Effects of Presentation and Electrocardiogram ... - Wiley Online Library

CLINICAL PRACTICE

Effects of Presentation and Electrocardiogram on Time to Treatment of Hyperkalemia Kalev Freeman, MD, PhD, James A. Feldman, MD, MPH, Patricia Mitchell, RN, Jacqueline Donovan, BS, K. Sophia Dyer, MD, Laura Eliseo, MD, MPH, Laura Forsberg White, PhD, Elizabeth S. Temin, MD, MPH

Abstract Objectives: To assess the time to treatment for emergency department (ED) patients with critical hyperkalemia and to determine whether the timing of treatment was associated with clinical characteristics or electrocardiographic abnormalities. Methods: The authors performed a retrospective chart review of ED patients with the laboratory diagnosis of hyperkalemia (potassium level > 6.0 mmol ⁄ L). Patients presenting in cardiac arrest or who were referred for hyperkalemia or dialysis were excluded. Patient charts were reviewed to find whether patients received specific treatment for hyperkalemia and, if so, what clinical attributes were associated with the time to initiation of treatment. Results: Of 175 ED visits that occurred over a 1-year time period, 168 (96%) received specific treatment for hyperkalemia. The median time from triage to initiation of treatment was 117 minutes (interquartile range [IQR] = 59 to 196 minutes). The 7 cases in which hyperkalemia was not treated include 4 cases in which the patient was discharged home, with a missed diagnosis of hyperkalemia. Despite initiation of specific therapy for hyperkalemia in 168 cases, 2 patients died of cardiac arrhythmias. Among the patients who received treatment, 15% had a documented systolic blood pressure (sBP) < 90 mmHg, and 30% of treated patients were admitted to intensive care units. The median potassium value was 6.5 mmol ⁄ L (IQR = 6.3 to 7.1 mmol ⁄ L). The predominant complaints were dyspnea (20%) and weakness (19%). Thirty-six percent of patients were taking angiotensin-converting enzyme (ACE) inhibitors. Initial electrocardiograms (ECGs) were abnormal in 83% of patient visits, including 24% of ECGs with nonspecific ST abnormalities. Findings of peaked T-wave morphology (34%), first-degree atrioventricular block (17%), and interventricular conduction delay (12%) did not lead to early treatment. Vital sign abnormalities, including hypotension (sBP < 90 mmHg), were not associated with early treatment. The chief complaint of ‘‘unresponsive’’ was most likely to lead to early treatment; treatment delays occurred in patients not transported by ambulance, those with a chief complaint of syncope and those with a history of hypertension. Conclusions: Recognition of patients with severe hyperkalemia is challenging, and the initiation of appropriate therapy for this disorder is frequently delayed. ACADEMIC EMERGENCY MEDICINE 2008; 15:239–249 ª 2008 by the Society for Academic Emergency Medicine Keywords: hyperkalemia, electrocardiography, emergencies, critical care, resuscitation

From the Department of Surgery, University of Vermont College of Medicine (KF), Burlington, VT; the Department of Emergency Medicine, Boston University School of Medicine (JAF, PM, JD, SD, LE, EST), Boston, MA; the Department of Biostatistics, Boston University School of Public Health (LF), Boston, MA; and the Department of Emergency Medicine, Harvard Medical School (EST), Boston, MA. Received June 18, 2007; revision received November 15, 2007; accepted November 15, 2007. Presented at the The Annual New England Regional Society for Academic Emergency Medicine (SAEM) Meeting, Shrewsbury, MA, March 2006, and at the SAEM Annual Meeting, San Francisco, CA, May 2006. Address for correspondence and reprints: Kalev Freeman, MD, PhD; e-mail: [email protected].

ª 2008 by the Society for Academic Emergency Medicine doi: 10.1111/j.1553-2712.2008.00058.x

I

n Ontario, Canada, thousands of hospitalizations for hyperkalemia were recently attributed to increased use of spironolactone after publication of the Randomized Aldactone Evaluation Study (RALES).1 While there is no direct evidence for an increasing prevalence of hyperkalemia in the United States, the combination of a rising prevalence of renal failure, coupled with changes in the pharmacologic treatment of hypertension and heart failure, is expected to lead to a surge in hyperkalemia cases.2–5 From 1990 to 2001, the number of cases of chronic renal failure in the United States nearly doubled, with most due to diabetes or hypertension.2 The number of patients with diabetes with new dialysis requirements

ISSN 1069-6563 PII ISSN 1069-6563583

239

240

Freeman et al.

•

PATIENT PRESENTATION AND ECG ON TIME TO TREATMENT OF HYPERKALEMIA

increased by 162% from 1990 to 2002.3 Hyperkalemia may be seen on routine predialysis testing in 10% of hemodialysis patients,4 and 3%–5% of deaths in dialysis patients are attributed to hyperkalemia;5 thus, the increasing number of patients living with chronic renal failure and hemodialysis will likely lead to increases in cases of hyperkalemia. Similarly, an evolving pharmacologic approach to hypertension and heart failure might further increase the prevalence of hyperkalemia. Angiotensin-converting enzyme (ACE) inhibitors cause hyperkalemia in approximately 10% of outpatients within a year of starting therapy.6–9 Paradoxically, patients with conditions such as type 2 diabetes and congestive heart failure may benefit most from ACE inhibitor therapy, but these comorbidities also place them at increased risk for development of significant hyperkalemia. Drug–drug interactions, including the combination of an ACE inhibitor and potassium-sparing diuretic, also contribute to hospitalizations for hyperkalemia.10 The lethal toxicity of potassium was linked to its cardiac effects nearly 100 years ago.11 It is now known that at the tissue level, hyperkalemia reduces myocardial conduction velocity and accelerates the repolarization phase, producing well-described changes on the surface electrocardiogram (ECG), including a narrow, symmetrical T wave, prolonged PR interval, diminished P-wave amplitude, QRS widening, and ultimately a sinusoidal ‘‘QRST’’ that terminates in asystole or ventricular fibrillation.12–17 Other ECG changes that have been associated with hyperkalemia include conduction blocks (fascicular block, bundle branch block, seconddegree heart block, and complete heart block) and block of bypass tracts with loss of the delta wave in patients with Wolff-Parkinson-White syndrome.18–20 Although few maneuvers are necessary for the initial management of hyperkalemia, prior studies in hospitalized patients have shown an average delay of 2.1 hours after laboratory notification before initiation of treatment for severe hyperkalemia.21 To our knowledge, there are no prior studies describing clinical characteristics or treatment patterns for emergency department (ED) patients with hyperkalemia. We sought to describe the clinical features and treatment patterns of ED patients with hyperkalemia, including those patients who did not receive specific treatment for hyperkalemia. We hypothesized that clinical and ECG findings would be associated with early initiation of treatment for hyperkalemia. The primary goal was to determine the length of time between presentation and treatment for patients with clinically significant hyperkalemia discovered in the ED. The secondary goal was to identify clinical findings associated with failure to treat hyperkalemia or with treatment delay. METHODS Study Design We conducted a retrospective medical record review designed using published guidelines for chart reviews in emergency medicine research.22,23 The institutional review board of our medical center approved the study design.

