Evaluating heart disease presciptions-filled as a proxy for heart disease prevalence rates.
Prevalence studies (Epidemiology)
Heart diseases (Statistics)
Cossman, Ronald E.
Cossman, Jeralynn S.
James, Wesley L.
Thomas, Richard K.
Pol, Louis G.
Cosby, Arthur G.
Mirvis, David M.
|Publication:||Name: Journal of Health and Human Services Administration Publisher: Southern Public Administration Education Foundation, Inc. Audience: Academic Format: Magazine/Journal Subject: Government; Health Copyright: COPYRIGHT 2008 Southern Public Administration Education Foundation, Inc. ISSN: 1079-3739|
|Issue:||Date: Spring, 2008 Source Volume: 30 Source Issue: 4|
|Topic:||Event Code: 680 Labor Distribution by Employer|
|Geographic:||Geographic Scope: United States Geographic Code: 1USA United States|
Heart disease is the leading cause of death in the U.S. Yet, prevalence rates are not reported at the county level. Not knowing how many have the disease, and where they are, may be a knowledge barrier to effective health care interventions. We use heart disease drug prescriptions-filled as a proxy measure for prevalence of heart disease. We test the correlation to the Behavioral Risk Factor Surveillance System (BRFSS) and find positive, statistically significant correlations. Next we illustrate the geographic patterns revealed using the county-level prevalence estimate maps. This information can be used to provide a better understanding of sub-state variations in disease patterns and subsequently target the delivery of health resources to small areas in need.
Diseases of the heart are the leading cause of death in the U.S., responsible for 29 percent of deaths (Anderson & Smith, 2001). In 2003, cardiovascular disease was an underlying cause of death in 37 percent of all cases and an underlying or contributing cause of death in 58% of all deaths (Thom et al., 2006a). The prevalence of cardiovascular diseases is estimated to be 34 percent of the total population (American Heart Association, 2004a). Coronary heart disease is a subset of the more general category of heart diseases. The estimated prevalence of coronary heart disease (in 1999-2000) is 7 percent of the total population (Anderson & Smith, 2001; American Heart Association, 2004b). Thus the prevalence of the broad category of heart disease may be roughly 33 percent to 34 percent of the population, while the prevalence of coronary heart disease is estimated in 7 percent of the population.
These national prevalence rates are estimated from sample-based surveys such as the National Health and Nutrition Examination Survey (NHANES); they are also reported at four census region levels. For state level data, researchers use the Behavioral Risk Factor Surveillance System (BRFSS) (Thom et al., 2006c), and more recently at the MSA level. However, this leading cause of death is not tracked in every state in every year via the BRFSS. Only 24 states administered this the heart disease question in 2003, leaving the rest of the states unmonitored from that source. Though survey data are useful for many research questions, national survey data cannot be used to estimate prevalence in sparsely populated areas, particularly in rural areas where individuals are not surveyed. Additionally, surveys must be explicitly undertaken for fairly long periods, which prescription data are already routinely collected. Thus health policy makers and health care providers do not have prevalence estimates for the leading cause of death at the small geographic scale (i.e., counties). Local data are necessary to inform the analysis of differences in the same state (e.g., Delta versus metropolitan Mississippi), the ability to reaggregate counties into self-defining regions that cross state borders, and the ability to reaggregate along hospital referral patterns (Dartmouth, 1998). Armed with this information, interventions at either the health policy or health delivery level can be tailored to meet the area's need.
In this paper we validate the use of heart disease prescriptions-filled as a proxy measure for heart disease prevalence in the U.S. We show that there is a sufficiently high correlation between state-level BRFSS and state-level heart disease prescriptions-filled to suggest that this new data set is a valid proxy measure for prevalence rates. More significantly, these rates can be disaggregated to the county level, providing prevalence rate estimates for under-sampled rural areas, as well as reaggregated to clusters of counties.
REVIEW OF THE LITERATURE
From the perspective of allocating resources and directing treatments and interventions, spatial data are necessary. As Stan Openshaw, an early leader in exploratory data visualization wrote, "People DIE each year because no one BOTHERS to properly analyse DISEASE and DEATH data for unusual localised concentrations" (Openshaw & Turner, 1997, emphasis in original). Put another way, "... sometimes the answer to the question "Why?" can come from first solving a different puzzle: "Where?" (Mummolo, 2006, p. 12). Thus spatial data are fundamental to testing local conditions (e.g., environmental, socioeconomic, cultural, etc.).
