Issues in studying the effectiveness of health services for children

OBJECTIVES

To discuss issues in studying the effectiveness of health services for children, suggest areas in which more research is needed, and recommend strategies for future research.

PRINCIPAL FINDINGS

Issues that should be considered include the choice of perspective, which will help determine the interventions studied and the measures of effectiveness and cost-effectiveness chosen. Unique challenges in this area include the fact that serious measurable morbidity is relatively uncommon in children, that causal relationships between services and outcomes may be difficult to establish, and that standard measures of cost-effectiveness may fail to accurately measure important benefits, such as reduced parental anxiety. More research is needed on high-risk and health-promoting behaviors, on critical parent behaviors, on classifying children by vulnerability status, on modes of delivery of preventive care, and on violence prevention.

RECOMMENDATIONS

Group-randomized designs and observational research designs that take advantage of natural variations in practice may be increasingly useful in effectiveness studies. Parent- and patient-reported measures of health status and quality of life should be made briefer and more practical for routine use, and better measures of cost-effectiveness are needed. Future research efforts can best be supported by the concerted efforts of various constituencies, including health plans, providers, patients, researchers, and the government.

Jaundice in full-term and near-term babies who leave the hospital within 36 hours. The pediatrician's nemesis

There has been an increase in hyperbilirubinemia in the newborn population and, perhaps, an increase in bilirubin encephalopathy. The early discharge of newborns from hospital has made it necessary for us to reorient our thinking about bilirubin levels in the first 24 to 48 hours of life and alter our approach to follow-up. The pediatrician must evaluate and follow infants who have risk factors for the development of severe hyperbilirubinemia, paying particular attention to the breast-feeding, near-term infant.

Association between Cryptosporidium infection and animal exposure in HIV-infected individuals

Cryptosporidium causes a debilitating illness in immunocompromised individuals, yet the source of sporadic Cryptosporidium infections is unknown. Because early cases of cryptosporidiosis were associated with animals, and pets are a source of companionship to individuals with AIDS, determination of the risk of cryptosporidiosis associated with pets is important. To assess this risk, we conducted a case-control study of HIV-infected individuals with and without cryptosporidiosis. No statistically significant difference in the rate of overall pet ownership, cat ownership, or bird ownership was found between the two groups. Dog ownership reached borderline statistical significance; 15 of 48 (31%) cases owned a dog, compared with 17 of 99 (17%) controls (odds ratio [OR] = 2.19; p = .05; 95% confidence interval [CI], 0.9-5.3). No statistically significant differences between cases and controls were found in the frequency of surface water ingestion, rural exposure, travel history, or contact with diarrheic individuals. Our findings indicate that pets do not represent a major risk factor for acquisition of Cryptosporidium for HIV-infected individuals. Whether dog ownership presents a minimal risk for cryptosporidiosis needs further investigation.

Evaluation of the erythrocyte sedimentation rate in children presenting with limp, fever, or abdominal pain

An erythrocyte sedimentation rate (ESR) is commonly ordered as part of the evaluation of patients with nonspecific but potentially serious symptoms. To investigate the performance of ESR in this setting, we used a computerized database and medical chart review to identify children (n=299) with ESR done for a previously undiagnosed condition. Medical records were reviewed to determine symptoms at presentation, referral status, and subsequent diagnoses, which were classified as serious (n=93) or benign (n=206). We found that serious underlying disease was about 7 times as likely in patients with ESR>50 mm/hr (57/102) than in patients with ESR<20 mm/hr (7/89). Although the prevalence of serious disease was higher among referral patients, the likelihood ratios were similar for referral and primary-care patients. An erythrocyte sedimentation rate greater than 50 mm/hr was most informative in patients presenting with limp (likelihood ratio [LR] =8.2) and abdominal pain (LR=6.0) and least informative in patients presenting with fever (LR=2.5). On the other hand, an ESR<20 mm/hr is reassuring in patients presenting with fever (LR=0) or limp (LR=0.3), but not in patients presenting with abdominal pain (LR=0.8). An ESR between 20 and 50 mm/hr (23% of the patients) provided little information (LR 1.2-1.5) in each of the three groups. These results suggest that the ESR often provides useful information about the likelihood of serious illness among children presenting with worrisome but nonspecific symptoms, in particular in patients presenting with limp.

