. Author manuscript; available in PMC: 2010 Jul 12.
Published in final edited form as: Int J Canc Prev. 2009;2(6):403–413.
Abstract Objectives:The objective of the study was to examine the association between access to mammography facilities and utilization of screening mammography in an urban population.
Methods:Data on female breast cancer cases were obtained from an extensive mammography surveillance project. Distance to mammography facilities was measured by using GIS, which was followed by measuring geographical access to mammography facilities using Floating Catchment Area (FCA) method (considering all available facilities within an arbitrary radius from the woman's residence by using Arc GIS 9.0 software).
Results:Of 2,024 women, 91.4% were Caucasian; age ranged from 25 to 98 years; most (95%) were non-Hispanic in origin. Logistic regression found age, family history, hormone replacement therapy, physician recommendation, and breast cancer stage at diagnosis to be significant predictors of having had a previous mammogram. Women having higher access to mammography facilities were less likely to have had a previous mammogram compared to women who had low access, considering all the facilities within 10 miles (OR=0.41, CI=0.22-0.76), 30 miles (OR=0.52, CI=0.29-0.91) and 40 miles (OR=0.51, CI=0.28-0.92) radiuses.
Conclusions:Physical distance to mammography facilities does not necessarily predict utilization of mammogram and greater access does not assure greater utilizations, due to constraints imposed by socio economic and cultural barriers. Future studies should focus on measuring access to mammography facilities capturing a broader dimension of access considering qualitative aspect of facilities, as well as other travel impedances.
Keywords: Mammography, GIS, access, distance, breast cancer
IntroductionBreast cancer is one of the leading causes of death among women in the United States. The American Cancer Society estimated that 178,480 new cases and 40,460 deaths from breast cancer occurred among women in the United States in 2007 (American Cancer Society [ACS], 2007). Due to a lack of primary prevention of breast cancer, breast cancer mortality and morbidity reduction depends on secondary prevention, chiefly through screening mammography. Several randomized trials as well as population-based screening evaluations have indicated that early detection of breast cancer through screening mammography improves treatment options, the likelihood of successful treatment, and improved survival (William, Holladay, and Sheikh, 2003; Taber et al., 2003; Humphrey, Helfand, Chan, and Woolf, 2002; Duffy, Tabar, and Chen, 2002). A rise in mammography utilization is suggested by the observed trends (1987-1999) of an increase in breast cancer incidence confined to early stage breast cancer (Howe, et al., 2001; Edwards, et al., 2002; Blanchard, et al., 2004). A significant and substantial reduction in female breast cancer mortality has been observed in recent years because of screening mammography (Smith, et al., 2003; Duffy et al., 2006). However, the mortality rate from breast cancer is still too high, even though screening rates have increased and mortality decreased somewhat. The Healthy People 2010 target is 22.3 deaths per 100,000 women, but according to the American Cancer Society data the death rate is 26 per 100,000 women in 2007 (ACS, 2007).
Several researchers have explored barriers to obtaining mammograms, including the physical distance to mammography facilities and other barriers (Ann, Ronald, Raymond, and Gilligan, 2001; Jilda, Hyndman, and Holman, 2000a; Jilda, Hyndman, and Holman, 2000b). Understanding the geographical and social connections between the utilization of mammography and the locations of mammography facilities is critically important for developing effective programs to reduce breast cancer mortality. Health Education Promotion programs designed to increase mammography screening and produce subsequent reduction in breast cancer mortality may have opportunities to improve their effectiveness if they are able take barriers such as geographic distance to screening services into consideration. Health care decisions are strongly influenced by the type and quality of services available in the local area and the distance, time, cost, and ease of traveling to reach those services (Goodman, Fisher, Stukel, and Chang, 1997; Haynes, Bentham, Lovett, and Gale, 1999; Joseph and Phillips, 1984; Croner, Sperling, and Broome, 1996; Fortney, Rost, and Warren, 2003). The term ‘spatial accessibility’ is gaining more and more attention in the health care geography literature (Khan and Bhardwaj, 1994; Luo, 2004; Luo and Wang, 2003), which is a combination of dimensions of accessibility (travel impedance between patients and service points, that is measured in units of distance or travel time), and availability (refers to the number of local service locations from which a patient can choose). In this study, we focused on measuring access to mammogram facility by using GIS, considering both accessibility and availability dimensions. We also examined whether access to mammography facilities and other demographic variables influence utilization of mammography.