Most atlases of health only present the data in map form, without additional spatial analysis (Barnett et al., 2001, Dartmouth, 1998).  Some studies recognize regional variations in heart disease mortality, treatment, correlations and prevalence, but stop short of spatial analysis. Two studies of heart disease mortality across time found varying evidence that the geographic area with the highest heart disease mortality stretches from the Southeast coastline (pre-Stroke Belt) southwesterly toward the Mississippi River (Pickle & Gillum, 1999; MacEachren, 2000). A similar study examining trends in premature heart disease mortality found that the mortality rate declined more slowly in the rural South over time than in the rest of the nation (Barnett & Halverson, 2000). The literature on regional variations in heart disease treatment and outcomes contains one finding of spatial interest: that of a New England advantage (Krumholz, Chen, Rathore, Wany, & Radford, 2003; Ayanian, 2003).
Equally varied and extensive is the literature on the correlates and causal agents for heart disease. As Beaglehole and Magnus (2002) point out, the factors accounting for some 75% of causal agents for new cases of coronary heart disease including high blood cholesterol, high blood pressure, cigarette smoking and physical inactivity, although epidemiologists continue to search for additional explanations such as fine particulate matter (e.g., smoke or smog) (Brook et al., 2004; Levy, J. L., Houseman, E. A., Spengler, J. D., Loh, P. and Ryan, L., 2001; Levy, J. L., Greco, S. L., and Spengler, J. D., 2002), inequality (Cooper, R., Fortmann, S. P., Hogelin, G., Friedman, L., Cutler, J., Desvigne-Nickens, P., McGovern, P., Havlik, R. and Marler, J., 2000; Krause, 1979; Marmot, 2001; Smith, Harper, & Hillemeier, 2004), diet (Hajjar & Kotchen, 2003; Shaw, H. J., 2006; Pearson, T., Barker, M. E., Russell, J., and Campbell, M. J., 2005), or prior exposure to the 1918 influenza (Patterson, K. D. and Pyle, G. F., 1991; Azambuja & Duncan, 2002). Most, if not all, of these underlying factors have spatial concentrations; therefore, spatial analyses of cardiovascular diseases are warranted.
One shortcoming in existing studies of regional variation in heart disease prevalence is dependence on survey populations which largely sample in urban locations and sometimes with restricted populations, such as the BRFSS (Hahn, Heath, & Chang, 1998), the NHIS (Gillum, 1994), special populations such as the Jackson Heart Study (Taylor, Hughes, & Garrison, 2002), and surveys or registries of Native Americans (Levin, Lamar-Welch, Bell, & Caspar, 2002; Rith-Najarian et al., 2002). Though these surveys are critical for a variety of analyses, they are not viable for use to estimate morbidity at the sub-state level in rural areas. As Geiss noted, "Data from population-based studies are generally considered more reliable than data from selected groups within the population because the latter may not represent the community with respect to factors such as age and health status" (Geiss, William, & Smith, 1995, p.236).
A population-based measure of heart disease prevalence is available. While the full dimensions of the relationship between diagnosis and prescription drug treatment are unknown, a significant proportion of the diagnosed population can be identified through prescription use. An estimated 77 percent of adults diagnosed with heart disease reported taking prescription medication for their illness, while 97 percent of adults diagnosed with high blood pressure reported taking medication for their illness (Stagnitti & Pancholi, 2004), suggesting that heart disease related drug prescriptions are a promising proxy for estimating local heart disease prevalence rates in rural areas.
We know mortality rates are clustered in spatial patterns that have been persistent for more than 30 years (Cossman, Cossman, James, Campbell, Blanchard & Cosby, 2006; Cossman, Cossman, Jackson, & Cosby, 2003; Cossman & James, 2003; Cossman, Blanchard, James, Jackson-Belli, & Cosby, 2002; James, Cossman, Cossman, Campbell, & Blanchard, 2004), but we do not yet know if there are similar patterns associated with morbidity. In this manuscript, we illustrate the added value of small area heart disease prevalence estimates on health care delivery. This advance in spatial morbidity research will allow planners and policy makers to assess areas of the country with small-area spatial variation, ultimately getting a better understanding of the underlying causes of these spatial variations.
METHODS AND DATA
We use (1) a data set of office visits to select the prescription drugs of interest, (2) a data set of prescriptions-filled as a proxy measure for heart disease prevalence, (3) a data set consisting of a national medical care survey to perform an age truncation to our calculation, and (4) a data set of self-reported heart disease prevalence at the state level to use as our referent, calculating state-level correlations as a method of validating the prescription data measure.
IMS Health, Inc., collects prescription drug data from nearly 30,000 suppliers covering 225,000 sites, e.g., drug manufacturers, wholesalers, retailers, pharmacies, mail order, long-term care facilities and hospitals.  To determine the prescription drug classes to be measured for heart disease we used IMS Health's National Disease and Therapeutic Index[TM] (NDTI), a database derived from an ongoing office-based physician panel providing national-level estimates of disease and treatment patterns for office-based physicians. The data in NDTI capture all medications associated with a patient visit for a particular treatment. The leading therapeutic classes (a.k.a. Uniform System of Classification or USC) used for heart disease were identified from this data set. The USC classes chosen for heart disease were: (USC 31100) Renin Angiotensin Systemic Antagonist, (USC 31400) Beta and Alpha blockers and (USC 32000) Cholesterol reducers and Lipotropics (IMS Health, 2004). 
