preliminary report

Dyspnea Differentiation Index*

A New Method for the Rapid Separation of Cardiac vs Pulmonary Dyspnea
Rajesh K. Ailani, MD; Keyvan Ravakhah, MD; Bruno DiGiovine, MD; Gordon Jacobsen, MS; Thaw Tun, MD, MRCP; Donald Epstein, MD, FCCP; and Burton C. West, MD

Study objective: To assess the utility of a new parameter in the differentiation of dyspnea of cardiac origin from dyspnea of pulmonary origin. Methods: The peak expiratory flow (PEF) rate and the partial pressure of oxygen in arterial blood (PaO2) were measured in 71 patients with the chief complaint of dyspnea. The patients were treated in the hospital, and the final diagnosis (cardiac or pulmonary) of the cause of dyspnea was made at discharge. We defined a new measure, the dyspnea differentiation index (DDI), as (PEF PaO2)/ 1,000. We performed a receiver operating characteristic (ROC) curve analysis of the data to define the measure that best distinguished cardiac from pulmonary dyspnea. The curves also allowed us to establish an optimal cut-off point to distinguish between cardiac and pulmonary dyspnea. Results: Patients with pulmonary dyspnea had a significantly lower mean PEF than patients with cardiac dyspnea (144 66 vs 267 97 L/min, respectively; p < 0.001). They also had a lower DDI than patients with cardiac dyspnea (8.4 4.0 vs 18.4 7.9 L-mm/min, respectively; p < 0.001). These two measures, PEF and DDI, also best distinguished pulmonary from cardiac dyspnea. PEF was able to diagnose the correct cause of dyspnea in 72% of patients, and DDI was correct in 79% of patients. This compares favorably to the performance of the emergency department physicians, who were able to predict the correct diagnosis in only 69% of patients. Conclusion: These results demonstrate that the PEF by itself is useful in differentiating between cardiac and pulmonary causes of dyspnea, but that the calculation of DDI is superior in this regard. (CHEST 1999; 116:11001104)
Key words: cardiac dyspnea; differentiation of dyspnea; dyspnea differentiation index; peak expiratory flow rate; pulmonary dyspnea Abbreviations: AUC area under the curve; CHF congestive heart failure; DDI dyspnea differentiation index; %DDI %PEF Pao2/1,000; ED emergency department; PEF peak expiratory flow; %PEF measured PEF/ predicted PEF 100; ROC receiver operating characteristics

clinical practice, it is often difficult to distinguish In between cardiac and pulmonary causes of dyspnea.1 Both exhibit similar physical findings, such as
*From the Department of Medicine (Drs. Ravakhah, Tun, Epstein, and West), Meridia Huron Hospital, Cleveland, OH; and the Department of Biostatistics (Mr. Jacobsen) and Division of Pulmonary and Critical Care Medicine (Drs. Ailani and DiGiovine), Henry Ford Hospital, Detroit, MI. Abstracts of this work were presented to the American College of Physicians at the Ohio Chapter Meeting, Columbus, OH, September 27, 1996, and at the Annual Meeting, Philadelphia, PA, March 21, 1997. Manuscript received July 31, 1998; revision accepted April 29, 1999. Correspondence to: Keyvan Ravakhah, MD, Meridia Huron Hospital, 13951 Terrace Rd, Cleveland, OH 44112
1100

pulmonary rales, wheezing, and jugular venous distention, and radiologic findings. Nonetheless, an accurate diagnosis is crucial because the treatment of these conditions is quite different. A diuretic used to treat cardiac dyspnea can cause acid-base disturbances and hypotension in patients with a pulmonary cause of dyspnea.2 Likewise, sympathomimetic amines and -agonists can induce angina and arrhythmia in patients with dyspnea related to to congestive heart failure.3 Although sophisticated and invasive tests such as Swan-Ganz catheterization can distinguish between cardiac and pulmonary causes of
Preliminary Report

