Average Annual Age-Adjusted Cancer Incidence Rates, 2000-2004, at the Delaware Sub-County Level Delaware Health and Social Services Division of Public Health Health Promotion and Disease Prevention, Bureau of Chronic Diseases Phone: 302-744-1040 Fax: 302-739-2545 April, 2008 Average Annual Age-Adjusted Cancer Incidence Rates, 2000-2004, at the Delaware Sub-County Level Summary This report was published by the Delaware Division of Public Health (DPH) to supplement cancer incidence data previously published at the state and county level. Incidence rates describe the extent to which new cancer cases are diagnosed in a population over a period of time; they are typically presented as the number of new cancer cases diagnosed per year per 100,000 individuals. This report focused on incidence rates for all cancers combined, as well as for the four most frequently occurring cancers (breast, colorectal, lung, and prostate cancers). Incidence rates were calculated from the most recently available five-year period of data (2000-2004). Additionally, rates were age-adjusted to reflect the unique age breakdown of residents in different areas of Delaware. Incidence rates presented in this report should be interpreted as the average number of new cancer cases diagnosed per 100,000 residents each year from 2000-2004. The report focuses on the sub-county, or Census County Division (CCD), level. CCDs are county subdivisions established cooperatively by the United State Census Bureau and state and local authorities. In general, CCDs are defined in such a way that boundaries follow visible features and census tracts. Delaware is divided into 27 CCDs, ranging in size from 6,093 to 83,884 people. DPH chose the CCD level of analysis because it provided cancer rate information for more distinct geographical areas than whole counties. Additionally, CCD was selected as the level of analysis because of the difficulty of drawing conclusions from analyses of cancer rates in areas with smaller populations. Analyses determined if incidence rates at the CCD level were significantly higher or lower than the incidence rate at the state level. Statistical significance was evaluated by comparing the upper and lower 95% confidence intervals of the rates. A confidence interval is sometimes called the "margin of error." A 95% confidence interval means that we are 95% sure that the incidence rate is between the lower confidence interval and upper confidence interval. If the confidence intervals from two incidence rates do not overlap, we call the rates significantly different. When incidence rates are significantly different, it means that the difference between these two rates is larger than would be expected by chance alone. On the other hand, if incidence rates from two populations have confidence intervals that overlap, we cannot say that the rates are significantly different. Even at the CCD level, the confidence intervals for rates are so large that, in many instances, they overlap. A non-significant difference between incidence rates is commonly interpreted as "no meaningful difference" between rates. Determining if two incidence rates are significantly different from one another is an important step in investigating suspected cancer clusters. The Centers for Disease Control and Prevention (CDC) and the National Cancer Institute (NCI) define a cancer cluster as a greater-than-expected number of cancers diagnosed among a population in a particular geographic area, over a given period of time. Some cancer clusters occur simply by chance. In these situations, clusters are not the result of a single, external cause; instead, the cluster simply reflects coincidental spatial clustering among individuals who have been diagnosed with cancer. Other cancer clusters may be due to environmental exposure. Health officials are more easily able to trace the origin of a cancer cluster to an environmental exposure if the cluster involves one or more of the following three traits: (1) the cluster contains a large number of one type of cancer, as opposed to several cancer types; (2) the cluster involves a rare cancer type, as opposed to more commonly diagnosed cancers; and (3) the cluster involves a large number of cases of a type of cancer in an age group that is usually not affected by that cancer type. Cancer clusters may also reflect better access to health care. Residents from one geographic area may be more likely to have cancer screening services compared to residents from another area. In these situations, cancer clusters exist because more cases of cancer are being diagnosed than in other areas. As such, these cancer clusters do not reflect a truly elevated cancer risk in a geographic area. Finally, cancer clusters may be due to clustering of lifestyle behaviors. Tobacco use, regular physical activity, diet, and other behaviors strongly