Paul S. Heckerling

Clinical Prediction Rule for Pulmonary Infiltrates

OBJECTIVE To derive and validate a clinical rule for predicting pneumonic infiltrates in adult patients with acute respiratory illness. DESIGN Prevalence studies in three settings. SETTING Emergency departments of the University of Illinois Hospital at Chicago, the University of Nebraska Medical Center at Omaha, and the Medical College of Virginia at Richmond. PATIENTS Symptoms, signs, comorbidity data, and chest roentgenogram results were recorded for 1134 patients from Illinois (the derivation set), 150 patients from Nebraska, and 152 patients from Virginia (the validation sets). All patients presented to the emergency department and had a chest roentgenogram to evaluate fever or respiratory complaints. MEASUREMENTS AND MAIN RESULTS Within the training set, temperature greater than 37.8 degrees C, pulse greater than 100 beats/min, rales, decreased breath sounds, and the absence of asthma were identified as significant predictors of radiographically proved pneumonia in a stepwise logistic regression model (P = 0.001). The logistic rule discriminated patients with and without pneumonia in the training set with a receiver operating characteristic (ROC) area of 0.82. In the validation sets, the rule discriminated pneumonia and nonpneumonia with ROC areas of 0.82 and 0.76 after adjusting for differences in disease prevalence (P greater than 0.2 compared with the training set). The predicted probability of having pneumonia for patients with different clinical findings corresponded closely with the incidence of pneumonia among patients with such findings in the three settings. CONCLUSIONS Among adults presenting with acute respiratory illness, a prediction rule based on clinical findings accurately discriminated patients with and without radiographic pneumonia, and was used in two other samples of patients without significant decrement in discriminatory ability. This rule can be used by physicians to develop more effective strategies for detecting pneumonia and for helping to determine the need for radiologic study among patients with acute respiratory disease.

Do physicians know when their diagnoses are correct? Implications for decision support and error reduction.

AbstractOBJECTIVE: This study explores the alignment between physicians’ confidence in their diagnoses and the “correctness” of these diagnoses, as a function of clinical experience, and whether subjects were prone to over-or underconfidence. DESIGN: Prospective, counterbalanced experimental design. SETTING: Laboratory study conducted under controlled conditions at three academic medical centers. PARTICIPANTS: Seventy-two senior medical students, 72 senior medical residents, and 72 faculty internists. INTERVENTION: We created highly detailed, 2-to 4-page synopses of 36 diagnostically challenging medical cases, each with a definitive correct diagnosis. Subjects generated a differential diagnosis for each of 9 assigned cases, and indicated their level of confidence in each diagnosis. MEASUREMENTS AND MAIN RESULTS: A differential was considered “correct” if the clinically true diagnosis was listed in that subject’s hypothesis list. To assess confidence, subjects rated the likelihood that they would, at the time they generated the differential, seek assistance in reaching a diagnosis. Subjects’ confidence and correctness were “mildly” aligned (k=.314 for all subjects, .285 for faculty, .227 for residents, and .349 for students). Residents were overconfident in 41% of cases where their confidence and correctness were not aligned, whereas faculty were overconfident in 36% of such cases and students in 25%. CONCLUSIONS: Even experienced clinicians may be unaware of the correctness of their diagnoses at the time they make them. Medical decision support systems, and other interventions designed to reduce medical errors, cannot rely exclusively on clinicians’ perceptions of their needs for such support.

Predictors of Occult Carbon Monoxide Poisoning in Patients with Headache and Dizziness

Headache and dizziness occur at carboxyhemoglobin levels of greater than 10%. We studied 89 patients with headache or dizziness for evidence of carbon monoxide exposure. The mean carboxyhemoglobin level was 2.87%. Number of cigarettes smoked per day (r = 0.471; p less than 0.00002), use of gas kitchen stoves for heating purposes (r = 0.252, p less than 0.02), problems with the home heating system (r = 0.278, p less than 0.01), and cohabitants with concurrent headache or dizziness (r = 0.427, p less than 0.01) correlated with carboxyhemoglobin levels. Multiple regression analysis identified number of cigarettes smoked daily, use of stoves for heat, and concurrently symptomatic cohabitants as significant predictors of carboxyhemoglobin level (F = 13.939 [3, 85]; p less than 0.01). Obtaining carboxyhemoglobin levels from patients who used stoves for heat or had similarly affected cohabitants identified 4 of 4 patients with levels greater than 10% (sensitivity, 100%) and excluded 47 of 85 patients with lower levels (specificity, 55.3%). The 4 patients with carboxyhemoglobin levels in excess of 10% may represent occult carbon monoxide poisoning in this population.

