View larger version (13K): [in this window] [in a new window]   Figure 1. Relation between admissions to the intensive care unit (ICU) (bar graph) and the heat index (line chart) during the height of the heat wave. Admissions increased 24 hours after the peak heat index on 13 July 1995. The peak admission rate occurred 2 days later, on 15 July 1995. Intensive care unit admissions coincided with heat-related deaths reported by the Cook County Medical Examiner's office [1].

Heat stroke has been classified as exertional or classic. Exertional heat stroke is precipitated by heavy exertion in very hot and humid climates and is usually seen in otherwise healthy young persons [3]. Classic heat stroke results from unabated exposure to high temperatures and humidity. Elderly persons with premorbid conditions are likely to experience classic heat stroke. Despite the high incidence of heat-related deaths reported annually in the United States [4], the literature on the clinical features of classic heat stroke is limited. Lactic acidosis, renal failure, rhabdomyolysis, and disseminated intravascular coagulation, well described in exertional heat stroke, are reported as uncommon or of minor consequence in classic heat stroke [3, 5-7].

The numerous persons who presented with near-fatal classic heat stroke during the Chicago heat wave provided an opportunity to analyze the clinical course of this condition. We sought to describe the course of patients admitted to the intensive care unit with classic heat stroke. Excessive functional disability identified in survivors of the acute phase of the illness prompted a 1-year follow-up to assess delayed functional outcome and mortality.

Methods

Top Methods Results Discussion Author & Article Info References

Medical Record Review

Twelve of 24 area hospitals permitted us access to patient records. We obtained institutional review board approval, which allowed us to contact patients for 1-year follow-up. Patients were identified by review of emergency department records, intensive care unit logs, and medical records discharge codes. Records of all patients admitted to each intensive care unit from 12 July to 20 July 1995 were reviewed. Enrollment criteria were 1) absence of vigorous physical activity during the heat wave; 2) alteration of mental status [coma, delirium, lethargy, disorientation, or seizures documented by paramedics and emergency department physicians]; 3) recorded core body temperature of more than 40.6 °C or documented evidence of cooling before the first recorded temperature and a reliable history of compatible environmental exposure [1, 5, 8, 9]; and 4) hot, dry, flushed skin.

Data were recorded by using a predesigned data collection form. Major components of this form were demographic characteristics, Glasgow coma score, temperature, and clinical condition at presentation; maximum Acute Physiology and Chronic Health Evaluation (APACHE) II score in the first 24 hours; length of intensive care unit and hospital stay; functional level and disposition at discharge; laboratory data; and results of radiography, electrocardiography, and echocardiography. Data on the presence or absence of air conditioning was obtained from paramedic encounter sheets. Information about when the patient was last seen was obtained from paramedic reports or emergency department history. The medical history was obtained from admission history and physical examination and was verified by the medical records of the patient's personal physician, when available, or outpatient prescription medication, as recorded on the paramedic report.

Organ system dysfunction was assessed by evaluation of serial laboratory results. The chart reviewer assessed neurologic impairment at discharge from documentation of the patient's neurologic status at discharge.

Follow-up

A modified Stanford Health Assessment Questionnaire was used to score functionality [10, 11]. Components of this tool assess performance of activities of daily living. Survivors were contacted 1 year after discharge and were interviewed by telephone to complete the modified Stanford Health Assessment Questionnaire. If the patient was unable to respond, a caretaker estimated their level of function by using the same questionnaire. If telephone contact could not be made, tracking was attempted through mail, health care providers, or acquaintances identified in the medical record. Deaths were investigated in the aforementioned manner and by a search of the Cook County Medical Examiner logs. Data that allowed completion of the modified Stanford Health Assessment Questionnaire at the time of hospital discharge were abstracted from the charts of patients whose outcomes were known at 1 year.