impact cancer risk. If residents in one geographic area are more likely to engage in unhealthy lifestyle behaviors, the cancer incidence rate for that area may be elevated compared to other areas. Based on 2000-2004 data, eight of the 27 CCDs in Delaware had statistically significant elevated incidence rates for one or more cancer types compared to the state. DPH will be arranging meetings with the communities in each of the eight areas with statistically higher rates, working with the respective legislators, seeking input from the Delaware Cancer Consortium, and, if appropriate, conducting formal epidemiologic studies to further investigate elevated cancer incidence rates. CCD Statistically Elevated CCD Incidence Rate Compared to State Incidence Rate 1. Central Pencader * All Cancer Combined (23) 2. Kenton * All Cancer Combined (6) 3. Lower Christiana * All Cancer Combined (21) * Lung Cancer (9) 4. Middletown-Odessa * Colorectal Cancer (8) 5. Millsboro * Lung Cancer (5) 6. New Castle * All Cancer Combined (65) * Lung Cancer (16) * Prostate Cancer (21) 7. Upper Christiana * Prostate Cancer (9) 8. Wilmington * All Cancer Combined (36) * Lung Cancer (10) * Prostate Cancer (13) Introduction Purpose This report was published by the Delaware Division of Public Health (DPH) to supplement cancer incidence data previously published at the state and county level. The current report focused on the sub-county, or Census County Division (CCD), level. CCDs are county subdivisions established cooperatively by the United State Census Bureau and state and local authorities. In general, CCDs are defined in such a way that boundaries follow visible features and census tracts. Census County Divisions Delaware is divided into 27 CCDs, ranging in size from 6,093 to 83,884 people. Figure 1 is a map of Delaware showing the CCDs used in this report. Incidence rates for all cancers combined, as well as for the four most frequently occurring cancers (breast, colorectal, lung, and prostate cancers) are presented in this report. DPH chose the CCD level of analysis because it provided cancer rate information for more distinct geographical areas than whole counties. Additionally, CCD was selected as the level of analysis because of the difficulty of drawing conclusions from analyses of cancer rates in areas with smaller populations (such as neighborhoods). Calculation of Rates A cancer incidence rate is defined as the number of new cancer cases diagnosed in a population over a period of time. Incidence rates are typically presented as the number of new cancer cases diagnosed per year per 100,000 individuals. This report calculated annual average incidence rates using five years of data (2000-2004). Because Delaware is a small state, five years of data are combined and then averaged to reduce the likelihood of statistical "flukes" that might occur if only one year of data was used. The period 2000-2004 is the most recent for which a complete set of 5-year data is available. The cancer incidence rates are age-adjusted, which is a statistical procedure that allows a rate to be compared to other populations that may be younger or older. Because the most significant influence on the risk of cancer is a person's age, age-adjustment is necessary to make sure that we "control" the effect of age when comparing a population's risk of cancer. Statistical Significance The 95% confidence interval for each cancer rate is also presented. A confidence interval is sometimes called the "margin of error." A 95% confidence interval means that we are 95% sure that the incidence rate is between the lower confidence interval (LCI) and upper confidence interval (UCI). When confidence intervals are small, we are more certain that the incidence rate is a good estimate. When confidence intervals are large, we are less certain that the incidence rate is a good estimate. One factor that determines the width of the confidence interval is the size of the population from which the rate is calculated. Rates calculated from small populations have larger confidence intervals than rates calculated from big populations. This means that we are less certain of the incidence rate when the rate is calculated for smaller populations. If the confidence intervals from two incidence rates do not overlap, we call the rates significantly different. When the difference between two incidence rates is statistically significant, it means that the difference between these two rates is larger than would be expected by chance alone. On the other hand, if rates from two populations have confidence intervals that overlap, we cannot say that the rates are significantly different. A non-significant difference between incidence rates is commonly interpreted as "no meaningful difference" between rates. Sometimes the incidence rate for a CCD is highly elevated compared to the incidence