Methods

We performed an analysis of the ED component of the 2001 to 2010 NHAMCS. The NHAMCS data set is a nationally representative sample of US ED visits obtained by the National Center for Health Statistics (NCHS) branch of the Centers for Disease Control and Prevention. NHAMCS uses a 4-stage sampling strategy, covering geographic primary sampling units, hospitals within primary sampling units, EDs within hospitals, and patient visits within EDs. The ED visit is the basic sampling unit and represents a larger number of samples based on the inflation factor called the ED patient weight. This weighting is based on 4 factors: the reciprocal of the probability of selection, nonresponses adjustment, population ratio adjustment, and weight smoothing.

All visit sampling and data collection were performed by hospital staff, and review of data collection was performed by a US Census Bureau field supervisor. The data abstraction forms include information pertaining to the sampled visit including demographic information, 3 patient “reason-for-visit” fields, triage acuity, initial vital signs, ED tests and procedures performed, 3 International Classification of Diseases, ninth revision (ICD-9) ED discharge diagnoses, and, starting in 2005, 1 hospital discharge diagnosis. Further data collection methods and sampling design are described in detail on the NCHS Web site ( http://www.cdc.gov/nchs ). This study was exempted from review by our institutional review board.

From the 2001 to 2010 NHAMCS database, we selected all ED visits that had a primary reason for visit (RFV) of “1260.0 Abnormal pulsations and palpitations; includes rapid heartbeat, slow heartbeat, irregular heartbeat, fluttering, jumping, racing, skipped beat” coded using Reason for Visit Classification for Ambulatory Care, a standardized sourcebook used in NCHS studies. ED visits with this RFV as secondary or tertiary complaints were not included.

We collected the demographic characteristics of the patients including age, gender, race, ethnicity, insurance status, metropolitan statistical area, and geographic region. We recorded clinical data such as vital signs, triage acuity, and mode of arrival, diagnostic testing data (i.e., laboratory tests, electrocardiograms, cardiac monitoring, and x-ray imaging), and ED therapy and procedures. Additionally, ED consultations, dispositions (i.e., admit to hospital, admit to observation unit, transferred to outside hospital, and discharged), and short-term mortality in ED or in hospital were examined. We also recorded the 3 ED discharge diagnoses provided for every ED visit and the single hospital discharge diagnosis for admitted patients.

The NHAMCS data form varies in content from year to year. For example, cardiac enzyme ordering and hospital discharge diagnosis were recorded starting in 2005, respiratory rate and pulse oximetry starting in 2007. We included in our analyses only data that were available without using imputation other than what was already done by the NCHS. For simplicity, all ED visits were categorized into 2 classes: high acuity, comprising patients needing to be seen in ≤1 hour, and low acuity, comprising patients needing to be seen in 1 to 24 hours, as has been done in previous NHAMCS analyses. We created new variables to examine the frequency of abnormal initial vital signs. We defined tachycardia, bradycardia, fever, hypoxia, tachypnea, hypotension, and hypertension using standard age-adjusted clinical cutoffs used in previous NHAMCS vital sign analyses (see Appendix 1 ).

We also recorded whether the ED visits contained a cardiac ICD-9 discharge diagnoses including dysrhythmias (e.g., cardiac dysrhythmias, atrial flutter or fibrillation, and ventricular fibrillation or flutter), structural heart disease (e.g., congestive heart failure, aortic valve disorder, and endocarditis), ischemic heart disease (e.g., angina pectoris and acute myocardial infarction), and other cardiac diagnoses (e.g., complication of heart transplant, cardiac murmurs, and premature beats). Such classifications have been used in previous NHAMCS analyses. We created a new variable, “cardiac diagnosis,” defined as positive for any visit in which at least 1 of the 3 possible ICD-9 ED discharge diagnoses included a cardiac diagnosis. A complete list of recorded cardiac diagnoses is presented in Appendix 2 .

We separated all primary diagnoses into 4 categories: cardiac, psychiatric, medication or substance related, and other diagnoses or symptoms. Each diagnosis was classified independently by 2 investigators (MAP and HKK), blinded to the disposition and outcome, with a third (JRH) serving as arbitrator in cases of disagreement. Interrater agreement was assessed using a kappa statistic. We defined hospital admission as a disposition of “admit to hospital” or “transferred to outside hospital.” Admissions to the observation area and discharges were considered to be nonadmissions.

We performed all statistical analyses with Stata (version 12.1; StataCorp LP, College Station, Texas) using a standard method for analyzing survey-weighted data by way of the survey command. The survey program from Stata takes into account the multilevel sample design when producing national estimates. We determined point estimates and 95% confidence intervals (CIs) for demographic and clinical characteristics of all ED visits with a primary RFV of palpitations. We additionally tabulated frequencies summarizing resource utilization with regard to ED testing, treatment, and hospital admission. We explored regional variation in clinical management with regard to testing and admission rates. Nationally representative estimates were determined using NCHS-assigned patient weights. Estimates based on <30 sample records were excluded as they are considered to be unreliable because of high relative standard errors.

Finally, using hospital admission as our binary outcome, we selected 29 candidate binary patient variables (see Appendix 3 ) based on construct validity and used survey-weighted chi-square recursive partitioning to identify factors associated with admission. Compared with logistic regression, this nonparametric technique is resistant to outliers, does not suffer from missing data, and does not rely on the independence of the explanatory variables. It involves successive univariate chi-square analyses for each of the candidate variables. The variable with the greatest discriminating power (i.e., highest chi-square value) is identified as the first criterion. Visits showing this variable are removed from further analysis, leaving a contracted database. Chi-square analyses were then performed on the contracted database to identify a second criterion from among the remaining variables, and so on. Within the recursive partitioning analysis, age was studied as a binary variable at 10-year cut-off intervals starting at age 20, and we excluded variables with a nonresponse rate of >30% threshold or with <30 sample records as per NHAMCS instructions.

Results

The complete data set contained a total of 357,681 ED visits from 2001 to 2010, representing an estimated 118 million visits. From this sample, we found 1,998 visits with a primary RFV of palpitations, representing an estimated 684,177 visits nationally. The nationally estimated prevalence of palpitations as a chief complaint in the ED was 5.8 per 1,000 patient visits (95% CI 5.2 to 6.4). Further demographic characteristics of patients who visit the ED with palpitations are provided in Table 1 . Most patients were considered high acuity on triage (87.8%, 95% CI 85.6 to 90.0). Further clinical and testing information is presented in Table 2 .

Stay updated, free articles. Join our Telegram channel

Join