Data Analysis

Data are presented as the mean ±SD except when they did not fit tests for normality and the mean did not accurately describe the data. In these instances, data are presented as the median and interquartile range. We used n to denote the number of values averaged when data were not available on all 58 study patients. Nominal data were compared by using chi-square analysis or the Fisher exact test of contingency tables. Continuous data were compared and correlations were made by using analysis of variance, the t-test, the Spearman rank test, and the Wilcoxon signed-rank test. Admission laboratory values and follow-up values were compared, and correlations were performed between laboratory tests to determine whether abnormalities were related. To determine the effect of functional disability and age on survival, survival analysis was performed by using Kaplan-Meier plots with the Mantel-Cox log-rank test and the Cox proportional-hazards model. Statistical analysis was performed by using StatView 4.5 (Abacus Concepts, Berkeley, California).

Role of Funding Source

Neither of our funding sources, the Park Ridge Health Foundation or the University of Chicago Clinical Research Center, participated in collection, analysis, and interpretation of the data or in preparation of the manuscript. The decision to publish this report was not influenced by the funding sources.

Results

Top Methods Results Discussion Author & Article Info References

Epidemiology

Fifty-eight patients met the inclusion criteria. Patients who met the case definition of classic heat stroke but were not admitted to the intensive care unit were excluded from further analysis.

Demographic characteristics of the patients are presented in Table 1. Forty-seven patients (83%) were found at home, and 77% were last seen within 12 hours of rescue (range, 2 to 42 hours; n = 39). The environments in which patients were found lacked air conditioning in 40 instances (69%). Three patients had air conditioners that were turned off, and 1 patient had the unit on at the time of rescue. In 14 cases, the presence or absence of air conditioning was not documented. Forty-three of 50 patients had medical insurance (Medicare [44%], private insurance [18%], private insurance and Medicare [14%], Medicaid [10%], and self-pay [14%]). Insurance data were not available for eight patients.

Chronic medical conditions were documented in 52 patients (Table 2). The most common comorbid conditions were hypertension and alcohol abuse. Before admission, 47 patients were taking at least 1 of 18 medications (Table 2). Sixteen patients were found to be taking medications that predispose to the development of heat stroke (diuretics [13 patients], phenothiazines [2 patients], or both [1 patient]). An additional 2 patients tested positive for cocaine, a substance known to increase heat production.

Presenting temperature, time to cooling, and APACHE II scores are shown in Table 3. Methods of cooling in the emergency department included antipyretic agents (15 patients), water or fan (24 patients), groin and axillary ice packs (34 patients), cooling blanket (19 patients), gastric lavage with cool saline (6 patients), cold intravenous fluids (17 patients), and ice bath (1 patient). The median hospital stay was 9 days (interquartile range, 13 days), and the median intensive care unit stay was 4 days (interquartile range, 5 days). Twelve patients died (20.7%).

Organ System Dysfunction

Cardiovascular Findings

Presenting electrocardiography showed sinus tachycardia (71%), normal sinus rhythm (8.5%), supraventricular tachycardia (8.5%), atrial fibrillation (8.5%), and multifocal atrial tachycardia (3%). Hypovolemic shock was suggested by physical examination in 17 of 58 patients. In the first 24 hours, 3.8 ± 2.1 L (range, 0.6 to 13.5 L) of fluid was infused. Six patients had central venous pressure measured within the first 24 hours; the mean was 12 ± 6 cm H₂O (range, 7 to 20 cm H₂O).

Fourteen of 43 patients underwent echocardiography early in the hospital course. Among these 14 patients, findings included normal left ventricular systolic function (7 patients), concentric left ventricular hypertrophy (4 patients), mild left ventricular dysfunction (1 patient), moderate to severe left ventricular dysfunction (4 patients), left-sided chamber enlargement (4 patients), and mitral regurgitation (3 patients). Six patients had one of the following findings: right ventricular hypertrophy, aortic insufficiency, tricuspid regurgitation, pericardial effusion, paradoxical septal-wall movement, and aneurysm. Two patients underwent follow-up echocardiography; one had normalization of previously noted moderate to severe left ventricular dysfunction and resolution of pericardial effusion. This patient also underwent cardiac catheterization, which showed normal coronary vasculature. The second patient showed interval development of severe apical hypokinesia with new wall-motion abnormalities.