Use of commercial record linkage software and vital statistics to identify patient deaths

We evaluate the ability of a microcomputer program (Automatch) to link patient records in our hospital's database (N = 253,836) with mortality files from California (N = 1,312,779) and the U.S. Social Security Administration (N = 13,341,581). We linked 96.5% of 3,448 in-hospital deaths, 99.3% for patients with social security numbers. None of 14,073 patients known to be alive (because they were subsequently admitted) was linked with California deaths, and only 6 (0.1%) of 6,444 were falsely identified as dead in the United States file. For patients with unknown vital status but items in the database likely to be associated with high 3-year mortality rates, we identified death records of 88% of 494 patients with cancer metastatic to the liver, 84% of 164 patients with pancreatic cancer, and 91% of 126 patients with CD4 counts of less than 50. Hospital data can be accurately linked with state and national vital statistics using commercial record linkage software.

Carcinogenicity of lipid-lowering drugs

OBJECTIVE

To review the findings and implications of studies of rodent carcinogenicity of lipid-lowering drugs.

DATA SOURCES

Summaries of carcinogenicity studies published in the 1992 and 1994 Physicians' Desk Reference (PDR), additional information obtained from the US Food and Drug Administration, and published articles identified by computer searching, bibliographies, and consultation with experts.

STUDY SAMPLE

We tabulated rodent carcinogenicity data from the 1994 PDR for all drugs listed as "hypolipidemics." For comparison, we selected a stratified random sample of antihypertensive drugs. We also reviewed methods and interpretation of carcinogenicity studies in rodents and results of clinical trials in humans.

DATA SYNTHESIS

All members of the two most popular classes of lipid-lowering drugs (the fibrates and the statins) cause cancer in rodents, in some cases at levels of animal exposure close to those prescribed to humans. In contrast, few of the antihypertensive drugs have been found to be carcinogenic in rodents. Evidence of carcinogenicity of lipid-lowering drugs from clinical trials in humans is inconclusive because of inconsistent results and insufficient duration of follow-up.

CONCLUSIONS

Extrapolation of this evidence of carcinogenesis from rodents to humans is an uncertain process. Longer-term clinical trials and careful postmarketing surveillance during the next several decades are needed to determine whether cholesterol-lowering drugs cause cancer in humans. In the meantime, the results of experiments in animals and humans suggest that lipid-lowering drug treatment, especially with the fibrates and statins, should be avoided except in patients at high short-term risk of coronary heart disease.

Carcinogenicity of lipid-lowering drugs

OBJECTIVE

To review the findings and implications of studies of rodent carcinogenicity of lipid-lowering drugs.

DATA SOURCES

Summaries of carcinogenicity studies published in the 1992 and 1994 Physicians' Desk Reference (PDR), additional information obtained from the US Food and Drug Administration, and published articles identified by computer searching, bibliographies, and consultation with experts.

STUDY SAMPLE

We tabulated rodent carcinogenicity data from the 1994 PDR for all drugs listed as "hypolipidemics." For comparison, we selected a stratified random sample of antihypertensive drugs. We also reviewed methods and interpretation of carcinogenicity studies in rodents and results of clinical trials in humans.

DATA SYNTHESIS

All members of the two most popular classes of lipid-lowering drugs (the fibrates and the statins) cause cancer in rodents, in some cases at levels of animal exposure close to those prescribed to humans. In contrast, few of the antihypertensive drugs have been found to be carcinogenic in rodents. Evidence of carcinogenicity of lipid-lowering drugs from clinical trials in humans is inconclusive because of inconsistent results and insufficient duration of follow-up.