Parametric three-way receiver operating characteristic surface analysis using Mathematica

Three-way receiver operating characteristic (ROC) surface analysis involves the calculation of a volume under an ROC surface (VUS), which is a measure of discriminatory accuracy of 2 diagnostic tests for 3 diseases. Nonparametric methods for calculating VUS and its standard error have been developed. The author presents the code for roc3D, a Mathematica computer program for performing parametric ROC surface analysis. roc3D calculates VUS assuming a multinormal distribution of test results in the 3 diseased populations, provides user-specified pointwise confidence limits for VUS, and displays a 3-dimensional plot of the ROC surface. Limitations of the roc3D program are discussed.

Occult carbon monoxide poisoning: A cause of winter headache

Headache is an early symptom of carbon monoxide (CO) poisoning, occurring at carboxyhemoglobin (COHb) levels of greater than 10%. We investigated 37 patients presenting to an emergency department during the winter-heating season with a complaint of headache for evidence of CO exposure. Seven of the 37 patients (18.9%) had COHb levels in excess of 10%, with a mean of 14.0%. Three of these seven patients (42.9%) used gas space heaters or kitchen stoves to heat their homes, and three (42.9%) had cohabitants at home with headache. Of the 30 patients with COHb levels of less than 10% (mean level, 3.2%), only four (13.3%) used gas space heaters and kitchen stoves for heat (P = NS), and none (0%) had cohabitants with headache (P = .0045). In six of the seven patients with elevated COHb levels (85.7%), gas company investigation or historical information obtained retrospectively revealed a definite or probably toxic CO exposure. In none of these patients had the diagnosis of CO poisoning been suspected by the emergency department physician or the patient prior to obtaining the result of the COHb level. Occult CO poisoning may be an important cause of winter headache.

Predictors of urinary tract infection based on artificial neural networks and genetic algorithms

BACKGROUND Among women who present with urinary complaints, only 50% are found to have urinary tract infection. Individual urinary symptoms and urinalysis are not sufficiently accurate to discriminate those with and without the diagnosis. METHODS We used artificial neural networks (ANN) coupled with genetic algorithms to evolve combinations of clinical variables optimized for predicting urinary tract infection. The ANN were applied to 212 women ages 19-84 who presented to an ambulatory clinic with urinary complaints. Urinary tract infection was defined in separate models as uropathogen counts of > or =10(5) colony-forming units (CFU) per milliliter, and counts of > or =10(2) CFU per milliliter. RESULTS Five-variable sets were evolved that classified cases of urinary tract infection and non-infection with receiver-operating characteristic (ROC) curve areas that ranged from 0.853 (for uropathogen counts of > or =10(5) CFU per milliliter) to 0.792 (for uropathogen counts of > or =10(2) CFU per milliliter). Predictor variables (which included urinary frequency, dysuria, foul urine odor, symptom duration, history of diabetes, leukocyte esterase on urine dipstick, and red blood cells, epithelial cells, and bacteria on urinalysis) differed depending on the pathogen count that defined urinary tract infection. Network influence analyses showed that some variables predicted urine infection in unexpected ways, and interacted with other variables in making predictions. CONCLUSIONS ANN and genetic algorithms can reveal parsimonious variable sets accurate for predicting urinary tract infection, and novel relationships between symptoms, urinalysis findings, and infection.

Occult carbon monoxide poisoning: Validation of a prediction model

Headache and dizziness are early symptoms of carbon monoxide poisoning, occurring at carboxyhemoglobin levels of greater than 10 percent. Previously, it was shown that among patients presenting to an emergency department during the winter with headache or dizziness, an algorithm for obtaining carboxyhemoglobin levels on patients who used gas stoves for heating purposes or who had similarly affected cohabitants correctly identified all patients with carboxyhemoglobin levels greater than 10 percent. To test the validity of this retrospectively derived rule, 65 patients were studied who were unaware of any carbon monoxide exposure and who presented during the winter of 1986-1987 with headache or dizziness. The algorithm correctly identified three of four patients with carboxyhemoglobin levels greater than 10 percent (sensitivity = 75 percent) and correctly excluded 45 of 61 patients with lower levels (specificity = 74 percent). The presence of symptomatic cohabitants alone was an equally sensitive (75 percent) but more specific (90 percent) marker for elevated carboxyhemoglobin levels. When data from the two cohorts were combined, stepwise multiple regression identified number of cigarettes smoked daily (F = 8.66) and concurrently symptomatic cohabitants (F = 34.71) as significant predictors of the carboxyhemoglobin level. It is concluded that a retrospectively derived rule correctly identified most cases of occult carbon monoxide poisoning when applied prospectively, and that the presence of similarly affected cohabitants was the most reliable marker for a carbon monoxide-mediated illness.