Study Setting and Population The study took place in an ED with 128,000 annual visits. The ED serves a 547-bed urban academic teaching hospital and a Level 1 trauma center. To identify patients with clinically significant hyperkalemia, we searched electronically captured laboratory records of all chemistry values for adult patients (age > 21 years) reported to the ED in the 12-month period from January 1, 2004, through December 31, 2004. A computer search of this database generated a list of medical record numbers for patients with critically elevated serum or whole blood potassium values, defined in our laboratory as > 6.0 mmol ⁄ L. Samples with visible hemolysis, as documented by the laboratory technician in the initial laboratory report, were dropped from this initial list. The medical record numbers were then cross-matched to identify patient charts and ECGs in the hospital’s electronic database. In some cases, multiple ECGs were available; for each case, the initial ED ECG for each patient visit, prior to treatment for hyperkalemia, was used for the purposes of this study. Study Protocol Interventions considered as possible treatment for hyperkalemia included administration of intravenous calcium gluconate, calcium chloride, bicarbonate, insulin and glucose, or furosemide; nebulized albuterol; oral exchange resin (sodium polystyrene sulfonate or Kayexalate); or hemodialysis. Methods of Measurement. Potassium values were determined either from venous or arterial whole blood gas analysis or from serum, as part of the chemistry panel. Lab specimens were obtained by ED nursing staff and then sent to the main hospital laboratory. The hospital laboratory posts the clinical results on a computer database, which is immediately accessible in the ED. Additionally, laboratory personnel call the ED by telephone to directly inform the responsible physician or nurse of critical values within 5 minutes of the laboratory result. Data Collection and Processing. Clinical data from the medical record were abstracted by four physicians (KF, KSD, LE, EST). ECG data were abstracted from the ED ECG by a single physician (KF). Prior to study review, physicians were trained in data abstraction for this project using sample medical records. The hospital laboratory provided a list of medical record numbers for all ED patient visits for which a critical potassium value was reported. Predetermined study inclusion criteria were a nonhemolyzed potassium level > 6.0 mmol ⁄ L and patient age > 21 years. Predefined criteria excluded patients referred to the ED for hyperkalemia or dialysis, patients with paced rhythms on ECG, if hyperkalemia was not treated due to advance directives (‘‘do not attempt resuscitation’’ or ‘‘comfort measures only’’), if the patient left the hospital prior to receiving treatment, or if a second measurement showed resolution of hyperkalemia without treatment. Patients with cardiac arrest prior to arrival in the ED were also excluded because closed-chest

ACAD EMERG MED • March 2008, Vol. 15, No. 3

•

www.aemj.org

cardiopulmonary resuscitation has been shown to cause secondary hyperkalemia.24,25 The medical records for each patient visit were accessed electronically. A standardized, closed-ended, data abstraction form included mode of arrival, demographics, chief complaint, vital signs, medications, previous medical history, laboratory results, interventions, and disposition. Time parameters included the time of triage, the time at which the laboratory result appeared on the computer, the time at which the result was called to the ED, and the times of physician ordering and nursing administration of medications. Demographic information was always taken from registration information, which includes patient-reported ethnicity and primary language. Chief complaints were uniformly abstracted from the physician notes, rather than triage or nursing progress notes. The medical history and medications were limited to the data available at that particular ED visit. Medications administered in the ED were identified from nursing progress notes; medications administered after admission to the hospital were identified from inpatient electronic pharmacy records. Vital signs were obtained from nursing records as documented in triage notes and in progress notes during the ED course. A second standardized, closed-ended data abstraction form was used to review all ECGs for the presence of abnormalities associated with hyperkalemia. A research assistant (JD) entered all data into a Microsoft Excel database (Microsoft Corp., Redmond, WA), which was checked for accuracy by the research coordinator (PM). Time to treatment was determined by calculating the difference between time of triage and the time of the first ED administration for any treatment possibly targeting hyperkalemia. Explanatory variables included mode of arrival; chief complaint; vital signs; previous medical history; medications; ECG findings; and laboratory results for potassium, blood urea nitrogen, and creatinine. Vital signs were dichotomized by relevant clinical cutoffs, and abnormal triage vital signs were defined as bradycardia, heart rate (HR) < 60 beats ⁄ min; tachycardia, HR > 100 beats ⁄ min; or hypotension, systolic blood pressure (sBP) < 90 mmHg. Hypotension during the ED course was defined as any documented sBP < 90 mmHg. ECG diagnoses of abnormal T-wave morphology (either nonspecific T-wave abnormality, peaked T waves, or ST elevation), abnormal impulse generation (sinus bradycardia < 50 beats ⁄ minute, sinus arrest, or wide complex tachycardia > 100 beats ⁄ minute), and abnormal impulse conduction (first-degree atrioventricular block > 200 msec and either nonspecific interventricular conduction delay or bundle branch block > 120 msec), based on standardized cardiology consensus definitions for terminology and interpretation published by the American College of Cardiology Task Force.26 The World Health Organization Task Force criteria for conduction abnormalities was used for differentiating right or left bundle branch from nonspecific interventricular conduction delay.27 We used standard cardiology morphology criteria for defining the characteristic T-wave morphology of hyperkalemia, referred to in this study as ‘‘peaked T waves.’’ While the term ‘‘tall peaked T waves’’ has also been used

