How to improve the management of chest pain: hospitalists and use of prediction rules.

Background: Three percent of patients with acute myocardial infarction are still missed despite the excess number of admissions with chest pain. The purpose of this study was to review the characteristics of patients who were admitted with chest pain, to evaluate the appropriateness of admissions and the outcomes. We also discuss whether use of a prediction rule could have made a difference in the management of these cases.

Methods: We performed retrospective chart review on all patients admitted to the hospitalist service with a diagnosis of chest pain. Each patient was risk stratified using Diamond and Forrester algorithm for probability of coronary artery disease (CAD), retrospectively. Results were analyzed using [chi square] test or exact test and Student's t test.

Results: Of 260 patients admitted with chest pain to the hospitalist service, only 24 (9.2%) received the final diagnosis of acute coronary syndrome (ACS). The patients in the ACS group were older and more likely to be male and to have a history of hyperlipidemia, CAD, peripheral vascular disease, cerebrovascular disease and percutaneous coronary intervention (PCI). Of 34 patients who underwent cardiac catheterization, 20 (58.8%) had occlusive CAD and 14 of them received PCI. Risk stratification of patients, retrospectively, revealed 28.3% of the total patient population was high risk, while 6.6% of them were low risk. The number of ACS cases was highest in the high risk group, while none was detected in the low risk group.

Conclusions: Our study demonstrated that using a prediction rule could have prevented about 6% of the chest pain admissions. Therefore, the use of risk stratification methods should be encouraged to decrease cost and improve efficiency of care.

Key Words: hospitalist, chest pain, acute coronary syndrome, prediction rules

**********

Heart disease is still the leading cause of death in the United States among both men and women, causing more than 710,000 deaths per year. (1) Each year, over 6 million Americans present to emergency departments (ED) with complaints of chest discomfort. Although approximately half of them are admitted to the hospital, only 20 to 25% eventually receive the diagnosis of acute coronary syndrome/coronary artery disease (ACS/CAD). (2,3) The cost of care of patients unnecessarily admitted to the coronary care unit has been estimated to be nearly $3 billion annually; whereas the total cost of liberal admissions to rule out acute myocardial infarction (AMI) in low risk patients has been more than $13 billion. (4) Unfortunately, despite the excess number of admissions, 1.2 to 3% of patients with AMI were reported to be missed. (3) Missed AMI ranks as the highest single diagnosis in terms of dollars paid and third highest in terms of frequency of claims in malpractice against ED physicians. (5) Efforts to develop accurate, reliable risk stratification systems and the implementation of chest pain observation units in EDs are both preventative measures to ensure proper diagnosis of AMI while still being cost-effective.

Despite the presence of multiple risk stratification systems, to date there is no single method to discharge patients home safely. As hospital medicine makes a name for itself, hospitalists are being asked to take on more and more responsibility for triaging/admitting chest pain patients. We conducted a retrospective study to review our workload, performance, quality of care and outcomes on patients admitted with chest pain.

Methods

We performed a retrospective study on patients with chest pain who were admitted to the hospitalist service in a community hospital between January 1 and June 30, 2005. All consecutive admissions with any chest pain-related diagnoses were included in the study using International Classification of Disease (ICD-9) codes of 786.5x, 410.x, 411.8, and 413.x as final discharge diagnoses. Patients who had ST-segment elevation on initial ECG were excluded.

Request for cardiology consultation, choice of medications, and further testing, including cardiac stress test, were all up to the admitting hospitalist's individual preference. One of the physicians and the clinical care coordinator RN of the hospitalist service reviewed the charts for demographic characteristics, clinical data, in-hospital treatment, final diagnosis (ACS) and outcomes using standardized reporting guidelines as recommended elsewhere. (6)

All patients were admitted to monitored beds (in intensive care unit or telemetry unit) and underwent serial cardiac enzymes (CK-total, CK-MB, troponin) and ECGs. The decision of starting the patient on Plavix, heparin, low molecular weight heparin (LMWH) or GPIIb/IIIa receptor inhibitor was again based on the admitting physician's decision and individual preference without using any universal risk stratification system. Some patients with low cardiac risk were discharged after 2 or 3 negative sets of serial cardiac enzymes, which were checked every 6 hours with a follow-up appointment with their primary care physicians. Most patients who were ruled out for AMI by cardiac enzymes underwent some type of cardiac stress test. Some of them were discharged if they preferred to have the cardiac stress test as an outpatient at their primary care physicians' office. All patients with elevated cardiac enzymes and/or changes in serial ECGs received a cardiology consultation. Patients were transferred to cardiology service in a nearby tertiary care center following cardiac catheterization if they required PCI. Therefore, length of stay in hours is the time patients stayed under hospitalist service care, and it does not include the time of stay after PCI.

