Asthma and Medicaid : An Overview of Issues and an Analysis of Asthma Education in Regard to NationalHospital Ambulatory Medical Survey
Christopher H. Purdy,PhD student, Biostatistics, New York State University at Buffalo
Leader - Angela Wisniewski, PharmD, Research Assistant Professor
Dan Morelli, MD, Clinical Professor of Family Medicine
School of Family Medicine, ECMCMedicalCenter
Faculty Liaison: Dr. John Taylor, MD
Sally Speed, Director, NYS Medicaid Training System
Funding for this research project was provided by Department of Health Training Resource System, Contract year 2005: Project 1044887, Award 34963, through the Center for Development of Human Services, College Relations Group, Research Foundation of SUNY, Buffalo State College.
Table of Contents
I. Introduction and Description of Dataset
II. Demographic Frequencies for the NHAMCS Data
III. Demographic Charts for the NHAMCS Data
IV. Logistic Regression Results in Table Form
V. Discussion of Logistic Regression Results
VI. Literature Review in Table Form
VII. Website Review in Table Form
VIII. Bibliography
IX. Annotated Bibliography (Medicaid, Asthma, Medication emphasis)
X. Power Point Presentation
I. Introduction to the Discussion of Medicaid, Asthma, and Asthma Education :
This paper addresses the issues of Medicaid, asthma, and asthma education. There exists an issue with both state Medicaid agencies and with private insurers, as regards the continuing care for asthmatic sufferers. As asthma is both an acute and chronic illness, prone to visits to both outpatient and emergency departments, there exists much information in the datasets used for this analysis (the datasets from 2003 from the NationalCenter for Health Statistics, titled the National Ambulatory Medical Care Survey and the National Hospital Ambulatory Medical Care Survey). These datasets are publicly available from the NationalCenter for Health Statistics website which is easily found and navigated. Obviously, insurers, whether they be government or private wish to control costs and to improve health outcomes.
The issue of asthma management is complex and can be viewed from an economic, patient, provider, and insurer point of view. This paper addresses several of these topics in brief. Also, this paper includes a SAS analysis of the variable asthma education as an outcome variable. Certainly, providers and insurers would most like asthma patients to be as well educated as possible in order to minimize severe attacks, improve patient satisfaction, reduce emergency room visits, and minimize costs. Hence, there exists a motivation to attempt to educate patients as well as possible in order to minimize costs.
Unfortunately, the information available in the NHAMCS dataset is somewhat limited. The medication information is detailed, but there is a possibility that some medications may not be included if the medication is not renewed at the time of the office visit. Also, the information in regard to asthma education itself as an outcome is limited in the sense that the type of education or intervention is not detailed, nor is there any indication on whether or not the patient actually completed the program. Also, a major limitation of this dataset is that it is not longitudinal, it is a visit based dataset that does not track patient identifiers in any capacity. The sample is conducted in a four stage probability sample, that divides the United States into geographical units, called primary sampling units, hospitals within those primary sampling units, clinics within those hospitals, and patients within those clinics. Also, available from the NCHS’s website are programs which incorporate sampling information such as weighting, clustering, stratification, and primary sampling unit, so this information can be included in the SAS programs in order obtain proper confidence intervals. The point estimates will be the same, but the estimates of the standard errors will be affected by the non-randomness of the sampling procedures. Also, it is important to take the weighting into account in terms of creating estimates that are applicable to the Unites States population as a whole.
The Nhamcs dataset is discussed in more detail in the following two sections, as well as the discussion of the logistic regression in terms of the asthma education, is also discussed in the following section. The relationship between certain demographic variables and the probability of receiving education in regard to their condition will attempt to be at least partly elucidated by the SAS analysis. Also the demographics of the dataset will be included in table form.
There is obviously much literature in regard to these issues of asthma, medication, Medicaid, management, and the costs associated with these issues. Here is a highlight of some of the research in regard to these issues. A more complete treatment of these issues can be found by perusing some of the articles listed in the bibliography or the websites also listed later in this paper.
Apter’s 1997 article in the Annals of Allergy Asthma and Immunology concludes that insurance was the most significant predictor of asthma care and morbidity from asthma related conditions. This study was primarily conducted in terms of pediatric patients. Balkrishnan’s 1998 paper concluded that the introduction of inhaled corticosteroid therapy for Medicaid patients was associated with an improvement in care and a decrease in cost. Berg’s 2004 paper concluded that low-income families of Latino origin were more likely not to have the resources for good asthma management and hence were associated with more acute episodes and emergency department visits. Similar to the analysis that this paper conducts, there is an argument for investing resources in education and controller medications, and prevention in the sense of prevention of acute episodes, as an attempt to increase patient satisfaction, decrease costs, and decrease tragic patient morbidity, in the sense that it is an avoidable morbidity.