241

colloquially to describe the morphology of the ST segment in hyperkalemia,28 T-wave amplitude is a nonspecific finding that is not helpful in distinguishing hyperkalemia from normal variants.16,29 Both low- and high-amplitude peaked T waves are seen in hyperkalemia, and the consensus definition for the ECG diagnosis of hyperkalemia depends on T-wave morphology, characterized by a ‘‘pinched,’’ narrow, symmetrical T wave, independent of amplitude.26,30,31 Physician abstractors reviewed a limited patient chart, which contained the ED documentation (triage, nursing, and physician notes), pharmacy records, and electronic laboratory records. A separate chart containing only the initial 12-lead ECG, the patient’s age, and an identification number provided the source for ECG data abstraction. The physician who interpreted all ECGs (KF) was blinded to the computer-generated ECG interpretation, the specific potassium level of the patient, and the associated medical record. To establish interrater reliability for abstraction of clinical variables, all four reviewers abstracted key variables from a random sample of 30 patient charts, and the results were compared to calculate of interrater reliability coefficients. Because a single abstractor scored all ECGs, interrater reliability for abstraction of ECG data was determined by comparison with scores obtained independently by an outside expert. A random sample of 20 ECGs was reviewed by a board-certified cardiologist, unaware of the information obtained by the first reviewer (KF). Both reviewers were aware that the ECGs included tracings with hyperkalemia, but they were blinded to the specific potassium values of each patient. Standardized ECG data abstraction forms were used, and the results were compared to determine interrater reliability. Measures The primary outcome for analysis was time from triage to initiation of treatment for hyperkalemia as a function of clinical presentation. Secondary outcomes, including dysrhythmias and death, were also recorded. Data Analysis All analyses were performed with SAS Version 9.1 (SAS Institute Inc., Cary, NC) using the GENMOD procedure, fitting a generalized linear model with a log link, and using the gamma distribution. Some patients were seen with hyperkalemia more than once during the study period. Correlation due to patients with multiple visits was accounted for by the use of generalized estimating equations (GEE).32,33 Because there is no particular time point that has been determined to represent ‘‘early treatment’’ for hyperkalemia, we performed an analysis that considered time to treatment as a continuum, using a logarithmic conversion to fit a generalized linear model. Univariate analyses were initially performed on data from the patient visit records to identify factors associated with time to treatment. Variables of primary interest, as well as those deemed significant in the univariate analysis at the 0.20 level of significance, were included in a multivariate generalized linear model. Specific variables related to time to treatment at the

242

Freeman et al.

•


p < 0.20 level by univariate analysis were: triage vital signs of hypotension, hypotension in the ED, medical history of hypertension, nonspecific abnormality on the ECG, transportation by emergency medical services (EMS), chief complaint of dyspnea, history of dialysis, potassium level > 7.5 mmol ⁄ L, or history of ACE inhibitor use. Further backward selection techniques were used to select the final model. As this was an exploratory analysis, covariates that were close to the 0.05 level of significance were included in the final model. Exponentiated parameter estimates from this model are interpreted as the ratio of the average time to treatment for a patient with the covariate versus a patient without that particular covariate, adjusting for all other terms in the model. Prior studies of hyperkalemia have used specific cutoffs based on the level of potassium elevation.26 In a similar fashion, we recoded potassium values into categorical values, defined as 6.0–6.4, 6.5–7.4, or 7.5 mmol ⁄ L and above, for analyzing the relationship between potassium level and time to treatment. The quasi-likelihood under the independence model criterion (QIC) statistic34 was used to assess the adequacy of the correlation structure and to aid in selecting the final model. The QIC is analogous to the familiar Akaike’s information criterion, used for comparing model’s fit with likelihood-based methods. Since the GEE method is not a likelihood-based method, the AIC statistic is not available. The QIC is calculated for each model being considered, and the model with the smallest value is deemed the most desirable in terms of parsimony and the amount of variance explained. In this model the exchangeable correlation matrix, which implies that multiple observations on the same patient are equally correlated to one another, was selected using the QIC criteria. RESULTS Interrater Reliability Interrater reliability for abstraction of clinical parameters was determined by comparison of key data points including time values for a set of 30 random charts. We found either ‘‘substantial’’ or ‘‘nearly perfect’’ agreement between reviewers, with kappa values of 0.64– 0.97. Interrater reliability for abstraction of ECG data was determined by comparison with scores obtained independently by an outside expert, using a random sample of 20 ECGs. Kappa values ranged from 0.77 for nonspecific interventricular conduction delay, 0.89 for peaked T waves, to 1.00 for bundle branch block. These kappa values indicate substantial to nearly perfect agreement between independent reviewers, providing confidence for the overall analysis. Main Results A computer search of electronically captured hospital laboratory records for nonhemolyzed, critical potassium values (‡ 6.0 mmol ⁄ L) yielded 291 unique patient visits. Chief complaint chart review identified 116 ineligible patients (Figure 1). The remaining 175 patient visits (76% of eligible visits) were included in our study. Hyperkalemia was specifically treated in 168 of these 175 patient visits. In the remaining 7 cases, no specific

Figure 1. Flow chart showing patient visits with hyperkalemia included in study population. ECG = electrocardiogram; ED = emergency department.

treatment for hyperkalemia was given. Details of the ED records for these cases are summarized in Table 1. In 2 of the 7 cases, a repeat potassium value obtained on the hospital floor showed improvement in hyperkalemia, without specific treatment. There was a single case in which the patient left prior to receiving treatment. The remaining 4 of the 7 patients were neither diagnosed nor treated for hyperkalemia in the ED, but instead they were discharged home. The ED notes do not indicate vital sign abnormalities or mention of arrhythmias prior to the time of discharge. Two patients died in the ED despite initiation of treatment for hyperkalemia within 1 hour of triage. Hyperkalemia may have caused the cardiac dysrhythmias that ultimately resulted in cardiac arrest, in both of these cases. In the first case, a 46-year-old female patient with a history of hepatitis C, human immunodeficiency virus nephropathy, and renal failure on dialysis presented with a presyncopal episode prior to dialysis. She was transferred to the dialysis clinic, with hypotension and altered mental status. The patient’s potassium level in the ED was 7.7 mmol ⁄ L, and treatment with calcium chloride, insulin, glucose, and bicarbonate was initiated. Despite this treatment, the patient remained hyperkalemic, and her HR dropped to 30 beats per minute. The patient died after approximately 120 minutes of


•

www.aemj.org

243

Table 1 Clinical Characteristics of Emergency Department Patients Not Treated for Hyperkalemia Chief Complaint

Age (Years)

K+ (mmol ⁄ L)