Methods Data CollectionThe data for this study were obtained from the Colorado Mammography Project (CMAP). CMAP was a National Cancer Institute funded project that was in operation from 1994-2004. CMAP was one member of a seven-site consortium, and obtained data on mammograms from approximately half of all mammography facilities in the six-county Denver metropolitan area of Colorado. For this study, information on mammograms for women from 1999-2001 was analyzed. The CMAP database included demographic data (age, race/ethnicity, education, and insurance status), data on mammogram results, previous mammogram history, family history, use of hormone replacement therapy, physician recommendation, and the zip codes of women's residences. Addresses of mammography facilities participating in CMAP were obtained for this study from the Colorado Department of Public Health and Environment. There were 46 facilities on the list that were operating during the time period (1999-2001) and were considered as the possible facilities that women might use to obtain a mammogram.
Calculation of Access to Mammography FacilityWe used the “Floating catchment area” (FCA) method by Luo and Wang (2004) to calculate access, that considered all available facilities within an arbitrary radius around a woman's address. Forty-six mammography facilities were geocoded using the ArcGIS System and placed in a separate file. Zip code centroids were obtained from a Zip code polygon file and compared to the database of patients. All Zip code centroids that had no patients from the sample were discarded, and then the numbers of patients were summed for each Zip code centroid and placed in a separate file. Mammography facility points and Zip code centroid points were connected to the regional street and highway network. Point-to-Point mileages were computed in a separate shortest path utility embedded within the GIS. The mileages were outputted in the form of a distance matrix. The distance matrix between Zip code centroids and mammography facilities was then imported into an Excel spreadsheet. Minimum distance that a woman would be willing to travel to get to a mammogram was considered 10, 20, 30, 40, or, 50 miles and following operations were performed for each of these arbitrary radius. For each specified radius, the number of women among all Zip codes within the specified radius was summed for each of the 46 mammography facilities identified within the study area. Then the inverse of these sums were computed to calculate the availability of that facility. Now, for each woman's Zip code within a specified radius, the availability for all facilities was summed to obtain the FCA index, representing her access to mammogram facility. Finally, indices for five different arbitrary radii 10, 20, 30, 40, and 50 miles were computed to calculate access to mammography facility.
Statistical Analysis to Examine Relationship between the VariablesTo further explore the association between the variables, logistic regression was performed. The dependent variable entered into the logistic model was whether the woman has had a previous mammogram or not (coded as yes=1 and No=0). Women who had a previous record of mammogram in the CMAP database or answered, “yes” on their patient information form when asked about their previous mammogram history at their index examination were considered as having had a previous mammogram (Figure 1 displays the distribution of the study population that did not have a previous mammogram in the six county areas).
Figure 1.Distribution of Population Who did not have had a Previous Mammogram (by Zip Code).
A series of categorical variables were created and entered into the logistic model such as, age, race/ethnicity, education, insurance status, family history, hormone replacement therapy, physician recommendation, and breast cancer stage at diagnosis along with access to mammography facilities. Among the independent variables, the ‘physician recommendation’ variable was divided into two broad categories: ‘diagnostic’ that included all the diagnostic procedures (such as, biopsy, needle localization, and ultrasound) and ‘evaluative’ that included the rest of the categories, such as, follow up, physical examination, surgical consultation etc. Breast cancer stage at diagnosis was also condensed into two categories: non-advanced breast cancer stage at diagnosis included carcinoma in situ, and localized tumors, which are malignant and invasive but confined to the organ of origin; and advanced stage of breast cancer at diagnosis included regional neoplasm that have extended beyond the organ of origin into surrounding tissues, involving regional lymph nodes, or both, and distant tumors that have spread to remote parts of the body from the primary site.