Blixen’s 1999 article argues that health insurance alone may not be sufficient to ensure proper asthma management. It may be necessary to educate persons so that they utilize available services appropriately and so they manage their own case properly. Cabana’s 2005 article argues that there exists a relative dearth of standardized reliable information to guide physicians for the most appropriate standards for pediatric asthma care. Chabra’s 1998 article addresses the inequities in regard to African American and Latino young children in regard to asthma care and concomitant hospitalizations, and argues for social welfare programs to address these disparities of care. Cooper’s 2001 article addressed the surprising reality that the majority of Tennessee children who had an acute asthmatic episode failed to fill an oral corticosteroid following an emergency department visit. This paper also concluded that there existed a statistically significant relationship between demographics and the probability of this negative outcome, and the paper argued for more extensive research to better elucidate this problem and possible solutions.
David’s 2004 article cites a positive reaction to the publication of NIH guidelines, that children in Florida had a marked increase in the rate of filling a prescription for a daily asthma controller medication, however the article also cites long term problems with adherence. Davidson’s 1994 article suggests that certain demographics are strongly associated with the likelihood of using emergency room services as opposed to primary care physician. This paper also suggests that patients with Medicaid were four times more likely to use the emergency room services than those with insurance or who were self-pay. Emerman’s 1994 paper found that increased case management and education services lead to better pediatric asthma control and less use of emergency room services. Evans’ 1999 paper found that the optimal site for inner city asthmatic patient management was in fact an inner city hospital. By aggregating all of the patients in one location, this resulted in better care, and also allowed the integration of emergency services with the services of management and education.
In conclusion there exists a wealth of research in regard to asthma management and control for both patients and providers. The NIH guidelines are certainly a benchmark for all providers, patients, and their families as the best place to start.
But there exist a wealth of research and websites that are readily available to address certain specific concerns of both medical and financial interest. The literature review and bibliography contain many citations that could be useful to anyone who has a particular interest in the issues of asthma management. Also, the logistic regression model that follows addresses the specific issue of the probability of receiving asthma education.
II. Demographic Frequencies for NHAMCS 2003 Data :
Table 1 : Frequencies for Demographic Variables to Describe Nhamcs and Namcs data from 2003 (Total Number of Persons equals 100,033).
Variable / Category / Number of Persons / PercentageAsthma Education / No / 59,003 / 98.70
Yes / 777 / 1.30
White (Race) / Yes / 77,329 / 77.30
Black (Race) / Yes / 18,614 / 18.61
Hispanic (Race) / Yes / 14,225 / 14.22
Asian (Race) / Yes / 2841 / 2.84
Medicaid / No / 59,961 / 59.94
Yes / 40,072 / 40.06
Private Insurance / No / 60,185 / 60.17
Yes / 39,848 / 39.83
Gender / Female / 56,774 / 56.76
Male / 43,259 / 43.24
III. Demographic Charts and Tables for the NHAMCS 2003 dataset:
Chart 1 : Racial Breakdown of the NHAMCS 2003 dataset :
Chart 2 : Insurance Breakdown of the NHAMCS 2003 dataset:
Chart 3 : Gender Breakdown of the NHAMCS 2003 dataset:
IV. Logistic Regression SAS Output Results:
The actual Logistic Regression Model:
Table 1: Type 3 Analysis of Effects
Variable / Degrees of Freedom / P ValueAge / 1 / 0.0001
White / 1 / 0.8017
Black / 1 / 0.2249
Hispanic / 1 / 0.1027
Asian / 1 / 0.2090
Medicaid / 1 / 0.0025
Private Insurance / 1 / 0.0357
Gender / 1 / 0.5201
Table 2 : Parameter Estimates
Parameter / Estimate / P ValueAge / 0.0313 / 0.0001
White / -0.0427 / 0.8017
Black / 0.2088 / 0.2249
Hispanic / 0.0788 / 0.1027
Asian / 0.2369 / 0.2090
Medicaid / 0.1775 / 0.0025
Private Insurance / 0.1248 / 0.0357
Gender / 0.0237 / 0.5201
Appendix : Discussion of Logistic Regression : For non-statisticians who are interested in a brief discussion of the theory behind logistic regression here is a short summary condensed from the website :
Given a binary respose variable with probability of success , the logistic regression is a non-linear regression model with the following modelequation:
where is the product of the transpose of the columnmatrixof explantory variables and the unknown column matrix of regression coefficients. Rewriting this so that the right hand side is , we arrive at a new equation
The left hand side of this new equation is known as the logit function, defined on the openunitintervalwith range the entirerealline:
where
Note that the logit of is the same as the natural log of the odds of success (over failures) with the probability of success = . Since is a binary response variable, so it has a binomial distribution with parameter (probability of success) , the logistic regression model equation can be rewritten as
/ (1)Logistic regression is a particular type of generalized linear model. In addition, the associated logit function is the most appropriate and natural choice for a link function. By natural we mean that is equal to the natural parameterappearing in the distribution function for the GLM (generalized linear model). To see this, first note that the distribution function for a binomial random variable is
where is the number of trials and is the event that there are success in these trials. , the parameter, is the probability of success. Let there be iid binomial random variables each corresponding to trials with probability of success. Then the joint probability distribution of these random variables is simply the product of the individual binomial distributions. Equating this to the distribution for the GLM, which belongs to the exponential family of distributions, we have:
Taking the natural log on both sides, we have the equality of log-likelihood function in two different forms:
Rearranging the left hand side and comparing term , we have
so that .