Clinical Notes

Disposition

Reason Not Treated

Increased pulse rate

59

6.1

Admitted

Resolved

Weakness

84

6.6

Admitted

Resolved

Chest pain

62

6.1

AMA

AMA

Abdominal pain

49

7.3

Discharged

Missed diagnosis

Stubbed toe

72

6.1

Discharged

Missed diagnosis

Vomiting

34

6.4

Discharged

Missed diagnosis

Cough, fever

51

6.6

Nursing home patient with pulse rate of 137 prior to arrival. IV fluids given, laboratory studies repeated, repeat K+ 5.4 without specific treatment. Patient recently diagnosed with metastatic colon cancer. Admitted for IV hydration and pain control, laboratory studies repeated. Repeat K+ 4.8 without specific treatment. Chest pain resolved prior to return of lab values, patient left AMA. Patient with end-stage renal disease, polysubstance abuse, diagnosed with gastritis. Laboratory value not reported prior to discharge. Diabetic patient, with toe injury ⁄ nail avulsion repaired by vascular surgery. Patient with suprapubic catheter for neurogenic bladder, admitted for small bowel obstruction, improved on the day following admission. Patient discharged without repeat labs. Renal transplant patient, with fever, cough, and normal CXR. Diagnosed with acute bronchitis.

Discharged

Missed diagnosis

AMA = left against medical advice; CXR = chest x-ray; IV = intravenous; K = potassium.

resuscitative efforts, including intubation, vasoactive agents, and transcutaneous pacing for continued bradycardia. The second patient was a 57-year-old male with hepatocellular cancer and chronic abdominal pain requiring narcotics. He presented with altered mental status and was found to be poorly responsive, with pinpoint pupils and abdominal pain. His mental status deteriorated despite naloxone, and he was intubated with a rapid sequence protocol including succinylcholine. His laboratory studies revealed combined respiratory and metabolic acidosis, with a potassium level of 7.3 mmol ⁄ L. Initial treatment included calcium chloride and bicarbonate, but the patient’s blood pressure dropped, with a rhythm progressing from pulseless electrical activity to ventricular fibrillation and asystole. Despite further treatment including bicarbonate, epinephrine, atropine, and amiodarone, the patient died after approximately 65 minutes of resuscitation. The cohort of 168 patient visits analyzed for time from triage to initiation of treatment included 129 unique patients, with 39 patients who had multiple ED visits (ranging from two to eight visits) and critical hyperkalemia during 2004. The median potassium value for this cohort was 6.5 mmol ⁄ L (interquartile range [IQR] = 6.3 to 7.1 mmol ⁄ L), the range was 6.0–8.5 mmol ⁄ L (Figure 2). We performed a chi-square statistical analysis comparing visit and patient-specific attributes for 61 patient

visits that were excluded from time-to-treatment analysis, to the 168 patient visits included in this analysis. Those excluded had a median K+ value of 6.4 mmol ⁄ L (IQR = 6.3 to 7.4 mmol ⁄ L). There were no significant differences in any patient-specific or visit-specific attribute between included and excluded patients except disposition (excluded patients were less likely to have been admitted to a bed with telemetry monitoring compared to included patients; p < 0.05, chi-square test). Analysis of Time to Treatment The median time to initial treatment of hyperkalemia was 117 minutes (IQR = 59 to 196 minutes). The distribution of time to initial treatment is shown in Figure 3. Twenty (12%) patients were treated within 30 minutes of triage, and 46 (28%) patients received treatment within 1 hour; 26 (16%) patients were treated before the potassium value was reported by the lab. Many patients experienced treatment delays. Of the 142 patients treated after laboratory results became available in the hospital database, the median time from laboratory result to initial treatment was 67 minutes (IQR = 49 to 82 minutes). A total of 105 were not treated until after the laboratory called the ED with the critical result; for these patients, the median time from documentation of critical result to initial treatment was 42 minutes (IQR = 20 to 95 minutes). The initial ED treatment was insulin and glucose in 54 (33%) patient

244

Freeman et al.

•


Table 2 Clinical Characteristics of Emergency Department Patients Treated for Hyperkalemia

Characteristics

Figure 2. Distribution of potassium values for 168 patient visits.

Figure 3. Proportion of patients receiving therapy for hyperkalemia as a function of time from triage. The individual curves represent exponentiated parameter estimates from the multivariate model, showing the ratio of the average time to treatment for a patient with the covariate versus a patient without that particular covariate, adjusting for all other terms in the model. EMS = emergency medical services.

visits; an intravenous calcium infusion (calcium gluconate, 37; calcium chloride, 5) was administered initially in 42 (25%) patients. Seventy-three percent of treated patients were transported to the ED by ambulance, 15% of patient visits had documented sBP < 90 mmHg, and 33% resulted in admission to intensive care units (Table 2). Four percent of patients were unresponsive at presentation. The

Frequency

Visit-specific attributes (n = 168) Transport Ambulance 116 (73.4%) Chief complaint Shortness of breath 33 (19.8%) Weakness 31 (18.6%) Mental status change 13 (7.8%) Syncope 9 (5.4%) Unresponsive 7 (4.2%) Other 72 (43.1%) Triage vital signs Tachycardia (HR > 100) 38 (22.6%) Bradycardia (HR < 60) 15 (8.9%) Hypotension (sBP < 90) 12 (7.1%) ED vital signs Hypotension (sBP < 90) 25 (14.9%) Disposition Floor with telemetry 65 (41.4%) Intensive care unit 51 (32.5%) Floor 20 (12.7%) Transfer ⁄ left ⁄ AMA 15 (9.6%) Death 2 (1.3%) Patient attributes (n = 129) Gender Male 80 (62.0%) Ethnicity African American 68 (52.7%) White 31 (24.0%) Hispanic 21 (16.3%) Asian 1 (0.8%) Other 7 (5.4%) Age (yr) 21–30 8 (6.3%) 31–40 8 (6.2%) 41–50 28 (21.7%) 51–60 32 (24.5%) 61–70 22 (17.1%) 71–80 19 (14.7%) >80 11 (8.5%) Medical history Hypertension 79 (61.7%) Diabetes 57 (44.2%) Renal failure 44 (34.1%) Dialysis 37 (28.7%) Congestive heart failure 24 (18.6%) Renal insufficiency 15 (11.6%) CAD 12 (9.3%) Medications ACE inhibitors 46 (35.7%) Loop diuretic 25 (19.4%) AR blockers 4 (3.1%) Spironolactone 4 (3.1%) Potassium 3 (2.3%)

