory care after a hospital visit. We also repeated these analyses using the secondary outcome of diagnosis at the revisit. Because large sample sizes increase the probability of a type I error, we considered results statistically significant if the p -value was < 0.01 and the revisit rates differed by 10% or more. We included 1,005,866 index visits to 746 hospitals for nonspecific chest pain. Most index visits occurred in the ED (638,302) with about half as many index observation visits (325,770) and fewer index inpatient visits (41,794). Compared with patients with index observation visits, patients with index inpatient visits were more likely to be older, have Medicare insurance, and have more documented coexisting conditions, whereas patients with index ED visits were more likely to be younger, have private insurance, and have fewer coexisting conditions (Table 1 ). There were 86,692 revisits (62,629 following index ED visits, 20,472 following index observation visits, and 3591 following index inpatient visits). Total unadjusted 7-day revisit rates after index visits for ED care, observation care, and inpatient care were 9.8, 6.3, and 8.6%, respectively (Fig. 1 ). Regardless of the type of index visit, over half of 7-day revisits had been discharged from the ED without observation or inpatient care. The rates at which patients returned for inpatient care were 1.5% among patients with an index visit for ED care, 1.2% among patients with index observation visits, and 2.7% among patients with index inpatient visits. Adjusted revisit rates following an index visit for nonspecific chest pain involving ED care, observation care, and inpatient care were 9.7% (95% CI: 9.7–9.8), 6.5% (95% CI: 6.4–6.6), and 7.7% (95% CI: 7.5–8.0), respectively (Fig. 2 ). The risk-adjusted 7-day revisit rate was 3.2 (95% CI: 3.1–3.3) and 1.2 (95% CI, 1.0–1.5) percentage points lower, respectively, following an index observation visit than following an index ED ( p < 0.001) or inpatient ( p < 0.001) visit. Unadjusted 7-Day Revisit Rate, by Type of Index Visit, 2013 and 2014. Abbreviation: ED, emergency department. Source: Agency for Healthcare Research and Quality, Healthcare Cost and Utilization Project, State Inpatient Databases, State Emergency Department Databases, and State Ambulatory Surgery and Services Databases, 10 states (Georgia, Iowa, Maryland, Nebraska, Nevada, South Carolina, South Dakota, Tennessee, Vermont, Wisconsin), 2013 and 2014 Results were similar across payers (Fig. 2 ). Relative to the risk-adjusted 7-day revisit rate following an index ED visit, the rate following an index observation visit was 5.1 (95% CI: 4.9–5.3) percentage points lower among patients with Medicare ( p < 0.001), 3.7 (95% CI: 3.3–3.9) percentage points lower among patients with Medicaid ( p < 0.001), 1.9 (95% CI: 1.7–2.1) percentage points lower among patients with private insurance ( p < 0.001), and 2.6 (95% CI: 2.2–2.8) percentage points lower among patients with no insurance ( p < 0.001). Relative to the risk-adjusted 7-day revisit rate following an index inpatient visit, the rate following an index observation visit was 1.1 (95% CI: 0.7–1.5) percentage points lower among patients with Medicare ( p < 0.001), 0.8 (95% CI: 0.3–1.4) percentage points lower among patients with private insurance ( p < 0.01), and 1.7 (95% CI: 0.9–2.5) percentage points lower among patients with no insurance ( p < 0.001). Among patients with Medicaid, revisit rates were similar after index visits for observation and inpatient care. Risk-Adjusted 7-Day Revisit Rate, by Payer Population and Type of Index Visit, 2013 and 2014. Abbreviation: ED, emergency department. Statistically significant at p < 0.05 and ≥ 10% difference between revisit rates for index observation visits and index ED or inpatient visits. Source: Agency for Healthcare Research and Quality, Healthcare Cost and Utilization Project, State Inpatient Databases, State Emergency Department Databases, and State Ambulatory Surgery and Services Databases, 10 states (Georgia, Iowa, Maryland, Nebraska, Nevada, South Carolina, South Dakota, Tennessee, Vermont, Wisconsin), 2013 and 2014 Revisits for acute coronary syndromes were rare (0.3 to 0.5% of index visits), while revisits for nonspecific chest pain and conditions potentially associated with chest pain were common (4.7 to 7.2% of index visits), irrespective of the type of index visit (Fig. 3 ). The risk-adjusted rate at which patients returned within 7 days with acute coronary syndromes was 0.2 (95% CI: 0.1–0.2) and 0.1 (95% CI: 0.0–0.2) percentage points lower, respectively, after an index observation visit than after an index ED visit ( p < 0.001) and after an index inpatient visit ( p < 0.001). Additionally, the risk-adjusted rate at which patients returned within 7 days with nonspecific chest pain or conditions potentially associated with chest pain were 2.5 (95% CI: 2.4–2.6) and 0.9 (95% CI: 0.6–1.1) percentage points lower, respectively, after an index observation visit than after an index ED ( p < 0.001) or inpatient ( p < 0.001) visit. Diagnosis at Revisit: Risk-Adjusted 7-Day Revisit Rates by Type of Index Visit, 2013 and 2014. Abbreviation: ED, emergency department. Statistically significant at p < 0.05 and ≥ 10% difference between revisit rates for index observation visits and index ED or inpatient visits. Source: Agency for Healthcare Research and Quality, Healthcare Cost and Utilization Project, State Inpatient Databases, State Emergency Department Databases, and State Ambulatory Surgery and Services Databases, 10 states (Georgia, Iowa, Maryland, Nebraska, Nevada, South Carolina, South Dakota, Tennessee, Vermont, Wisconsin), 2013 and 2014 Patients most frequently returned to the hospital with nonspecific chest pain or conditions potentially associated with chest pain. The risk-adjusted 7-day revisit rate following an index ED visit was 9.7% (95% CI: 9.7–9.8), including a 7.2% (95% CI: 7.2–7.3) revisit rate for nonspecific chest pain or conditions potentially associated with chest pain (i.e., 74.3% of total revisits). The results from Fig. 3 were largely consistent across payers (not shown).