CONCLUSIONS

Extrapolation of this evidence of carcinogenesis from rodents to humans is an uncertain process. Longer-term clinical trials and careful postmarketing surveillance during the next several decades are needed to determine whether cholesterol-lowering drugs cause cancer in humans. In the meantime, the results of experiments in animals and humans suggest that lipid-lowering drug treatment, especially with the fibrates and statins, should be avoided except in patients at high short-term risk of coronary heart disease.

Kernicterus in otherwise healthy, breast-fed term newborns

OBJECTIVE

To document the occurrence of classical kernicterus in full-term, otherwise healthy, breast-fed infants.

METHODS

We reviewed the files of 22 cases referred to us by attorneys throughout the United States during a period of 18 years, in which neonatal hyperbilirubinemia was alleged to be responsible for brain damage in apparently healthy, nonimmunized, full-term infants. To qualify for inclusion, these infants had to be born at 37 or more weeks' gestation, manifest the classic signs of acute bilirubin encephalopathy, and have the typical neurologic sequelae.

RESULTS

Six infants, born between 1979 and 1991, met the criteria for inclusion. Their peak recorded bilirubin levels occurred 4 to 10 days after birth and ranged from 39.0 to 49.7 mg/dL. All had one or more exchange transfusions. One infant had an elevated reticulocyte count (9%) but no other evidence of hemolysis. The other infants had no evidence of hemolysis, and no cause was found for the hyperbilirubinemia (other than breast-feeding).

CONCLUSIONS

Although very rare, classic kernicterus can occur in apparently healthy, full-term, breast-fed newborns who do not have hemolytic disease or any other discernible cause for their jaundice. Such extreme elevations of bilirubin are rare, and we do not know how often infants with similar serum bilirubin levels escape harm. We also have no reliable method for identifying these infants early in the neonatal period. Closer follow-up after birth and discharge from the hospital might have prevented some of these outcomes, but rare, sporadic cases of kernicterus might not be preventable unless we adopt an approach to follow-up and surveillance of the newborn that is significantly more rigorous than has been practiced. The feasibility, risks, costs, and benefits of this type of intervention need to be determined.

Problems with the report of the Expert Panel on blood cholesterol levels in children and adolescents

An Expert Panel convened by the National Cholesterol Education Program has recommended selective screening and treatment of children for high blood cholesterol levels, based on family history of cardiovascular disease or high blood cholesterol. This recommendation is problematic for several reasons. First, the recommended diets are likely to cause only a slight decrease in low-density lipoprotein cholesterol levels, the projected benefits of which will be offset by a similar decrease in high-density lipoprotein cholesterol levels. Lack of efficacy of the recommended diets could lead to use of more restrictive diets or to cholesterol lowering drugs. Second, even under optimistic assumptions, beneficial effects of cholesterol intervention will be small and delayed for many decades. As a result, childhood cholesterol-lowering efforts will not be cost-effective. Third, the Expert Panel's recommendations do not address important gender differences. Girls have higher average cholesterol levels than boys. They will therefore qualify for more dietary and drug treatment despite their lower age-adjusted risk of heart disease and the lack of association between cholesterol levels and cardiovascular mortality in women. Finally, recent evidence from randomized trials, cohort studies, and animal experiments suggests that cholesterol lowering may have serious adverse effects. This evidence was not discussed in the Expert Panel's report. Given current evidence, any screening and treatment of children for high blood cholesterol levels is, at best, premature.

Screening for major depression in Vietnamese refugees: a validation and comparison of two instruments in a health screening population

OBJECTIVES

1) Using standard cutoffs, to determine the accuracy of two Vietnamese-language depression screening instruments for major depression in a nonpsychiatric setting, 2) to examine the utility of other cutoffs, and 3) to compare the instruments' overall accuracies.

DESIGN

1) A research assistant administered the Vietnamese Depression Scale (VDS) and the Indochinese Hopkins Symptom Checklist Depression Subscale (HSCL-D) to all subjects. 2) The "gold standard" was determined by a native Vietnamese-speaking psychiatrist using a written translation of a standard semistructured clinical interview.

SETTING

A health screening clinic at a large public hospital.

PATIENTS

A convenience sample of 206 newly arrived adult Vietnamese refugees undergoing routine, mandatory health screening.