Treated in < 1 hr

40 (34.5%) 11 11 3 2 4 12

(33.3%) (35.5%) (23.1%) (22.2%) (57.1%) (16.7%)

15 (39.5%) 6 (40.0%) 5 (41.7%) 11 (44.0%) 15 20 2 1 1

(23.1%) (40.8%) (10.0%) (6.7%) (50.0%)

21 (26.3%) 14 7 6 1 3

(20.6%) (22.6%) (28.6%) (100.0%) (37.5%)

2 1 7 9 5 3 3

(25.0%) (12.5%) (25.0%) (28.1%) (22.7%) (15.8%) (27.3%)

15 14 7 7 6 2 7

(19.0%) (24.6%) (15.9%) (18.9%) (25.0%) (13.3%) (58.3%)

10 7 0 1 1

(21.7%) (28.0%) (0.0%) (25.0%) (33.3%)

ACE = angiotensin-converting enzyme; AMA = against medical advice; AR = adrenoreceptor; CAD = coronary artery disease; HR = heart rate; sBP = systolic blood pressure; ED = emergency department.

most common chief complaints were dyspnea (20%), weakness (19%), and altered mental status (8%). The chief complaints for many patient visits (43%) were either unique to that patient visit or occurred infrequently, in the overall study population. Characteristics of the 129 unique patients in the study population are shown in Table 2.


•

www.aemj.org

Table 3 Retrospective Interpretation of Initial Emergency Department Electrocardiogram (ECG)

245

Table 4 Results from the Multivariate Generalized Linear Model with a Log Link and Gamma Distribution

Frequency (n = 168) T-wave morphology Peaked T wave Nonspecific ST abnormality ST segment elevation Impulse generation Bradycardia (HR < 50) Sinus arrest Wide complex tachycardia Conduction abnormalities First-degree AV block Interventricular conduction delay (> 120 msec) Bundle branch block (left or right) Retrospective ECG interpretation Any abnormality consistent with hyperkalemia Nonspecific ST abnormality No signs of hyperkalemia ⁄ normal ECG Peaked T wave and first-degree AV block Peaked T wave and interventricular conduction delay Interventricular conduction delay and first-degree AV block

58 (34.5%) 56 (33.3%) 7 (4.2%) 7 (4.2%) 3 (1.8%) 1 (0.6%) 28 (16.7%) 19 (11.3%) 10 (6.0%) 83 (50.0%) 40 28 9 7

(23.8%) (16.7%) (5.4%) (4.2%)

5 (3.0%)

AV = atrioventricular; HR = heart rate.

The initial ECG was abnormal in 83% of patient visits. However, 24% of ECGs showed only nonspecific ST abnormalities. The most frequent findings are listed in Table 3. Patient visit records were analyzed to identify clinical characteristics associated with early treatment (within 1 hour of triage). As part of initial data exploration, we calculated the proportion of patients with each characteristic treated within 1 hour of triage (Table 2). Of note, fewer than half of hypotensive patients (12 of 26, 46%) received early treatment for hyperkalemia. Patient-specific attributes, including age, ethnicity, and primary language, were not associated with early time to treatment. There was no difference in time to treatment based on potassium level. None of the ECG abnormalities, including the presence of peaked T-wave morphology, were associated with early treatment. Furthermore, vital sign abnormalities were not associated with early treatment. Factors independently associated with treatment delay for hyperkalemia in the multivariate model included transport by private vehicle (not by ambulance), chief complaint of syncope, and history of hypertension (Table 4). Patients with the chief complaint of ‘‘unresponsive,’’ ‘‘dyspnea,’’ or ‘‘weakness’’ had a trend toward early treatment that was not statistically significant in multivariate analysis. A graphical representation of the time to initiation of treatment for patients with each of these attributes is shown in Figure 3. DISCUSSION Our study indicates that the recognition of patients with severe hyperkalemia is challenging, and the initiation of appropriate therapy for this disorder is frequently delayed. One of our most striking findings was the

Medical history Hypertension Chief complaint Unresponsive Dyspnea Weakness Syncope Transport Not by ambulance

Estimated Ratio of Time to Treatment

95% CI

p-Value

1.38

1.04,1.83

0.03

0.46 0.74 0.76 1.73

0.19, 0.53, 0.56, 1.16,

1.08 1.04 1.04 2.59

0.07 0.09 0.09 0.01

1.44

1.11, 1.83

0.01

Estimates give the ratio of the time to treatment for a patient with the factor versus one without the factor. For instance, the time to treatment for a patient who is unresponsive is expected to be 0.46 times that of a patient without that chief complaint. CI = confidence interval.

observation that the mean time from triage to treatment in patients with critical hyperkalemia was nearly 2 hours. The American Heart Association recommends ‘‘immediate therapy’’ for serum potassium > 6.0 mEq ⁄ L,35 but we found that even when starting from the time of laboratory report of critical hyperkalemia, the median time to treatment was 42 minutes. Delays in treatment for hyperkalemia have previously been reported in hospitalized patients, with a mean time of 2.1 hours from laboratory notification of potassium > 6.5 mmol ⁄ L before initiation of treatment.21 Although a median delay of 67 minutes between the initial laboratory result of critical hyperkalemia and initiation of specific treatment may seem long for the ED setting, this delay was only for a subset of our population; the 141 patients who were treated after the laboratory results were actually logged in the hospital computer system. The median time from actual telephone notification of the critical potassium level until treatment was 42 minutes. On the other end of the spectrum in treatment patterns, 16% of patients were treated empirically, before the potassium value even returned from the laboratory. This observation suggests that empiric treatment is possible and this may be considered in the many clinical settings where immediate quantitative potassium values are not available. We studied the population of patients in whom the diagnosis of hyperkalemia was identified after ED arrival. The variety of complaints and frequent nonspecific ECG abnormalities reported here may have contributed to some diagnostic uncertainty prior to the return of laboratory results. We noted some patterns in reviewing the charts, which would explain delays in treatment. In many cases, the initial physician response to being alerted of an unexpectedly elevated potassium level was to obtain intravenous access, send a repeat potassium level, and acquire an ECG. In some cases, difficulties in intravenous access were documented as reasons for treatment delay, which is not surprising for a population including 28% dialysis patients. However, in

246

Freeman et al.

