Previous Next

PREOPERATIVE AND PREPROCEDURE TESTING

Most hospitals, many anesthesia departments, many outpatient surgical centers, and now many office-based operatories have rather arbitrary rules and recommendations regarding tests that should be performed before elective surgery. With good intentions, anesthesiologists tried to follow those rules, and problems began. The inexpensive


947
multiphasic screening batteries of tests subsequently developed by the Kaiser Hospitals and Health Plan seemed to be the answer to this confusing and arbitrary process. [170] Physicians believed they could now order inexpensive batteries of tests and thus efficiently screen for disease. However, physicians were still trying to determine which tests to order before surgery, and what to do with the unexpectedly abnormal result on the morning of surgery. Kaiser found that this system of preoperative multiphasic screening was not practical.[171] The system produced so many false-positive and false-negative results that the subsequent harm vastly outweighed any possible benefit. Nevertheless, the notion of "the more testing, the better" is still with us.

The Low Predictive Value of an "Abnormal" Laboratory Test Result

Understanding what constitutes an "abnormal" laboratory test result requires an appreciation of the way "normal" values are determined. A normal range is based on the typical distribution of the gaussian curve.[172] For example, assuming a gaussian distribution and hemoglobin values of 13.5 to 16.7 g/dL ("reference range") for healthy men, one can expect 5% of "normal" men to have a test result outside that range.

Of prime importance in preoperative evaluation is knowing the percentage of abnormal laboratory test values that truly indicates disease. If the anesthetic management of a patient is altered because of a test abnormality, that abnormality should indicate a condition (1) that poses a significant risk of preoperative morbidity that can be lessened by preoperative treatment, (2) that cannot be discovered through history-taking and physical examination, and (3) that is sufficiently prevalent in the population to justify the risk of performing the follow-up test. To be cost-efficient, the test should be sufficiently "sensitive" (have "positivity in disease") and sufficiently "specific" (have "negativity in health"). That is, test results should be positive if the patient has disease and negative if the patient is healthy.[172] [173] [174] In fact, what a clinician really wants to know is what a positive or negative test means for the individual patient in front of him or her. The values representing the predictive ability of a positive or negative test ("positive predictive values" and "negative predictive values") depend on the pretest population probability.[172] [173] [174]

We shall now consider the significance of false-positive and false-negative results and the prevalence of disease in the test population in relation to abnormal laboratory test results. (I am indebted to Drs. Jorgen Hilden and Anders Hald, whose comments have helped make this section clearer.) For example, patients with pneumonia would have the notation "pneumonia" (or some significant abnormality) written as the diagnosis on their chest radiograph reports. Let us assume that the specificity of a test (its negativity in health) is 98.3%; that is, 983 of 1,000 people who actually do not have asymptomatic pneumonia will have a comment such as "without evidence of pneumonia" or "normal" written on their chest radiograph reports. Next, let us assume that 0.5% of the asymptomatic population under age 40 who are about to undergo routine elective surgery has pneumonia. Given the preceding assumptions, what is the likelihood that a person whose chest radiograph report reads "pneumonia" would actually have pneumonia?

If we test 100,000 asymptomatic persons, and 0.5% are assumed to be diseased, this means that 500 people would have undetected pneumonia. If the sensitivity of chest radiographs to pneumonia is assumed to be 75%, 375 of these people would have abnormal radiographs. Then, if specificity is assumed to be 98.3%, 97,809 of the 99,500 healthy people would have normal results on chest radiographs. This means that 1,691 (1.7%) would have abnormal radiographic results. Thus, of 2,066 patients having a diagnosis of pneumonia based on chest radiographs, 1,691 (82%) of the results would be falsely positive. Therefore, it is entirely possible that 82% of the chest radiographs indicating "infiltrate compatible with pneumonia" in otherwise asymptomatic persons would actually be radiographs of totally healthy people. Expressed in another way, when the above assumptions discussed are applied, the likelihood that an asymptomatic person would actually have pneumonia when the chest radiograph contains that notation is only 18%; that is, for this patient group, the predictive value of a positive test (the "positive predictive value") is only 18%.

