Distribution of serum potassium values.
Acker CG, Johnson JP, Palevsky PM, Greenberg A. Hyperkalemia in Hospitalized PatientsCauses, Adequacy of Treatment, and Results of an Attempt to Improve Physician Compliance With Published Therapy Guidelines. Arch Intern Med. 1998;158(8):917-924. doi:10.1001/archinte.158.8.917
Hyperkalemia is a common, potentially life-threatening disorder. Electrocardiograms are considered to be sensitive indicators of the presence of hyperkalemia. Since the treatment of hyperkalemia involves relatively few maneuvers and because its success can be objectively scored, we investigated how physicians manage this disorder and how successful their prescribed therapy is. We also sought to determine whether treatment could be improved by providing the treating physicians with therapy guidelines on a real-time basis.
Consecutive patients with hyperkalemia were identified by review of laboratory records. During the observation-only phase of the study, demographic data, contributing causes, electrocardiogram findings, treatments used, compliance with prescribing guidelines, and patient outcome were recorded. During the subsequent notification phase of the study, treatment recommendations were sent to the patient's ward when the elevated potassium value was noted. The same outcome data were collected.
There were 127 episodes of hyperkalemia during the observation-only phase and 115 during the notification phase. No patients died or had life-threatening cardiac arrhythmias. Electrocardiographic abnormalities consistent with hyperkalemia were observed in only 14% of episodes. Renal failure (77%), drugs (63%), and hyperglycemia (49%) contributed to most episodes. Treatments used were exchange resin (51%), insulin (46%), calcium (36%), bicarbonate (34%), and albuterol (4%). The agents were equally efficacious. The time to first treatment was shorter in patients with potassium levels of 6.5 mmol/L or more than in patients with lower values (2.1±2.2 vs 2.8±2.4 hours; P<.05). Treatment was better in the intensive care unit than on regular wards. Only 39% of episodes during the observation-only period met the predetermined criteria for monitoring and diagnosis, initial treatment, and follow-up. During the notification period, physician performance was no better; only 42% of episodes met all criteria. The laboratory transmitted a copy of the guidelines to the patient's ward only 38% of the time. In a separate analysis of these episodes, there was no improvement in treatment. Physicians who did not receive the notification fulfilled all treatment criteria more often than physicians who did (50% vs 30%; P<.05).
Although treatment of hyperkalemia was frequently suboptimal, no serious arrhythmias and no deaths complicated management of 242 episodes of severe hyperkalemia. A narrowly targeted effort to improve physician management of a disorder with discrete treatment options did not improve therapy.
HYPERKALEMIA is a frequent occurrence in hospitalized patients, with a reported incidence of 1.1 to 10 patients per 100 hospitalized.1- 3 Without warning, hyperkalemia may cause nearly any dysrhythmia.4 Estimates of deaths caused by hyperkalemia in the general population are unavailable; in patients with end-stage renal disease in the United States, hyperkalemia accounted for 1.9% of mortality in 1993.5 Higher rates have been reported in a comparable European population.6
Accepted treatments for hyperkalemia include (1) stabilization of electrically excitable membranes by administration of calcium; (2) shift of potassium from the extracellular to the intracellular compartment by means of sodium bicarbonate, insulin, or albuterol; and (3) removal of potassium from the body by sodium polystyrene exchange resins or dialysis. The efficacy of alkali therapy has been questioned,7- 10 and current treatment recommendations differ from those promulgated 10 years ago.11- 14
Renal failure and drugs are the principal predisposing factors for development of hyperkalemia.1,3,15,16 More widespread use of angiotensin-converting enzyme inhibitors and immunosuppressants, such as cyclosporine and tacrolimus, have undoubtedly altered the spectrum of causes of hyperkalemia.