•


patients with difficult intravenous access, other treatments can be considered, depending on the patients’ clinical status. These include high-dose inhaled beta agonist therapy or direct intravenous or shunt injection of therapies in a life-threatening situation. Also, in the 18 patients (10.8%) treated directly with dialysis from the ED (and no other specific treatment), the time from laboratory results until initiation of dialysis often exceeded the median of 42 minutes. We were also surprised to find that patients with vital sign or ECG abnormalities did not receive early treatment for hyperkalemia. Fifty-four percent of those with a sBP < 90 mmHg while in the ED did not receive treatment within the first hour from triage. Blood pressure measurements are inadequate determinants of tissue perfusion, and auscultatory blood pressure measurements may be inaccurate. Nevertheless, the presence of sBP < 90 mmHg in the context of critical hyperkalemia suggests hemodynamic instability, and a delay of more than 1 hour from triage for initiation of treatment in these patients is a particular concern. The ultimate treatment delay might be considered failure to treat due to a missed diagnosis. While the ED notes do not indicate arrhythmias or other abnormalities at the time of discharge, we are unable to determine retrospectively whether any of the four undiagnosed patients ultimately had an adverse outcome due to the failure to diagnose and treat hyperkalemia. Nonetheless, these patient visits represent a particular risk management concern. To our knowledge, this is the first study to report the clinical presentation and timing of treatment for hyperkalemia in the ED setting. Hyperkalemia is often of multifactorial origin6,21,36 and we did not attempt to attribute a specific cause to the cases reviewed here. Nonetheless, risk factors for potassium toxicity in our ED patients mirrored those commonly reported in the general population, including a history of hypertension (62%), diabetes (45%), renal failure (34%), dialysis (28%), and congestive heart failure (19%). Thirty-six percent of patients were taking ACE inhibitors. A recent population analysis of 1.3 million adults in Ontario, Canada demonstrated an abrupt increase in hyperkalemia-associated hospitalization and death associated with increased rates of spironolactone prescriptions since the publication of the RALES.1 In our population, only 3% of patients with hyperkalemia had a history of spironolactone use. We observed peaked T-wave morphology in 57 (34%) of cases. Other ECG changes were less frequently observed (Table 3). Previous studies reported that the frequency of peaked T waves in ECGs of hospitalized patients with hyperkalemia was 36%; other abnormalities reported included 11% with first-degree heart block, 8% with widened QRS complex, and 4% with junctional rhythm.21,37 A previous report of ECG findings in ED patients with hyperkalemia compared to those with renal failure and normal potassium levels found that peaked T waves on the ECG had a poor sensitivity for predicting hyperkalemia, ranging from 38% to 46%, with a specificity of 85%.33 Attempts to use quantitative criteria for defining hyperkalemia on the

ECG have not been validated. For example, in one study, 23 patients with hyperkalemia were used to develop initial quantitative criteria for the diagnosis of hyperkalemia and then, of a 20,000 ECG database, an additional 21 patients with hyperkalemia and similar ECG abnormalities were identified.29 Normal ECGs were used as controls. The pool of 44 hyperkalemic patient ECGs was used to develop the quantitative criteria of T-wave amplitude ⁄ duration > 4.5 mv ⁄ msec, T wave duration < 135 msec, and T-wave amplitude > 0.4 mv. The stated sensitivity and specificity figures were never prospectively validated. The fact that the hyperkalemic population studied represents a nondirected collection of cases, and the control population consists only of patients with normal ECGs, brings up the question of selection bias in this study.29 While we found a high incidence of ECG abnormalities, most were nonspecific. This may be attributable in part to the fact that 28% of the patients in our study population were dialysis patients. It has previously been reported that hemodialysis patients with hyperkalemia are less likely to show ECG changes, despite the risk for suddenly developing arrhythmias including asystole or ventricular fibrillation without prior ECG abnormalities.15,38–40 None of the ECG abnormalities noted in our study patients, including peaked T-wave morphology, prompted early treatment. Thus, our hypothesis that ECG findings such as peaked T-wave morphology would lead to early treatment was not supported by our data. In our population, similar to prior reports, the ECG had a generally poor sensitivity for hyperkalemia. The utility of ECG in screening for hyperkalemia may be negligible in the modern era of rapid laboratory assessment of potassium levels. The presence of ECG abnormalities in the context of hyperkalemia may be nonspecific; in some cases, however, these abnormalities suggest impending circulatory collapse and should mandate aggressive treatment.41 Since in many settings (out of hospital, ED), the initial ECG remains the primary diagnostic tool available when immediate quantitative potassium levels are not available, research should examine the ability of clinicians to identify ECG findings consistent with hyperkalemia. Multivariate analysis identified several factors associated with early treatment (within 1 hour of triage). Chief complaints of unresponsive, weakness, or dyspnea were associated with early treatment, possibly due to earlier time seen by the emergency physician, earlier blood draws for chemistry analysis, or earlier ECG. Not surprisingly, patients brought in by private vehicle (not by EMS) were more likely to have delayed treatment. It is unclear why medical history of hypertension and chief complaint of syncope were associated with treatment delay. While we did find clinical variables that were associated with timing of treatment for hyperkalemia, we were somewhat surprised to find that certain findings, such as hypotension, history of renal failure or dialysis, and ACE inhibitor use, were not associated with early initiation of treatment. This could be due to confounding. For instance, there was a strong correlation between patients with renal failure and those with


