Ronald Fischbach, Elio Spinello

 

Using GIS to Analyze Geographic Patterns of Acute Myocardial Infarction in Los Angeles County

 

Abstract

 Target populations for preventative cardiology efforts can be identified geographically by understanding areas of highest incidence with respect to socio-economic characteristics, physician office locations, hospital and emergency room access, and morbidity and mortality rates. Using Atlas GIS and Arc View GIS, a comprehensive analysis of hospital discharges, death certificates, and medical facilities was performed for Los Angeles County. This study evaluates the morbidity and mortality rates of acute myocardial infarctions (MI) by combining hospital discharge data together with death certificate data for the same time period and geographic region. Concentrations of MIs are analyzed with respect to location, proximity to health care providers, public transportation routes, socio-economic and demographic characteristics. Based upon the above findings a prescriptive and narrowly targeted health education program addressing the issues of awareness of symptomatology and appropriate emergency response can be designed and implemented with the objective of reducing the incidence of MIs. The study was completed utilizing Atlas GIS version 3.03, ArcView GIS version 3.0, SPSS for Windows, and Visual Foxpro.

 

Introduction

Annually, approximately 1.5 million acute myocardial infarctions occur in the United States. Of these, approximately 500 thousand result in death (Med Help International, 1996) . The most common type of myocardial infarction occurs as a result of a coronary thrombosis when a clot (thrombus) blocks one or more of the blood vessels that nourish the heart muscle. A decrease in blood flow can also be cause by a narrowing of the coronary arteries leading to reduced blood supply to the heart. The reduced blood supply causes a lack of oxygenation to the myocardium, or heart muscle (Biger, J. Thomas, Jr., 1989).

 A relatively minor myocardial infarction can result in a short disruption in blood flow causing little or no permanent damage to heart tissue. A more serious incident can be caused by a prolonged disruption to the blood supply which can, in turn result in necrosis and ultimately permanent damage to the heart tissue. Damage to the muscle tissue can also result in a disruption in the heart's electrical function resulting in an irregular heartbeat or fibrillation. The degree of damage is generally related to the length of time during which the oxygenation of the muscle tissue is reduced. As more time lapses, more of the heart muscle is affected and more permanent damage occurs.

 Since one of the major causes of death from coronary thrombosis is the development of abnormal heart rhythms, emergency treatment usually centers on monitoring and stabilizing the heart rhythm. Treatment is also included to help relieve pain and prevent shock.

One of the most significant risk factors which has been found to increase the probability of an MI is an elevated level of HDL cholesterol. The National Heart, Lung and Blood Institute has found that for every 1 percent lowering in total blood cholesterol, individuals can lower their risk of heart attack by 2 percent. In a study of psychosocial factors by O'Conner, Manson, O'Conner, and Buring (1995), it was found that there is a possible association between type "A" personalities and the risk of heart attack. The same study found that there is no apparent relationship between the risk of heart attack and suppressed anger. The Framingham Study found that male gender, history of hypertension, and current smoking were associated with an increased risk of recurrent coronary events or death from cardiac disease (Wong, ND; Gardin JM; Kurosaki, T.; Anton-Culver, H.; Rosmeman, Sidney; Gidding, S, 1995).

In terms of long term risk, patients who have recovered from myocardial infarction and who have high cholesterol levels are at an increased long-term risk for new myocardial infarctions and are also at increased risk of mortality from other forms of coronary heart disease. One of the best preventative measures after a heart attack occurs is the on-going monitoring of cholesterol levels and lipid management (Wong, N.D., Wilson, P.W., Kannel, W.B., 1991). As a preventative measure, periodic electrocardiograms have been found to be a good indicator of long term risk for persons who have not had an MI and for MI patients long after recovery (Wong, ND; Levy, D; Kannel, WB, 1990).