Two patients underwent early monitoring with pulmonary artery catheters. In both, the pulmonary capillary wedge pressure was elevated (>20 mm Hg). In one of the two patients, cardiac output was high (7.2 to 7.7 L/min) and systemic vascular resistance was low (311 to 433 dynes/sec · cm⁵); in the other, cardiac output was low (2.0 to 3.0 L/min) and systemic vascular resistance was elevated (1508 to 2330 dynes/sec · cm⁵).

Pulmonary Findings

Thirty-five patients (60%) required intubation and mechanical ventilation for a median of 5 days (interquartile range, 7 days). The most common acid-base abnormalities were mixed non-anion gap metabolic acidosis and respiratory alkalosis (12 of 49 patients) and mixed positive anion gap metabolic acidosis and respiratory alkalosis (10 of 49 patients).

Initial chest radiography showed no active pulmonary disease in 29 of 54 patients. Eight patients had atelectasis, 8 had increased pulmonary vascular markings, 7 had infiltrates, and 2 had pleural effusions. Although 16 patients had a PaO₂:FIO₂ ratio less than 200, only 6 of these patients met the radiographic criteria for the acute respiratory distress syndrome. One of these patients had a pulmonary capillary wedge pressure greater than 18 mm Hg.

Renal Findings

Twenty-four patients developed moderate renal insufficiency (creatinine concentration 177 µmol/L) and 7 developed severe renal insufficiency (creatinine concentration 309 µmol/L) [12]. Of these 31 patients, 6 died, 18 recovered fully, 2 improved, and 5 had no follow-up data available. Although creatine phosphokinase levels were elevated on admission (11.0 ± 16.0 µkat/L [range, 0.75 to 101.0 µkat/L]; n = 46), the peak did not occur until the third hospital day (70.0 ± 139.0 µkat/L [range, 0.75 to 540 µkat/L]). Increased creatine phosphokinase levels were primarily due to skeletal muscle fraction according to measurement of isoenzymes. Six of 10 patients tested were positive for urine myoglobin, including 2 patients with creatinine concentrations of at least 707 µmol/L. Creatine phosphokinase values were not higher in patients with hypokalemia (potassium 3.2 mmol/L) on admission (unpaired t-test, P > 0.2; mean difference, 31.5 µkat/L [95% CI, –51.5 to 114.5 µkat/L]. No correlation was found between admission potassium levels and high creatine phosphokinase levels (Spearman rank correlation coefficient, = 0.05; P > 0.2). A correlation was found between the high creatinine and creatine phosphokinase ( = 0.50; P < 0.001). One patient was excluded from the analysis because the result of a urine toxicology screening test was positive for cocaine, a substance reported to be associated with rhabdomyolysis. On admission, the mean blood urea nitrogen-to-creatinine ratio was 5.0 ± 3.0 mmol/L (range, 1.4 to 21.1 mmol/L). No correlation was found between high creatinine concentrations and the admitting mean arterial pressure ( = –0.09;P > 0.2) or blood urea nitrogen-to-creatinine ratio ( = –0.22;P = 0.11).

Gastroenterologic Findings

Admission values of liver function tests were normal, except for mild elevations of aspartate aminotransferase, total bilirubin, and lactic dehydrogenase levels. Levels of aspartate aminotransferase, alanine aminotransferase, -glutamyl transpeptidase, total bilirubin, and lactic dehydrogenase increased during hospitalization; peak values occurred on day 3. Diarrhea or melena was present on admission in 10 patients. One patient, who had a history of recent antibiotic use, had a positive culture for Clostridium difficile. Another patient underwent colonoscopy and had clinical and pathologic findings consistent with focal ischemic colitis.