RESULTS

The psychiatrist diagnosed 7% of the refugees as having major depression. At standard cutoffs, the VDS had a 64% sensitivity, a 98% specificity, a 75% positive predictive value, and a 97% negative predictive value. Corresponding results for the HSCL-D were 86%, 93%, 48%, and 99%. More than half of the patients who had false-positive results had other clinical disorders. For each instrument, adjusting the cutoff improved sensitivity and positive predictive value. Receiver operating characteristic (ROC) curve analysis showed no difference in accuracy between the two instruments. Each instrument took approximately 5-10 minutes to administer.

CONCLUSIONS

These instruments accurately identified Vietnamese refugees with major depression and should be of use to clinicians in primary care settings. Standard cutoffs may need to be adjusted in nonpsychiatric settings.

Problems with serum IgM as a screening test for congenital infection

The purpose of this study was to determine the usefulness of the total serum IgM level as a screening test for congenital infection in asymptomatic or mildly symptomatic infants. A retrospective medical record review was performed on 168 infants in whom the serum IgM was measured as a screen for congenital infection. The indications for testing, the yield of testing, and the adequacy of follow-up of abnormal values were examined. Only one infant was diagnosed with a congenital infection which was not specifically suspected prior to screening; this was a case of congenital cytomegalovirus (CMV). Inappropriate screening was frequently performed in infants in whom indications for specific evaluation were present. Appropriate follow-up testing was performed in only 30% (seven of 23) of the infants with elevated serum IgM who received their pediatric care at our institution. Because of the low yield (< 1%) and lack of follow-up shown in this study, as well as poor sensitivity, serum IgM was not a useful screening test for congenital infection in our institution.

Yield of reticulocyte counts and blood smears in term infants

Reticulocyte counts and blood smears are commonly recommended to evaluate jaundice in newborns. To investigate the results and diagnostic yield of these tests, we reviewed a computerized database and medical records of term newborns who had reticulocyte counts (n = 799) or blood smears (n = 781) within the first week after birth at two hospitals. Nearly a threefold difference was noted in reticulocyte counts between the two hospitals (median 8.0% vs 2.8%; P < .0001), apparently due to differences in laboratory methods. Among the patients with "abnormal" reticulocyte counts or blood smears (n = 192), isoimmunization was diagnosed in 54, presumed hemolysis of unknown etiology in two, G6PD deficiency in one, and pyropoikilocytosis in one. We conclude that better standardization of reticulocyte counts is needed. When ordered as screening tests for hemolysis in jaundiced infants, reticulocyte counts and blood smears seldom lead to diagnoses of hemolysis other than isoimmunization.

Obstacles and approaches to clinical database research: experience at the University of California, San Francisco

With increasing availability of clinical data in machine-readable form, and decreasing cost of storing and manipulating that data, retrospective research using clinical databases has become more feasible. Nonetheless, much of the potential for clinical research using these data remains unrealized. Obstacles to clinical database research include difficulty accessing data, difficulty using retrospective data to draw valid inferences about medical tests and treatments, and a shortage of investigators trained and interested in using a clinical database to answer their questions. At the University of California, San Francisco, we have developed a Clinical Database Research Program (CDRP) to try to overcome these obstacles. The CDRP maintains a relational database of patient data obtained from diverse sources and a small staff dedicated to providing such data to researchers. The CDRP staff also provides support for design and analysis of studies using the database--the development of methods for such studies is our primary research interest. Finally, to increase the number of investigators using the database for research, we are integrating training in clinical epidemiology and clinical research methods into residency and fellowship training, and offering an elective in clinical database research for trainees who wish to undertake a specific project.

Neonatal hyperbilirubinemia and long-term outcome: another look at the Collaborative Perinatal Project

OBJECTIVE

To examine the association between neonatal bilirubin levels and subsequent neurodevelopmental outcome.

DESIGN

Prospective cohort study.