Typical chest pain was described as a deep, poorly localized chest or arm discomfort that is reproducibly associated with physical exertion or emotional stress and is relieved promptly with rest and/or the use of sublingual nitroglycerin. Pain presenting with two of the above features was considered to be atypical, whereas one and no above features was noncardiac chest pain.

ECGs were recorded as normal; abnormal but nonspecific/nondiagnostic if there were nonspecific ST-T wave changes, sinus tachycardia, prolonged PR, QRS, QTc intervals, bundle branch blocks, or LVH with strain; ischemianew if there were ST-segment depression more than 0.1 mV measured 40 to 60 milliseconds, inverted T waves more than 0.1 to 0.3 mV or Q waves at least 30 milliseconds in duration; ischemia-old if changes were known to be old; acute MI if there was new left bundle branch block or ST-segment elevation in two or more contiguous leads.

ACS was defined as angina at rest or prolonged/recurrent angina within the preceding 24 hours along with ST segment depression [greater than or equal to]0.05 mV, transient ST segment elevation [greater than or equal to]0.1 mV, T-wave inversion [greater than or equal to]0.3 mV in at least 2 leads or elevated troponin/CK-MB.

Retrospectively, the patients were stratified into low (<5%), intermediate (5-65%), and high (>65%) risk, based on Diamond and Forrester method for probability of CAD (appendix 1).

Categorical variables were compared using a [chi square] test or an exact test. Continuous variables were expressed in means/medians and compared with Student's t test. A 2-sided p value <0.05 was considered to be statistically significant. All analyses were performed using SPSS statistical software (version 13.0). This study was approved by the Institutional Review Board of Carolinas HealthCare System.

Results

Of 260 patients admitted with chest pain to the hospitalist service, only 24 (9.2%) received the final diagnosis of ACS. The characteristics of patients are detailed in Tables 1 and 2. The patients in the ACS group were older than the non-ACS group (62.0 versus 50.9, P < 0.001). Despite the predominance of female sex and African-American race in our patient population, 66.6% of patients with ACS were male (P = 0.042) and Caucasian (although not statistically different). One fifth of patients were self-pay, with no primary care physicians, while 26.2% had Medicare [+ or -] Medicaid. No significant differences were found between the ACS and non-ACS groups with respect to family history for CAD, history of diabetes mellitus, hypertension, renal insufficiency and tobacco use. The number of patients with hyperlipidemia was higher in the ACS group (P = 0.005).

History of CAD/prior AMI, PCI, peripheral vascular disease (PVD), cerebrovascular accident (CVA), as well as typical chest pain and pain lasting >20 minutes, were all predictive factors for ACS in our patient population. None of the clinical symptoms, including dyspnea, weakness, syncope, nausea/vomiting, palpitation, or diaphoresis, were found to be strongly suggestive of ACS (P > 0.05).

Only 29.2% of patients with ACS had ECGs with new ischemic changes, while 16.7% of them had normal ECGs (Table 3). Overall, 175 patients underwent cardiac stress testing of which 116 (66.3%) were performed by the hospitalists (P < 0.001). All patients with ACS received cardiology consultation during their hospital stay, although only 58.3% of consultations were requested at arrival. All patients with ACS received aspirin and heparin/LMWH. In the ACS group, [beta]-blocker use was 83.3%, while GP IIb/IIIa use was 69.6%, with one patient having contraindication due to active gastrointestinal system bleeding (Table 4). Of 34 patients who underwent cardiac catheterization, 20 (58.8%) had occlusive CAD and 14 of them received PCI without any complications; others were treated medically. Only one patient underwent CABG. Overall, no deaths occurred.

The rates of cardiology consultation and cardiac catheterization were similar for each of the race groups as described above. However, compared with females, more male patients received cardiology consultation, both at arrival and during hospitalization (P = 0.01).