Because hyperkalemia is a common, potentially lethal disorder whose causes and treatment have undergone recent change, we thought it worthwhile to revisit the causes and current treatment practice for this disorder. In our study, we prospectively examined a consecutive series of hyperkalemic patients at a general medical-surgical hospital. We ascertained which factors contributed to the development of hyperkalemia and evaluated how well prescribed therapy adhered to published guidelines for timeliness, choice of therapy, and monitoring. In the second phase of the study, these guidelines were distributed to the treating physicians at the time hyperkalemia was first identified in their patients. We then determined whether this notification improved compliance with the guidelines or led to a better outcome. Our results suggest that the treatment of hospital inpatients with hyperkalemia frequently fails to meet accepted criteria and is unimproved by the provision of management recommendations to the responsible physician at the time of diagnosis.
Using clinical chemistry laboratory records, we prospectively identified all adult (≥16 years of age) inpatients at the University of Pittsburgh Medical Center with severe hyperkalemia during the period February 15 through June 30, 1996. University of Pittsburgh Medical Center is a medical-surgical teaching hospital with 703 licensed beds. Although the upper limit of normal for serum potassium level in the hospital laboratory is 5.0 mmol/L, we enrolled only patients with potassium levels at 6.0 mmol/L or more. This level corresponded to the "critical" level at which hospital policy required telephone notification of the responsible physician on an emergency basis. This enrollment level was also selected because there is general agreement that hyperkalemia of this severity requires prompt attention.14
The study was conducted in 2 phases, which were timed to avoid the learning curve seen at house staff changeover on July 1; this permitted comparisons of physician cohorts with similar baseline knowledge of the management of hyperkalemia. During the initial observation-only phase of the study (9.5 weeks), the therapy and management for each patient was determined by the patient's primary service, without intervention by us. During the subsequent notification phase (10 weeks), recommendations for the management and treatment of hyperkalemia were sent by telefacsimile from the clinical chemistry laboratory to the nursing station on the ward where the patient was housed. Unit clerks were instructed to place these guidelines in the front of the patient chart. These guidelines (Table 1) had been developed by us at the request of the hospital quality assurance program and had last been disseminated 212 years earlier in a physician newsletter. Notification was endorsed by and implemented with the assistance of the hospital's Total Quality Council, its principal continuous quality improvement policy board. The study protocol was submitted to the University of Pittsburgh Institutional Review Board for Biomedical Research for review.
Recurrent episodes of hyperkalemia were considered to be separate events when they occurred more than 24 to 48 hours apart with an intervening serum potassium value less than 5.5 mmol/L. A total of 207 episodes of hyperkalemia were identified during the observation-only interval and 191 during the notification phase. Of these, 156 episodes were excluded because of specimen hemolysis (60 in the observation-only phase and 55 in the notification phase) or withdrawal of therapy for terminally ill patients (3 and 3, respectively), as well as hyperkalemia during reversal of cardioplegia after cardiopulmonary bypass or during reperfusion after hepatic bypass during liver transplant surgery (17 and 18, respectively). Therapy for postbypass hyperkalemia was usually given preemptively, before hyperkalemia was actually documented. We believed this specialized occurrence was not pertinent to the care of patients with usual causes of hyperkalemia. Thus, the study comprised 127 episodes in 104 patients during the observation-only period and 115 episodes in 102 patients during the notification phase.
The medical records of all study patients were evaluated by 1 reviewer (C.G.A.) with the use of prospectively established criteria. The history was reviewed for potential causative factors, including administration of an external potassium load, release from the intracellular pool, and renal failure, as well as other causes of impaired excretion. Values for serum electrolytes, serum urea nitrogen, creatinine, glucose, complete blood cell count, arterial pH, and drug levels (tacrolimus, cyclosporine, and digoxin) were recorded when obtained by the patient's physicians. Drugs were considered potentially contributory if administered within 24 to 36 hours before the onset of hyperkalemia. Drugs that induce hyperkalemia by impairing renal excretion were not considered contributory if the patient was already dialysis dependent. Hyperglycemia was deemed a contributing factor when the blood glucose level was 19 mmol/L (350 mg/dL) or greater at the time of the hyperkalemic event. Inadequate removal of potassium during hemodialysis because of access recirculation was recorded when an appropriate evaluation was performed within 72 hours of the index event. Electrocardiographic (ECG) monitoring, either at the time of the event or as a continuous modality, was identified. For 22 observation-only and 31 notification episodes, baseline 12-lead ECGs obtained before or after the hyperkalemic episode and within 4 to 16 hours of a serum potassium value of 5.0 mmol/L or less were available along with ECGs obtained at the time of hyperkalemia ("paired ECGs"). For 14 observation-only and 5 notification episodes, only an ECG obtained with hyperkalemia was available ("solitary ECGs"). The 12-lead tracings obtained with the hyperkalemic episodes were examined for the presence of rhythm disturbances, peaked T waves, or prolongation of PR (>0.20 seconds) or QRS (>0.12 seconds) intervals.