•

www.aemj.org

hypertension. Improved emergency physician recognition of risk factors or clinical findings suggestive of critical hyperkalemia might improve the time to treatment. We found that 26 (16%) patients received empiric treatment for hyperkalemia prior to laboratory confirmation of the diagnosis. In the context of the current era of widespread ACE inhibitor use and an increasing prevalence of hyperkalemia, our finding of treatment delays in patients with critical hyperkalemia may support early, empiric treatment for patients with possible hyperkalemia, particularly during cardiac resuscitation. Current Emergency Cardiovascular Care (ECC) algorithms for pulseless arrest, bradycardia, and tachycardia recommend that the provider ‘‘search for and treat’’ possible contributing factors, including hypo- or hyperkalemia; there is no specific mention of intravenous calcium or other agents for treatment of hyperkalemia in these resuscitation algorithms. Calcium chloride is listed in the current guidelines only as a medication for the treatment of known hyperkalemia, with the warning ‘‘do not use routinely in cardiac arrest.’’ The addition of a suggestion for empiric treatment for hyperkalemia with intravenous calcium to future ECC protocols might improve early treatment. Because this study was performed at a single institution, we have carefully considered institution-specific factors that might explain the treatment delays. At the time of this study, laboratory specimens in our institution were obtained by ED nursing staff, and then sent to the main hospital laboratory via a pneumatic tube system. We have recently developed a ‘‘Stat lab’’ in the ED, which avoids the need to use the pneumatic tub for processing specimens; this intervention may decrease the time for diagnosis and treatment of hyperkalemia. Despite the standing protocol for the lab to directly inform providers of critical results, we still identified four cases of missed diagnosis in which hospital records clearly note that the critical results were called to the ED. There were delays both in physician ordering and in nursing administration of treatment for hyperkalemia. Because the times of physician ordering were not always documented, we are unable to clearly differentiate the delay due to ordering versus administration of medications. Technologic advancements may also lead to earlier diagnosis of hyperkalemia. The time required to measure serum potassium has decreased with automated labs and most recently, with the availability of rapid point-of-care testing using whole-blood samples. It is now possible to check electrolytes within 1–2 minutes using a finger-stick sample.42 Point-of-care testing in the out-of-hospital setting, in triage, or at the bedside during resuscitation might expedite the recognition and treatment of critical hyperkalemia. LIMITATIONS The retrospective design of this study may have led to missed cases of hyperkalemia; we were unable to identify patients who did not have laboratory determination of potassium level performed during the ED visit. Prospective institution of a rigorous abstractor-training

247

process, uniform data collection, use of consistent definitions, and detailed periodic review of abstracted charts support the validity and integrity of the data set. Nevertheless, like all single-center studies, the database used in this study has potential limitations related to the validity of the data and sampling bias. The theoretical model of our research is limited by the assumption that early treatment for critical hyperkalemia impacts patient outcome and by the use of the arbitrary clinical cutoff of a potassium level > 6.0 mmol ⁄ L. The serum potassium level does not predict development of hemodynamic instability in any particular patient, because the effects of potassium depend on the rate of increase and on the associated metabolic milieu, including the pH and calcium level.35,41 The American Heart Association defines hyperkalemia as a serum potassium > 5.0 mEq ⁄ L, but notes that ‘‘it is moderate (6 to 7 mEq ⁄ L) and severe (> 7 mEq ⁄ L) hyperkalemia that are life-threatening and require immediate therapy.’’35 Other sources recommend ‘‘emergent treatment for all patients with electrocardiographic evidence of hyperkalemia, regardless of serum potassium levels,’’ but use a cutoff of 6.5 mEq ⁄ L for treatment of asymptomatic patients without ECG changes.43 We used 6.0 mmol ⁄ L, because this is considered critical in our hospital laboratory and because it is consistent with American Heart Association guidelines for immediate therapy. Etiologies for hyperkalemia are multifactorial, and we did not attempt to identify a specific cause for each case in our study. However, differences in the specific etiology for hyperkalemia may have contributed to variation in time to treatment and limit our data analysis. Another limitation is our treatment of the initial ECG as an isolated tracing, without reference to prior ECGs or notation of dynamic ECG changes during the ED course. In the actual ED course, comparison of the initial ECG with prior ECGs may have affected the treating physician’s time to administration of therapy. An additional limitation is that the timing of ECG acquisition may not correlate exactly with the time of blood draws for laboratory studies. Both the timing of medication order and the medication administration were measured, when available, and in our preliminary analysis, we found that the time between physician ordering and nursing administration of medications was inconsequential. However, the primary reason for exclusion of patient visits from analysis was lack of documentation of time points for medication administration and, overall, documentation of physician time points (time seen by physician, time of medication order) was less complete than documentation of nursing-specific time points (time of triage, time of medication administration). We therefore used the more consistently documented nursing time points for calculation of time to treatment to maximize the number of patient visits included in the study. This methodology limits our ability to identify other potential causes for treatment delay, such as an extended time spent in the waiting room prior to being seen, or a delay between physician order and nursing administration of treatment.

248

Freeman et al.

•


CONCLUSIONS The majority of patients with critical hyperkalemia waited for approximately 2 hours from triage to initial treatment. Even after laboratory diagnosis of critical hyperkalemia, the mean time from laboratory result to treatment was more than 1 hour. Time to treatment was unaffected by hypotension or ECG abnormalities. It is not surprising that many cases were not immediately recognized as attributable to a problem with the serum potassium, in the setting of a busy ED, with a potentially unstable patient whose history may be unknown, and with a complaint such as dyspnea or weakness that suggests a broad differential diagnosis. Possible ways to improve early intervention in hyperkalemia include consideration of rapid point-of-care testing at the bedside. If the prevalence of severe and life-threatening hyperkalemia continues to increase, the benefit of a trial of calcium administration in cardiac resuscitation may begin to outweigh its risks, particularly in patients with documented risk factors for hyperkalemia. The authors thank Casey Rebholz, MPH, and Supriya D. Mehta, PhD, for their valuable assistance with the statistical analysis. They also thank Sheilah Bernard, MD, for electrocardiogram interpretation to establish interrater reliability and, moreover, for her cheerful commitment to cross-departmental collegiality.

11.

12.

13. 14.

15.

16.

17.

18.

References 1. Juurlink DN, Mamdani MM, Lee DS, et al. Rates of hyperkalemia after publication of the Randomized Aldactone Evaluation Study. N Engl J Med. 2004; 351:543–51. 2. Centers for Disease Control and Prevention. Statespecific trends in chronic kidney failure–United States, 1990–2001. MMWR Morb Mortal Wkly Rep. 2004; 53:918–20. 3. Centers for Disease Control and Prevention. Incidence of end-stage renal disease among persons with diabetes–United States, 1990–2002. MMWR Morb Mortal Wkly Rep. 2005; 54:1097–100. 4. Tzamaloukas AH, Avasthi PS. Temporal profile of serum potassium concentration in nondiabetic and diabetic outpatients on chronic dialysis. Am J Nephrol. 1987; 7:101–9. 5. Ahmed J, Weisberg LS. Hyperkalemia in dialysis patients. Semin Dial. 2001; 14:348–56. 6. Palmer BF. Managing hyperkalemia caused by inhibitors of the renin-angiotensin-aldosterone system. N Engl J Med. 2004; 351:585–92. 7. Rimmer JM, Horn JF, Gennari FJ. Hyperkalemia as a complication of drug therapy. Arch Intern Med. 1987; 147:867–9. 8. Perazella MA. Drug-induced hyperkalemia: old culprits and new offenders. Am J Med. 2000; 109:307– 14. 9. Ahuja TS, Freeman D Jr, Mahnken JD, Agraharkar M, Siddiqui M, Memon A. Predictors of the development of hyperkalemia in patients using angiotensin-converting enzyme inhibitors. Am J Nephrol. 2000; 20:268–72. 10. Juurlink DN, Mamdani M, Kopp A, Laupacis A, Redelmeier DA. Drug-drug interactions among

19.