Survival rates for myocardial infarction patients is often directly related to the length of time between the onset of symptoms and treatment. In general, of patients who get treatment within the first 70 minutes almost 99% survive, of those who get treatment between 70 and 180 minutes about 92% survive, and only 50% survive if 3 to 4 hours lapse before treatment. Immediate access to care is the most important factor to survival (Archives of Internal Medicine Archives of Internal Medicine, 1992).

Given the relationship between survival and the length of time before which medical treatment is initiated, this study examines the hypothesis that the average length of time before treatment is accessed is related to the distance between the patient and the nearest emergency medical facility.

 Profiling Myocardial Infarction Deaths

 An analysis of the hospital discharge and death certificate data for the State of California in 1994 shows that about 78% of identified myocardial infarction victims survive. Since the database identifies cases and not individuals, it is likely that the survivor segment includes some individuals who experienced multiple MIs during the year. This has the effect of somewhat understating the fatality rates. Of the MI victims who do not survive approximately 42% of deaths occur in a hospital after the patient has been admitted. Of those remaining, 20% occur in a hospital emergency room or are dead on arrival (DOA).

One of the important segments of the MI victim population is the 8% of patients who died away from a medical facility. These are individuals who, presumably, did not recognize the symptoms, waited too long to get medical treatment, or simply did not attempt to get treatment. When broken down by sex (see Appendix 1), it appears that roughly the same percentage of male and female MI deaths occurred after being admitted to hospitals and at home. However, a significantly larger percentage of male MI deaths occurred in emergency rooms while a larger percentage of female deaths tended to occur at "other locations", a category primarily made up of nursing homes. Although approximately 52% of all MI deaths were males, just over 65% of MI deaths which occurred in emergency rooms were males.

 

Given the 1994 data, females are more likely to die from an MI prior to reaching a medical facility. While 32.9% of male victims died before making it to a medical facility 42.3% of females died away from a medical facility. A breakdown of the patients who died away from a hospital shows that most (50.4%) died at home or in a nursing home ( 41.4%) and about 8.2% died at other locations.

An analysis of those who died in nursing homes shows that 37.2% of female deaths occurred in nursing homes versus 17.2% of male deaths. This may be due, in part, to a larger female population among nursing home residents. The percentages of males and females who die at home were fairly similar.

Summary: Older people were more likely to die after being admitted to the hospital. Deaths of younger victims were more likely to occur away from a medical facility or home. Females who died away from a hospital were generally in a nursing home. Males were more likely to be at home.

Although further study is needed to determine the reasons why victims in the relatively younger age groups were more likely to die away from a medical facility, a similar hypothesis to the one explaining why females are more likely to die away from a medical facility may hold. Younger individuals may be less likely to expect a heart attack and, for that reason, be less likely to recognize the symptoms and warning signs causing them to wait longer before seeking medical attention.

A breakdown by education shows that the likelihood of death occurring at home seems to increase somewhat as education increases. Thirty and two tenths percent of those with less than a high school education died at home versus 35.5% of those who were college graduates. This seems to offset an opposite trend with respect to deaths which occur in nursing homes. As education increases, the likelihood of an MI death occurring in a nursing home decreases. Finally, the likelihood of an MI death occurring in an emergency room increases as education increases (see Appendix 3 - Location of Death by Education). By combining the hospital discharge file together with the death certificate file an analysis of case fatality rates can be generated by age and by sex (the only two demographic variables common to both files and available for this study). When broken down by gender the data show that females are less likely to survive an MI. While the fatality rate for males was 18.5%, the fatality rate for women was 26.2%.

A breakdown by age shows that, as would be expected, the fatality rate started at 9.8% for persons between the ages of 25 to 34 and increased steadily to 34.1% for those over 75. The fatality rate for the group under 25 years of age, however, was a rather high 22.2% While this cannot be readily explained from the data, it is possible that some miscoded and uncoded ages fall into this category and may partially distort the results (see Appendix 4). Finally, an analysis of mortality by age by sex indicates that, given the 1994 data, the fatality rates for females tended to be higher than for males across the board for all age groups. The most significant differences, however, were in the younger age groups. For example, the fatality rate for males aged 25 to 34 was 8.7% versus 12.9% for females aged 25 to 34. The gap closed, somewhat when comparing the older age groups. One hypothesis is that there is a likelihood of MIs being initially misdiagnosed among younger females resulting in inadequate treatment. Further study is needed to better understand the reasons.