Next, setting the natural link function logit of the expected value of , which is , to the linear portion of the GLM, we have
giving us the model formula for the logistic regression.
Remarks.
- Comparing model equation for the logistic regression to that of the normal or Gaussianlinear regression model, we see that the difference is in the choice of link function. In normal liner model, the regression equation looks like
/ (2)
The link function in this case is the identity function. The model equation is consistent because the linear terms on the right hand side allow on the left hand side to vary over the reals. However, for a binary response variable, Equation (2) would not be appropriate as the left hand side is restricted to only within the unit interval, whereas the right hand side has the possibility of going outside of . Therefore, Equation (1) is more appropriate when we are dealing with a binary response data variable.- The logit function is not the only choice of link function for the logistic regression. Other, ``non-natural'' link functions are available. Two such examples are the probit function, or the inverse cumulative normal distribution function and the complimentary-log-log function . Both of these functions map the open unit interval to .
V. Discussion of Logistic Regression Results:
The logistic regression revealed the relationship between asthma education and various demographic variables and insurance variables. As indicated by tables one and two some of the variables had a statistically significant relationship with asthma education and some did not. The variables age, Medicaid insurance, and Private Insurance had a statistically significant relationship with asthma education. The variables white, black, Hispanic, Asian, and gender did not have a significant relationship with asthma education. An increase in age is associated with an increase in the probability of receiving asthma education. If an individual has either Medicaid or Private Insurance, this is associated with an increase in the probability of receiving asthma education. Some things should be noted. The SAS output notes that the convergence criteria is satisfied. Also, the likelihood ratio test, the score test, and the Wald test indicate that there exists statistical evidence to reject the global null hypothesis.
In terms of commentary on the results of this analysis, these results tend to indicate that an increase in age is associated with an increased probability of receiving asthma education, this result is possibly confounded by a secondary relationship between age and severity of pulmonary illness. Also, this information (severity of illness) is not available in the data. Also, Medicaid insurance and private insurance, are both associated with an increase in the probability of receiving asthma education, which may be indicative of providers may be more likely to order asthma education for individuals who do not have to private pay. Note that among the demographic variables, those in the white demographic group were slightly less likely to receive asthma education, and in terms of the gender variable, males were slightly more likely to receive asthma education than females, although it should be noted that the p-values were not statistically significant at the standard 0.05 level. No obvious explanation for the relationship with the white racial group and asthma education exists, and a similar comment can be made for the relationship between gender and asthma education. Restated, there does not exist an obvious explanation for why males would be more likely to receive asthma education than females would.
VI. Literature Review for Asthma, Medicaid, Beta-Agonists, Control, Cost:
Author/Journal / Objective/Background / Methods / Results/ConclusionsApter, A. J., S. T. Reisine, et al. (1997). "Demographic predictors of asthma treatment site: outpatient, inpatient, or emergency department." Ann Allergy Asthma Immunol79(4): 353-61. / OBJECTIVE: To identify the demographic predictors of asthma treatment site: outpatient clinic, emergency department, or hospital. / METHODS: From the November 1993 to July 1995 claims data of the University of Connecticut Health Center, asthmatic patient sex, age, racial/ethnic group, address, and health insurance status were examined to identify predictors of treatment site. Patient addresses generated maps and census data. / CONCLUSION: Although not population-based, this group of asthmatic patients represents a group diverse in socioeconomic status and racial/ethnic background. Insurance category was the most influential factor predicting asthma treatment site, suggesting that economic status may be the most important determinant of higher morbidity. Children were treated predominantly in acute care settings.
Balkrishnan, R., L. M. Nelsen, et al. (2005). "Outcomes associated with initiation of different controller therapies in a Medicaid asthmatic population: a retrospective data analysis." J Asthma42(1): 35-40. / BACKGROUND: Outcomes in asthmatic patients may vary depending on the controller medication used. Observational studies of outcomes of asthma therapy are needed to understand the implications of choice of controller in different populations. / METHOD: Using data from the North Carolina Medicaid program, we compared continuously enrolled asthmatic patients starting either fluticasone propionate 44 microg (FP44), an inhaled corticosteroid (ICS) (n = 312), or montelukast 5 and 10 mg, an oral leukotriene modifier (LM) (n = 398) between the years 1998 and 1999. A secondary analysis compared continuously enrolled asthmatic patients already using ICS as controller therapy initiating either salmeterol (long-acting beta-agonist) (n = 97) or montelukast (n = 101) in the year 1998. / CONCLUSION: There were no cost and major health care use differences between the two primary or secondary controller therapies in the postinitiation year. Although FP was associated with lower rate of controller switch, montelukast use was associated with significantly better treatment adherence in patients with treatment persistence in this population of Medicaid-enrolled asthmatic patients. The addition of salmeterol as additional controller was associated with a significant decrease in inhaled corticosteroid use, suggesting decreased adherence in patients on the two-drug regimen.