Infectious Disease Findings

Forty sites of infection were present in 33 patients. Only specimens obtained within the first 24 hours of admission were evaluated and, therefore, were accepted as concurrent with the diagnosis of heat stroke. Infection was defined as a positive blood or urine culture, a catheterized urine specimen with pyuria and bacteruria, a positive sputum culture with a concurrent positive chest radiograph, or substantial physical or radiographic evidence of closed-space infection. The sites of infection were blood (16 patients), urinary tract (15 patients), lungs (3 patients), skin (2 patients), sinuses (2 patients), and colon (1 patient). Blood cultures grew Staphylococcus species (13 patients), Fusobacterium organisms (2 patients), and Salmonella organisms (1 patient). Organisms cultured from the urinary tract included Escherichia coli (4 patients), Staphylococcus species (2 patients), enterococci organisms (2 patients), Citrobacter organisms (1 patient) and Streptococcus group B organisms (1 patient). Urinary tract infection was diagnosed by positive urinalysis in five patients with no culture results available. Sputum cultures showed Staphylococcus species (2 patients) and Klebsiella organisms (1 patient). In 1 patient, C. difficile grew in the gastrointestinal tract. Soft-tissue cultures were positive for Staphylococcus species (1 patient) and Pseudomonas organisms (1 patient). Three patients had evidence of sinusitis on axial computed tomography (CT). One patient with no identifiable bacterial infection tested positive for HIV.

Hematologic Findings

The median admission platelet count was 157 x 10⁹ cells/L (interquartile range, 87 x 10⁹ cells/L [n = 56]). Follow-up platelet counts decreased to 61 x 10⁹ cells/L (interquartile range, 83 x 10⁹ cells/L [n = 49]; Wilcoxon signed-rank P < 0.001 compared with admission counts), with the nadir occurring, on average, on day 3 of hospitalization. A total of 39 patients had a decrease in platelet count of more than 25%, and 30 of 39 had a nadir count of less than 80 x 10⁹ cells/L. Twenty-six of 39 patients (45%) had evidence of disseminated intravascular coagulation, defined as a more than 25% decrease in platelet count and at least one of the following conditions: prothrombin time more than 25% of normal, increased fibrin degradation products, or increased D-dimer value. Six of 39 patients had a negative evaluation for disseminated intravascular coagulation, and 7 of 39 patients were not evaluated. All 5 patients who required platelet transfusions had evidence of disseminated intravascular coagulation. An increased mortality rate was seen in patients with disseminated intravascular coagulation (chi-square P = 0.045; relative risk by odds ratio, 4.15). Four of 5 patients who met the criteria for the acute respiratory distress syndrome had evidence of disseminated intravascular coagulation compared with 26 of the total group (Fisher exact test, P = 0.16; relative risk by odds ratio, 5.61).

The mean hemoglobin value at admission was 8.0 ± 1.43 mmol/L (range, 5.27 to 10.97 mmol/L). Eight patients required erythrocyte transfusion during the first week. Seven patients had leukocytosis on admission (>20 x 10⁹ cells/L). The leukocyte count was not higher in patients with documented infection (unpaired t-test, P > 0.2; mean difference, 0.931 x 10⁹/L [CI, –4.18 to 2.32 x 10⁹/L]).

Neurologic Findings

On admission, patients had the following alterations in mental status: coma (33 patients), delirium (2 patients), lethargy (8 patients), disorientation (6 patients), and seizures (9 patients). Neurologic findings included flaccid muscle tone (14 patients), hyperreflexia (1 patient), hypertonia (3 patients), and posturing (2 patients). None of the patients with hyperreflexia or hypertonia were taking any medications; however, one had a positive result on a urine screening test for cocaine. Neuroradiologic imaging was performed in 35 patients. On the 34 axial CT scans of the brain, findings included atrophy (15 patients), old cerebral infarction (6 patients), old cerebellar infarction (2 patients), cerebral edema (1 patient), ischemic demyelination (1 patient), and periventricular white matter disease (1 patient). Twelve CT scans were normal, and follow-up CT scans were unchanged in 6 patients. Magnetic resonance imaging in 5 patients showed ischemic changes (2 patients), atrophy (2 patients), and suprasellar adenoma (1 patient). Magnetic resonance imaging added information in 2 of the 5 patients who underwent both CT and magnetic resonance imaging. Sixteen patients underwent electroencephalography within 24 hours of admission; no focal abnormalities were noted.