SETTING

12 US medical centers from 1959 (first births) to 1974 (last follow-up).

PARTICIPANTS

41,324 singleton white or black infants with birth weight > or = 2500 g who had neonatal bilirubin measurements recorded and survived at least 1 year.

MAIN OUTCOME MEASURES

Wechsler Intelligence Scale for Children Intelligence Quotient (IQ) at age 7 years, blinded neurologic examination at age 7 years, and sensorineural hearing loss at age 8 years.

RESULTS

There was no association between IQ and bilirubin. For example, comparing children who had maximum bilirubin levels > or = 342 mumol/L (20 mg/dL) with those who had lower bilirubin levels, adjusted mean IQs were 105.0 and 103.4 in whites (difference + 1.6; 95% confidence interval [CI]: -0.4 to +3.5) and 91.0 and 93.3 in blacks (difference -2.3; 95% CI: -4.8 to +0.2). Abnormal neurologic examination results were reported in 12 of 268 children (4.5%) with bilirubin > or = 342 mumol/L (20 mg/dL) compared with 1249 of 33,004 children (3.8%) with lower levels (relative risk [RR] = 1.2; 95% CI: 0.7 to 2.1). The frequency of abnormal or suspicious neurologic examination results increased in a stepwise fashion with increasing bilirubin level (P < .001), from 4346/29,258 (14.9%) of those with bilirubin levels < 171 mumol/L (10 mg/dL) to 60/268 (22.4%) of those with bilirubin levels. > or = 342 mumol/L (20 mg/dL), apparently due to increasing minor motor abnormalities at higher bilirubin levels. Sensorineural hearing loss was not associated with high bilirubin levels (RR = 1.0; 95% CI: 0.3 to 3.0).

CONCLUSIONS

Neonatal bilirubin levels seem to have little effect on IQ, definite neurologic abnormalities, or hearing loss. Higher bilirubin levels are associated with minor motor abnormalities, but the clinical importance of this finding is limited by the weakness of the association, the mild nature of the abnormalities, and the lack of evidence that they are prevented by treatment.

Should we be measuring blood cholesterol levels in young adults?

Should we measure blood cholesterol levels in all adults, or only in those at high risk of coronary heart disease (CHD)? Most men under the age of 35 years and women under the age of 45 years--roughly half the adult population--are at very low short-term risk of CHD. One consequence is that drug treatment to lower high blood cholesterol levels in the average young adult is an extremely expensive means of prolonging life; the estimated $1 million to $10 million per year of life is 100 to 1000 times the cost of other approaches. Individualized dietary treatment is somewhat cheaper but relatively ineffective. Another consequence of the low CHD risk in young adults is the greater likelihood that intervention may have harmful effects that outweight the benefits. Meta-analysis of primary prevention trials in middle-aged men reveal an increase in non-CHD deaths among those randomized to cholesterol interventions, an unexpected finding that is more substantial than the decrease in CHD deaths. This raises the possibility that one or more of the cholesterol interventions could have very serious adverse effects among young adults, whose risk of non-CHD death is normally 100 times their risk of CHD death. We conclude that the policy of screening and treating high blood cholesterol levels in young adults is neither cost-effective, nor does it satisfy ethical standards requiring strong evidence that preventive interventions do more good than harm. Fortunately, cholesterol screening in young adults is also not necessary: most CHD events associated with high blood cholesterol levels in this population will not occur for decades and can be prevented by treatment that is begun in middle age. Cholesterol screening and treatment in young adults should be limited to individuals with known coronary disease or other unusual factors that place them at high short-term risk of CHD death.