Based on the results from the NDTI, we purchased monthly, county-level prescriptions-filled data for 1999-2003 for the aforementioned drugs classes contained in IMS Health's Xponent[TM] database. Total U.S. prescriptions-filled estimates are determined from IMS Heath's prescription database (RxD), obtained from pharmaceutical chain organizations and pharmacy software vendors, and reported at the prescription transaction level by individual outlet level. The RxD encompasses roughly 72 percent of the national retail prescription sale volume (as of February 2006), a rate that has been stable since 1998. To estimate the remaining 28 percent, IMS Health's data from the RxD are weighted to generate estimates representing total dispensed prescription volume at the national, sub-national and prescriber level. IMS uses outlet purchase, volume data and pharmacy distance measures to determine applicable weights for sample pharmacies to estimate the dispensed prescription volume for each non-sample pharmacy. Weights are derived through a proprietary, patented geo-spatial methodology. According to IMS Health estimates, retail pharmacies account for 67 percent of total national prescription activity. The remainder consists of mail-order (23 percent), clinics, long-term care, prisons, universities and non-federal hospitals (8 percent) and Federal facilities (e.g., V.A) (2 percent) (IMS Health, 2005).
The prescription data set did not contain patient demographics. To place the prescription data on an equal footing to the BRFSS, which surveys only adults, we sought to determine the percentage of individuals under age 18 who were receiving diabetes medications (the chronic illness medication most likely to be taken by children). We used the National Ambulatory Medical Care Survey (NAMCS), a national probability sample survey of visits to office-based physicians conducted by the National Center for Health Statistics at the CDC. In addition to patient demographics, the NAMCS collects data on the therapeutic class of drug prescribed. We aggregated 2000-2002 files to the four Census regions level. Among patients under age 20 (the closest age cut-point), the usage of diabetes medications was always below 1 percent, indicating that children could be dropped from the base population for medications under consideration here. A 12-month average was calculated to smooth the rates, and to account for multiple-month prescription fills. The refined rate of prescriptions-filled at the state level was calculated as follows. 
Estimated prescriptions filled in a year in the state / 12 months / Age 18 and over resident state population (in 100s)
In order to validate the IMS Health prescription data with the BRFSS, we aggregated the IMS Health data to the state-level, as BRFSS cannot be examined at the sub-state level in rural areas. The BRFSS, a state-level monthly telephone survey of adults about behaviors associated with health risks, and the incidence of medical conditions provides our reference point. BRFSS data are routinely used to create state-level estimates of chronic illness and health risk prevalence rates for the nation and, more recently, for metropolitan and micropolitan areas; however, due to the survey design, the BRFSS continues to under-sample rural residents. For compiling BRFSS prevalence estimates we used the following survey question "Have you ever been told by a doctor, nurse or other health profession that you have coronary heart disease?" from the 1999-2003 BRFSS.
In a test of correlations we paired appropriate drug classes with BRFSS questions. The statistically significant correlations between the BRFSS prevalence rate and the prescription rate (only for years with 51 states reporting, which does not include heart disease) ranged from a low of .658 (stroke in 2001) to a high of .774 (diabetes in 2001). The r values were statistically significant at the 0.01 level (2-tailed test). A total of 10 correlations were tested and all were statistically significant.
Direct comparisons between the state-level BRFSS disease prevalence rates and the state-level IMS Health prescription rates are restricted by the number of BRFSS reporting states in each year. While IMS Health reports all 51 states on a monthly basis, the BRFSS question about coronary heart disease has never been a core question (asked in all 50 states plus the District of Columbia); the number of states including the coronary heart disease BRFSS question has ranged from 7 to 24.
Due to the wording of the heart disease question in the BRFSS survey, we must compare coronary heart disease to the broader category of heart disease prescriptions-filled. When comparing the state-level BRFSS coronary heart disease prevalence rates with the state-level IMS Health heart disease prescription rates, the correlations ranged from 0.438 to 0.662 (p < .01, See Table 1.). These correlations are based on data from 20 to 24 states. [5,6] The BRFSS mean state-level self-reported adult prevalence rate for coronary heart disease ranges from 4.27 percent of the adult population (1999) to 4.48 percent (2003); the mean state-level adult heart disease prescriptions-filled rate ranged from 15.19 percent (1999) to 18.74 percent (2003), meaning that almost one-in-six adults are filling orders for heart disease prescription drugs. This is half of the estimated 33 percent national heart disease prevalence rate. When prescriptions-filled are reported at the county-level, the county average ranged from 13.13 percent (1999) to 15.61 percent (2003). The ratio of the prescriptions-filled rates to BRFSS prevalence rates ranges from a low of 1.57 (in 2001) to 2.25 (in 2003).