Ethylene glycol poisoning with a normal anion gap caused by concurrent ethanol ingestion: Importance of the osmolal gap

Ethylene glycol poisoning classically presents as a metabolic acidosis with an increased anion gap. Metabolism of ethylene glycol to organic acids, and increased production of lactate, are responsible for the increased gap. We report the case of an alcohol user who consumed ethanol and ethylene glycol concurrently, and presented without acidosis, with a normal anion gap. Several hours later, when his serum ethanol level had declined, he developed severe acidosis with an elevated anion gap. An increased osmolal gap, not accounted for by the serum ethanol level, proved to be an important clue to the diagnosis. In this patient, ingestion of ethanol inhibited the hepatic metabolism of ethylene glycol to organic acids, obscuring the diagnosis. In intoxicated alcohol users, even in the absence of metabolic acidosis, serum osmolality measurements and calculation of the osmolal gap may facilitate the rapid diagnosis of ethylene glycol poisoning.

Preferences of pregnant women for amniocentesis or chorionic villus sampling for prenatal testing: Comparison of patients' choices and those of a decision-analytic model

Decision analytic models have suggested that the choice of amniocentesis or chorionic villus sampling for prenatal genetic testing is a utility-driven decision. We compared preferences for prenatal testing among 156 pregnant women who had chosen either amniocentesis (n = 82) or chorionic villus sampling (n = 74) for the indication of maternal age. We also compared their choices with those of a decision-analytic model based on their preferences, and age-specific rates of spontaneous abortion and chromosomal abnormalities. Preferences were assessed using written scenarios describing potential outcomes of prenatal testing, and were recorded on linear rating scales. The differences in preference ratings for first- vs second-trimester prenatal diagnosis of a normal child (4.2 vs -1.6, p = 0.0004), and for first- vs second-trimester abortion of an abnormal fetus (4.4 vs -1.6, p = 0.01), were significantly greater among women choosing chorionic villus sampling than among women choosing amniocentesis. There were no significant differences between chorionic villus sampling and amniocentesis patients in their preference ratings for test-related miscarriage, disconfirmed results at pregnancy termination, or maternal morbidity from therapeutic abortion. After adjusting for demographic and obstetric factors, the difference in preferences for early vs late prenatal diagnosis was an independent predictor of the choice of chorionic villus sampling in a multivariate model. Among women whose decision analyses selected amniocentesis, 56.8% had chosen amniocentesis, and among women whose analyses selected chorionic villus sampling, 63.2% had chosen chorionic villus sampling (p = 0.05). We conclude that the preferences of pregnant women for the outcomes of prenatal testing were associated with their choice of amniocentesis or chorionic villus sampling. In addition, the choice of prenatal test made by the majority of women was concordant with that of a decision-analytic model that incorporated their preferences. Nevertheless, because many women made choices that were discordant with their decision-analytic results, further research into the bases for their choices is warranted.

Occult carbon monoxide poisoning in patients with neurologic illness

To investigate occult carbon monoxide poisoning in patients with neurologic illness, we prospectively studied 168 patients who presented to the emergency department between December 1987 and February 1988 with neurologic symptoms for evidence of carbon monoxide exposure. Patients with known carbon monoxide poisoning were excluded. The mean carboxyhemoglobin level was 3.1 percent; there were no significant differences in carboxyhemoglobin between categories of neurologic illness (F(5,162) = 1.35; p less than 0.25). Five patients (3 percent) had a carboxyhemoglobin greater than 10 percent, with levels ranging from 11.7 percent to 29.5 percent. After controlling for the effects of active and passive exposure to cigarette smoke, problems with the home heating system (odds ratio 9.6; p less than 0.03) and the presence of cohabitants with concurrent headache or dizziness (odds ratio 21.6; p less than 0.0001) were associated with an increased risk of a carboxyhemoglobin greater than 10 percent. A rule for obtaining carboxyhemoglobin tests only on patients who used gas stoves for heat or who had symptomatic cohabitants would have correctly identified all patients with carboxyhemoglobins greater than 10 percent, correctly excluded 77 percent of patients with lower levels, and eliminated the need for testing in 75 percent of cases. We conclude that unrecognized carbon monoxide poisoning occurs in a small but important fraction of patients with wintertime neurologic illness and can be identified by a characteristic risk factor profile.