MCHB/EPI Miami Conference — December 7 - 9, 2005

The National Survey for Children's Health: From Data to Action in Two States — Transcript

MARY BETH ZENI: Thank you, Michael, I guess. Thank you. The title of the presentation is Florida 's Children in Need of a Consistent Medical Provider.

I would like to thank and acknowledge my co-authors, Dr. Bill Sappenfield, Mr. Dan Thompson and Mr. Hae Won Chen. Also, the initial research was funded through the Florida State University , Council on Research Creativity.

First, I'd like to start with the background to the study. A consistent medical home for children is a national goal shared by public health professionals and their partners in health care, including professional organizations. The concept, consistent medical home, is noted in Healthy People 2010 and various professional standards. An important component of consistent medical care is a personal health care provider. Previous studies have indicated that a personal provider of pediatric care is better able to assess the child and family over time, and to provide comprehensive care. In addition, studies have noted that having a personal provider is associated with better access to care.

From an evaluation framework, one, to propose that personal provider is one process indicator of a larger outcome measure. Access to consistent quality care. The national survey of children's health has a specific question addressing if a child has a personal health care provider.

The current research for the study was--what are the characteristics of children in Florida without a personal provider. And a second research question, are there similar characteristics among children in Florida without a personal care provider?

Logistic regression was selected as the methodology. The dependent variable was personal provider defined in the survey as an MD or a nurse. Independent variables were identified and based on an access to medical care conceptual framework. The framework, sometimes referred to as a model, is an adaptation of the Anderson and Newman medical care utilization framework. The access to care framework includes the following components: Health policy such as the financing, education and human resources aspect. Characteristics of the health care delivery system. For example, resources and organization, Utilization of health services, consumer satisfaction and characteristics of the population at risk. This study addressed selected characteristics of the population at risk. Specifically, children without a personal health care provider.

In the framework, characteristics of the population at risk include the following three determinants: Predisposing, enabling and need. The following 12 independent variables for the regression model were selected based on the Anderson framework and reflected in the following slides: For predisposing, age, race, ethnicity (defined in the survey as Hispanic and non-Hispanic) and gender. Enabling: Poverty level, the child's insurance status, highest education level in the hues hold, and family structure.

In the need, we have health rating, health condition, emotional/behavioral condition, medication longer than 12 months.

A specific question in the survey asked respondents the following question, and this is S5 Q01 in the survey: Do you have one or more persons you think of as the selected child's personal doctor or nurse? This is a weighted sample. 83 percent of the U.S. and 80 percent of Florida reported yes. 17 percent of the U.S. and 20 percent of Florida reported no. Our first analysis examined if there was a difference between the Florida and U.S. percents.

And in this slide, we see the results of binomial Z, significant at .05 level. The percent of children who do not have a personal provider is significantly higher in Florida compared to the U.S.

Our next analysis examined the prevalence aspect of children without a consistent provider. The next set of slides provides a weighted comparison between the Florida and the U.S. for the fine variables that emerged from the regression model. And race is also included in this slide set.

The first, age. If you look on this chart and the bar charts, Florida is yellow and the United States is blue. And you can see comparison then. And the tendency that is the older age groups are more likely to be without a personal provider. And you can see that trend for both the U.S. and Florida .

Next, race. Although race was not significant in our final model, we included race in this slide set to note the black children are more likely compared to children in other age groups to not have a personal provider. And you can see this in the bar charts for both the U.S. and Florida .

In the survey, ethnicity was defined as Hispanic an non- Hispanic. In both the U.S. and Florida , Hispanic children at a higher percent than non-Hispanic children for not having a personal provider. And this is illustrated in this bar chart.

Next we looked at poverty level. For both U.S. and Florida , there are higher percents and a lower poverty level groups. And there's some differences noted between Florida and the United States .

For both the U.S. and Florida , differences in health insurance status noted and there's some variations between the U.S. and Florida . As you can see from this slide, that children without health insurance had a higher prevalence than children with health insurance, for both the U.S. and Florida .