For purposes of assessment of its adequacy, the management of hyperkalemia was divided in 3 phases: (1) determination of the cause and initiation of appropriate monitoring for life-threatening complications, (2) initial treatment, (3) and follow-up and subsequent therapy. Adequacy of management for each phase was based on the criteria in Table 2 and was judged by consensus of the investigators with pass/fail grading. The minimum requirements for a passing grade for each phase are listed in Table 2. Timing of events was determined by review of the computerized database for laboratory values and ECG acquisition as well as chart review.
Data were expressed as mean±SD and analyzed by means of the t test or χ2 testing (NCSS, Kaysville, Utah, or GreyMatter International, Cambridge, Mass, respectively). Results were considered significant when P<.05. Analysis of patients in the observation-only or notification group was based on intention to treat.
The distribution of potassium values in the observation-only and notification groups is shown in Figure 1. The mean serum potassium values were 6.4±0.5 and 6.5±0.6 mmol/L, respectively. The serum potassium value was 6.5 mmol/L or more in an equal fraction of observation-only (39/127 [31%]) and notification (43/115 [37%]) episodes. The mean frequencies of hyperkalemic episodes were 1.94±1.47 per day during the observation-only phase and 1.61±1.40 per day during the notification phase. As the hospital census was 556±38 (range, 466-634) and 477±45 (range, 400-551) during the 2 phases, respectively, the incidence of new cases of hyperkalemia was similar during the 2 periods at 2.4% for the observation-only phase, 2.1% for the notification phase, and 2.3% overall.
Patients frequently had multiple causes for hyperkalemia (Table 3). As expected, renal failure was present in the majority of patients. Hyperglycemia was the second most common contributor to hyperkalemia. Overall, drugs contributed to development of hyperkalemia in 63% of cases, 81 (64%) of 127 for controls and 72 (63%) of 115 intervention patients. The medications most often implicated were the immunosuppressive drugs tacrolimus and cyclosporine; less commonly, angiotensin-converting enzyme inhibitors, β-blockers, digoxin, and nonsteroidal anti-inflammatory drugs were responsible.
No patient deaths were directly attributable to hyperkalemia and no patient suffered a significant morbid event, such as exchange resin or sorbitol–related intestinal necrosis or insulin-related hypoglycemia.
During 85 (83%) of observation-only episodes and 76 (75%) of notification episodes, patients were placed on ECG monitors. As there was a lower threshold for concluding that T waves were peaked when a comparison ECG was available, there was a tendency toward the more frequent occurrence of peaked T waves among the patients for whom paired 12-lead ECGs were available (22/53 ) when compared with patients for whom only solitary 12-lead tracings were available (4/19 ). However, this difference was not significant, and there were no differences between patients in the observation-only or notification phases. Consequently, ECG results were pooled for analysis. Cardiac dysrhythmias noted with the initial ECG included first-degree atrioventricular block in 8 episodes (11%) and a junctional rhythm in 3 (4%). Six patients had a widened QRS complex (8%), and 26 had peaked T waves (36%). Overall, 33 tracings (46%) had any change suggestive of hyperkalemia, on the basis of physician review. The prevalence of ECG changes was independent of the severity of hyperkalemia. In patients with potassium values less than 6.8 mmol/L, 23 (43%) of 54 tracings showed changes, whereas 10 (55%) of 18 tracings in patients with higher potassium values had characteristic hyperkalemic changes. Computer-generated interpretations were not useful in diagnosing hyperkalemia. In the 33 episodes in which they were available, one suggested that "an electrolyte imbalance" might be present; no other report made reference to hyperkalemia.