20.

21.

22.

23.

24.

25.

26.

27.

elderly patients hospitalized for drug toxicity. JAMA. 2003; 289:1652–8. Mathison G. The effects of potassium salts upon the circulation and their action on plain muscle. J Physiol. 1911; 42:471–94. Winkler AW, Hoff HE, Smith PK. Electrocardiographic changes and concentration of potassium in serum following intravenous injection of potassium chloride. Am J Physiol. 1938; 124:478–83. Thomson WA. Potassium and the T wave of the electrocardiogram. Lancet. 1939; 1:808–11. Tarail R. Relation of abnormalities in concentration of serum potassium to electrocardiographic disturbances. Am J Med. 1948; 5:828–37. Currens J, Crawford J. The electrocardiogram and disturbance of potassium metabolism. N Engl J Med. 1950; 243:843–50. Braun HA, Surawicz B, Bellet S. T waves in hyperpotassemia; their differentiation from simulating T waves in other conditions. Am J Med Sci. 1955; 230:147–56. Dreifus LS, Pick A. A clinical correlative study of the electrocardiogram in electrolyte imbalance. Circulation. 1956; 14:815–25. Przybojewski JZ, Knott-Craig CJ. Hyperkalaemic complete heart block. A report of 2 unique cases and a review of the literature. S Afr Med J. 1983; 63:413–20. Kuvin JT. Images in clinical medicine. Electrocardiographic changes of hyperkalemia. N Engl J Med. 1998; 338:662. Diercks DB, Shumaik GM, Harrigan RA, Brady WJ, Chan TC. Electrocardiographic manifestations: electrolyte abnormalities. J Emerg Med. 2004; 27:153– 60. Acker CG, Johnson JP, Palevsky PM, Greenberg A. Hyperkalemia in hospitalized patients: causes, adequacy of treatment, and results of an attempt to improve physician compliance with published therapy guidelines. Arch Intern Med. 1998; 158:917– 24. Gilbert EH, Lowenstein SR, Koziol-McLain J, Barta DC, Steiner J. Chart reviews in emergency medicine research: where are the methods? Ann Emerg Med. 1996; 27:305–8. Lowenstein SR. Medical record reviews in emergency medicine: the blessing and the curse. Ann Emerg Med. 2005; 45:452–5. Niemann JT, Cairns CB. Hyperkalemia and ionized hypocalcemia during cardiac arrest and resuscitation: possible culprits for postcountershock arrhythmias? Ann Emerg Med. 1999; 34:1–7. Martin GB, Nowak RM, Cisek JE, Carden DL, Tomlanovich MC. Hyperkalemia during human cardiopulmonary resuscitation: incidence and ramifications. J Emerg Med. 1989; 7:109–13. Surawicz B, Uhley H, Borun R, et al. The quest for optimal electrocardiography. Task Force I: standardization of terminology and interpretation. Am J Cardiol. 1978; 41:130–45. Willems JL, Robles de Medina EO, Bernard R, et al. Criteria for intraventricular conduction disturbances


28.

29.

30.

31.

32.

33.

34. 35.

•

www.aemj.org

and pre-excitation. World Health Organizational ⁄ International Society and Federation for Cardiology Task Force Ad Hoc. J Am Coll Cardiol. 1985; 5:1261– 75. Goldberger AL. Clinical Electrocardiography: A Simplified Approach. 6th ed. St. Louis: Mosby, 1999, p 117. Carlisle DM, Niemann JT, Laks MM. The use of the computer to develop new electrocardiogram criteria for hyperkalemia. IEEE Computers Cardiol. 1984; 317–20. Marx JA (editor). Rosen’s Emergency Medicine: Concepts and Clinical Practice. 6th ed. Philadelphia: Mosby, 2006, p 1940. Zipes DP (editor). Braunwald’s Heart Disease: A Textbook of Cardiovascular Medicine. 7th ed. Philadelphia: Elsevier, 2005, pp 142–3. Liang KY, Zeger SL. Longitudinal data analysis using generalized linear models. Biometrika. 1986; 73:13–22. Zeger SL, Liang KY. Longitudinal data analysis for discrete and continuous outcomes. Biometrics. 1986; 42:121–30. Pan W. Akaike’s information criterion in generalized estimating equations. Biometrics. 2001; 57:120–5. American Heart Association. Advanced challenges in resuscitation: Life threatening abnormalities. Circulation. 2000; 102(Suppl I):217–22.

249

36. Halperin ML, Kamel KS. Potassium. Lancet. 1998; 352(9122):135–40. 37. Wrenn KD, Slovis CM, Slovis BS. The ability of physicians to predict hyperkalemia from the ECG. Ann Emerg Med. 1991; 20:1229–32. 38. Szerlip HM, Weiss J, Singer I. Profound hyperkalemia without electrocardiographic manifestations. Am J Kidney Dis. 1986; 7:461–5. 39. Aslam S, Friedman EA, Ifudu O. Electrocardiography is unreliable in detecting potentially lethal hyperkalaemia in haemodialysis patients. Nephrol Dial Transplant. 2002; 17:1639–42. 40. Yu AS. Atypical electrocardiographic changes in severe hyperkalemia. Am J Cardiol. 1996; 77:906–8. 41. Surawicz B, Chlebus H, Mazzoleni A. Hemodynamic and electrocardiographic effects of hyperpotassemia. Differences in response to slow and rapid increases in concentration of plasma K. Am Heart J. 1967; 73:647–64. 42. Backer HD, Collins S. Use of a handheld, batteryoperated chemistry analyzer for evaluation of heat-related symptoms in the backcountry of Grand Canyon National Park: a brief report. Ann Emerg Med. 1999; 33:418–22. 43. Evans K, Reddan DN, Szczech LA. Nondialytic management of hyperkalemia and pulmonary edema among end-stage renal disease patients: an evaluation of the evidence. Semin Dial. 2004; 17:22–9.