At discharge, 11 patients did not have neurologic impairment (24%), 20 had minimal impairment (43%), and 15 had moderate to severe impairment (33%). A strong correlation was found between age and modified Stanford Health Assessment score of function at the time of discharge (r = 0.59 [CI, 0.271 to 0.793]; P < 0.001; z = 3.32).

Follow-up

One-year follow-up was attempted in all 46 survivors. Fourteen patients could not be located. Five patients were verified as alive but could not be contacted. Contact was made with the remaining 27 patients or their immediate caretakers. Nine of these patients had died within the year; most of the deaths occurred in the first 12 weeks after discharge.

Survival analysis revealed a significant difference in survival between patients who were discharged with minimal or moderate disability scores compared with those who were discharged with severe disability scores (Mantel-Cox chi-square P = 0.04) (Figure 2). Estimation of proportional hazard showed that the relative hazard of death increased 10.1% for each one-point increase in functional score over the 1-year follow-up (regression coefficient, 0.096; chi-square P = 0.03, exponential [coef] per point, 1.10 [CI, 1.010 to 1.198]).

View larger version (10K): [in this window] [in a new window]   Figure 2. Kaplan-Meier plot showing 1-year survival based on the level of functional disability at the time of discharge, as determined by the modified Stanford Health Assessment score. Black circles represent survival in patients with minimal disability, white box represents patients with moderate disability, and black triangles represent patients with severe disability. Patients with severe disability had significantly decreased survival (Mantel-Cox log-rank P = 0.04).

When both in-hospital death and delayed death were analyzed, age was not found to be an independent risk factor despite an increase in the relative hazard of death of 12.7% for each decade of life over the 1-year follow-up (regression coefficient, 0.012; chi-square P > 0.2; exponential [coef] per decade, 1.127 [CI, 0.86 to 1.48]). This result is supported by Kaplan-Meier analysis (Mantel-Cox log-rank chi-square P > 0.2) (Figure 3). Although age was not found to be independently associated with mortality, it was an important covariant with functional score, increasing the relative hazard to from 10.1% to 14.8% for each one-point increase in functional score (the regression coefficient increased from 0.096 to 0.138).

View larger version (10K): [in this window] [in a new window]   Figure 3. Kaplan-Meler plot showing survival from the time of presentation through 1 year of follow-up on the basis of age. Patients in category 1 were younger than 60 years of age (circles); patients in category 2 were 60 to 69 years of age (triangles); and patients in category 3 were older than 70 years of age (squares). Survival did not differ among the groups (Mantel-Cox log-rank chi-square P > 0.2).

The remaining 18 patients were recategorized for current functionality by using the modified Stanford Health Assessment Questionnaire. No patient changed functional category during follow-up (Table 4). The 1-year mortality rate was 28% (9 of 32 patients); this percentage includes patients who were verified as alive but were not contacted.

Discussion

Top Methods Results Discussion Author & Article Info References

Our study of 58 patients with near-fatal heat stroke showed an acute mortality rate of 21% and an additional mortality rate of 28% at 1 year. Functional impairment at discharge persisted at 1 year and was predictive of 1-year mortality. Patients had multiple organ system dysfunction, and a high prevalence of infection was found on admission. Only 1 of 58 patients was cooled within an acceptable time frame (body temperature < 38.9 °C within 30 minutes of presentation to the emergency department).

Heat stroke is the most severe of the heat-related illnesses. It is defined as a condition caused by an excessive increase in body temperature caused by overload or failure of the thermoregulatory system during exposure to heat stress [3, 13, 14]. Reported risk factors, including the use of certain medications, are similar to those described in our patients [3, 14-16]. Most of our patients had medical insurance, suggesting that concern over medical cost may not have been a risk factor for severe heat-related illness. The time frame of emergency department visits, deaths, and hospital admissions during the Chicago heat wave were predictable from reports of other heat stroke epidemics [17, 18].