The modalities chosen for medical management of hyperkalemia are shown in Table 4 and Table 5. Most patients received therapy or combinations of therapy that included insulin, sodium polystyrene resin, or calcium. Sodium bicarbonate was used in more than 25% of episodes; less than 10% were treated with albuterol. On a total of 99 occasions, only 1 modality was used (54 and 45 episodes in the observation-only and notification phases, respectively): insulin (4 and 8 episodes), albuterol (0 and 3), sodium bicarbonate (4 and 2), exchange resin (23 and 22), and dialysis (23 and 9). In addition, monotherapy with calcium was administered to 1 patient in the notification phase. The average number of drugs used per patient in the observation-only and notification phases was 1.7±1.2 and 1.9±1.2 (P=.09), respectively. Patients with more severe hyperkalemia received calcium salts, insulin, or sodium bicarbonate more frequently than patients with less severe hyperkalemia (Table 5). There was no difference in use of medications between the 2 phases except for a greater use of sodium bicarbonate at potassium levels of 6.5 mmol/L or more (Table 5).
We wanted to determine whether the decision to use dialysis as definitive therapy might delay institution of temporizing methods. Therefore, we separately analyzed times to first treatment maneuver in dialyzed and nondialyzed patients. In the observation-only phase, the time to first medical treatment was 3.4±3.8 hours vs 2.7±2.4 hours for dialyzed vs nondialyzed patients (P=.11). In the notification phase, the times were 3.1±3.0 vs 2.4±2.3 hours (P=.11). These trends did reach statistical significance when the groups were pooled: 3.3±3.5 vs 2.6±2.3 hours (observation-only vs notification, P=.03). Times to therapy for first drug and individual drugs are shown in Table 5 for patients who did not receive dialysis. There was no significant difference in times to therapy when the observation-only and notification phases were compared or when the patients with potassium values less than 6.5 mmol/L were compared with those with more severe elevations of serum potassium level. When analyzed according to time to first medication given, however, patients with worse hyperkalemia overall received treatment at a shorter interval. Patients in the intensive care unit (ICU) also received faster treatment than patients on the ward (1.3±1.4 hours vs 3.4±2.4 hours; P<.001). A similar difference was observed during both study phases when analyzed separately.
Each of the maneuvers used had similar success. Among patients who received only 1 agent, initial follow-up potassium values were 5.5 mmol/L or less after therapy with insulin in 8 (75%) of 12 instances; sodium bicarbonate, 3 (50%) of 6; exchange resin, 25 (56%) of 45; and albuterol in 2 (67%) of 3. For the 6 patients who received monotherapy with sodium bicarbonate, the mean potassium level noted at initial follow-up was 5.1±1.3 mmol/L. The results in the 32 patients who were treated with dialysis alone and who had a follow-up potassium value were similar; in 19 (59%) the first potassium value obtained during or after dialysis was 5.5 mmol/L or less.
Results of the analysis of clinician performance are shown in Table 6. In the observation-only period, only 39% of episodes met the criteria in all 3 areas assessed: diagnosis, treatment, and follow-up. The pass rate was higher in patients whose hyperkalemia developed in an ICU than in patients on regular hospital wards. Only 4% of the ICU patients failed to meet criteria for diagnosis. This principally reflects the routine use of continuous ECG monitoring in ICU patients. Even allowing for this difference, care on the ward was less satisfactory. The ward fail rates for the treatment and follow-up phases were more than double the rates for the ICU. In the notification phase, similar differences between ICU and ward care were observed. There were no differences in how any of the 3 segments scored when the observation-only and notification phases were compared.