Initially we mapped and compared state-level BRFSS to state-level prescriptions-filled rates. To map the relative differences in heart disease prevalence rates we used quintiles to facilitate comparisons between the BRFSS and prescriptions-filled maps. The legend contains four categories (solid black is the highest 20 percent, hashed black is the second highest 20 percent, solid gray is the lowest 20 percent and hashed gray is the second lowest 20 percent) that diverge from the middle quintile (white), which is centered on the mean.  Figure 1 displays BRFSS data at the state-level and contains a 6th category, "States not surveyed in BRFSS." The self-reported coronary heart disease prevalence ranges from 2.06 percent to 8.72 percent of the adult population. The map contains 5 states in each category except for 4 in the solid black (top 20 percent) category.
[FIGURE 1 OMITTED]
This map shows Arkansas, Alabama, Kentucky and West Virginia ranked "very high" in coronary heart disease prevalence rates (5.32 percent and higher); the remainder of the South ranks as an area of "high" or "average" prevalence (4.17 percent-5.31 percent). The majority of the Midwest and West did not report this question, although 4 of the 8 "low" and "very low" prevalence states are in the Midwest and West (Minnesota, North Dakota, Montana and Colorado). The BRFSS data displayed at the state-level suggests coronary heart disease prevalence rates are divided into high prevalence Southeast zone and low prevalence Midwest and West zones, with a cluster of very high states that roughly follow the Appalachian mountain region.
In comparison, the IMS Health state-level prescription data map (Figure 2) is interesting in two ways. First, it covers the entire U.S., completing the health picture of the Midwest and West and, second, it reinforces the same cluster of high rates in the Southeast, while low rates are clustered in the Midwest and West.
Arkansas falls into the "very high" quintile of coronary heart disease but is ranked "average" for heart disease prescriptions-filled, perhaps suggesting residents are under-treated via drugs for heart disease. In contrast, five states shifted by two or more categories. Hawaii was in the lowest category of self-reported coronary heart disease but ranked in the second highest quintile of heart disease prescriptions-filled, while North Carolina was in the second lowest 20 percent of self-reported coronary heart disease and then ranked in the highest 20 percent of heart disease prescriptions-filled. Compared to other states, it seems as if a higher proportion of residents in Hawaii and North Carolina are filling heart disease prescriptions for treatment of their chronic disease. Since these categories are based on relative (state-to-state) measures, it is not possible to link actual need for prescriptions (prevalence rates) to prescription-filled rates in each state. Rather, these results suggest which states warrant further investigation.
[FIGURE 2 OMITTED]
The Figure 3 map shows a much more varied distribution of heart disease prevalence expressed by prescription-fill rates and a possible disconnect between prescription-fill rates and the geographic estimates of coronary heart disease prevalence from mapped BRFSS data in Figure 2. The range of rates is expanded to zero percent--87.9 percent, although the two counties in which the rate is 66 percent or more are interpreted as regional distribution centers.  Striking inter-county differences in prescription fill rates are found within the "very high" prevalence states, as well as the "very low" prevalence states. The most informative geographic patterns are (1) low prescription-fill rate counties in states with overall high prevalence rates, (2) high-prescription-fill rate counties in states with overall low prevalence rates, and (3) "very high" or "high" prescription-fill rate counties contiguous to "low" or "very low" prescription-fill rate counties.
[FIGURE 3 OMITTED]
The advantage of using the prescriptions-filled data is that it reports for all counties in the U.S., allowing for an inspection of intrastate and regional differences in crude prescription rates. Initial statistical tests are encouraging; however further spatial statistical testing is necessary. At a minimum, the current tests indicate vast variation across counties in prescription fill rates.
POTENTIAL LIMITATIONS AND DIFFICULTIES
Proxies, by definition, are imperfect measures. In 2003, the zero-order correlation of roughly 60% indicates that nearly 40 percent of the variation between the two data sets has not been captured, likely due to (1) the comparison of self-reported coronary heart disease in BRFSS and the broader category of heart disease drug prescriptions-filled (2) the telephone sampling method of BRFSS which over-represents densely populated metropolitan areas while our prescription data are based on a geographically distributed count (which comprises 72% of all prescriptions filled at retail outlets) and an estimate of the remainder; and (3) a discrepancy between the rate of self-reported prevalence and the rate of those in drug therapies.