Previous studies have shown that decreasing the core body temperature to less than 38.9 °C within 30 minutes of presentation improves survival. The accepted methods of cooling are immersion in ice water, immersion in cold water, and use of a body cooling unit [9, 14]. One of 58 patients in our study was cooled within an acceptable time frame. The one patient who received an ice water bath did so after maintaining a core temperature of 41.6 °C for longer than 1 hour. Cooling may have been delayed in our patient population secondary to being sent for CT of the head, which did not contribute to diagnosis or management.

The reported mortality rate and estimated annual number of deaths associated with classic heat stroke vary by data source [4, 19]. We found that patients who required intensive care unit admission incurred substantial additional mortality. Although we were able to substantiate some of the clinical features of classic heat stroke described in the literature, our data are contrary to other reports in terms of the severity of renal and hematologic disease and prevalence of infection.

Tachyarrhythmia and hypotension have been documented in patients with heat stroke [20-22]. Our patient population confirmed these findings. In addition, one patient had reversible myocardial depression, defined by pulmonary artery catheter data and echocardiography. Cardiac catheterization did not find any cause for depressed left ventricular function, and follow-up echocardiography yielded normal findings. Other cases of reversible myocardial dysfunction have been reported in patients with heat stroke [23, 24].

Renal dysfunction, well documented in exertional heat stroke, has been attributed to numerous factors, including direct thermal injury, prerenal insult, rhabdomyolysis, and disseminated intravascular coagulation [15, 25-28]. Unlike the patients in the prevailing literature on classic heat stroke, most of our patients also developed substantial renal insufficiency caused in part by rhabdomyolysis [3, 6, 16, 26, 29-31]. We were unable to identify hypokalemia as a risk factor for rhabdomyolysis. Although creatine phosphokinase levels were not as elevated as those seen in exertional heat stroke, it is important to note that this occurs despite a lack of history of exercise.

Contrary to other reports, we found that the coagulation abnormalities described in exertional heat stroke were also prevalent in persons with near-fatal classic heat stroke [6, 14, 28, 32]. The primary event seems to be related to direct thermal injury of vascular endothelium, initiating platelet aggregation and activation of the clotting cascade along with increased fibrinolysis and disseminated intravascular coagulation [3, 22, 32-37]. Markers of endothelial cell activation and injury have been shown to be elevated in heat stroke [36]. Bone marrow examination has shown a decrease in megakaryocytes, and liver failure may also contribute to decreased production of clotting factors. Our data confirm the association between disseminated intravascular coagulation and mortality [3, 28, 35, 37] and agree with the association between disseminated intravascular coagulation and the acute respiratory distress syndrome, reported elsewhere [38].

A high prevalence of elevated liver aminotransferase levels has been reported in experimental and exertional heat stroke and has been attributed to ischemia and direct thermal injury [20, 39, 40]. It seems that persons with near-fatal classic heat stroke also have elevated aminotransferase levels. Studies have shown that even though thermal stress results in an increased cardiac output, splanchnic flow is reduced because of increased splanchnic vascular resistance [20, 21]. The result may be gut ischemia with focal necrosis, a speculated cause of gastrointestinal disturbances reported by marathon runners who become hyperthermic during extreme exercise stress [41]. We speculate that a similar mechanism contributed to the diarrhea and melena in our patients. Ischemia of the intestinal mucosa was supported by biopsy results in one patient.

Concurrent infection has been suggested as a predisposing factor for the development of heat stroke [3]. Most series of heat stroke cases have not reported culture results [5-7, 15]. The high prevalence of infection at the time of admission among our patients suggests that infections play a role in developing heat stroke, perhaps by contributing to dehydration or thermal load. Heat stroke itself may predispose to early infection through such mechanisms as altered immune status related to heat [42], altered distribution of peripheral blood leukocytes [43], endotoxemia, and bacterial translocation due to gut ischemia [44-46]. Bacterial translocation may have been the source of Salmonella bacteremia seen in one of our patients.