With the access ratio for five different radii (such as, 10 miles, 20 miles, 30 miles, 40 miles, and 50 miles) five different logistic regression models were developed. Both univariate and multivariate analyses were conducted and on the basis of analysis of maximum likelihood estimates, significant interaction terms were identified and there was no significant interaction between the variables.
Results Demographic CharacteristicsThe data from the Colorado Cancer Registry included 2042 individuals diagnosed with breast cancer during the period of 1999 to 2001. Descriptive statistics for the study population are summarized in Table 1.
Table 1.Characteristics of the Study Population
Factors Number (N=2042) Percentage (%) Age Below 40 years 121 6 40-49 503 24 50-59 609 30 60-69 364 18 70 years and above 445 22 Race/Ethnicity White 1811 91 Black 45 2 Asian 17 <1 American Indians and others 5 <1 Hispanic 104 5 Ethnicity Hispanic 104 5 Non Hispanic 1898 95 Education Less than High School Graduate 46 6 High School Graduate 198 26 Some College 251 32 College, or Post Graduate 280 36 Insurance Status Medicaid and/or Medicare 392 21 No Insurance 15 <1 Other (Private insurance, managed care and others) 1417 78 Stage of Breast Cancer In situ 320 16 Localized 1080 56 Regional 498 26 Direct 32 2 Previous mammogram Yes 1560 91 No 155 9 Family History Yes 190 17 No 953 83 Hormone replacement therapy Yes 440 42 No 608 58 Physician Recommendation Diagnostic 497 46 Evaluative 585 54The breast cancer cases ranged in age from 25 to 98 years with 30% being 50-59 years of age and nearly all were Caucasian (91%). Twenty one percent reported having Medicaid and/or Medicare and 78% also had private insurance.
Among those with data on family history, 17% had a positive family history of breast cancer. Among those with data on hormone replacement therapy, 42% were on hormone replacement therapy at the time of the initial mammogram. Nearly all (91%) of the women in the database had a previous mammogram.
Table 2 presents the odds ratios for the factors influencing having had a previous mammogram. Access to mammography facilities was negatively associated with having had a previous mammogram in the adjusted model developed for 10 miles, 30 miles, and 40 miles radius. Women who had greater access to mammography facilities were 59% less likely and women who had medium access to such facilities were 58% less likely respectively of having had a previous mammogram, compared to women who had low access to mammography facilities; and these findings were significant in both the crude and adjusted models for the 10 miles radius measure.
Table 2.OR for Factors Predicting Women who had a Previous Mammogram
Factors Crude OR 95% CI Adjusted OR 95% CI Age Group Below 40 years 0.13 0.07-0.22* 0.11 0.06-0.22* 40-49 1.00 1.00 50-59 3.24 1.81-5.80* 1.63 0.80-3.32 60-69 2.79 1.44-5.40* 1.72 0.77-3.90 70 years and above 0.73 0.47-1.14 1.02 0.50-2.09 Race/Ethnicity White 1.00 1.00 Black 0.45 0.18-1.10 0.68 0.20-2.19 Asian 1.57 0.21-11.89 2.75 0.21-35.65 Hispanic 0.37 0.21-0.66* 0.51 0.06-4.69 Ethnicity Hispanic 1.00 1.00 Non Hispanic 2.03 1.17-3.52* 1.55 0.19-12.82 Education Less than High School Graduate 1.00 1.00 High School Graduate 0.33 0.20-0.54* 0.66 0.34-1.26 Some College 0.32 0.21-0.51* 0.65 0.36-1.19 College, or Post Graduate 0.38 0.24-0.60* 0.82 0.44-1.51 Insurance Status Medicaid and/or Medicare 0.38 0.19-0.75* 0.48 0.20-1.17 No insurance 1.00 1.00 Other (Private insurance, managed care and others) 0.64 0.34-1.22 0.95 0.43-2.13 Family History Yes 1.00 1.00 No 0.18 0.11-0.28* 0.37 0.19-0.69* Hormone replacement therapy Yes 1.00 1.00 No 0.09 0.06-0.14* 0.15 0.08-0.27* Physician Recommendation Diagnostic 1.00 1.00 Evaluative 1.23 0.85-1.77 2.00 1.24-3.23* Breast cancer stage at diagnosis Non-advance stage 1.00 1.00 Advance stage 0.58 0.41-0.83* 0.69 0.43-1.10 Access to mammogram facilities Within 10 miles radius High access 0.46 0.27-0.77* 0.41 0.22-0.76* Medium access 0.42 0.25-0.71* 0.42 0.23-0.76* Low access 1.00 1.00 Within 20 miles radius High access 0.60 0.38-0.96* 0.58 0.34-1.00 Medium access 0.81 0.49-1.35 0.72 0.39-1.31 Low access 1.00 1.00 Within 30 miles radius High access 0.68 0.42-1.10 0.52 0.29-0.91* Medium access 0.67 0.41-1.09 0.85 0.49-1.49 Low access 1.00 1.00 Within 40 miles radius High access 0.58 0.35-0.97* 0.51 0.28-0.92* Medium access 0.79 0.47-1.32 0.68 0.37-1.25 Low access 1.00 1.00 Within 50 miles radius High access 0.91 0.50-1.63 0.82 0.41-1.61 Medium access 0.81 0.46-1.44 0.78 0.40-1.51 Low access 1.00 1.00For the 30 miles radius access measure, women who had high access to mammography facilities was 48% less likely of having had a previous mammogram when compared to women who had less access to mammography facilities (OR = 0.52, 95% CI = 0.29-0.91).