We are aware of the coverage limitations of this prescriptions-filled data set. There are several categories of possible exclusion. First, there are people outside the medical system, whether they are excluded at the point of the physician (those who do not seek medical attention, those not diagnosed when examined or those diagnosed but not treated), or at the point of the pharmacy (people who do not fill their prescription for any reason). The second category of exclusion is drugs used in a clinical setting, such as a hospital, clinic (e.g., chemotherapy) or doctor's office (e.g., samples). The third category of exclusion reflects non-participation by providers in the program that tracks prescription use, either by the pharmacy (this is the largest source of retail exclusion) or by the patient (e.g., mail-order purchases, or outside the U.S.). We also realize that all prescriptions are not filled in the county of patients' residence. Some preliminary work has been done by other researchers using Primary Care Service Areas to assess the operational coverage of primary care; although no test has yet been made of what percentage of prescriptions are filled out-of-county (Goodman et al., 2003).
Our comparisons to other data sets confirm that state-level prescriptions-filled rates are statistically significantly correlated to state-level self-reported prevalence of chronic disease; therefore, we conclude that the prescriptions-filled data set can be used, with caution, as a proxy measure of state-level disease prevalence rates. More importantly, the data can be disaggregated to the county-level, providing an estimated disease prevalence rate that is not otherwise available elsewhere, particularly for sparsely populated rural areas. Further, prescription data can be plotted over time on a monthly basis to reveal trends in prescription usage. We do not want to overstate the methodology's results or precision. However, despite the inherent limitations of crude rates based on a combination of count and estimation methodology, this methodology can be valuable in identifying relative-levels of prescription prevalence rate, and perhaps the associated disease's prevalence rate as well. Areas that exhibit unusually high or low prescription rates compared to their neighbors would suggest geographic areas for further investigation. Our next step in the calibration of this model is to translate the "statistically significant" highs and lows into "medically significant" highs and lows by determining a medically appropriate prescription rate, given the chronic disease prevalence and demographics of the resident population. Additionally, we plan on calculating spatial statistics to statistically quantify the geographic patterns that are seen in these maps.
With a valid proxy for morbidity rates, health policy makers and administrators can make more informed decisions concerning the health of their local populations. Interventions to target the population at risk can be developed and implemented at a much smaller geographic level than previously available. This methodology can be used for a variety of illnesses, but will work particularly well for diseases that have high rates of prescription treatment, e.g., heart disease, asthma, attention deficit disorder, or diabetes. This methodology will allow health services personnel to devise targeted programs, improving care in the long run and reducing spatial health disparities.
American Heart Association. (2004) "Statistical Fact Sheet--Populations: At-a-glance Summary Tables Men and Cardiovascular Diseases." 16 August 2006
Anderson R. N. and Rosenberg H. M. (1998) "Age Standardization of Death Rates: Implementation of the Year 2000 Standard." National Center for Health Statistics. National Vital Statistics Reports 47(3).
Anderson R. N. and Rosenberg H. M. (1998) "Report on the Second Workshop on Age Adjustment." National Center for Health Statistics. National Vital Health Statistics 4(30).
Anderson R. N. and Smith B. L. (2005) "Table C: Deaths and Percentage of Total Deaths for the 10 Leading Causes of Death: United States, 2001-02." National Vital Statistics Reports, 53(17):7. 16 August 2006
Ayanian J. Z. (2003) "Is Geography Destiny? Illuminating the Survival Advantage of Elderly Patients in New England after Acute Myocardial Infarction." American Heart Journal. 146(2):207-209.
Azambuja M. I. and Duncan B. B. (2002) "Similarities in Mortality Patterns from Influenza in the First Half of the 20th Century and the Rise and Fall of Ischemic Heart Disease in the United States: A New Hypothesis Concerning the Coronary Heart Disease Epidemic." Cadernos de Saude Pubica. 18(3):557-566.
Barnett E. and Halverson J. (2000) "Disparities in Premature Coronary Heart Disease Mortality by Region and Urbanicity among Black and White Adults Ages 35-64, 1985-1995." Public Health Reports. 115(1):52-64.
Barnett E, Casper M.L., Halverson J.A., Elmes G.A., Brahan V.E., Majeed Z.A., Bloom A.S. and Stanley S. (2001) Men and Heart Disease: An Atlas of Racial and Ethnic Disparities in Mortality. Office for Social and Environment and Health Research, West Virginia University: Morgantown, WA. December 9, 2006 < http://0-www.cdc.gov.mill1.sjlibrary.org/dhdsp/library/map s/cvdatlas/atlas_mens/mens_download.htm>
Beaglehole R. and Magnus P. (2002) "The Search for New Risk Factors for Coronary Heart Disease: Occupational Therapy for Epidemiologists?" International Journal of Epidemiology. 31:1117-1122.
Blalock, H. M. (1979) Social Statistics, Revised 2nd Edition, Upper Saddle River, New Jersey: McGraw-Hill, Inc.
Brewer C. (2003) Color Brewer: Web Tool for Selecting Colors for Maps, Pennsylvania State University, 16 August 2006
Brook R. D., Franklin B., Cascio W., Hong Y., Howard G., and others. (2004) "Air Pollution and Cardiovascular Disease: A Statement for Healthcare Professionals from the Expert Panel on Population and Prevention Science of the American Heart Association." Circulation. 109(21):2655-2671.