Position statement: cholesterol screening in children is not indicated, even with positive family history

The same reasons for the policy of rejecting universal blood cholesterol screening in all children also lead us to reject selective screening in those with a family history of premature CHD or high blood cholesterol. It is probably true that this large group--roughly 25% of all children, 12 million in the United States and many tens of millions worldwide--is at somewhat higher risk than the other 75% of children of eventually dying of CHD many decades later. But the small size and remoteness of any benefit achieved by cholesterol intervention is illustrated by the projection that we would need to treat 300 girls in the top cholesterol quintile, and treat them effectively for 50-60 years, in order to defer just one premature CHD death before age 65. The benefits of cholesterol screening and treatment in children are not only very small and remote, they are also uncertain; we have no firm evidence that they exist. The benefits are not needed, given the clinical trial evidence that most of the risk associated with high blood cholesterol among adults is reversible, even when intervention is not begun until middle age. The harmful effects of cholesterol screening and treatment are more firmly established than the benefits, and some are very serious. Such a program would be expensive orders of magnitude from conventional criteria for cost effectiveness. It would cause malnutrition in some children, and have the adverse consequences of labelling.(ABSTRACT TRUNCATED AT 250 WORDS)

Evaluation and treatment of jaundice in the term newborn: a kinder, gentler approach

Standard recommendations for evaluating and treating jaundice in term babies include following all babies closely for jaundice, obtaining several laboratory tests in those with early jaundice or bilirubin levels more than 12 to 13 mg/dL (205 to 222 mumol/L), using phototherapy to try to keep bilirubin levels below 20 mg/dL (342 mumol/L), and doing exchange transfusions if phototherapy fails, regardless of the cause of the jaundice. These recommendations are likely to lead to unnecessary testing and treatment of many jaundiced term infants. Because most jaundiced infants have no underlying illness, and the generally recommended laboratory tests lack sensitivity and specificity, they are seldom useful. In most babies, the only blood tests needed to evaluate jaundice are the blood type and group (of baby and mother) and a direct Coombs' test. A determination of direct bilirubin level should be added if jaundice is prolonged (greater than 2 to 4 weeks) or the baby has other signs of illness. Bilirubin toxicity is rare in term babies without hemolysis. In this low-risk group, the risks and cost of identifying and treating high bilirubin levels may exceed the benefits. Such infants need not be closely followed for jaundice. If significant jaundice is nonetheless found, treatment should be deferred to relatively high levels of serum bilirubin, with a goal of keeping bilirubin levels below 400 to 500 mumol/L (23.4 to 29.2 mg/dL). Babies with hemolytic disease should be followed more closely, and their bilirubin levels kept below 300 to 400 mumol/L (17.5 to 23.4 mg/dL). These recommendations should be reevaluated as new data become available. In the meantime, currently available data justify an approach to the jaundiced term infant that is less aggressive than previously recommended.

Direct bilirubin measurements in jaundiced term newborns. A reevaluation

To investigate the usefulness of measuring direct bilirubin in jaundiced term newborns, we reviewed the outcome of 5255 such measurements on 2877 term (37 weeks' gestation) newborns in two hospitals. Direct bilirubin tests were ordered 15 times as often per infant at the University of California, San Francisco, as at Stanford (Calif) University, and the reported results were more than twice as high. In most of the 149 infants with high (greater than 95th percentile) direct bilirubin levels, the high levels remained unexplained (52% of cases) or were due to apparent laboratory errors (21% of cases). Forty infants (27%) had conditions sometimes associated with high direct bilirubin levels. Elevation of direct bilirubin levels contributed to the diagnosis in only four of these infants. All had minor laboratory abnormalities that resolved spontaneously. Because of their low yield and poor specificity, direct bilirubin tests are seldom helpful in evaluating jaundice in term newborns.

The draft lottery and AIDS: evidence against increased intravenous drug use by Vietnam-era veterans

To investigate whether intravenous drug use begun during military service might affect risk of acquired immunodeficiency syndrome (AIDS), the authors compared AIDS cases in men eligible to be drafted with those in men who were exempt in the Vietnam-era draft lottery of 1970-1972. Draft-eligible men were less likely to develop AIDS attributed to intravenous drug use than were draft-exempt men (relative risk = 0.87; 95% confidence interval 0.80-0.95; p = 0.001). Other human immunodeficiency virus exposure categories showed no difference between the two groups. These results argue against increased intravenous drug use by Vietnam-era veterans.