The comparison of observation-only and notification phases was based on intention-to-treat analysis. To examine the efficacy of notification more carefully, a separate analysis was carried out for those patients whose physicians actually received the treatment suggestions. In only 44 (38%) of 115 notification episodes was the facsimile notification transmitted to the patient ward. Physicians who received the notification received a passing score for all 3 segments in 13 (30%) of 44 instances. Physicians who did not receive the notice did better; they were successful in 35 (50%) of 71 instances (P<.05, χ2).
In this study, we prospectively examined the physician diagnosis and management of hyperkalemia. We determined what caused the hyperkalemia, how it was treated, and whether the treatment was conducted appropriately. To our knowledge, this is the first study to grade hyperkalemia treatment efforts before and after use of an intervention designed to enhance compliance with published treatment guidelines.
The causes of hyperkalemia identified in this study mirror those noted in earlier studies, although differences in classification methods preclude direct comparison. Previous reports tended to assign a single cause for the elevation of serum potassium level.1- 3,15 We found that hyperkalemia was usually of multifactorial origin; assignment of only a single cause would usually have been arbitrary. Nonetheless, it is clear that renal failure, which underlay 77% of our episodes, contributed in a similar majority in other studies.2,3,15 Shift of potassium from cells because of hyperglycemia or insulin lack in diabetes mellitus also featured prominently, occurring in 49% in our patients and 6% to 38% previously.2,3,15 Drugs were a factor in 63% of our patients and 27% to 69% in other studies.1- 3,15 Potassium supplements contributed in 15% of our patients, substantially less than the 31% to 40% previously noted.1- 3 Potassium-sparing diuretics, digoxin, ß-blockers, and angiotensin-converting enzyme inhibitors were other minor causes, accounting for 12% to 14% of our cases. Tacrolimus and cyclosporine were the most commonly implicated drugs in our series. Pentamidine caused only 2 episodes. These findings are likely the result of the emphasis on transplantation and the relative paucity of human immunodeficiency virus–infected patients at our institution.
Previous reports have suggested that ECGs are a highly sensitive tool for detecting severe hyperkalemia.17 Twelve-lead ECGs were available in only 30% of episodes in the present study, limiting the scope of our conclusions. Of these, however, only 46% displayed typical hyperkalemic changes. Only 1 computer-generated ECG reading was consistent with hyperkalemia. In 1 report, all patients with potassium values above 6.8 mmol/L had hyperkalemic changes, although those with intermediate elevations had a lower prevalence of ECG changes.18,19 Only 56% of tracings obtained in our patients with like potassium values were abnormal. This is consistent with more recent observations emphasizing the insensitivity of ECGs as a tool to diagnose even profound hyperkalemia.20
A principal goal of our study was to determine how hyperkalemia was managed. This question has been examined previously, in a study that determined what treatments would be recommended by nephrology training program directors. That study did not provide actual data on current practice.11 The hypothetical patient for whom treatment recommendations were solicited had peaked T waves and a widened QRS complex, and the initial treatments recommended were calcium (60%) and sodium bicarbonate (32%). For resistant hyperkalemia, insulin with glucose (34%) was advised. In our study, few episodes were accompanied by hyperkalemic ECG changes, and our results are therefore more reflective of a pattern of treatment targeted at lowering the serum potassium value rather than mitigating cardiotoxic effects. At our hospital, 51% of episodes were treated with exchange resins, 46% with insulin, 36% with calcium, 34% with sodium bicarbonate, and 4% with albuterol. Studies from several groups have questioned the efficacy of sodium bicarbonate in patients with end-stage renal disease.7- 9,21- 23 Their findings are reflected in the evolution of treatment recommendations.14,24 Nonetheless, in our patients, there was no difference when the efficacy of monotherapy with sodium bicarbonate was compared with that of insulin. However, our patients differed in 2 important respects from those in the studies mentioned above; the patients did not all have end-stage renal disease and they were hyperkalemic. The studies that have shown a lack of potassium-lowering effect of sodium bicarbonate were all performed in normokalemic subjects with end-stage renal disease. These may be important distinctions.25,26
Recent studies have shown inhaled albuterol to be highly efficacious in lowering serum potassium level, an effect augmented by coadministration of insulin, but not bicarbonate.27- 29 Our guidelines were developed before authoritative recommendations adopted the use of albuterol,24 and we did not emphasize its use in lieu of sodium bicarbonate. As such, the latter was selected in 34% of episodes while albuterol was used in only 4%. As might be expected, higher levels of potassium often prompted more aggressive treatment. Drug use for each modality was higher when the serum potassium level exceeded 6.5 mmol/L. In addition, treatment was started more than 30 minutes earlier in the high-potassium group.