Centers for Disease Control and Prevention. CDC Wonder. 16 August 2006
Centers for Disease Control and Prevention. (2004) Behavioral Risk Factor Surveillance System Survey, Atlanta, Georgia: U.S. Department of Health and Human Services, Centers for Disease Control and Prevention. 16 August 2006
Centers for Disease Control and Prevention. (1999) Behavioral Risk Factor Surveillance System Survey Questionnaire. Atlanta, Georgia: U.S. Department of Health and Human Services, Centers for Disease Control and Prevention.
Centers for Disease Control and Prevention. (2001) Behavioral Risk Factor Surveillance System Survey Questionnaire. Atlanta, Georgia: U.S. Department of Health and Human Services, Centers for Disease Control and Prevention.
Centers for Disease Control and Prevention. (2003) Behavioral Risk Factor Surveillance System Survey Questionnaire. Atlanta, Georgia: U.S. Department of Health and Human Services, Centers for Disease Control and Prevention.
Cooper, R., Fortmann, S. P., Hogelin, G., Friedman, L., Cutler, J., Desvigne-Nickens, P., McGovern, P., Havlik, R. and Marler, J. (2000) "Trends and Disparities in Coronary Heart Disease, Stroke, and Other Cardiovascular Diseases in the United States: Findings of the National Conference on Cardiovascular Disease Prevention." Circulation. 102 (25):3137-3147.Cossman, Jeralynn S., R. E. Cossman, W. L. James, C.R. Campbell, T.C. Blanchard and A. G. Cosby. "Persistent Clusters of Mortality in the U.S." American Journal of Public Health. Forthcoming. Accepted for publication August 2006.
Cossman R. E., Cossman J. S., Jackson R., and Cosby A. (2003) "Mapping High or Low Mortality Places Across Time in the U.S.: A Research Note on a Health Visualization and Analysis Project." Health & Place 9(4)361-369.
Cossman R. E. and James W. L. (2003) "Characteristics of "Healthy and Unhealthy" Places in the U.S." Presentation made to the Southern Demographic Association Annual Meeting. October 23-25, 2003. Arlington, VA.
Cossman R. E., Blanchard T. C., James W. L., Jackson- Belli R., and Cosby A. (2002) "Healthy and Unhealthy Places in America: Are These Really Spatial Clusters?" Proceedings of the 22nd Annual ESRI International User Conference. ESRI Library 16 August 2006
Dartmouth Medical School. Wennberg, J.E. Series Editor. (1998) The Darmought Atlas of Health Care 1998. Center for the Evaluative Clinical Science. Dartmouth Medical School, Hanover, New Hampshire.
Geiss L. S., William W. H., and Smith P. J. (1995) "Chapter 11: Mortality in Non-Insulin-Dependent Diabetes." In Diabetes in America, 2nd edition. National Diabetes Data Group, National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Disease (NIH Publication No. 95-1468). p. 236. 16 August 2006
Gillum R. F. (1994) "Prevalence of Cardiovascular and Pulmonary Disease and Risk Factors by Region and Urbanization in the United States." Journal of the National Medical Association. 86(2):105-112.
Goodman D. C., Mick S. S., Bott D., Stukel T., Chang C., Martin N., and others. (2003) "Primary Care Service Areas: A New Tool for the Evaluation of Primary Care Services." Health Services Research 38(1-p1):287-309.
Hajjar I. and Kotchen T. (2003) "Regional Variations of Blood Pressure in the United States are Associated with Regional Variations in Dietary Intakes: The NHANES-III data." The Journal of Nutrition. 133:211-214.
Hahn R. A., Heath G. W., and Chang M. H. (1998) "Cardiovascular Disease Risk Factors and Preventive Practices among Adults--United States, 1994: A Behavioral Risk Factor Atlas." MMWR CDC Surveillance Summary. 47(5):35-69.
IMS Health, IMS National Disease and Therapeutic Index[TM], Plymouth Meeting PA. (2004) "Heart Disease Therapeutic Classes from 1999-2003 Database." October 2004. Database.
IMS Health, IMS Xponent[TM], Plymouth Meeting PA. (2004) "Heart Disease, Stroke and Diabetes therapeutic classes from 1999-2003 Database." December 2004. Database.
IMS Health. (2005) U.S. Prescription Activity by Channel, 2004. National Prescription Audit Plus, 2005. 16 August 2006
James W. L., Cossman R. E., Cossman J. S., Campbell C., and Blanchard T. (2004) "A Brief Visual Primer for the Mapping of Mortality Trend Data." International Journal of Health Geographics 3(7). 16 August 2006
Krause, R. M. (1979) "The Influence of Infection on the Geography of Heart Disease." Circulation. 60 (5):972-976.