 The majority of those that died away from a hospital or ER were usually at home or in a nursing home. Those who died at home were most likely male, relatively younger and somewhat better educated. Those who died in nursing homes were most likely females with somewhat less education. Overall, fatality rates increase for older patients. Generally, men are somewhat more likely to survive an MI compared to women. Young women (under 25) much less likely to survive than young men.

Analyzing the Relationship Between Hospital Distance and Survival

Given the relationship between survival and the amount of time which passes before treatment, a key hypothesis is that the distance which the patient must travel for treatment is related to the likelihood of survival. After the onset of symptoms, a longer commute to a hospital translates to a longer time period which must pass before medical intervention can begin.

To test this hypothesis, a sample of myocardial infarction cases were selected from Los Angeles county and geocoded by the population centroid of the zip code in which the patients reside. Using Atlas GIS, the distance was then calculated between the geocoded location and the location of the hospital to which the patient was admitted. Cases were also coded as to whether they were discharged normally or if they died while in the hospital. Cross-tabs were then produced so that the survival rates could be viewed with respect to the distance from the hospitals. Through this approach it was determined that the mean distance between patients' homes and the hospitals to which they were admitted was 4.3 miles and a median of approximately 2.8 miles. Approximately 75% of all patients lived within 5 miles of the hospitals to which they were admitted.

It should be noted that this approach holds several inherent limitations. The first limitation is that the lack of addresses in the discharge records meant that the zip code centroid had to be used for point geocoding, resulting in location accuracy limitations. To help minimize the error, the population centroids of zip codes were used versus the geographic centroids. Another limitation relates to the fact that there is no way to determine where the infarction occurred. Cases which did not occur in the patients' homes would not have accurate distance calculations to the hospitals. While all of these limitations were expected to create some noise in the data, it was expected that the overall trends would still surface.

The analysis of survival by distance from the hospital actually produced an almost linear inverse relationship between distance and survival. Patients who were relatively closer to their hospital had consistently lower survival rates than those who were further away. In fact, the lowest survival rates were found to be within one mile of the hospitals. A hypothesis for the possible reason for the relationship was that demographic groups which cluster geographically may explain the trend.

To test this, a cross-tab was produced which plotted patient age by survival status. The cross-tab was consistent with studies which showed decreasing likelihood of survival among older patients. Next, the patient age was plotted against the distance from the hospital to which they were admitted. Using this approach it was found that there is a consistent pattern -- the oldest patients tend to cluster closest to their hospitals. Younger patients tend to live further away from the hospitals. This is consistent with the observations that many larger hospitals tend to be located in relatively urban areas with higher population density while suburban areas tend to have relatively younger residents.

Another possible explanation for the lower survival rates of patients living closer to hospitals revolved around socio-economic status. If hospitals tend to be located in urban areas with relatively lower affluence, there may be a more significant financial barrier to preventative care experienced by those patients. While it is difficult to test this hypothesis without primary research, an analysis was done to determine if the areas in which hospitals tend to be located are of a lower socio-economic status. Using Atlas GIS, the hospitals from which the patient samples were discharged were analyzed on the basis of the demographic patterns. The average household income of the general population was calculated from 1 to 10 miles from the hospitals to determine if there was a general socio-economic pattern. Using this analysis, it was found that while incomes tend to increase somewhat as one moves from 1 to 3 miles away from the hospitals. The average household income for residents within a 1 mile radius of the hospitals averaged $63 thousand and rose to $68 thousand at a 2 mile radius before leveling off. While this pattern alone is probably not enough to explain the survival rate trend, it does help substantiate the theory that affluence increases some distance away from the hospitals in this analysis.