An additional goal of our study was to determine how well clinicians adhered to the hospital standard for treatment (Table 1) and whether adherence could be improved by providing clinicians with a timely synopsis of these recommendations. Consequently, we scored 3 aspects of the care of each episode: diagnosis, initial treatment, and follow-up. The results were sobering. In the initial and purely observational phase of the study, only 39% of episodes satisfied the published criteria in all 3 phases. Treatment was more attentive in the ICU, where two thirds of all episodes were treated appropriately; only a quarter of ward cases were similarly well managed. Times to treatment were also shorter in the ICU. Overall, however, treatment of patients with more severe hyperkalemia did not have a higher pass rate than treatment of patients with a lesser degree of hyperkalemia.
In the second phase of our study, recommended guidelines for treatment were to be faxed to the responsible house officer at the time the laboratory first detected critical hyperkalemia. Although extensive efforts to coordinate the notification system were undertaken, its implementation was disappointing. We could confirm that the faxes reached the ward in only 38% of episodes. Based on intention-to-treat analysis, the intervention did not improve any aspect of management. It seems doubtful that more complete notification would have improved outcome, however, as a separate analysis comparing episodes in which notification occurred with those where it did not disclosed that treatment was better when the reminder was not received on the ward.
In a previous study, we showed that hypernatremia was often iatrogenic and that its management was frequently suboptimal.30 While we suggested that treatment could be improved with physician education efforts, traditional educational efforts such as lectures or promulgation of guidelines have a variable success rate. They may certainly improve treatment,31,32 but providing physicians with patient-specific recommendations is more effective.33 Other such system approaches may be applied.34,35 These include reduction of reliance on physician memory, improved information delivery, standardization of treatment, and forcing systems. Examples of successful use of the last have been reported.
By delivering therapeutic guidelines to clinicians at the time treatment was indicated, our method focused on reduction of reliance on memory. Although its principal virtue was ready acceptance, results of a meta-analysis also supported this approach. Guidelines are most likely to be effective when they are of low complexity and easy to implement.36 Although forcing systems could have been designed, imposing pharmacy restrictions on administration of nonsteroidal anti-inflammatory drugs or angiotensin-converting enzyme inhibitors to patients with elevated serum creatinine levels had the potential to be too restrictive. They would have little effect on hyperkalemia induced by hyperglycemia or immunosuppression. Our results highlight the pitfalls of reliance on a time-sensitive notification program. Despite a dedicated research and quality improvement effort focused on a very discrete problem that should have been conspicuous to treating physicians because of its potential to be life threatening, our intervention was not successful. In part, it failed because the hospital information system proved ill suited to deliver the message. This deficiency is amenable to correction by automation. However, it is unclear that an improvement in hyperkalemia treatment would result, since physicians did not reliably follow the advice even when they received it. Implementation of standing orders for hyperkalemia management seems inappropriate since the causes and required treatments vary widely. An alternative would be to have a physician or pharmacist contact the treating physician directly. Whether this labor-intensive and therefore expensive approach will succeed is unknown. The present study demonstrates a method for testing it. Additional studies would also provide an opportunity to confirm that so few patients with severe hyperkalemia develop cardiac arrhythmias. On the basis of this study alone, we believe it would be premature to advocate less vigilance for hyperkalemia or reliance on a less vigorous treatment approach.
Accepted for publication September 12, 1997.
This study was presented in part at the Annual Meeting of the American Society of Nephrology, New Orleans, La, November 5, 1996.
Reprints: Arthur Greenberg, MD, Division of Nephrology, Duke University Medical Center, Box 3014, Durham, NC 27710.