Overall, length of stay ranged between 4 to 97 hours with a mean of 26 [+ or -] 15.4 hours (median = 22.5 hours). Compared with patients who received cardiology consultation at arrival, those receiving cardiology consultation later during the hospitalization were likely to stay longer with mean length of stay 41.28 [+ or -] 22.85 hours versus 32.44 [+ or -] 20.48. However, the difference was not statistically significant (P = 0.133).

Fourteen (5.4%) patients had positive urine drug screen with cocaine [+ or -] marijuana, with none having ACS.

Documentation of diet modification and smoking cessation recommendations were 80% (207/260) and 70% (61/87), respectively.

During retrospective CAD risk stratification and related statistics, 34 patients (8 of them having ACS) were excluded from the analysis due to chronological age being out of the age range of the Diamond-Forrester method. Of 226 patients, the cardiac risk was low, intermediate, and high in 6.6%, 65% and 28.3%, respectively (Fig.). About 68% of ACS cases were within the high risk group, while none were detected in the low risk group. None of the low risk patients had a positive stress test or received cardiac catheterization. The rates of positive stress test, cardiology consultation, cardiac catheterization and PCI were more prevalent in the high risk group.

Discussion

Our study demonstrated that using a simple prediction rule could have prevented about 6% of the chest pain admissions, along with unnecessary stress tests. On the other hand, more patients with high risk could have been referred initially for cardiac catheterization rather than stress testing. Therefore, use of risk stratification methods should be encouraged to improve cost and care efficiency.

There are multiple prediction rules and models to detect the pretest probability of coronary artery disease. (7,8) Traditional risk factors from the Framingham study have been shown to perform poorly in acute cardiac ischemia in the inpatient setting. (4) Acute cardiac ischemia time-insensitive predictive instrument (ACI-TIPI), which is a computer derived program using age, sex, the presence of chest pain or left arm pain, pathologic Q waves or the presence and degree of ST segment elevation or depression and T-wave elevation or inversion, was shown to reduce hospitalization among emergency department patients without acute cardiac ischemia. (9) The American College of Cardiology/American Heart Association (ACC/AHA) guidelines for UA/NSTEMI provide a very good evidence-based approach to the care of patients with ACS. (10-12) The Goldman prediction rule requires knowledge of patient age, ECG findings, length of symptoms, radiation of symptoms, reproducibility of chest pain by palpation, and character of pain. (13) Here, we used the Diamond-Forrester method because of its simplicity and generally accepted accuracy. It uses patient age, sex and basic symptom description. Addition of diabetes, hypertension, hyperlipidemia, smoking, family history of coronary artery disease, obesity and estrogen status in women in a recent scoring system was found to be as accurate as the Diamond-Forrester method and may be more valuable in asymptomatic patients. (7)

Unfortunately, there is no one single method to decide if a patient can be discharged from the ED or hospital safely. To avoid the missed AMIs, we will continue to admit more patients with low risk. Social issues, including lack of insurance, not having a primary care physician, inability of arranging an early outpatient follow-up, patients' anxiety level and requests for further testing will also continue to be the reason of some unnecessary admissions.

The major limitation to this study is the lack of follow-up. As hospitalists, we are not able to follow patients after discharge, nor can we learn about the long term outcomes. The patients who were discharged after 3 sets of negative cardiac enzymes or those who requested to decide about stress testing or cardiac catheterization after being seen by their PCPs, may eventually have received the diagnosis of cardiac disease on further workup. We also assumed the chest pain was noncardiac if stress test was negative. Therefore, it is possible that actual coronary artery disease prevalence in our patient population is higher than we detected as we cannot determine the false negative results due to lack of follow-up. Ideally, at least a 3 to 6 month follow-up would be expected to follow the outcomes. Unfortunately, lack of follow-up will continue to be a pitfall in studies conducted by hospitalist groups which we will be seeing more and more in the near future. Approximately 70% of our patients in the high cardiac risk group underwent cardiac stress test with about 60% of them having negative results and were discharged home. However, in a tertiary care center with a more aggressive invasive cardiology service, these patients could have undergone cardiac catheterization directly with possible detection of higher rates of cardiac disease. Overall, our study results cannot enhance the disposition protocols for chest pain patients. However, it does support the fact that the use of risk stratification methods can help to identify the high risk patients who should be worked-up aggressively. The communication between hospitalists and primary care physicians is vital when handing over the patients who may have false-negative results despite an uneventful hospitalization.