Krumholz H. M., Chen J., Rathore S. S., Wany Y., and Radford M. J. (2003) "Regional Variation in the Treatment and Outcomes of Myocardial Infarction: Investigating New England's Advantage." American Heart Journal 146(2):242-249.
Levin S., Lamar-Welch V. L., Bell, R. A. and Casper M. L. (2002) "Geographic Variation in Cardiovascular Disease Risk Factors among American Indians and Comparisons with the Corresponding State Populations." Ethnicity & Health. 7(1):57-67.
Levy, J. L., Houseman, E. A., Spengler, J. D., Loh, P. and Ryan, L. (2001) "Fine Particulate Matter and Polycyclic Aromatic Hydrocarbon Concentration Patterns in Roxbury, Massachusetts: A Community-Based GIS Analysis." Environmental Health Perspectives. 109 (4): 341-347.
Levy, J. L., Greco, S. L., and Spengler, J. D. (2002) "The Importance of Population Susceptibility for Air Pollution Risk Assessment: A Case Study of Power Plants Near Washington, D.C." Environmental Health Perspectives. 110 (12): 1253-1260.
MacEachren A. (2000) "What Is a Map?" Online Research Penn State. 21(2). 16 August 2006
Marmot M. (2001) "Inequalities in Health." New England Journal of Medicine. 345(2):121.
Mummolo J. "Health: Mapping a Disease." Newsweek Magazine. August 21, 2006. p.12.
National Center for Health Statistics. (2000) Compressed Mortality File: 1968-88. Centers for Disease Control and Prevention, U.S. Department of Health and Human Services, Hyattsville, Maryland.
National Center for Health Statistics. (2001a) Compressed Mortality File: 1989-98. Centers for Disease Control and Prevention, U.S. Department of Health and Human Services, Hyattsville, Maryland.
National Center for Health Statistics. (2001b) Documentation for the Compressed Mortality File: 1989-98. Centers for Disease Control and Prevention, U.S. Department of Health and Human Services, Hyattsville, Maryland.
National Center for Health Statistics. (2005) NAMCS Description. Centers for Disease Control and Prevention, U.S. Department of Health and Human Services, Hyattsville, Maryland. 16 August 2006
Openshaw S. and Turner A. (1997) GIS in Health and Crime Analysis. School of Geography, University of Leeds. 16 August 2006
Patterson, K. D. and Pyle, G. F. (1991) "The Geography and Mortality of the 1918 Influenza Pandemic." Bulletin of Historical Medicine. 65 (1):4-12.
Pearson, T., Barker, M. E., Russell, J., and Campbell, M. J. (2005) "Do Food Deserts Influence Fruit and Vegetable Consumption?--A Cross-Sectional Study." Appetite. 45 (2):195-197.
Pickle L. W. and Gillum R. F. (1999) "Geographic Variation in Cardiovascular Disease Mortality in US Blacks and Whites." Journal of the National Medical Association. 91(10):545-556.
Rith-Najarian S. J., Gohdes D. M., Shields R., Skipper B., Moore K. R., and others. (2002) "Regional Variation in Cardiovascular Disease Risk Factors among American Indians and Alaska Natives with Diabetes." Diabetes Care. 25(2):279-283.
Shaw, H. J. (2006) "Food Deserts: Towards the Development of a Classification." Geografiska Annaler Series B: Human Geography. 88 (2):231.
Smith, L. J., Harper S., and Hillemeier M. (2004) "Is Income Inequality a Determinant of Population Health?" Part 2. The Milbank Quarterly. 82(2):355-400.
Stagnitti M. N. and Pancholi M. (2004) Chronic Conditions and Outpatient Prescription Medicines for Persons 18 and Older in the U.S. Civilian Noninstitutionalized Population, 1987 and 2001. Statistics Brief #58, Agency for Healthcare Research and Quality. 16 August 2006
Stuchlak W. Senior Principal, IMS Management Consulting, Plymouth Meeting, PA < WStuchlak@us.imshealth.com>. Re: IMS Data, email to Ronald E. Cossman. March 31, 2006.
Taylor H. A., Hughes G. D., and Garrison R. J. (2002) "Cardiovascular Disease among Women Residing in Rural America: Epidemiology, Explanations, and Challenges." American Journal of Public Health. 92(4): 548-551.
Thom T., Haase N., Rosamond W., Howard V., Rumsfeld J., and others. (2006a update) "Heart Disease and Stroke Statistics." A Report from the American Heart Association Statistics Committee and Stroke Statistics Subcommittee. Circulation. p. 3. 11 January 2006
Thom T., Haase N., Rosamond W., Howard V., Rumsfeld J, and others. (2006b update) "Heart Disease and Stroke Statistics." A Report from the American Heart Association Statistics Committee and Stroke Statistics Subcommittee. Circulation. p. 2. 11 January 2006
(1.) To illustrate this point, in the Dartmouth Atlas of Health Care 1998, the phrase "geographic pattern" is used once and "spatial pattern" is never used.