rate for the state of Delaware; but, if the CCD incidence rate has a very wide confidence interval, it still may not be significantly different from the state incidence rate. When incidence rates are computed for an entire geographic area based on a very small number of cases, rates are estimated with a larger degree of uncertainty. This uncertainty is represented by a very wide confidence interval. When confidence intervals are wide, they are more likely to overlap with the confidence intervals of incidence rates from other areas; this means that it is more difficult to establish a significant difference between incidence rates. For this reason, incidence rates are not calculated at smaller levels (such as neighborhoods). Incidence rates for small geographic levels are estimated with such a high degree of uncertainty that few, if any, conclusions can be drawn from the data. In addition to all cancer sites combined, this report focused on calculating incidence rates for breast, colorectal, lung and prostate cancers because they are the four most commonly occurring cancers in Delaware and the United States. Incidence rates for rare cancers were not calculated at the CCD level because of the high degree of uncertainty associated with rates calculated from a very small number of cases. Determining if two incidence rates are significantly different from one another is an important step in investigating suspected cancer clusters. The Centers for Disease Control and Prevention (CDC) and that National Cancer Institute (NCI) define a cancer cluster as a greater-than-expected number of cancers diagnosed among a population in a particular geographic area, over a given period of time. Reasons Why Cancer Clusters Occur Some cancer clusters occur simply by chance. In these situations, clusters are not the result of a single, external cause; rather, the cluster reflects coincidental spatial clustering among individuals who have been diagnosed with cancer. Other cancer clusters may be due to environmental exposure. Health officials are more easily able to trace the origin of a cancer cluster to an environmental exposure if the cluster involves one or more of the following three traits: (1) the cluster contains a large number of one type of cancer, as opposed to several cancer types; (2) the cluster involves a rare cancer type, as opposed to more commonly diagnosed cancers; and (3) the cluster involves a large number of cases of a type of cancer in an age group that is usually not affected by that cancer type. Cancer clusters may also reflect better access to health care. Residents from one geographic area may be more likely to have cancer screening services compared to residents from another area. In these situations, cancer clusters exist because more cases of cancer are being diagnosed earlier than in other areas. As such, these cancer clusters do not reflect a truly elevated cancer risk in a geographic area. Finally, cancer clusters may be due to clustering of lifestyle behaviors. Tobacco use, regular physical activity, diet, and other behaviors strongly impact cancer risk. If residents in one geographic area are more likely to engage in unhealthy lifestyle behaviors, the cancer incidence rate for that area may be elevated compared to other areas. Differences in Terms of Number of Cases For those CCDs with significantly higher rates of cancer than all of Delaware, we calculated the difference in terms of the number of cases. To do this, the age-adjusted rate for the state was multiplied by the CCD?s population. Next, the age-adjusted rate for the CCD was multiplied by the CCD?s population. The first value was subtracted from the second value to yield the difference between the CCD and the state, in terms of additional cancer cases. These differences should be interpreted as the ?extra? number of cases per year that are expected to occur in the CCD as a result of the CCD having a higher incidence rate compared to the state. For example, the 2000-2004 average annual age-adjusted lung cancer incidence rate for Delaware was 75.2 per 100,000. If this rate is applied to the New Castle CCD population, the resulting number of expected lung cancer cases is 62. The lung cancer incidence rate for the New Castle CCD is 94.0 per 100,000. When this second rate is applied to the New Castle CCD population, the resulting number of expected lung cancer cases is 78. Subtracting the expected number of lung cancer cases in the New Castle CCD based on the state incidence rate (62) from the number of expected lung cancer cases based on the New Castle CCD incidence rate (78) yields 16. This means that the New Castle CCD experienced an average of 16 more lung cancer cases per year compared to the state as a whole. Appendix A of this report provides details about how the populations, CCDs, rates, and confidence intervals were