Identifying Geographic Regions of High Incidence

In order to identify a target region for an intervention, a thematic map of Los Angeles County zip codes was first created for purposes of depicting the characteristics of MI cases. The maps below depict zip codes in terms of:

· Incidence rate - MI cases per 1,000 Persons

· Mortality Rate - The percentage of MI cases resulting in death

· The number of MI cases per square mile

The incidence rate was used to identify areas with the highest incidence High incidence areas were defined as the top 20% of all zip codes in Los Angeles County on the basis of cases per 1,000 residents. Contiguous high incidence zip codes were then aggregated together to form regions. In grouping together the specific zip codes other factors were also considered such as natural boundaries between zip codes and natural "community" breaks. For example, although Studio City and Beverly Hills zip codes are contiguous and both were considered incidence, the two were separated due to the natural boundary of the Santa Monica Mountains and the fact that Studio City is a distinctly separate community from Beverly Hills. In some cases zip codes which were not in the top 20% but were located between ones that are were selected to form a single contiguous region to avoid "holes" inside of regions.

Using this approach, 22 individual regions were created throughout Los Angeles County. Map 1 illustrates Los Angeles County zip codes, the incidence rate by zip code, and the regions into which the zip codes were combined. The region containing portions of Santa Monica and West Los Angeles surfaced as having the highest incidence rate at 4.46 cases per 1,000 persons.

Map 1

Once the high incidence regions were created they were screened on the basis of MI cases per square mile. This measure was used to identify the region with the largest number of cases in the smallest area. Regions ranged from a low of .09 cases per square mile (12 cases) in the Lake Hughes area of Northern Los Angeles County to 46.3 per square mile in the Santa Monica/West Los Angeles area (1,122 cases). Using this method, the Santa Monica/West Los Angeles area was identified as the region with the highest concentration of cases per square mile. Map 2 depicts the high incidence regions themed by the number of cases per square mile. The yellow indicates that the highest single region in Los Angeles County was the Santa Monica/West Los Angeles area in terms of cases per square mile.

Map 2

Given that some regions had a relatively high incidence per square mile but few cases, the regions were then ranked by total number of cases. The region which contained the largest number of cases (regardless of area or population) was again, the Santa Monica/West Los Angeles region. Map 3, below, depicts the regions in terms of the number of 1994 cases in each region. The yellow area represents the region with the largest number of diagnosed MI cases in the county.

 

Map 3

Using this approach, the Santa Monica/West Los Angeles region stands out as having not only the highest incidence rate per 1,000 persons but also the highest concentration per square mile and the largest number of cases overall.

Profile of the Region

The region defined by the methodology consists of 14 zip codes in the areas of Beverly Hills, Santa Monica, and West Los Angeles (see Appendix 7). The total area is approximately 24.2 square miles and is bordered by Santa Monica Blvd. to the north, Venice and Culver City to the south. It extends from the Pacific Ocean to just west of West Hollywood.

 Access to Health Care

The region contains 7 hospitals and approximately 2,648 physician office locations as of 1994. Approximately 3.6% (94) of the physicians in the region specialize in cardiovascular disease compared to 2.4% for all Los Angeles County physicians.

The average distance from all MI victims in the region to any closest hospital was approximately .84 miles. This was derived by assigning each MI case a latitude and longitude based on the population centroid of the zip code in which it was located. The distance from the centroid to the nearest hospital was then calculated and assigned to each case. The average distance to the closest emergency room for all victims was .99 miles. Emergency rooms were identified as hospitals with an 1994 emergency room expenditures of greater than zero dollars.

Findings and Conclusions

Target Population

Based on the approach used in this study, the Santa Monica/West Los Angeles region surfaces as having a very high incidence of myocardial infarction within Los Angeles County. Based on the 1994 data, and assuming that the pattern of 1994 MIs is indicative of future activity, this area appears to be one of the based locations in the county to target for an MI intervention program. The specific zip codes which would be targeted are identified in Appendix 7.

