Automated serial ECG comparison improves computerized interpretation of 12-lead ECG.

BACKGROUND: Interpretation of a patient's 12-lead ECG frequently involves comparison to a previously recorded ECG. Automated serial ECG comparison can be helpful not only to note significant ECG changes but also to improve the single-ECG interpretation. Corrections from the previous ECG are carried forward by the serial comparison algorithm when measurements do not change significantly. METHODS: A sample of patients from three hospitals was collected with two or more 12-lead ECGs from each patient. There were 233 serial comparisons from 143 patients. 41% of patients had two ECGs and 59% of patients had more than two ECGs. ECGs were taken from a difficult population as measured by ECG abnormalities, 197/233 abnormal, 11/233 borderline, 14/233 otherwise-normal and 11/233 normal. ECGs were processed with the Philips DXL algorithm and then in time order for each patient with the Philips serial comparison algorithm. To measure accuracy of interpretation and serial change, an expert cardiologist corrected the ECGs in stages. The first ECG was corrected and used as the reference for the second ECG. The second ECG was then corrected and used as the reference for the third ECG and so on. At each stage, the serial comparison algorithm compared an unedited ECG to an earlier edited ECG. Interpretation accuracy was measured by comparing the algorithm to the cardiologist on a statement by statement basis. The effect of serial comparison was measured by the sum of interpretive statement mismatches between the algorithm and cardiologist. Statement mismatches were measured in two ways, (1) exact match and (2) match within the same diagnostic category. RESULTS: The cardiologist used 910 statements over 233 ECGs for an average number of 3.9 statements per ECG and a mode of 4 statements. When automated serial comparison was used, the total number of exact statement mismatches decreased by 29% and the total same-category statement mismatches decreased by 47%. CONCLUSION: Automated serial comparison improves interpretation accuracy in addition to its main role of noting differences between ECGs.

Development of a serial comparison program for conduction defects, acute myocardial infarction, and the use of additional leads.

Serial comparison of electrocardiograms (ECGs) is a useful tool in clinical diagnostic ECG and an enhancement to computer ECG analysis. When an analysis algorithm is modified, the corresponding serial comparison program needs to be updated accordingly. The new Philips diagnostic algorithm increased the number of leads in the ECG from the traditional 12 leads to 16, making it possible to diagnose right ventricular infarct/injury based on right-sided lead V4R. To keep pace with the widespread reperfusion therapy for acute myocardial infarct, the serial comparison program was revised to recognize the rapid ECG changes in patients with ST-elevation myocardial infarct following successful reperfusion therapies. The serial comparison program was also enhanced to split "combined" statements in the category of ventricular conduction delay (includes incomplete ventricular conduction delay and bundle-branch blocks) and compare each of the statements separately.

Computerized serial comparison of electrocardiograms: program performance in myocardial infarction.

Serial comparison of electrocardiograms (ECGs) can provide a useful clinical function by reporting to the editing cardiologist the diagnostic changes that have occurred since the previous ECG. This program detects "significant measurement differences" in each of the diagnostic categories to detect these changes. We evaluated the accuracy and use of this serial comparison program by comparing the diagnostic results of the program with those of an expert cardiologist using a database of ECGs obtained from patients with symptoms admitted to the hospital and other laboratory results consistent with acute myocardial infarction. We found that the level of agreement between the computer and the cardiologist was much higher when a current ECG was compared with a previous that had been edited by the cardiologist than when that same ECG was analyzed in isolation.

Standardized risk and description of results from multivariable modeling of a binary response.

Descriptions of significant associations found from a logistic regression analysis typically are based on adjusted odds ratios. Unfortunately, odds ratios provide no information about the prevalence of response. In this paper, we justify and recommend using standardized risks, i.e., standardized probabilities, which do provide information about prevalence, in addition to adjusted odds ratios, for pairwise comparisons of the levels of a significant factor. We illustrate the advantages of generally reporting standardized risk estimates, in the context of assessing the effect of blood lead levels during the preschool years on occurrence of academic problems in kindergarten. Results are more meaningfully interpreted when accompanied by standardized risk estimates.