calculated. Appendix B of this report provides the population for the CCDs. Figure 1: New Castle County Delaware Census County Divisions (not shown) Figure 1 (continued): Kent and Sussex Counties, Delaware Census County Divisions (not shown) Limitations As mentioned above, we have less confidence in rates calculated for small geographic areas. CCDs are larger than other geographic units (such as neighborhoods or census tracts); however, even at the CCD level, the confidence intervals for rates are so large that, in many instances, they overlap. Rates for different race/ethnic groups, for males and for females, and for infrequently-occurring cancers were not computed. Calculating rates for these subgroups would result in incidence rates with very wide confidence intervals, making it impossible to determine with certainty if one incidence rate is different from another. Many factors contribute to cancer risk, including age, heredity, tobacco use, nutrition, and exposures to cancer-causing agents. Because of this, these analyses cannot determine the reason why CCD incidence rates are higher or lower in certain geographic areas. Cancers resulting from exposure develop slowly in people, usually appearing 5 - 40 years after exposure to a cancer-causing agent. Migration within the state further complicates the issue; people are constantly moving in and out CCDs, making it hard to link exposure to a cancer causing agent to where a person lives. Therefore, a map, alone, cannot prove that something in the environment is responsible for elevated cancer rates. It is also important to note that just because a person lives in an area where there is a statistically higher rate of cancer does not mean that he or she, personally, is more likely to get cancer than someone who lives in an area with a lower rate. A person's risk depends on many things including lifestyle (smoking, diet), family history, and contact with cancer causing substances (sunlight, x-rays, tobacco smoke, and chemicals). Results For each of the five cancer types considered here ((a) all cancer sites combined; (b) breast cancer; (c) colorectal cancer; (d) lung cancer; and (e) prostate cancer), average annual age-adjusted incidence rates for CCDs are presented in both table and graph format. For comparison purposes, tables and graphs also include respective incidence rates for the state of Delaware. Incidence rates are arranged in descending order. In graphical format, incidence rates for CCDs are represented by blue bars. State incidence rates are represented by red bars. The black lines that accompany each bar represent the lower and upper confidence interval for that particular incidence rate. Table 1: Average Annual Age-Adjusted Cancer Incidence Rate, All Cancer Sites Combined, Delaware, 2000-2004, by Census County Division Lower Upper Difference Rate Confidence Confidence Statistical in Terms of Interval Interval Significance+ "Extra Cases"# Kenton 595.44 496.81 694.08 Greater than State Rate 6 New Castle 564.76 539.83 589.69 Greater than State Rate 65 Central Pencader 554.60 508.42 600.78 Greater than State Rate 23 Lower Christiana 543.99 512.82 575.16 Greater than State Rate 21 Wilmington 534.94 510.53 559.36 Greater than State Rate 36 Upper Christiana 531.79 483.76 579.81 Lewes 511.19 480.40 541.98 Smyrna 510.41 457.39 563.43 Middletown-Odessa 508.98 470.47 547.49 Georgetown 500.99 444.11 557.88 Harrington 493.38 434.85 551.91 STATE OF DELAWARE 486.34 479.72 492.96 Millsboro 481.98 448.31 515.66 Bridgeville-Greenwood 479.79 420.67 538.92 Felton 477.99 394.85 561.12 Greater Newark 477.91 452.24 503.57 Brandywine 471.69 452.89 490.49 Dover 470.84 447.37 494.31 Seaford 465.43 428.80 502.07 Piedmont 464.22 433.40 495.03 Pike Creek- Central Kirkwood 462.11 435.13 489.09 Laurel-Delmar 454.83 416.18 493.48 Central Kent 436.89 390.45 483.33 Milton 436.06 389.08 483.04 Milford North 432.06 376.81 487.30 Selbyville-Frankford 425.29 397.84 452.74 Less than State Rate Milford South 374.78 337.20 412.37 Less than State Rate Red Lion 317.34 251.78 382.90 Less than State Rate All rates are per 100,000 and age-adjusted to the 2000 U.S. standard population +"Greater than State Rate" indicates CCD rate is statistically significantly greater than the state rate +"Less than State Rate" indicates CCD rate is statistically significantly lower than the state rate # Numbers should be interpreted as the "extra" number of cases per year that are expected to occur in the CCD as a result of the CCD having a higher incidence rate compared to the state Rates statistically different from the state rate are shown in italics (in original document) Source: Delaware Division of Public Health, Cancer Registry Figure 2: Average Annual Age-Adjusted Cancer Incidence Rate, All Sites Combined, Delaware, 2000-2004, by Census County Division (not shown) Table 2: Average Annual