Only patients with abnormal test results who actually have disease ("true positives") benefit from laboratory testing. Let us assume that 2.2% of the chest radiographs in the under-age-40 population are positive, and that, for each true positive, perioperative mortality decreases by 50%. If we use the 82% false-positive rate derived previously, the number of patients benefiting per 1,000 radiographs is 3.9 (true positives per 1,000 = all positives − false positives; i.e., [2.2% × 1,000] − [82% percent − 2.2 percent × 1,000] = 3.9 patients). Therefore, a reduction in operative mortality of 50%, or 1 per 20,000 (i.e., 3.9 − 0.5 × 0.00005), gives 0.000095 fewer deaths per 1,000 operations when preoperative chest radiographs are obtained.

Hilden and Hald (personal communication) question whether this is a logical calculation. For example, one study analyzed outcomes for asymptomatic patients.[175] The death rate for this group was 1 in 10,000. Of these deaths, more than 90% were caused by avoidable problems unrelated to unknown diseases. Therefore, at most, fewer than 10% of the deaths of asymptomatic individuals could be attributed to preoperative conditions, and probably fewer than 10% of those deaths (<1% of the total, or 1 in 1,000,000) would be related to pneumonia. Thus, the intuitive value is similar to what we have estimated. Furthermore, from an examination of the data in Table 25-5 and Table 25-10 , it is evident that few chest radiographs in the under-40 asymptomatic population led to benefit. In addition, the harm from these tests has not yet been considered.

Translating this figure into the present value for years of life saved per 1,000 chest radiographs yields the following: 0.000095 fewer deaths per 1,000 operations × 22.62 years saved per life saved = 0.0022 years of life (22.62 is the present value of 60 more years of life for a 20-year-old, per Neuhauser[172] ). At the University of Chicago, the 0.0022 years of life saved would cost $78,000 (an anteroposterior and lateral chest radiograph cost $78, not including fees for consultations, repeated radiographs, or other laboratory tests or procedures). Therefore, each year of life saved by


948

TABLE 25-10 -- Screening chest radiographs: Incidence of abnormal test results, the discovery of which may change management of anesthesia
Age(y) Series Patients Examined (n) Abnormalities * (%) New Abnormalities (%)
0–14 Farnsworth et al[176] 350 8.9  0.3
0–18 Brill et al[177] 1,000 1.9  0.7
0–19 Sagel et al[178] 521 0    0
0–19 Sane et al[179] 1,500 5.4  2.2
0–19 Wood and Hoekelman 749 4.7  1.2
1–20 Rees et al[180] 46 0    0
20–29 Sagel et al[178] 894 1  
21–30 Rees et al[180] 62 3  
30 Loder[181] 437 10.1  0.2
30 Hubbell et al[182] 12 0    0
30 Maigaard et al[183] 1,256 4.5  0
0–39 Umbach et al[166] 305 3.0
30–39 Sagel et al[178] 942 2.3
40 Catchlove et al[184] 29 0    0
40 Collen et al[170] 15,978 2.1
40 Sagel et al[178] 2,357 1.3  1.3
40 McKee and Scott[112] , 26 7.7  3.9
40 Combined[112] [176] [177] [178] [179] [181] [182] [183] , § 6,787 4.0  0.8
40–49 Sagel et al[178] 928 7.1
<40 Velanovich[185] 3.1
40–49 Umbach et al[166] 290 6.2
40–59 Collen et al[170] 21,489 7.4
41–50 Hubbell et al[182] 28 17.9  0
41–50 Rees et al[180] 119 19  
41–50 McKee and Scott[112] , 53 17    0
30–69 Loder[181] 515 6.0
31–40 Rees et al[180] 93 13  
31–40 Hubbell et al[182] 22 22.5  4.5
40–60 Velanovich[185] NA 23.9
≧40 Sagel et al[178] 3,689 23.9  6.0
≧40 Catchlove et al[184] 50 0    0
≧40 Thomsen et al[186] 1,823 2.3  0.2
50–59 Umbach et al[166] 247 10.9
50–59 Sagel et al[178] 733 20.3
51–60 Hubbell et al[182] 87 36.8  4.4
51–60 Rees et al[180] 121 40.0
51–60 McKee and Scott[112] , 85 40.0  0
>55 Gupta et al[187] , 346 5.5  0.9
60–69 Umbach et al[166] 202 13.3
60–69 Sagel et al[178] 977 29.7
61–70 Boghosian and Mooradian[188] 78 49  
>60 Boghosian and Mooradian[188]