Racial and gender differences in the viability of extremely low birth weight infants: a population-based study.

OBJECTIVE: The purpose of this study is to provide a race- and gender-specific model for predicting 1-year survival rates for extremely low birth weight (ELBW) infants by using population-based data. METHODS: Birth and death certificates were analyzed for all children (N = 5076) with birth weights between 300 g and 1000 g who were born in Florida between 1996 and 2000. Semiparametric, multivariate, logistic regression analysis was used to model 1-year survival probabilities as a function of birth weight, gestational age, mother's race, and infant's gender. Estimated survival rates among different race/gender groups were compared by using odds ratios (ORs). RESULTS: One-year survival rates for 5076 ELBW infants born between 1996 and 2000 did not change during the 5-year period (60-62%). The survival rate at < or = 500 g was < or = 14% (n = 716). Survival rates at 501 to 600 g and 601 to 700 g were 36% and 62%, respectively. The survival rate reached > 85% for infants of > 800 g. Modeling indicated a survival advantage for female infants, compared with male infants (OR: 1.7; 95% confidence interval: 1.5-1.9), and for black infants, compared with white infants (OR: 1.3; 95% confidence interval: 1.1-1.5). Black female infants had 2.1 greater odds of survival than did white male infants. CONCLUSIONS: This population-based study highlights the significant race and gender differences in 1-year survival rates for ELBW infants, as well as the interactions of these 2 factors. These findings can assist obstetricians and neonatologists not only in the care of ELBW infants but also in frank discussions with families.

The risk of birth defects in multiple births: a population-based study.

OBJECTIVES: To determine if multiple births have higher risks of birth defects compared to singletons and to identify types of birth defects that occur more frequently in multiple births, controlling for seven sociodemographic and health-related variables. METHODS: A retrospective cohort study was conducted of all resident live births in Florida during 1996-2000 using data from a population-based surveillance system. Birth defects were defined as in the 9th edition of the International Classification of Diseases-Clinical Modification (ICD-9-CM) code for the 42 reportable categories in the Centers for Disease Control and Prevention (CDC) Birth Defects Registry list and eight major birth defects classifications. Relative risks (RR) before and after adjusting for control variables and 95% confidence intervals (95% CI) were calculated. The control variables included mother's race, age, previous adverse pregnancy experience, education, Medicaid participation during pregnancy, infant's sex and number of siblings. RESULTS: This study included 972,694 live births (27,727 multiple births and 944,967 singletons). Birth defects prevalence per 10,000 live births was 358.50 for multiple births and 250.54 for singletons. After adjusting for control variables, multiple births had a 46% increased risk of birth defects compared to singletons. Higher risks were found in 23 of 40 birth defects for multiple births. Five highest adjusted relative risks for birth defects among multiple births were: anencephalus, biliary atresia, hydrocephalus without spina bifida, pulmonary valve atresia and stenosis, and bladder exstrophy. Increased risks were also found in 6 out of 8 major birth defects classifications. CONCLUSIONS: Multiple births have increased risks of birth defects compared to singletons.

Sex differences in birth defects: a study of opposite-sex twins.

BACKGROUND: Sex differences in structural birth defects are often confounded by environmental risk factors. Opposite-sex twins provide a unique model for detecting sex differences in birth defects while maximally controlling environmental risk factors in a natural setting. METHODS: Population data from the Florida Birth Defects Registry were analyzed. A total of 4,768 pairs of twins who were discordant for sex and born between 1996 and 2001 were analyzed. The McNemar test was used to compare the differences between a male twin and his twin sister for the risk of developing specific defects and organ-system defects. RESULTS: Of 4,768 twin pairs, 225 males (4.72%) and 175 females (3.67%) had birth defects. Among opposite-sex twin pairs, males had a 29% higher risk for birth defects than their twin sisters. Compared to their twin sisters, males had a 5.4 times higher risk for pyloric stenosis and a 2.4 times higher risk for obstructive genitourinary defect, but only one-tenth the risk for congenital hip dislocation. CONCLUSIONS: Sex differences in birth defects exist between opposite-sex twins.