Age-Adjusted Incidence Rate, Breast Cancer, Delaware, 2000-2004, by Census County Division Lower Upper Difference Rate Confidence Confidence Statistical in Terms of Interval Interval Significance+ "Extra Cases"# Greater Newark 139.27 120.61 157.93 Upper Christiana 138.17 106.69 169.65 Piedmont 135.95 113.01 158.88 Milford North 135.09 92.69 177.49 Central Kent 131.38 98.14 164.62 Lewes 129.79 107.48 152.10 Georgetown 128.83 89.87 167.80 Wilmington 127.77 111.81 143.74 New Castle 127.63 112.03 143.24 Brandywine 127.34 113.89 140.80 Pike Creek- Central Kirkwood 127.29 108.15 146.42 Central Pencader 124.70 97.37 152.03 Lower Christiana 124.55 103.85 145.26 STATE OF DELAWARE 123.10 118.56 127.64 Harrington 120.32 81.01 159.62 Dover 119.71 103.63 135.78 Middletown-Odessa 119.58 95.66 143.50 Laurel-Delmar 118.18 90.88 145.48 Millsboro 115.25 91.95 138.54 Milton 113.21 79.37 147.04 Seaford 112.67 87.98 137.35 Bridgeville-Greenwood 111.61 71.67 151.55 Milford South 104.27 76.71 131.83 Smyrna 95.45 63.83 127.08 Selbyville-Frankford 93.24 75.06 111.43 Less than State Rate Felton * * * Kenton * * * Red Lion * * * All rates are per 100,000 and age-adjusted to the 2000 U.S. standard population * Rates are omitted because they are based on fewer than 25 cases + "Greater than State Rate" indicates CCD rate is statistically significantly greater than the state rate + "Less than State Rate" indicates CCD rate is statistically significantly lower than the state rate # Numbers should be interpreted as the "extra" number of cases per year that are expected to occur in the CCD as a result of the CCD having a higher incidence rate compared to the state Rates statistically different from the state rate are shown in italics (in original document) Source: Delaware Division of Public Health, Cancer Registry Figure 3: Average Annual Age-Adjusted Incidence Rate, Breast Cancer, Delaware, 2000-2004, by Census County Division (not shown) Table 3: Average Annual Age-Adjusted Incidence Rate, Colorectal Cancer, Delaware 2000-2004, by Census County Division Lower Upper Difference Rate Confidence Confidence Statistical in Terms of Interval Interval Significance+ "Extra Cases"# Middletown-Odessa 77.08 61.33 92.84 Greater than State Rate 8 Smyrna 71.20 51.46 90.93 Bridgeville-Greenwood 65.71 43.94 87.48 Dover 63.39 54.73 72.04 Wilmington 63.01 54.65 71.37 New Castle 59.90 51.55 68.24 Upper Christiana 58.29 41.45 75.14 Milford South 57.59 43.02 72.17 Harrington 56.88 36.53 77.24 Seaford 56.27 43.70 68.83 Lower Christiana 55.70 45.89 65.50 Central Kent 54.18 36.23 72.13 STATE OF DELAWARE 53.23 51.03 55.42 Lewes 52.81 43.03 62.59 Central Pencader 51.85 37.33 66.37 Millsboro 50.43 39.64 61.21 Milford North 50.14 31.57 68.71 Piedmont 49.61 39.79 59.44 Laurel-Delmar 47.99 35.42 60.55 Greater Newark 46.64 38.53 54.75 Brandywine 45.48 39.79 51.18 Milton 43.90 29.35 58.44 Pike Creek- Central Kirkwood 43.12 34.95 51.29 Selbyville-Frankford 41.36 32.62 50.11 Less than State Rate Felton * * * Georgetown * * * Kenton * * * Red Lion * * * All rates are per 100,000 and age-adjusted to the 2000 U.S. standard population * Rates are omitted because they are based on fewer than 25 cases + "Greater than State Rate" indicates CCD rate is statistically significantly greater than the state rate + "Less than State Rate" indicates CCD rate is statistically significantly lower than the state rate # Numbers should be interpreted as the "extra" number of cases per year that are expected to occur in the CCD as a result of the CCD having a higher incidence rate compared to the state Rates statistically different from the state rate are shown in italics (in original document) Source: Delaware Division of Public Health, Cancer Registry Figure 4: Average Annual Age-Adjusted Incidence Rate, Colorectal Cancer, Delaware, 2000-2004, by Census County Division (not shown) Table 4: Average Annual Age-Adjusted Incidence Rate, Lung Cancer, Delaware, 2000-2004, by Census County Division Lower Upper Difference Rate Confidence Confidence Statistical in Terms of Interval Interval Significance+ "Extra Cases"# Kenton 110.13 69.34 150.92 Lower Christiana 99.76 86.55 112.97 Greater than State Rate 9 Harrington 99.47 73.42 125.52 Smyrna 99.06 75.85 122.26 Millsboro 97.61 82.98 112.24 Greater than State Rate 5 New Castle 94.04 83.68 104.41 Greater than State Rate 16 Bridgeville-Greenwood 91.87 66.15 117.59 Wilmington 88.35 78.39 98.32 Greater than State Rate 10 Georgetown 84.34 60.96 107.72 Central Pencader 80.55 61.40 99.69 Dover 79.28 69.64 88.91 Seaford 78.18 63.30 93.06 Laurel-Delmar 76.39 60.78 92.00 STATE OF DELAWARE 75.22 72.63 77.81 Upper Christiana 74.67 56.08 93.26 Selbyville-Frankford 74.63 63.54 85.72 Middletown-Odessa 72.08 57.27 86.89 Pike Creek- Central Kirkwood 71.48 60.95 82.01 Central Kent 70.68 51.28 90.07 Milford North 67.42 45.98 88.86 Lewes 67.36 56.65 78.07 Greater