Downloaded From: https://journal.publications.chestnet.org/ on 04/10/2013

dyspnea, a simple, practical bedside test is needed to distinguish between the two. There has been only one study looking at the utility of peak expiratory flow rate (PEF) alone in the differentiation of cardiac vs pulmonary dyspnea.4 The authors concluded that PEF might be a useful adjunctive tool in the differentiation of acute dyspnea of cardiac vs pulmonary origin. However, this study did not compare PEF to the clinical evaluation of dyspnea. There has been another study looking at the utility of end-tidal CO2 and PEF in the differentiation of cardiac and obstructive causes of dyspnea.5 In that study, the authors found no significant difference in PEF between the cardiac and the pulmonary groups, although end-tidal CO2 differed significantly between the two groups. Although there have been many studies looking into the degree of hypoxia and change in pulmonary function in patients with myocardial ischemia/infarction and congestive heart failure (CHF),6 8 no study has been done to compare Pao2 in patients with cardiac vs pulmonary dyspnea. We hypothesized that PEF alone or with the addition of Pao2 would be superior to a clinical evaluation in distinguishing between cardiac and pulmonary causes of dyspnea. We conducted a prospective trial to test this hypothesis.

Airmed; Harlow-Essex, UK). Each patient was asked to attempt PEF three times, and the highest PEF was recorded. A medical resident or the attending ED physician was in charge of the procedures. A board-certified physician with one or two physician assistants and one or two medical residents staffs our ED. Prior to admission, the attending ED physician made a final diagnosis as to the cause of dyspnea. Calculation of DDI To combine the effects of PEF and Pao2, a new parameter, the product of PEF Pao2 divided by 1,000, was calculated and called the dyspnea differentiation index (DDI). Final Diagnosis The final determination of the cause of dyspnea was based on on the attending physicians discharge diagnosis. The attending physicians were not aware of the study results or the study protocol. Based on the attending physicians diagnosis, the patients were divided into three categories: (1) patients with a cardiac cause of dyspnea (cardiac group); (2) patients with any of the pulmonary causes of dyspnea, eg, exacerbation of COPD, asthma, or pneumonia (pulmonary group); and (3) patients with combined causes of dyspnea (both cardiac and pulmonary). Statistical Analysis For each group, we calculated the mean SD of Pao2, PEF, and DDI. We calculated the predicted PEF for each patient based on height and age.10 In addition, we calculated the percentage of the measured PEF to the predicted PEF for each patient (%PEF measured PEF/predicted PEF 100). We also calculated DDI for each patient by using the %PEF, rather than the absolute value of the PEF, and denoted it as %DDI, that is, %DDI %PEF Pao2/1,000. Comparisons of means were made using the two-tailed Students t test for unpaired data. The alpha level for all statistical tests was 0.05. We did a receiver operating characteristics (ROC) curve analysis for PEF, %PEF, DDI, and %DDI for all patients.11 Here, an ROC curve is a graph of sensitivity vs 1 specificity for diagnosing cardiac dyspnea, corresponding to each possible cut-off for the diagnostic test result. For the purpose of calculation of ROC curve, the patients with both cardiac and pulmonary causes of dyspnea had to be excluded, as the test is designed only for bivariate outcomes. We selected an optimal cut-off point on the ROC curve that corresponds to the best sensitivity and specificity (maximum sum of sensitivity and specificity). We also calculated the area under the curve (AUC) on ROC analysis for PEF, %PEF, DDI, and %DDI. A perfect diagnostic test would have an AUC on ROC analysis of 1.0. A good discrimination corresponds to AUC levels of 0.80 to 0.85. Based on the optimal cut-off points, we calculated the sensitivity, specificity, and positive and negative predictive values for PEF, %PEF, DDI, and %DDI. In calculating these values, the patients with diagnoses of either both cardiac and pulmonary causes of dyspnea or neither were included. To have the most conservative assessment of our tests, the test was considered incorrect when predicting the cause of dyspnea in these patients. To evaluate the real world utility of our test, we compared the diagnosis made by the ED physicians and the diagnosis made by our test. In this analysis, the gold standard was the final diagnosis made at discharge by the attending physician. For patients with a combined cardiac and pulmonary cause of dyspnea, our test was scored as unable to provide correct diagnosis. The ED physicians made their diagnosis of cardiac, pulmonary, or both cardiac and pulmonary causes of dyspnea after clinical evaluation.
CHEST / 116 / 4 / OCTOBER, 1999