Risk factors for infant maltreatment: a population-based study.

CONTEXT: Of the approximately 900,000 children who were determined to be victims of abuse or neglect by US child protective services in 2002, the birth-to-3 age group had the highest rate of victimization (1.6%) and children younger than 1 accounted for the largest percentage of victims (9.6%). OBJECTIVE: To identify perinatal and sociodemographic risk factors associated with maltreatment of infants up to 1 year of age. DESIGN AND SETTING: Observational cohort study. PARTICIPANTS: 189,055 children born in 1996 in Florida. MAIN OUTCOME MEASURE: Infant maltreatment, defined as a verified report of abuse, neglect, or threatened harm that occurred between day 3 of life and 1 year. RESULTS: 1,602 children (.85%) of the 1996 birth cohort had verified instances of maltreatment by age 1. Of 15 perinatal and sociodemographic variables studied, 11 were found to be significantly related to infant maltreatment. Five factors had adjusted relative risks (RR) of two or greater: Mother smoked during pregnancy (RR 2.8); more than two siblings (RR 2.7); Medicaid beneficiary (RR 2.1); unmarried marital status (RR 2.0); low birth weight infant (RR 2.0). Infants who had four of these five risk factors had a maltreatment rate seven times higher than the population average. CONCLUSIONS: Data on nearly all risk factors found to be significantly associated with infant maltreatment are available on the birth certificate. Such information can be incorporated into a population-based risk-assessment tool that could identify subpopulations at highest risk for infant maltreatment. Because resources are limited, these groups should be given priority for enrollment in child abuse prevention programs.

Maternal and infant factors associated with excess kindergarten costs.

OBJECTIVE: To estimate the excess educational costs at kindergarten from infant and maternal factors that are reported routinely at birth. METHODS: Birth and school records were analyzed for all children who were born in Florida between September 1, 1990, and August 31, 1991, and entered kindergarten from 1996 through 1999 (N = 120,554). Outcome measure was cost to state, derived from base allocation for students in regular classrooms plus multiplier weights for those who were assigned to 8 mutually exclusive special education categories or who repeated kindergarten. RESULTS: More than one quarter of the study cohort was found to be assigned to special education classes at kindergarten. Regression model estimates indicated that children who were born at <1000 g (n = 380) generated 71% higher costs in kindergarten than children who were born at >or=2500 g. Children who were born at 1000 to 1499 g (n = 839) generated 49% higher costs. Other birth conditions, independent of birth weight, were associated with higher kindergarten costs: family poverty (31%), congenital anomalies (29%), maternal education less than high school (20%), and no prenatal care (14%). Because of their prevalence, family poverty and low maternal education accounted for >75% of excess kindergarten costs. If 9% of infants who weighed between 1500 and 2499 g (n = 1027) could be delivered at 2500 g, then the state of Florida potentially could save 1 million dollars in kindergarten costs. Savings of a similar magnitude might be achieved if 3% of mothers who left school without a diploma (n = 1528) were to graduate. CONCLUSIONS: Any policy recommendation aimed at reducing education costs in kindergarten must take into consideration 3 factors: the prevalence of risk conditions whose amelioration is desired, the potential cost savings associated with reducing those conditions, and the costs of amelioration. Projecting these costs from information that is available at birth can assist school districts and state agencies in allocating resources.

A population-based study of the effects of birth weight on early developmental delay or disability in children.