Pharmacotherapies, such as IV GP IIb/IIIa inhibitors, LM WH, and clopidogrel, have demonstrated incremental benefits for patients with non-ST segment elevating (NSTE) ACS, therefore reducing the composite of death or MI. Results from the CRUSADE (Can Rapid risk stratification of Unstable angina patients Suppress Adverse outcomes with Early implementation of the ACC/AHA guidelines) Quality Improvement Initiative, which is an observational multicenter study with high risk NSTE MI patients, showed use of aspirin, any heparin, GPIIb IIIa inhibitor, [beta]-blocker was 93.8%, 88.8%, 50.9%, and 82.3% respectively, in the early invasive care group. (14) Our hospitalists provided quality comprehensive care for the ACS group with 100% use of ASA and any heparin as well as 69.6% GPIIb IIIa inhibitor use. [beta]-blocker use rate (83.3%) was also comparable.

A prospective randomized trial comparing a chest pain center with traditional in-hospital evaluation revealed a significant decrease in mean total cost ($1,539 versus $2,095) and length of stay (33.1 versus 44.4 h). (15) Our hospitalist service length of stay (median = 22.5 h) is comparable and may be even shorter as compared with the above chest pain center length of stay, indicating we do practice cost-effectively.

A study using GUARANTEE (Global Unstable Angina Registry and Treatment Evaluation) Registry data showed involvement of cardiologists in the management of ACS could improve cost and care efficiency as compared with family physicians/internists. (16) However, there are no reports on the level of care hospitalist services provide.

Conclusion

Our study shows hospitalists are reasonably cost-effective with appropriate use of life saving medications and relatively short length of stay. Further studies comparing the care given by hospitalists versus cardiologists and by hospitalists versus traditional internists are required to determine the difference hospitalist programs bring to the hospitals. In the meantime, the clinical evaluation of potential cardiac ischemia will continue to be challenging with excessive admissions. We recommend use of prediction/risk stratification rules in the management of chest pain patients to improve cost and care efficiency.

Appendix: Likelihood of significant coronary disease based on
Diamond-Forrester method

                                   Atypical  Typical
Sex/age  Asymptomatic  Nonanginal  angina    angina

Men
60-69    0.12          0.28        0.67      0.94
50-59    0.10          0.22        0.59      0.92
40-49    0.06          0.14        0.46      0.87
30-39    0.02          0.05        0.22      0.70

Women
60-69    0.08          0.19        0.54      0.91
50-59    0.03          0.08        0.32      0.79
40-49    0.01          0.03        0.13      0.55
30-39    0.00          0.01        0.04      0.26

Adopted from: Diamond GA, Forrester SF. Analysis of probability as an
aid in the clinical diagnosis of coronary artery disease. NEJM
1979;300(24):1350-1358.

References

1. Miller CD, Lindsell CJ, Khandelwal S, et al. Is the initial diagnostic impression of noncardiac chest pain adequate to exclude cardiac disease? Ann Emerg Med 2004;44:565-574.

2. Blomkalns AL, Gibler WB. Development of the chest pain center: rationale, implementation, efficacy, and cost effectiveness. Prog Cardiovasc Dis 2004;46:393-403.

3. Chandra A, Rudraiah L, Zalenski RJ. Stress testing for risk stratification of patients with low to moderate probability of acute cardiac ischemia. Emerg Med Clin North Am 2001;19:87-103.

4. Wilkinson K, Severance H. Identification of chest pain patients appropriate for an emergency department observation unit. Emerg Med Clin North Am 2001;19:35-66.

5. Kamineni R, Alpert JS. Acute coronary syndromes: initial evaluation and risk stratification. Prog Cardiovasc Dis 2004;46:379-392.

6. Hollander JE, Blomkalns AL, Brogan GX, et al. Standardized reporting guidelines for studies evaluating risk stratification of emergency department patients with potential acute coronary syndromes. Ann Emerg Med 2004;44:589-598.

7. Morise AP. Comparison of the Diamond-Forrseter method and a new score to estimate the pretest probability of coronary disease before exercise testing. Am Heart J 1999;138:740-745.