Substantial acute morbidity was incurred as a result of the functional consequences of classic heat stroke; this effect persisted at the time of hospital discharge and 1 year later. We used a modification of the Stanford Health Assessment Questionnaire, a tool for functional assessment in rheumatologic diseases research that has been adapted for use with other diseases [47]. The modified questionnaire focused on functional disability [11]. The ability of this questionnaire to predict mortality at 1 year suggests that patients died of factors related to the degree of disability. The relation between functional disability and mortality has been proposed for other diseases [48-50]. The 1-year mortality rate for heat stroke is similar to the 1-year mortality rate reported for cerebral vascular accidents [51].

In this group of patients with heat stroke necessitating intensive care, we found a new subset of patients with heat-related illness who present differently from previously described patients with classic heat stroke. In some ways, they are more similar to patients with exertional heat stroke, although the severity of organ system dysfunction is not as profound.

Our study had several limitations. Although we attempted to include hospitals that sampled a representative cross-section of Chicago, most of the hospitals that permitted us access were located in south Chicago. In addition, the lack of premorbid data to establish baseline functional status made it difficult to causally relate the residual functional disability found after heat stroke. However, given the resolution of the abnormalities found in other organ systems (such as disseminated intravascular coagulopathy and hepatic and renal abnormalities), it seems reasonable to assume that they were not preexistent premorbid conditions. Despite these and other limitations inherent to retrospective data collection, such as missing data points, our study presents a new perspective on the multisystem dysfunction seen in near-fatal classic heat stroke.

In conclusion, the Chicago heat wave of 1995 caused more than 600 excess deaths and an epidemic of heat-related illnesses, of which near-fatal heat stroke was the most harmful. We show that in addition to a high acute in-hospital mortality rate, heat stroke results in permanent loss of independent function in most survivors. Patients with severe disability had a high 1-year mortality rate. Our data on near-fatal classic heat stroke suggest multisystem organ dysfunction similar to that reported for exertional heat stroke. Intercurrent infection was prevalent; however, its relation to classic heat stroke is unclear and warrants further study.

Future studies should stratify classic heat stroke by severity. This will allow a better understanding of the types of heat stroke and lead to identification of management strategies that may improve survival and functional outcomes. Survivors should also be stratified according to a standard functional disability index, such as the modified Stanford Health Assessment Questionnaire, because the disabilities found are likely to be permanent and affect short-term survival.

Heat stroke is inherently preventable, and the recurrence of such an epidemic is likely to be avoided only with the aggressive implementation of a community-wide disaster plan. Our study shows the current limitations of medical care once the condition has advanced to near-fatal heat stroke. This emphasizes the need for education at all levels of medical care in conjunction with an aggressive prehospital prevention and rescue plan when faced with this type of environmental catastrophe.

From Michael Reese Hospital and Medical Center, University of Illinois at Chicago, Resurrection Medical Center, and Pritzker School of Medicine, University of Chicago, Chicago, Illinois; and the Centers for Disease Control and Prevention, Atlanta, Georgia.

Dr. O'Mara: Resurrection Medical Center, Critical Care Unit, 7435 West Talcott Avenue, Chicago, IL 60631.

Ms. Buescher: University of Chicago, Pritzker School of Medicine, Dean's Students' Office, Biology and Anatomy, 924 East 57th, Suite 104, Chicago, IL 60637.

Dr. Whitney: Centers for Disease Control and Prevention/C09, 1600 Clifton Road, NE B-1 Room 4045B, Atlanta, GA 30333.

Drs. Forsythe and Ndukwu: Section of Pulmonary and Critical Care Medicine, University of Chicago, 5481 South Maryland Avenue, MC 6076, Chicago, IL 60637.

Dr. McNamee: Beth Israel Medical Center, Health Care Department, 330 Brookline Avenue, Boston, MA 02215.

Dr. Adiga: Medclinic, Inc., 803 Highway 71W, Savannah, MO 64485.