Chest pain: myocardial infarction: a clinical prediction rule can help identify high- or low-risk patients.

Clinical bottom line (level 1a)

  1. A clinical prediction rule can help identify patients at high or low-risk for myocardial infarction, but cannot exclude an infarct by itself.
  2. An infarct is more likely if there is ST elevation or Q waves in two or more leads, not known to be old (LR+22) .
  3. An infarct is less likely if the patient has:
    • a previous history of angina or MI, without ST changes on ECG, and pain lasting <1 hour which is not as bad as previous infarcts or angina attacks (LR-0.076)
    • pain radiating to the neck, left shoulder or left arm, and reproduced on palpation (LR-0.093)
    • chest pain for 48 hours or more and no new changes on ECG (LR-0.12)
    • pain radiating to the neck, left shoulder or left arm, and stabbing in nature (LR-0.13)
    • no new changes on ECG (LR-0.15)
    • pain radiating to the neck, left shoulder or left arm, and aged <40 (LR-0.18)
Goldman et al: New England Journal of Medicine 1988; 318 (13): 797-803
Expires July 2003

The study

Setting: emergency departments, six university hospitals, USA

6149 patients (aged mean 56 years, 50% male) presenting with acute chest pain

Excluded if
  • discharged and refused to be followed up within 3 days
  • cardiac arrest
  • left against medical advice



Independent blinded reference standard, applied in all patients from a consecutive appropriate spectrum.
Reference standard:
    • characteristic enzyme rise of creatinine kinase and CK-MB
    • development of new pathological Q-wave and a 25% decrease in the following R-wave
    • Technetium-99m focal uptake in a scintiscan
    • sudden and unexplained death within 72 hours
Diagnostic test: clinical prediction rule- if clinical finding not present, go onto next question. See evidence for CPG
  • The study was a clinical decision analysis validated in an independent set of patients.
  • There were 1379 patients in the derivation set and 4770 in the validation set.
  • 100% followed for one week (at least by telephone conversation): 84% of discharged patents had repeat blood tests.

The evidence

pre-test probability of MI: 12%, (95% CI: 11% to 13%)

diagnostic test MI no MI LR
(95% CI)		 post-test probability
ST elevation or Q waves in 2 or more leads, not known to be old 262 88 22
(17 to 27) 		75%
chest pain began 48 or more hours ago~> ST-T changes of ischaemia or strain, not known to be old 32 117 2.0
(1.5 to 2.9) 		22%
chest pain began 48 or more hours ago~>no changes or old 17 1029 0.12
(0.076 to 0.19) 		1.6%
previous history of angina or MI~>ST-T changes of ischaemia or strain, not known to be old 89 254 2.6
(2.0 to 3.2) 		26%
previous history of angina or MI~>longest pain episode <1 hour 16 388 0.30
(0.18 to 0.49) 		4.0%
previous history of angina or MI~>pain worse than prior angina or the same as a prior MI 34 270 0.92
(0.65 to 1.3) 		11%
previous history of angina or MI~>pain not as bad 2 192 0.076
(0.019 to 0.31) 		1.0%
pain radiates to neck, left shoulder or left arm~>aged <40 3 121 0.18
(0.058 to 0.57) 		2.4%
pain radiates to neck, left shoulder or left arm~>chest pain reproduced by palpation 1 78 0.093
(0.013 to 0.67) 		1.3%
pain radiates to neck, left shoulder or left arm~>pain radiates to back, abdomen or legs 5 60 0.60
(0.25 to 1.5) 		7.7%
pain radiates to neck, left shoulder or left arm~>chest pain stabbing 1 56 0.13
(0.018 to 0.94) 		1.8%
pain radiates to neck, left shoulder or left arm~>chest pain not stabbing 51 243 1.5
(1.1 to 2.0) 		17%
ST-T changes of ischaemia or strain, not known to be old 39 111 2.3
(1.7 to 3.3) 		26%
no changes or old 24 1194 0.15
(0.099 to 0.22) 		2.0%
total 576 4194


diagnostic test MI no MI LR+
(95% CI)		 post-test probability LR-
(95% CI)		 post-test probability
physician's prediction 506 2970 1.2
(1.2 to 1.3) 		15% 0.42
(0.33 to 0.52) 		5%
clinical prediction rule 507 3111 1.2
(1.2 to 1.2) 		14% 0.46
(0.37 to 0.58) 		6%
total 576 4194

  • Using table 1 - If clinical finding not present, go onto next question.
  • For the physician's prediction and the clinical prediction rule, a cut-off of 7% was used for 'significant risk of MI'.

Comments

  1. This clinical prediction rule makes MI less likely, but may not rule out unstable angina. Using a 7% cut-off, one patients in thirteen with a MI would be missed.

Citation

  1. Goldman L, et al: A computer protocol to predict myocardial infarction in emergency department patients with chest pain. New England Journal of Medicine 1988; 318 (13): 797-803
Search Terms:
Contributor: Nick Shenker and Chris Ball, July 2000
Reviewer:

Clinical Question.
    Patient chest pain
    Intervention or Exposure clinical features, ECG features
    Outcome MI, death