Materials and Methods

Design A prospective study of PEF and arterial blood gases in patients evaluated for moderate to severe dyspnea in the Emergency Department (ED) of Meridia Huron hospital, between July 1995 and February 1997. Patients Patients were eligible for our study if they were 18 years or older and if they presented to the ED with a chief complaint of moderate to severe dyspnea and were to be admitted to the hospital. Patients were excluded if they met the following exclusion criteria: (1) patients not requiring hospitalization; (2) clinically unstable, unable to discontinue oxygen; and (3) inability to perform PEF. Patient selection and treatment were performed according to a protocol approved by the hospitals Institutional Research and Review Committee. Data Collection PEF and Pao2 were measured prior to treatment or at least 15 min after oxygen therapy was stopped. All measurements were made before respiratory treatments were given. All of the patients were monitored for O2 saturation, respiratory rate, and heart rate. Arterial blood gases were drawn from radial, femoral, or brachial arteries. In cases of radial artery usage, an Allen test was performed to rule out insufficient collaterals.9 PEF was measured by the use of a peak flowmeter (mini-Wright;

1101

Downloaded From: https://journal.publications.chestnet.org/ on 04/10/2013

Results Seventy-one patients were enrolled in the study. There were 34 men and 37 women. Seven of the 71 patients (9.9%) had both cardiac and pulmonary causes of dyspnea. Of these, four patients had COPD and CHF as their final diagnosis, two patients had CHF and pneumonia, and one patient had CHF and bronchial asthma. In one patient, the cause of shortness of breath was mild diabetic ketoacidosis. No patients had pulmonary emboli as the final diagnosis. For the remaining 63 patients, a single diagnosis as to the cause of dyspnea (cardiac or pulmonary) was established. Twenty-four of the 71 patients (34%) had dyspnea of cardiac origin, and 39 patients (55%) had dyspnea of pulmonary origin. Nine of the 24 cardiac cases and 23 of the 39 pulmonary patients were women. All 24 patients with the final diagnosis of cardiac dyspnea had CHF. Two of the 24 patients also had a non-Q wave myocardial infarction. Of the 39 patients with the final diagnosis of pulmonary dyspnea, 18 patients had COPD exacerbation, 14 had pneumonia (1 had Pneumocystis carinii pneumonia), and 7 had asthma. The mean SD age for patients with cardiac and pulmonary cause of dyspnea was 58.8 16.8 and 59.8 16.2 years, respectively. The mean SD for height among patients with cardiac and pulmonary cause of dyspnea was 67.4 4.4 and 65.2 3.7 inches, respectively. The Pao2, PEF, %PEF, DDI, and %DDI for all patients in the cardiac and pulmonary groups are shown in Table 1. We found a statistically significant difference in all the parameters when we compared the cardiac and pulmonary groups. The ROC analysis of PEF, %PEF, DDI, and %DDI is summarized in Figure 1. All the param-