Improving medical treatment of extremely low-birth-weight infants over the last 20 to 30 years resulted in increased survival rates. The developmental sequela of salvaged infants is of great interest to perinatologists. The primary purposes of the current study were to assess the effect of birth weight (BW) on developmental delay or disability (DDD) in the first three years of life and determine whether there is a BW threshold below which all infants should be evaluated to determine if intervention services for children with DDD should be received. Three statewide databases were merged: 1998 Birth Vital Statistics; 1997-1998 Medicaid eligibility files; and 1998-2001 Children's Medical Services' Early Intervention Program (CMS-EIP) data. Infants who died within the first year of life and plural births were excluded. The final dataset consisted of 170,874 records. A child was determined to have a DDD if a developmental delay, or an established condition, such as sensory impairment, genetic, metabolic, neurological, or severe attachment disorders, was diagnosed through a multidisciplinary evaluation. Logistic regression models were used to relate BW to DDD, controlling for sociodemographic, behavioral, and perinatal variables. Adjusted odds ratios (OR) were calculated to describe the effects of BW on DDD. There was a significant effect of BW on DDD (Adjusted OR &equals 97.50, 40.01, 15.84, 3.29, 1.39, 1.00, 1.52 for BW categories 450-749, 750-999, 1000-1499, 1500- 2499, 2500-2999, 3000-4749, 4750-6050 g, respectively). In these categories, 70%, 56%, 36%, 11%, 4%, 3%, and 6% of surviving singleton infants, respectively, suffered a DDD in their first 3 years of life. Four medical, five sociodemographic, and two behavioral factors were significant in addition to BW. An equation for predicting the probability of DDD given these factors was obtained, and its use exemplified. BW is strongly associated with DDD. Over 60% of infants weighing < 1000 g and nearly half (46%) of those weighing < 1500 g at birth are diagnosed with a DDD before 3 years of age. The probability of DDD for a specific infant also varies by sociodemographic, other perinatal, and behavioral factors. The results of this paper suggest that all surviving infants of BW < 1000 g, and perhaps < 1500 g, should be automatically referred for evaluation.

A risk assessment screening test for very low birth weight.

OBJECTIVES: To develop a risk-assessment screening tool for very low birth weight (VLBW) and to compare our empirically derived tool to the nonempirically derived screening tool used by the State of Florida. METHODS: Birth records from the State of Florida Vital Statistics between 04/01/92 and 12/07/94 were matched with State Healthy Start prenatal records, reported from 04/01/92 through 03/31/94. Known and additional potentially important risk factors were identified from both sources. Generalized Linear Modeling techniques were used to estimate associations between risk factors and VLBW. A risk assessment system was then developed using the estimated model. The resulting screening test was compared with the one used by the Florida State Department of Health in terms of sensitivity and specificity on an independent validation sample. RESULTS: The proposed screening tool had comparable specificity to the Healthy Start screening tool but significantly better sensitivity. Both instruments are simple and easy to implement. CONCLUSIONS: Identification of women at high risk for VLBW would be improved using the model-based screening tool developed in this paper. Public health policy makers should use statistical methods in addition to expert opinion to improve existing risk assessment methods. The actual value of an improved screening instrument is dependent on the availability of effective intervention programs.

An empirical comparison of record linkage procedures.

We consider the problem of record linkage in the situation where we have only non-unique identifiers, like names, sex, race etc., as common identifiers in databases to be linked. For such situations much work on probabilistic methods of record linkage can be found in the statistical literature. However, although many groups undoubtedly still use deterministic procedures, not much literature is available on deterministic strategies. Furthermore, there appears to exist almost no documentation on the comparison of results for the two strategies. In this work we compare a stepwise deterministic linkage strategy with a probabilistic strategy, as implemented in AUTOMATCH, for a situation in which the truth is known. The comparison was carried out on a linkage between medical records from the Regional Perinatal Intensive Care Centers database and educational records from the Florida Department of Education. Social security numbers, available in both databases, were used to decide the true status of each record pair after matching. Match rates and error rates for the two strategies are compared and a discussion of their similarities and differences, strengths and weaknesses is presented.