Epidemiology concepts and applications

Low Birth Weight Babies and Asthma in Urban Populations

This paper is a critical review of the article by Lenna Nepomnyaschy and Nancy Reichman, published in the American Journal for Public Health. It will review the purpose of the study, its study design, and the sample selection. In addition, it will analyze what correlations there may be in a fairly complex series of confounding variables. Finally, it will review the practice implications of the study for the nurse.

Purpose and Objective of this Epidemiological Research

The authors make it clear that there have been a large number of studies in the past to attempt to correlate low birth weight with asthma in young children, but these studies did not, in the authors' opinions, control adequately for a number of additional factors which could have contributed to asthma in babies and children at the age of three.

The authors' conclusions are preliminary, but they contend that, even when normalized for other confounding conditions, such as maternal demographic, socioeconomic, medical and behavioral risk factors, the correlation remained strong between low birth weight and asthma at the age of three. They also found a statistically significant correlation between measures of neighborhood housing quality and poverty.

Study Design

The study attempted to count babies and toddlers from a representative group within urban populations, which is a partial nod to Gordon's dictum to "be able to identify the specific factors which put [high risk] groups at high risk and try to modify those factors (Gordon, 2004)." The study population did not mirror the general U.S. population, and it also differed from the large urban center population in important ways: according to the paper, the mothers were more likely to be non-marital birth mothers, and a significant number of surveyed mothers and their babies dropped out of the study from birth to age 3. The dropout numbers, as reported in the study, are as follows:

4,898 total included in the survey

3,319 interviewed 3 years later (68%)

2,994 for which medical records were available from the 4,898 (61%)

2,032 for which medical records were available from the 3,319 (61%)

The choice of sample population is quite complete, but leaves some questions open:

1. Which cities were chosen? The study shows that 20 cities were included, but does not indicate where they were in the U.S. This question of location could be important, particularly if the asthma results need to be controlled for climate. if, for example, there were more cities than proportionately in the U.S. In northern climes, such as Minneapolis and Chicago, there could be a greater number of children who contract asthma in whole or in part due to inadequate heating or exposure to the cold.

2. Why were multiple births excluded? The authors point out that much of low birth weight trends is due to multiple births. This variable would be helpful to understand.

3. What are the commonalities in those who did not have medical data available, or were not available for follow-up three years later? If the toddlers and mothers were not available, were there confounding variables, such as housing, drug use and/or low birth weight? If the demographic variables were similar for both groups -- those followed up and those not followed up -- this should be indicated in the study.

3. The importance of prenatal drug use and smoking pre- and post-natally was demonstrated to be relatively low in comparison to other factors, with little statistical significance. Since the drug use and smoking were self-declared, it would be helpful to understand the correlation between a subsample of mothers' lab results (for the 61% where medical records were available) against their self-declared behaviors. A representative sample may have been enough to indicate whether or not the self-declared behaviors were in fact accurate indicators of actual drug and/or tobacco use.

4. The authors claim that there may have been mothers who were less than 18 at the time of birth. Since 61% of the total had complete medical records, an estimate of that number would be important. The reader would also benefit by a split-out of the mothers less than 18 and the asthma rates of their babies, in order to establish if low birth weights were more common in such populations, and other confounding variables were different.

How was the sample selection done and what was the final sample size on which the data was analyzed? was there a control group?

The final sample size appeared to be 2,032 or 1,845, depending on the variables analyzed. That is because the patients in the 1,845 group had complete medical data available on all study variables, while the 2,032 included all those with some medical data. This means that there were 187 subjects for whom not all data was available. A subtraction of multiple-birth events brings us to 1,803 in the sample.

There was no control group per se. The objective of the study was to have enough patients to be able to run multiple correlate exercises and eliminate variables which are confounding. In this type of study, therefore, a classic 'control group' is not required.

What is a confounding variable, What confounding variable and biases were taken in consideration by the researcher in this study?

Confounding variables are those which may be correlated with the sample population, but bring little or no new information. An example might be socioeconomic level (i.e. how much income is earned in the household) with marital status, which is a confounding variable because it has been established that single mothers have lower incomes than married mothers. An additional confounding variable is vacancy rates in the neighborhood; if there are high vacancy rates, it is likely that there are more people with confounded low socioeconomic status as compared to the rest of the population.

The researchers included several variables, but the degree of confoundedness was intermingled. In the discussion of results, the authors indicated that low birth weight is associated with increased rates of asthma, and that predictors of asthma also included "census tracts with higher vacancy rates, fewer people per nonvacant household, and higher poverty rates." It would have been helpful to run a test of correlation between these three variables, in order to establish the degree of independence between them. Perhaps a better way to correlate low birth weight, asthma and socioeconomic status is to develop a socioeconomic status index, much as the authors referred to in the UK example, which includes these three confounding variables.