Next, family structure. There are also some slight differences between the U.S. and Florida and family structure. In the survey, the choices included two biological parents, one biological, one step-parent, single parent, and other.

Here, we see an interesting difference between the U.S. and Florida regarding children with a reported moderate to severe emotional/behavioral condition. Again, the Florida is yellow bar and the U.S. is blue.

And our final slide is the need for medication for some condition greater than 12 months. This variable was used to examine a possible condition that may benefit from monitoring by a personal care provider. And you can see a difference between Florida and the U.S.

The next two tables examine the entire U.S. and Florida samples, weighted for the seven variables in the final model and race. Somewhat equal representation is noted between the U.S. and Florida in age and race.

And you can see in Florida , for example, there is somewhat equal representation for black. There's 20 percent Florida , 15 percent U.S. And for the Hispanic ethnicity, you can see similar representation too. And somewhat similar with very slight rare variations in poverty. But looking at health insurance and family structure, that there is a difference again between the U.S. and Florida samples regarding health insurance. Again, this is the entire sample weighted.

And this information is presented, including the emotional/behavioral condition, moderate and severe for the entire sample, and medication greater than 12 months. Because sometimes one wonders, for selected variables, how the entire sample looks regarding U.S. and Florida .

Next, we have the results of the logistic regression. These are the variables that emerged in the final model. As noted previously, the initial variables for this study were identified from the Day and Anderson framework. I'd like to credit Dan Thompson for the creative visual display of these results. Just to go over this slide before we move onto competence intervals in the following slide, these are the adjusted odds ratios as noted on the Y axis, and labeled above each bar chart. The red line running horizontal to the X axis corresponds to the number one on the Y axis. The bar charts were labeled under the X axis and the asterisk denotes significance at the .05 level. Bar charts below the line could be viewed as protective factors. For example, a poverty level of 200 of 400 percent having health insurance, medications used longer than 12 months, and family structure could be viewed as protective. Family structure was defined in this study as a two-parent family and include both two-parent biological and a biological and step-parent. In contrast, not having health insurance could be viewed as a vulnerability.

Significant findings above the line include older age groups, Hispanic ethnicity, poverty less than a hundred percent, and emotional behavioral, moderate and severe. Two variables are not significant but are important for dummy variables. The reference variables for age was zero to one year and poverty greater than 400 percent.

Next, we see the results of the regression, with the confidence intervals. And now the bar charts have been listed by variable and the adjusted odds ratios presented next to confidence intervals.

Just the same information we just reviewed, with confidence intervals.

Selected limitations of the study. I'd like to point out a few in this study and the first involves the limitations with most survey methods. We are assuming the respondents understood the questions and the responses, including the question for the dependent variable. Second, the responses are perceived health conditions and validation is not available. Second, there are noted limitations with regression models. And third, in the sample, the U.S. percentage of respondents that are in metropolitan statistical areas, also known as MSAs, 73 percent, and 18 percent were not in an MSA. In Florida , the sample included 93 percent of the children in MSA, seven percent not in the MSA. While a majority of the Florida population is in urban areas, we just raised this as an interesting piece and it might be interesting to examine additional rural issues in a different way.

Hispanic as a classification and the term usually does not account for acculturation factors. So it may benefit to explore acculturation issues among Hispanics and the various cultural groups that the overall term captures. And especially in Florida , we know there are various groups and cultures within the label or the term Hispanic.

And of course, these limitations and ideas lead to additional studies and future research plans. And first, one is to expand the dependent variable to a derived variable, medical home. And again, as noted by Michael, on a website and a resource available in the data resource center, there are listings and new ones coming up on a consistent basis of derived variables so the exciting thing about this survey is that the analysis could include future and current derived variables. And that is to identify and study additional indicators based on the Day and Anderson framework, again using derived variables.

And back to the home territory of evaluative research. If a child did have a personal care provider and answered yes to the question on our dependent variable, there was an opportunity for a set of questions that the respondent could then evaluate, in a sense, the quality of care received from the provider. And this provides an opportunity for an additional study. And this new study would also address a component in the Day and Anderson model. Okay. So thank you.