(2.) For a fuller description of IMS Health's products, see Wysowski D. K., Armstrong G., and Governale L. (2003) "Rapid increase in the use of oral antidiabetic drugs in the United States, 1990-2001." Diabetes Care, 26(6) June 2003, pp. 1852-1855.
(3.) Specifically these include Angiotensin-Converting Enzyme (ACE) inhibitors (along, with diuretics and other), angiotensin II type I receptor antagonist (alone and in combination), peripheral vasodilators, calcium blockers, beta blockers, alpha-blockers, beta/alpha blockers (with diuretics), alpha blockers (alone and in combination), central acting agent (alone and in combination), antihypertensive (other), HMG-COA reductase inhibitor (3-hydroxy-3methylgluatryl coenzyme A reductase), bile acid sequestrants, fibric acid derivative, cholesterol absorption inhibitor, cholesterol red combination, lipotropics, and antihyperlipidemic agent (other).
(4.) There is a slight difference in the calculation of the 1999 prescriptions-filled rate and subsequent years, due to the way that the base population was reported by the Census bureau. The population denominator for 1999 is age 20 and older. The population denominator for 2000-2003 is age 18 and older.
(5.) There is a slight difference in the population base (denominator) used in the 1999 calculation compared to subsequent years. The denominator uses the population estimate for 1999, which was based on the 1990 census, and was subject to estimation error across nine years of population change. The change in percent from the 1999 state-level population estimate to the 2000 state-level population census was an average of 97 percent, with changes ranging from 88 percent to 1.04 percent for the 21 reporting states. In contrast, the change in percent from the 2000 state-level population to the 2001 state-level population was an average of 99 percent, with changes ranging from 97 percent to 99 percent for the same 21 states.
(6.) Correlations based on samples or populations of fewer than 30 may be unstable and should be viewed with caution (Blalock, 1979).
(7.) Color versions of all three maps are available at: http://www.ssrc.msstate.edu/publications/prescriptions/hdra tes.html The color maps use a legend that diverges from the mean, which is the same as the legend for the black and white maps. The color ramp is based on the ColorBrewer color diagnostic tool (Brewer, 2003).
(8.) Two outlier counties are noted. The reported heart disease prescription-fill rate is 87.94% for Adams County, North Dakota, and 73.55% for Montour County, Pennsylvania. Without further investigation, we assume these to be prescription distribution centers. The remainder of the counties falls under 66%.
RONALD E. COSSMAN
Mississippi State University
JERALYNN S. COSSMAN
Mississippi State University
WESLEY L. JAMES
Mississippi State University
Mississippi State University
RICHARD K. THOMAS
University of Tennessee Health Science Center
LOUIS G. POL
University of Nebraska, Omaha
ARTHUR G. COSBY
Mississippi State University
DAVID M. MIRVIS
University of Tennessee Health Science Center
Table 1. Descriptive Statistics and Correlations of State-Level BRFSS to State-Level Rx 1999 2000 2001 BRFSS Q: "Have you ever been told by a doctor that you have coronary heart 0.438 * -- 0.662 * disease?" compared to heart disease prescriptions filled N (States) 21 14 20 BRFSS Heart Disease: Mean (State-levels) ** 4.27% -- 4.47% Std. Dev. (State-level) 0.90% 1.08% N (States) 21 20 Prescriptions-Filled: Mean (State-level) 15.19% 16.04% 16.99% Std. Dev. (State-level) 3.44% 3.57% 3.64% Ratio of Rx/BRFSS rates 1.60 1.57 Prescriptions-Filled: Mean (County-level) 13.13% 13.37% 14.07% Std. Dev. (County-level) 8.14% 8.34% 8.69% N (counties) 3,103 3,103 3,103 2002 2003 BRFSS Q: "Have you ever been told by a doctor that you have coronary heart -- 0.613 * disease?" compared to heart disease prescriptions filled N (States) 7 24 BRFSS Heart Disease: Mean (State-levels) ** -- 4.48% Std. Dev. (State-level) 1.39% N (States) 24 Prescriptions-Filled: Mean (State-level) 18.02% 18.74% Std. Dev. (State-level) 3.83% 3.96% Ratio of Rx/BRFSS rates 2.25 Prescriptions-Filled: Mean (County-level) 14.91% 15.61% Std. Dev. (County-level) 9.20% 9.51% N (counties) 3,103 3,103 Note: Correlations with populations of less than 30 are unstable and should be viewed with caution. * Correlation is significant at the 0.01 level (2-tailed) ** Mean of the state-level rates, not a calculated national rate
|Gale Copyright:||Copyright 2008 Gale, Cengage Learning. All rights reserved.|