An analysis of the target region shows that it is relatively non-homogeneous with respect to demographic and socio-economic characteristics. A number of pockets exist in which the population varies considerably in terms of age, ethnicity, and income.

Given the variance in the makeup of the population throughout the region, a key recommendation is to split the region into two halves. Although the assumption throughout the study was that the region would be dealt with as whole, the diversity in the population will make it difficult to develop a health education program and communications approach which addresses the entire population. For example, while in zip code 90067 (Beverly Hills) Hispanics comprise only about 3.9% of the population (approximately 120 persons), zip code 90404 in Santa Monica is almost one third Hispanic (about 7,000 persons). Income also varies from an average household income of almost $146,000 in 90067 to $43,000 in 90404.

Given the makeup of the region, using the 405 freeway as a dividing line will divide the region into two subsets which can each be addressed individually for purposes of designing an intervention. This would result in southern and northern portions of the region.

The southern portion would include zip codes: 90025, 90064, 90401, 90403, 90404, 90405. This area is characterized by a larger percentage of Hispanics, lower overall income, but still above average for Los Angeles County, and a somewhat younger population. The remaining zip codes in the northern region contain a predominately white population, extremely high income, and a very high percent of persons over the age of 65.

The Relationship Between Survival and Distance Traveled

Based on the analysis, the distance between patients' homes and the hospital to which they were admitted does not appear to be a significant factor in survival rates. While distance cannot be specifically ruled out as a factor, age appears to be a more predictive variable. Given the limitations of the methodology a follow-up study is recommended which would specifically examine the distance from the exact location of the infarction to the hospital.

References

Archives of Internal Medicine. (1992). Inside Medicine Report, Mar. 1992.

Biger, J. Thomas, Jr. (1989). The Columbia University College of Physicians & Surgeons Complete Home Medical Guide. Edition 2, 1989, p388(1).

Bild, D.E., Fitzpatric, A., Fried, L.P., Wong, N.D., Haan, M.N., Lyles, M., Bovill, E., Polak, J.F., Schulz, R. (1993). Age-related trends in cardiovascular morbidity and physical functioning in the elderly; the cardiovascular health study. Journal of the American Geriatrics Society, Oct. 1993.

Med Help International. (1996). What Happens During a Heart Attack. Available online via HTTP: http://medhlp.netusa.net/index.html

Med Help International. (1996). What Happens During a Heart Attack, Available online via HTTP: http://medhlp.netusa.et/index.htm

O'Conner, N., Manson, JE, O'Conner, G.T., Buring, J.E. (1995). Psychosocial risk factors and nonfatal myocardial infarction. American Heart Association, Circulation,V. 92, No. 6, pp. 1458-1464, Sept. 1995.

Wong, ND; Gardin JM; Kurosaki, T.; Anton-Culver, H.; Rosmeman, Sidney; Gidding, S. (1995). Echocardiographic left ventricular systolic function and volumes in young adults. American Heart Journal. 1995 Mar.

Wong, N.D., Wilson, P.W., Kannel, W.B. (1991). Serum cholesterol as a prognostic factor after myocardial infarction. The Framingham Study, Annals of Internal Medicine. Nov, 1991.

Wong, ND; Levy, D; Kannel, W.B. (1990). Prognostic significance of the electrocardiogram after Q-wave myocardial infarction. The Framingham Study. Circulation. Mar 1990.

Yano, Katsuhiko, MacLean, C.J. (1989). The incidence of prognosis of unrecognized myocardial infarction in Honolulu, Hawaii, Heart Program Archives of Internal Medicine, July 1989, v149, n7.

 


Ronald Fischbach
Associate Professor, California State University, Northridge
18111 Nordhoff Street
Northridge, CA 91330-8285
Phone 818-677-3701
E-Mail ronald.fischbach@csun.edu

Elio Spinello
Partner, Retail Profit Management
17130 Devonshire Street Suite 205
Northridge, CA 91325
Phone 818-831-7607
E-Mail elio.spinello@csun.edu