Newark 63.70 54.24 73.17 Milton 61.55 44.14 78.96 Milford South 57.74 43.13 72.35 Less than State Rate Brandywine 57.42 50.99 63.86 Less than State Rate Piedmont 45.51 35.95 55.07 Less than State Rate Felton * * * Red Lion * * * All rates are per 100,000 and age-adjusted to the 2000 U.S. standard population * Rates are omitted because they are based on fewer than 25 cases + "Greater than State Rate" indicates CCD rate is statistically significantly greater than the state rate + "Less than State Rate" indicates CCD rate is statistically significantly lower than the state rate # Numbers should be interpreted as the "extra" number of cases per year that are expected to occur in the CCD as a result of the CCD having a higher incidence rate compared to the state Rates statistically different from the state rate are shown in italics (in original document) Source: Delaware Division of Public Health, Cancer Registry Figure 5: Average Annual Age-Adjusted Incidence Rates, Lung Cancer, Delaware, 2000-2004, by Census County Division (not shown) Table 5: Average Annual Age-Adjusted Incidence Rates, Prostate Cancer, Delaware, 2000-2004, by Census County Division Lower Upper Difference Rate Confidence Confidence Statistical in Terms of Interval Interval Significance+ "Extra Cases"# Kenton 263.23 165.73 360.73 Upper Christiana 244.70 192.98 296.42 Greater than State Rate 9 New Castle 221.40 197.30 245.51 Greater than State Rate 21 Piedmont 204.82 173.56 236.07 Wilmington 204.61 181.57 227.65 Greater than State Rate 13 Georgetown 190.94 139.04 242.84 Lower Christiana 188.28 160.92 215.63 Central Pencader 185.40 146.44 224.35 Brandywine 181.71 164.29 199.13 Greater Newark 181.28 157.04 205.51 Middletown-Odessa 176.40 141.11 211.68 Dover 170.26 149.24 191.28 STATE OF DELAWARE 169.01 163.20 174.82 Pike Creek- Central Kirkwood 161.08 137.28 184.88 Smyrna 158.43 114.95 201.92 Central Kent 156.56 115.92 197.20 Milford North 140.91 95.50 186.31 Selbyville-Frankford 135.11 113.63 156.59 Less than State Rate Lewes 134.36 112.26 156.46 Less than State Rate Millsboro 127.22 103.11 151.32 Less than State Rate Laurel-Delmar 125.62 96.19 155.05 Less than State Rate Milton 119.64 84.29 155.00 Less than State Rate Harrington 117.96 75.03 160.90 Less than State Rate Seaford 114.76 88.07 141.45 Less than State Rate Bridgeville-Greenwood 114.04 71.80 156.28 Less than State Rate Milford South 105.21 75.75 134.67 Less than State Rate Felton * * * Red Lion * * * All rates are per 100,000 and age-adjusted to the 2000 U.S. standard population * Rates are omitted because they are based on fewer than 25 cases + "Greater than State Rate" indicates CCD rate is statistically significantly greater than the state rate + "Less than State Rate" indicates CCD rate is statistically significantly lower than the state rate # Numbers should be interpreted as the "extra" number of cases per year that are expected to occur in the CCD as a result of the CCD having a higher incidence rate compared to the state Rates statistically different from the state rate are shown in italics (in original document) Source: Delaware Division of Public Health, Cancer Registry Figure 6: Average Annual Age-Adjusted Incidence Rates, Prostate Cancer, Delaware, 2000-2004, by Census County Division (not shown) A Need for Future Research As discussed under the ?Limitations? section, this analysis cannot determine why some areas of the state have higher cancer rates than others. Possible reasons could include coincidental spatial clustering, environmental exposure, migration of people into or out of the area, differences in lifestyle risk factors, and improved access to health care. Additional study of the areas where there are higher rates will be necessary to gain further insight into the possible reasons. Questions that could be asked are: * Are unusual cancers occurring that have known environmental causes? * Are the people who are getting cancer younger than we normally would expect? * Do these people have similar occupations? * Are people in the area more or less likely to be screened for cancer than in areas with lower rates? * Do areas with elevated rates contain disproportionate numbers of those at higher risk for cancer, including individuals in low income groups or in racial and ethnic groups known to be at high risk? * Do a disproportionate number of residents in areas with elevated rates engage in lifestyle risk factors that place them at higher risk for developing cancer (e.g., tobacco use, sedentary lifestyle)? To explore these questions and others, DPH will be arranging meetings with the communities in each of the areas with statistically higher rates, working with the respective legislators, seeking input from the Delaware Cancer Consortium, and, if appropriate, conducting formal epidemiologic studies to answer specific questions. Appendix A: Review of