8. Diercks DB, Hollander JE, Sites F, et al. Derivation and validation of a risk stratification model to identify coronary artery disease in women who present to the emergency department with potential coronary artery disease. Acad Emerg Med 2004;11:630-634.

9. Selker H, Beshansky JR, Griffith JL, et al. Use of the acute cardiac ischemia time-insensitive predictive instrument (ACI-TIPI) to assist with triage of patients with chest pain or other symptoms suggestive of acute cardiac ischemia: a multicenter, controlled clinical trial. Ann Intern Med 1998;129:845-855.

10. Gibler WB, Cannon CP, Blomkalns AL, et al. Practical implementation of the guidelines for unstable angina/NSTEMI in the emergency department. Ann Emerg Med 2005;46:185-197.

11. Pollack CV, Gibler WB. 2000 ACC/AHA guidelines for the management of patients with unstable angina and NSTEMI: Practical summary for emergency physicians. Ann Emerg Med 2001;38:229-240.

12. Braunwald E, Antman EM, Beasley JW, et al. ACC/AHA guidelines for the management of patients with unstable angina and NSTEMI: Executive summary and recommendations. Circulation 2000;102:1193-1209.

13. Sequist TD, Bates DW, Cook EF, et al. Prediction of missed MI among symptomatic outpatients without coronary artery disease. Am Heart J 2005;149:74-81.

14. Bhatt DL, Roe MT, Peterson ED, et al. Utilization of early invasive management strategies for high risk patients with non-ST segment elevation acute coronary syndromes: results from the CRUSADE Quality Improvement Initiative. JAMA 2004;292:2096-2104.

15. Storrow AB, Gibler WB. Chest pain centers: diagnosis of acute coronary syndromes. Ann Emerg Med 2000;35:449-461.

16. Gomberd-Maitland M, Murphy SB, Moliterno DJ, et al. Are we appropriately triaging patients with unstable angina? Am Heart J 2005;149:613-618.

Beril Cakir, MD and Kay Blue, RN

From Carolinas Medical Center-University, Charlotte, NC.

Reprint requests to Dr. Beril Cakir, Carolinas Medical Center-University, PO Box 560727, Charlotte, NC 28256. Email: [e-mail]

Accepted September 22, 2006.

RELATED ARTICLE: Key Points

* Missed acute myocardial infarction ranks as the highest single diagnosis in terms of dollars paid in malpractice against ED physicians.

* As hospital medicine makes a name for itself, hospitalists are being asked to take on more and more responsibility for triaging/admitting chest pain patients.

* Use of risk stratification methods/prediction rules should be encouraged to improve cost and care efficiency in the management of chest pain.

Table 1. Baseline demographics and characteristics of patients

                               ACS                 Non-ACS
Characteristics                % (n = 24)          % (n = 236)

Age (years) ([+ or -]st. dev)  62.0 [+ or -] 14.7  50.9 [+ or -] 12.4
Sex
  Male                         66.7                44.9
  Female                       33.3                55.1
Race
  African-American             29.2                53.4
  Caucasian                    66.7                38.6
  Hispanic                      4.2                 6.4
  Other                         0.0                 1.7
Health insurance
  Self pay                      8.3                22.9
  Medicare [+ or -] Medicaid   33.3                18.6
  Medicare + Private           16.7                 5.1
  Private                      41.7                53.4
Family history of CAD          54.2                50.0
Obesity                        25.0                49.2
Tobacco use                    45.8                39.0

                               Total
Characteristics                % (n = 260)         P value

Age (years) ([+ or -]st. dev)  51.9 [+ or -] 13.0  <0.001*
Sex
  Male                         46.9 (122)           0.042
  Female                       53.1 (138)
Race
  African-American             51.2 (133)           0.075**
  Caucasian                    41.2 (107)
  Hispanic                      6.2 (16)
  Other                         1.5 (4)
Health insurance
  Self pay                     21.5 (56)            0.022**
  Medicare [+ or -] Medicaid   20.0 (52)
  Medicare + Private            6.2 (16)
  Private                      52.3 (136)
Family history of CAD          50.4 (131)           0.697
Obesity                        46.9 (122)           0.024
Tobacco use                    39.6 (103)           0.513

*P value by independent samples' t test, **2-sided p value by Fisher
exact test, others by Chi-square test.
CAD, coronary artery disease: ACS, acute coronary syndrome.