eters were effective in discriminating between cardiac and pulmonary causes of dyspnea based on the AUC. Based on these graphs, we evaluated cut-off points to distinguish between cardiac and pulmonary dyspnea. Sensitivity, specificity, and positive and negative predictive values of the optimal cut-off points in the diagnosis of dyspnea are shown in Table 2. In terms of specificity and positive predictive values, these four measures were similar. However, DDI and %DDI had higher sensitivity and negative predictive value than PEF and %PEF. We compared the diagnosis of the ED physician with the diagnosis made by our tests. We were able to obtain records of the ED diagnosis for 70 of 71 patients. The ED physician was correct in predicting cardiac or pulmonary or both as the cause of dyspnea in 69% (48/70) of the study patients. By using PEF alone, the correct diagnosis was predicted in 72% of the patients (51/71), and by using DDI alone, the correct diagnosis was predicted in 79% of the patients (56/71). In addition, when the ED physician and the DDI agreed on the cause of dyspnea, they were correct in 94% (44/47) of the patients. Discussion We studied PEF, %PEF, DDI, and %DDI. Although there was a statistically significant difference between the two groups of patients for all of these parameters, no absolute cut-off value was found. At the optimal cut-off point (found by ROC analysis), for the diagnosis of pulmonary dyspnea, DDI and %DDI had a higher sensitivity and negative predictive value than PEF and %PEF. The most significant result of the study was that DDI was more accurate than PEF or the ED physician in predicting the correct cause of dyspnea. Our new parameter, DDI, the product of PEF Pao2/1,000, has not been used in any other study. We proposed this parameter in an effort to combine the similar effects of pulmonary disease on PEF and Pao2. We believed that both PEF and Pao2 would be lower in patients with pulmonary disease and, therefore, their product will provide better discrimination than either measure alone. As we will discuss, there is both theoretical and empiric evidence to support the idea that PEF and Pao2 can distinguish between patients with cardiac and pulmonary causes of dyspnea. Most of the patients with moderate to severe dyspnea secondary to a pulmonary etiology have low FEV1 and PEF.12 In contrast, in patients with a cardiac cause of dyspnea, although the functions of small airways (closing volume and frequency dependence of compliance) are abnormal, active bronchoconstriction is unusual, and large airway function and
Preliminary Report

Table 1PaO2 PEF, %PEF, DDI, and %DDI in Patients With Pulmonary vs Cardiac Causes of Dyspnea*
Patients with Cardiac Dyspnea n 24 68.0 12.2 (43.692.4) 267 97 (73461) 58 19 (2096) 18.4 7.9 (2.634.2) 4.0 1.4 (1.26.8) Patients with Pulmonary Dyspnea n 39 58.7 13.2 (32.385.1) 144 66 (12276) 35 17 (169) 8.4 4.0 (0.416.4) 2.1 1.0 (0.14.1)

Variables Pao2 PEF, L/min %PEF DDI %DDI

p Value 0.007 0.001 0.001 0.001 0.001

*Variables are shown as mean SD (95% confidence interval) ( 2 SD). 1102

Downloaded From: https://journal.publications.chestnet.org/ on 04/10/2013

Figure 1. ROC curves of PEF, %PEF, DDI, and %DDI, along with the AUC for each graph. AUC for each parameter is shown, along with the standard error.

PEF are not affected.6,13 Previous studies on pulmonary function in CHF demonstrated only a mild decrease in PEF and FEV1.6,14 There has been only one study using PEF primarily to differentiate between the cardiac and pulmonary causes of dyspnea.4 In that study, patients with pulmonary dyspnea had lower PEF than the cardiac patients. There was an overlap between the cardiac and pulmonary groups and no single cut-off point separated pulmonary and cardiac patients. That study concluded that PEF, particularly relatively high or low values, may be a useful adjunct to a rapid clinical assessment of patients with dyspnea. A PEF reading 150 L/min was suggestive of a cardiac cause of dyspnea

Table 2ROC Curve Analysis of PEF, %PEF, DDI, and %DDI for All Patients*
Optimal Positive Negative Comparison Cut-Off Sensitivity, Specificity, Predictive Predictive Variable Point % % Value, % Value, % PEF %PEF DDI %DDI 200 42 13 3.0 72 59 82 77 71 78 74 70 80 81 84 81 63 54 71 66

*Sensitivity, specificity, and positive and negative predictive values are in reference to a pulmonary cause of dyspnea at the optimal cut-off point.