  Without risk factor 44 34.0

  With risk factors 92 62.0
>60 Collen et al[170] 7,196 19.2
>60 Hubbell et al[182] 145 44.1  4.8
>60 McKee and Scott[112] , 163 44.2  1.9
>60 Velanovich[185] NA 31.9
61–70 Rees et al[180] 134 43.3
61–70 Hubbell et al[182] 94 36.2  5.4
>65 Sewell et al[189] 28 21  
≧69 Loder[181] 48 72.9
≧?70 Wolf-Klein et al[137] 500 1.9
≧70 Sagel et al[178] 832 41.7
≧70 Törnebrandt and Fletcher[180] 100 37   8.1?
≧70 Boghosian and Mooradian[188]



  Without risk factors 58 59  

  With risk factors 45 64  
≥71 Levinstein et al[156] 121 84.4  0.9
70–79 Umbach et al[166] 110 27.2
71–80 Rees et al[180] 76 61.8
71–80 Hubbell et al[182] 28 57.1  0
>80 Umbach et al[166] 21 33.3
>80 Hubbell et al[182] 23 60.9  8.7
20–89 Fink et al[191] 127 46  
74–97 Domoto et al[136] 69 72.5 33
>81 Rees et al[180] 16 68.8
?0–?90 Wiencek et al[192] 403 ≤25.1  2.4
0–90 Delahunt and Turnbull[139] 860  0
24–90 Tape and Mushlin[149] , 318 33    0
0–?90 Petterson and Janower[193] 1,530 9.8  1.3
0–?90 Turnbull and Buck[113] , 691 5.5
0–?90 Royal College[184] 3,052 3.8
0–?90 Rucker et al[195] , 371 0.3
0–?90 Blery et al[128] 2,765 0.7  0.1
0–90+ Weibman et al[196] , 734 5.0
0–90+ Muskett and McGreevy[115] 119 29.4  5.0
0–90+ Gagner and Chiasson[197] 1,000 7.4
*These data constitute an edited summary of the data presented by various articles, edited to select abnormalities that might change management of anesthesia.
†Abnormalities not already known or suspected by history or physical examination.
‡Patients were asymptomatic.
§Combined studies in younger-than-40 population excluding two studies, those by Rees et al[180] and Collen et al.[170]
‖0% changed treatment.
¶Tape and Mushlin[149] studied vascular surgery patients; Turnbull and Buck,[113] cholecystectomy patients; and Weibman and colleagues,[196] cancer patients.




obtaining chest radiographs costs about $35,500,000 (78,000 divided by 0.0022).

However, just as there are other costs (e.g., pursuing some false-positive chest shadows will result in computed tomographic needle biopsies and lobectomies in totally healthy patients[109] [110] [111] [112] ), there are other benefits (e.g., treatment of some patients having solitary nodules or mediastinal masses may prolong life). Let us arbitrarily assume that these costs and benefits are equal. One is forced to conclude that screening for an asymptomatic disease having a low prevalence rate is a very expensive and possibly risky procedure.

Previous Next