The odds of having had a previous mammogram for women who had high access to mammography facilities were 0.51(95% CI= 0.28-0.92) times compared to women who had less access to mammography facilities and these findings were significant for both the crude and adjusted models developed for the 40 miles radius access measure. The 50 miles radius access measure finding was not statistically significant in any model.
In Table 2, after adjustment for all other variables, women in the age group below 40 years were negatively associated with having had a previous mammogram when compared to women in the age group 40-49 years, which was statistically significant (adjusted OR = 0.11, 95% CI = 0.06-0.22). After adjustment for other variables, neither race nor ethnicity remained significantly associated with having had a previous mammogram when compared with White women. Women's educational attainment level and insurance status were not statistically significantly associated with having had a previous mammogram in the adjusted model. Not having a family history of breast cancer appeared as a negative predictor of having had a previous mammogram in the adjusted model, as it had in the univariate model, and remained statistically significant (adjusted OR = 0.37, 95% CI = 0.19-0.69). The odds of having had a previous mammogram for women who did not have a family history of breast cancer were about one-third as likely as women who had a positive family history of breast cancer. Hormone replacement therapy remained negatively associated with having had a previous mammogram after controlling for all other variables in the adjusted model (adjusted OR = 0.15, 95% CI = 0.08-0.27) and the finding was statistically significant. In the adjusted model after controlling for all other variables, the evaluative recommendation by physicians was found to be a significant predictor of having had a previous mammogram (OR = 2.00, 95% CI = 1.24-3.23).
DiscussionThe gravity model, a combined measure of accessibility and availability was used to evaluate the potential spatial interaction between any woman's location and all alternative mammography facilities within a reasonable distance. The relationship of geographical access and utilization of mammogram is noteworthy. In Denver metropolitan area most of mammography facilities are located close to the downtown where accessibility is higher. Women diagnosed with breast cancer without a previous mammogram also higher in this area (Figure 1). In another study we found women diagnosed with advanced stage of breast cancer are also higher in these areas (Rahman, et al., 2007). Several issues contribute in determining which mammography facility to be used to get a mammogram, or more specifically, why a woman would not use the nearest mammography facility or just one facility to obtain her mammograms. Factors such as the type of health insurance and their policies regarding reimbursement may have determined which mammography facility a woman must use to get a mammogram. A common physician practice is to recommend their patients to a specific mammography facility. Some women may prefer to go to a mammography facility that is closer to their work place rather than from their residence. Moreover, it is crucial to specify one mammogram facility that the woman used to measure the straight-line distance from her residence. Typically for a diagnosis of breast cancer a woman will have one or two mammograms and then an ultrasound, which will be followed by a biopsy and all of these examinations usually do not occur within the same clinic or on the same day. Both access and distance are equally important in considering barriers to overcome for screening mammogram and diagnostic testings for breast cancer. Being hindered in either way would likely result in a later stage of breast cancer at diagnosis. Taking into account all the above issues it seems more appropriate that we measure access to mammogram facility considering all the available facilities within an arbitrary radius, rather than the distance from the woman's residence to nearest facility or one specific facility.