with a positive predictive value of 82%, while a reading of 150 L/min suggested a pulmonary diagnosis with a positive predictive value of 83%. Utilizing absolute PEF and not PEF as a ratio of the predicted PEF (%PEF) was criticized, and the inability to clearly separate the two categories was a drawback.15 We had similar results in our study, and although we used both %PEF and absolute PEF, we were unable to get a single cut-off point. The results in our study were not different regardless of whether we used absolute PEF or %PEF. This may be related to the fact that our cardiac and pulmonary patients were closely matched in age and height. Besides the study by McNamara and Cionni,4 there has been one more study looking at use of end-tidal CO2 measurement along with PEF to differentiate between cardiac and pulmonary dyspnea.5 In that study, the authors found no significant difference in PEF between the cardiac and the pulmonary groups. Utilizing absolute PEF to compare two groups with a significant age difference (the cardiac group was much older than the pulmonary group) could have been the reason for the lack of a significant difference. The sensitivity, specificity, and positive predictive value of PEF in the diagnosis of COPD/ asthma were 0.52, 0.40, and 0.61, respectively. The flaw with the PEF is the effort-dependent nature of the test. A poor effort can result in a low and inaccurate result.16 We guarded against this by
CHEST / 116 / 4 / OCTOBER, 1999

1103

Downloaded From: https://journal.publications.chestnet.org/ on 04/10/2013

having a respiratory therapist obtain the measurements three times and recording the best result. In our study, the readings were obtained with a mini-Wright peak flowmeter, which correlates highly with the readings obtained from the larger peak flowmeter (Wright).17 There are reasons to believe that not only PEF, but also Pao2, may help distinguish between cardiac and pulmonary dyspnea. In patients with pneumonia, the end products of the inflammatory cascade lead to bronchoconstriction along with ventilation/perfusion mismatch, and consequently lead to a low Pao2.18 In patients with COPD, there is long-standing, ongoing pulmonary damage with changes in the pulmonary circulation, including ventilation/perfusion mismatch and hypoventilation, leading to hypoxia.19 Moreover, there is also a direct relationship between low FEV1 seen in COPD patients and low Pao2.20,21 In contrast, in patients with CHF, large airway function (PEF and FEV1) is unaffected.6 Thus, a greater degree of hypoxia might be expected in patients with a pulmonary cause of dyspnea than in patients with a cardiac cause of dyspnea, except for severe pulmonary edema. Although we found a statistically significant difference in Pao2 between the two groups, there was again a significant overlap, and Pao2 by itself was not useful in distinguishing between the two conditions. One potential limitation to our measurement is that, patients who had received oxygen were kept off oxygen for a period of only 15 min. Although the recommended time for stabilizing oxygen saturation is between 15 to 30 min,22 we chose the lower limit so as to minimize the time the patient is without oxygen. We could have used Pao2/fraction of inspired oxygen ratios, although it would have introduced more calculations and we doubt that it would have been any more accurate. In our study, DDI was able to correctly predict the diagnosis in more patients than the ED physicians. However, DDI does not have the ability to predict combined disease, and that is a significant limitation to the utility of the test. Another limitation is the fact that we did not identify any patients with pulmonary embolism and interstitial lung disease in our study. Therefore, we cannot conclude that it would be accurate in patients with pulmonary embolism or interstitial lung disease. The best way to answer these questions would be in a validation study with a larger number of patients. In conclusion, we have found that DDI, the simple product of PEF and Pao2, is more valuable than either PEF or Pao2 alone in the evaluation of dyspnea. Also, the DDI was more accurate than the ED physician in diagnosing a cause of dyspnea. Therefore, the use of DDI may be beneficial in the initial evaluation of dyspnea, as both Pao2 and PEF are easily obtained in the ED.
1104

ACKNOWLEDGMENTS: The authors acknowledge the contribution of Dr. Richard Frires and the Emergency Department of Meridia Huron Hospital for help in collecting cases for this study. The authors also thank Dr. Joseph E. Loewenstein for the help rendered in preparing this manuscript.