Methodology Preliminary Data Analyses Preliminary analyses were performed on five raw data files created for the Division of Public Health (DPH) by the Delaware Cancer Registry One file included records of all cancer cases diagnosed in Delaware between 2000 and 2004 (all cancer sites combined). The remaining four data files included all cancer cases, diagnosed in Delaware during the same time period, for the following four primary sites: (a) female breast; (b) colorectal; (c) lung / bronchus; and (d) prostate. Raw data files were sorted by tumor behavior code . Behavior codes are assigned using the following coding scheme : 0 = benign (tumor is growing in an isolated location and will not spread to surrounding tissue) 1 = uncertain (tumor displays borderline malignancy or uncertain malignancy potential) 2 = in situ (tumor has the potential to spread to surrounding tissue, but is currently contained in an isolated location) 3 = malignant (tumor has invaded surrounding tissue) Per reporting guidelines mandated by the Surveillance, Epidemiology, and End Results (SEER) Program of the National Cancer Institute, cancer incidence rates exclude benign tumors, non-urinary bladder in situ tumors, and basal and squamous cell cancers. However, state cancer registries may still collect data on these tumors for tracking purposes. Therefore, raw data files were analyzed to determine the proportion of cases involving benign, non-urinary bladder in situ tumors, and basal and squamous cell cancers; all cases involving benign tumors, non-bladder in situ tumors, and basal and squamous cell cancer were eliminated from files; all malignant tumors, as well as any tumors with an unknown behavior code, were retained for further analyses. Table A1 displays the distribution of tumor behaviors across the five raw cancer files. Table A1: Distribution of Tumor Behavior by Primary Site All Site Breast Colorectal Lung Prostate Total Cases 22,791 3,681 2,439 3,328 3,340 Benign 58 0 1 0 0 Unknown 24 0 0 0 0 In Situ 2,035 795 127 1 4 Urinary Bladder4 493 0 0 0 0 Non-Urinary Bladder 1,542 795 127 1 4 Malignant 20,774 2,886 2,311 3,327 3,336 Total Cases Retained 21,191 2,886 2,311 3,327 3,336 Source: Delaware Division of Public Health, Cancer Registry 4 In Situ bladder tumors indicated by primary site codes C670-C679 After retaining all appropriate cancer cases, raw data files were truncated to include the following 12 variables: (1) patient ID number; (2) gender; (3) age at diagnosis; (4) race; (5) Spanish / Hispanic origin; (6) census tract at diagnosis; (7) county of residence at diagnosis; (8) zipcode at diagnosis; (9) year of diagnosis; (10) primary tumor site; (11) tumor behavior code; (12) tumor histology code. Raw data files were converted into comma separated data files and imported into SAS for further analysis. Secondary Data Analyses Secondary analyses focused on creation of the Census County Division (CCD) variable. CCDs represent county subdivisions and are established, cooperatively, by the United State Census Bureau and state and local authorities. CCDs have been established in 21 states and represent a useful method of presenting statistics in areas where minor civil divisions (i.e., primary governmental or administrative sectors within counties) do not exist. In general, CCDs are defined in such a way that boundaries follow visible features and census tracts (where available). In Delaware, 27 CCDs are defined using census tracts. While the Delaware Cancer Registry collects census tract data, it is not mandated to record the CCD in which each new diagnosed cancer case exists; therefore, prior to any analyses involving CCDs, the CCD variable must be manually created using existing census tract data. As of the 2000 Census, Delaware is subdivided into 197 census tracts (please note that census tracts do not follow a consecutive numbering scheme). New Castle County includes tracts 1 through 169.02. Kent County is comprised of census tracts 401 through 431, and Sussex County includes census tracts 501.01 through 519. The Delaware Cancer Registry codes census tracts using a 5-digit coding scheme; the 5-digit coding system accounts for census tracts with values carried out to the tenths and hundredths positions following a decimal point (e.g., 169.02). Census tracts without digits following a decimal point are coded with trailing zeroes that serve as "place holder" values. For example, within the raw data, census tract 519 is coded as 51900. For census tracts lower than 100 (i.e., tracts 1 through 27), leading zeroes serve as "place holder" values in the 5-digit numbering scheme. For example, census tracts 1 and 6.01 are coded as 00100 and 00601, respectively, within the registry data. Importation of the raw data files from Excel into SAS results in the deletion of leading zeroes from the census tract variable. Therefore, the