Again, in the literature the arbitrary radius is usually considered as 30 miles for FCA method; however most of these studies are about primary care rather than preventive care. Assuming that the minimum distance a woman would be willing to travel to get a preventive service, such as, mammogram would be different, access ratio for several different radii (10, 20, 30, 40, and 50 miles) were measured and compared. While comparing access measures of different arbitrary radiuses in the FCA method, as the radius increased from 10 miles to 50 miles, the standard deviation of access measures decreased and also the range from minimum to maximum decreased (Table 3).
Table 3.Comparison of Accessibility Measures
Radius Total number Minimum Maximum Mean Std. Deviation 10 Mile 1745 .0000000 .1075926 .0263610 .0261128 20 Mile 1745 .0000000 .0768412 .0263610 .0136416 30 Mile 1745 .0000000 .0536146 .0263610 .0084047 40 Mile 1745 .0000000 .0377056 .0263610 .0056572 50 Mile 1745 .0011524 .0314764 .0263610 .0042359This indicates that the access measure with a higher radius had less variance, which led to stronger spatial smoothing, which is a manifestation of MAUP (modifiable areal unit problem). Access scores tended to increase with increasing radius, as one would have more access if she were permitted or capable of traveling further. As the radius increased from 10 miles to 50 miles, the population with high access also increased (Figure 2, Figure 3 and Figure 4).
Figure 2.Accessibility to Mammogram Facilities within 10 Miles Radius (by Zip Code).
Figure 3.Accessibility to Mammogram Facilities within 30 Miles Radius (by Zip Code).
Figure 4.Accessibility to Mammogram Facilities within 50 Miles Radius (by Zip Code).
However, if we look at the mean access measure for the population, it remains the same for all the measures with different radiuses (Table 3) as because increasing radius does not necessarily mean increasing access. Access depends on the distribution of supply of and demand for mammograms. In the method of calculating access to mammography facilities in the current study, the availability of the facility was considered only by the total number of women sharing that facility, which meant that all the mammography facilities were viewed as having equal capacity. When the radius increased, the number of women within that arbitrary radius increased as well, which acted to decrease the ultimate access to a mammography facility as more women shared that facility. To overcome this limitation, future research is needed that will consider the qualitative aspects of the mammography facilities, such as, the size, number of staff members, amount of equipment and other details that might have affected the capacity of a facility.
Several other limitations that were related to the access measure of the current study are as follows: First, the population data were geocoded by using women's Zip codes, as exact addresses were not available because of a requirement to maintain confidentiality. By using Zip codes, women were assigned to an area rather than assigned to a single point. This technique might have decreased the level of precision for the measure of access to mammography facility. Second, the current study was limited to only six county areas. A known limitation of the FCA method in measuring access is that people within a catchment area have equal access to all providers within that same catchment area, and all providers beyond the radius of the catchment area are inaccessible, regardless of any differences in distances (Luo, 2004; Luo and Wang, 2004;). Finally, absence of individual level data on income or socio-economic status and missing data on health insurance, education, hormone replacement therapy, family history, physician recommendation and previous mammogram were also a major limitation of the current study.
AcknowledgementDr. Fahui Wang, Department of Geography, Northern Illinois University.
ReferencesRetroSearch is an open source project built by @garambo | Open a GitHub Issue
Search and Browse the WWW like it's 1997 | Search results from DuckDuckGo
HTML:
3.2
| Encoding:
UTF-8
| Version:
0.7.3