References
1 Ingram RH Jr, Braunwald E. Dyspnea and pulmonary edema. In: Isselbacher KJ, Braunwald E, Wilson JD, et al, eds. Harrisons principles of internal medicine. 13th ed. New York, NY: McGraw-Hill, 1994; 174 178 2 Weitzenblum E. Acute respiratory failure in the patient with obstructive airways disease. In: Fishman AP, ed. Pulmonary diseases and disorders. 2nd ed. New York, NY: McGraw-Hill, 1988; 22872298 3 Yurchak PM. Cardiac problems in the pulmonary patient. In: Fishman AP, ed. Pulmonary diseases and disorders. 2nd ed. New York, NY: McGraw-Hill, 1988; 10871102 4 McNamara RM, Cionni DJ. Utility of the peak expiratory flow rate in the differentiation of acute dyspnea. Chest 1992; 101:129 132 5 Brown LH, Gough JE, Seim RH. Can quantitative capnometry differentiate between cardiac and obstructive causes of respiratory distress? Chest 1998; 113:323326 6 Hales CA, Kazemi H. Pulmonary function after uncomplicated myocardial infarction. Chest 1977; 72:350 358 7 Filmore SJ, Shapiro M, Killip T. Arterial oxygenation in acute myocardial infarction: serial analysis of clinical state and blood gas changes. Am Heart J 1970; 79:620 629 8 Enright PL, Kronmal RA, Smith VE, et al. Reduced vital capacity in elderly persons with hypertension, coronary heart disease, or left ventricular hypertrophy. Chest 1995; 107:28 35 9 Allen EV. Thromboanginitis obliterans: methods of diagnosis of chronic occlusive arterial lesions distal to the wrist with illustrative cases. Am J Med Sci 1929; 178:237244 10 Cherniack RM. Normal standards for ventilatory function. Am Rev Respir Dis 1972; 106:38 46 11 Beck JR, Colice Gl. Relative operating characteristic analysis applied to tests of pulmonary function. Semin Respir Med 1989; 10:211217 12 Ruppel G. Manual of pulmonary function testing. 6th ed. St Louis, MO: Mosby, 1994; 4354 13 Interiano B, Hyde RW, Hodges M. Interrelation between alterations in pulmonary mechanics and hemodynamics in acute myocardial infarction. J Clin Invest 1973; 52:1994 2006 14 Light RW, George RB. Serial pulmonary function in patients with acute heart failure. Arch Intern Med 1983; 143:429 433 15 Pollack CV. Utilization of the peak expiratory flow rate in evaluation of acute dyspnea of cardiac or pulmonary origin. Chest 1993; 103:13071308 16 van As A. The accuracy of peak expiratory flow meters. Chest 1982; 82:263 17 Perks WH, Tams IP, Thompson DA, et al. An evaluation of the mini-Wright peak flow meter. Thorax 1979; 34:79 81 18 Davidson FF, Glazier JB, Murray JF. The components of the alveolar-arterial oxygen tension difference in normal and obstructive lung disease. Am J Med 1972; 52:754 762 19 Seaton A, Seaton D, Leitch AG. Chronic bronchitis and emphysema. In: Seaton A, Seaton D, Leitch AG, et al, eds. Crofton and Douglass respiratory diseases. 4th ed. Oxford, UK: Blackwell Scientific Publications, 1989; 490 525 20 Rees HA, Miller JS, Donald KW. A study of the clinical course and arterial blood gas tensions of patients in status asthmatics. Q J Med 1968; 37:541561 21 McFadden ER, Lyons HA. Arterial blood gas tensions in asthma. N Engl J Med 1968; 278:10271032 22 Dantzker DR, MacIntyre NR, Bakow ED. Oxygen therapy. In: Comprehensive respiratory care. 1st ed. Philadelphia, PA: WB Saunders, 1994; 506 512
Preliminary Report

Downloaded From: https://journal.publications.chestnet.org/ on 04/10/2013