Table 2. Past medical history and clinical presentation of patients

                          ACS         Non-ACS      Total
Characteristics           % (n = 24)  % (n = 236)  % (n = 260)  P value

Positive history of
  CAD/AMI                 37.5        12.3         14.6 (38)     0.001
  Congestive heart         8.3         4.7          5.0 (13)     0.342*
    failure
  Hypertension            66.7        55.1         56.2 (146)    0.276
  Diabetes mellitus       29.2        24.2         24.6 (64)     0.587
  Hyperlipidemia          87.8        58.5         61.2 (159)    0.005
  Peripheral vascular     16.7         1.7          3.1 (8)      0.003*
    disease
  Renal insufficiency      8.3         2.5          3.1 (8)      0.162*
  Cerebrovascular         12.5         2.5          3.5 (9)      0.040*
    accident
  PCI                     20.8         3.4          5.0 (13)     0.003*
  CABG                    12.5         3.8          4.6 (12)     0.087*
Typical chest pain on     70.8        25.0         29.2 (76)    <0.001
  arrival
Duration of pain >20 min  83.3        49.2         52.3 (136)    0.001

*2-sided P value by Fisher exact test, others by Chi-square test.
ACS, acute coronary syndrome; CAD, coronary artery disease; AMI, acute
myocardial infarction; PCI, percutaneous coronary intervention; CABG,
coronary artery bypass graft.

Table 3. Diagnostic tests, procedures

                         ACS         Non-ACS      Total
Test/procedure           % (n = 24)  % (n = 236)  % (n = 260)  P value

ECG
  Normal                  16.7       48.3         45.4 (118)   <0.001
  Abnormal, not           45.8       47.0         46.9 (122)
    diagnostic
  Ischemia -- old          8.3        0.4          1.2 (3)
  Ischemia -- new         29.2        4.2          6.5 (17)
Stress test
  Not performed           66.7       29.2         32.7 (85)    <0.001*
  Performed               33.3       70.8         67.3 (175)
    Nondiagnostic         12.5        3.6          4.0 (7)
    Negative              12.5       92.2         88.5 (155)
    Positive              75.0        4.2          7.5 (13)
Cardiology consultation  100.0       18.6         26.2 (68)
Cardiac catheterization   83.3        5.9         13.1 (34)    <0.001
PCI                       58.3        0.0          5.4 (14)
CABG                       4.2        0.0          0.4 (1)

*2-sided P value by Fisher exact test, others by Chi-square test.
ACS, acute coronary syndrome; ECG, electrocardiogram; PCI, percutaneous
coronary intervention; CABG, coronary artery bypass graft.

Table 4. Medical management of patients

                           ACS         Non-ACS      Total
Medications started <24 h  % (n = 24)  % (n = 236)  % (n = 260)  P value

Aspirin                    100.0       97.0         97.3 (253)    1.0
Beta-blocker                83.3       30.1         35.0 (91)    <0.001
Nitroglycerin               91.7       71.0         72.7 (189)    0.029
ACE inhibitor/ARB           66.7       32.6         35.8 (93)     0.001
Calcium channel blockers    12.5       14.8         14.6 (38)     0.784*
Heparin                     33.3        0.0          3.1 (8)
LMWH                        66.7       29.2         32.7 (85)    <0.001
GPIIb/IIIa inhibitor        69.6        0.0          6.2 (16)

*2-sided P value by Fisher exact test, others by Chi-square test.
ACS, acute coronary syndrome; ARB, angiotensin receptor blocker; LMWH,
low molecular weight heparin.

                         Stress test*  C.Cath**  ACS***  PCI****

Low risk (n = 15)        53.3           0         0       0
Interm ediate (n = 147)  72.1           7.5       3.4     1.4
High risk (n = 64)       70.3          21.9      17.2    14.1

Fig. Percentage of patients in each cardiac risk group based on Diamond-
Forrester method with respect to performed diagnostic tests and final
diagnosis of ACS.
Thirty-four patients (8 with acute coronary syndrome) were excluded from
the analysis due to chronological age outside the age range of the
Diamond-Forrester method.
C.Cath, coronary catheterization; ACS, acute coronary syndrome; PCI,
percutaneous coronary intervention.
*P = 0.309; **P = 0.007; ***P = 0.02; ****P = 0.001 by Fisher exact
test.