In defense of standardized regression coefficients

The association between a risk factor and a disease can be expressed as a standardized regression coefficient derived from a logistic model. When exponentiated, this standardized coefficient equals the odds ratio associated with a one-standard-deviation change in the risk factor. Some epidemiologists have recently recommended that standardized regression coefficients not be used in epidemiologic research. We disagree and provide examples that demonstrate that, when a risk factor is continuous, standardized regression coefficients may be helpful for comparing variables measured in different units. Standardized regression coefficients may also be helpful for comparing the effect of the same risk factor in different populations. Misinterpretations can be avoided if the standard deviations of the variables of interest are also provided. There is no reason to abandon the use of standardized regression coefficients in epidemiologic analyses.

The case against childhood cholesterol screening

Because some authorities have proposed blood cholesterol screening for children to prevent coronary heart disease, we reviewed published studies to estimate the potential risks and benefits of such screening. Childhood cholesterol levels are a poor predictor of high cholesterol levels in young adulthood and will be an even poorer predictor of coronary heart disease later in life. There is no evidence that blood cholesterol levels can be lowered more easily in children than in adults, and it seems unlikely that cholesterol reduction in childhood will be much more effective at preventing coronary heart disease than cholesterol reduction begun in middle age. Screening and interventions to lower blood cholesterol levels for millions of children would be expensive, could lead to labeling and family conflicts, and may cause malnutrition and increased noncardiovascular mortality. Because the benefits of cholesterol screening are unlikely to exceed these risks, we conclude that children should not be screened for high blood cholesterol levels.

Does hyperbilirubinemia damage the brain of healthy full-term infants?

In the 1950s, exchange transfusion to keep the total serum bilirubin below 20 mg/dl was shown to be an effective way of preventing kernicterus in babies with erythroblastosis fetalis. For the last 15 to 20 years this level has also been used to determine the need for intervention in healthy full-term infants who do not have hemolytic disease. A critical review of all the available data including six studies from the collaborative perinatal project (more than 30,000 infants) and several smaller studies of term infants without hemolysis reveals essentially no evidence of adverse effects of bilirubin on IQ, neurologic examination, or hearing. The investigation and treatment of normal infants with jaundice is expensive and potentially harmful. We need to reassess our approach to hyperbilirubinemia in healthy full-term infants.

Laboratory evaluation of jaundice in newborns. Frequency, cost, and yield

Neonates with hyperbilirubinemia commonly undergo a battery of laboratory tests. We used a computerized database and medical records to study the frequency, cost, and yield of these tests in 2443 infants born at the University of California, San Francisco, between 1980 and 1982. Four hundred forty-seven (18%) of the infants met standard criteria for "nonphysiologic" hyperbilirubinemia; the incidence varied from 9% in blacks to 31% in Asian infants. About 55% of these 447 infants received a $125 "hyperbilirubinemia workup." Hospital discharge diagnoses on all 447 hyperbilirubinemic infants were reviewed. In 214 (48%), no cause of the jaundice was identified. An additional 145 (32%) had a possible cause apparent from history, physical examination, or initial hematocrit determination. The only diagnosis made as a result of routine investigations of hyperbilirubinemia was possible ABO or Rh isoimmunization in 75 infants (17%). Nonphysiologic hyperbilirubinemia may be more common than previously reported. The recommended tests are expensive and rarely lead to diagnoses other than ABO or Rh isoimmunization. Their routine use should be reevaluated.

Sample size and power based on the population attributable fraction

Most methods for calculating sample size use the relative risk (RR) to indicate the strength of the association between exposure and disease. For measuring the public health importance of a possible association, the population attributable fraction (PAF)--the proportion of disease incidence in a population that is attributable to an exposure--is more appropriate. We determined sample size and power for detecting a specified PAF in both cohort and case-control studies and compared the results with those obtained using conventional estimates based on the relative risk. When an exposure is rare, a study that has little power to detect a small RR often has adequate power to detect a small PAF. On the other hand, for common exposures, even a relatively large study may have inadequate power to detect a small PAF. These comparisons emphasize the importance of selecting the most pertinent measure of association, either relative risk or population attributable fraction, when calculating power and sample size.