first step in creating CCDs was to manually recode all census tract values back to their original 5-digit values. The CCD variable was created using operational CCD definitions provided by the Delaware Health Statistics Center. Table A2 shows the operational definitions (based on census tracts) for all 27 CCDs. For cancer cases which missing or invalid census tract data, CCD was labeled "Unknown". Table A2: Census County Divisions: Operational Definitions by Census Tract CCD Name Census Tracts Included 1 Central Kent 42000, 42100, 42201, 42202 2 Dover 40400 (if zip = 19901 or 19961), 40500, 40600, 40700, 40800, 40900, 41000, 41100, 41200, 41300, 41400, 41500, 41600, 41701, 41702, 41801, 41802, 41900 3 Felton 42800 4 Harrington 42900, 43000, 43100 5 Kenton 40100 6 Milford North 42400, 42500, 42600, 42700 7 Smyrna 40400 (if zip = 19977), 40201, 40202, 40203 8 Brandywine 10101, 10102, 10200, 10300, 10400, 10500, 10700, 10800, 10900, 11000, 11100, 11201, 11202, 11203, 11204, 11205, 11206, 11300, 11400, 11500, 11600, 11700 9 Central Pencader 14805, 14806, 14807, 14808 10 Greater Newark 13610, 13611, 13700, 14803, 14100, 14200, 14300, 14402, 11403, 14404, 14501, 14502, 14702, 14703, 14705, 14706 11 Lower Christiana 12000, 12100, 12200, 12300, 12400, 12500, 12600, 12700, 12900 12 Middletown-Odessa 16601, 16602, 16603, 16604, 16801, 16802, 16901, 16902 13 New Castle 14902, 14903, 14904, 14905, 15000, 15100, 15200, 15400, 15500, 15600, 15800, 15900, 16000, 16100, 16200, 16301, 16302, 16303 14 Peidmont 11800, 11900, 13501, 13503, 13504 15 Pike Creek-Central Kirkwood 13000, 13100, 13200, 13300, 13400, 13604, 13607, 13608, 13609, 13612, 13613 16 Red Lion 16401, 16402 17 Upper Christiana 13800, 13901, 13902, 14000 18 Wilmington 00100, 00200, 00300, 00400, 00500, 00601, 00602, 00700, 00800, 00900, 01000, 01100, 01200, 01300, 01400, 01500, 01600, 01700, 01800, 01900, 02000, 02100, 02200, 02300, 02400, 02500, 02600, 02700 19 Bridgeville-Greenwood 50301, 50302 20 Georgetown 50501, 50502 21 Laurel-Delmar 51701, 51702, 51801, 51802, 51900 22 Lewes 50900, 51001, 51002, 51003, 51100 23 Milford South 50101, 50102, 50103, 50200 24 Millsboro 50601, 50602, 50701, 50702 25 Milton 50801, 50802, 50803 26 Seaford 50401, 50402, 50403, 50404 27 Selbyville-Frankford 51200, 51301, 51302, 51303, 51304, 51400, 51500 Source: Delaware Division of Public Health, Delaware Health Statistics Center To evaluate the accuracy of geographic variables, follow-up analyses involved cross-tabulations using three geographic variables. The Delaware Cancer Registry collects several variables related to location of residence at the time of cancer diagnosis. Hospital registries report patients? street address, zip code, and county of residence, all at the time of diagnosis, to the Central Registry. Geocoding software is used at the Central Registry level to assign census tract codes to each new record; census tract assignment is based on street address and zip code. Data analysts are increasingly confident of the accuracy of the data when record-level zip code, census tract, and county code data indicate the same county of residence at the time of diagnosis. As mentioned previously, Delaware is comprised of 197 census tracts. Based on census tracts data, a new variable "ccode" was created to represent the county in which the cancer case would have been diagnosed assuming the census tract of residence at time of diagnosis was coded accurately. That is, the variable "ccode" answered the question, "If a person was diagnosed in this census tract, in what county would they have been diagnosed?" Table A3 shows the breakdown of counties by census tract. Table A3: Delaware Counties by Census Tract County Census Tracts Included New Castle County 00100, 00200, 00300, 00400, 00500, 00601, 00602, 00700, 00800, 00900, 01000, 01100, 01200, 01300, 01400, 01500, 01600, 01700, 01800, 01900, 02000, 02100, 02200, 02300, 02400, 02500, 02600, 02700, 10101, 10102, 10200, 10300, 10400, 10500, 10700, 10800, 10900, 11000, 11100, 11201, 11202, 11203, 11204, 11205, 11206, 11300, 11400, 11500, 11600, 11700, 11800, 11900, 12000, 12100, 12200, 12300, 12400, 12500, 12600, 12700, 12900, 13000, 13100, 13200, 13300, 13400, 13501, 13503, 13504, 13604, 13607, 13608, 13609, 13610, 13611, 13612, 13613, 13700, 13800, 13901, 13902, 14000, 14100, 14200, 14300, 14402, 14403, 14404, 14501, 14502, 14702, 14703, 14705, 14706, 14803, 14805, 14806, 14807, 14808, 14902, 14903, 14904, 14905, 15000, 15100, 15200, 15400, 15500, 15600, 15800, 15900, 16000, 16100, 16200, 16301, 16302, 16303, 16401, 16402, 16601, 16602, 16603, 16604, 16801, 16802, 16901, 16902 Kent County 40100, 40201, 40202, 40203, 40400, 40500, 40600, 40700, 40800, 40900, 41000, 41100, 41200, 41300, 41400, 41500, 41600, 41701, 41702, 41801, 41802, 41900, 42000, 42100, 42201, 42202, 42400, 42500, 42600, 42700, 42800, 42900, 43000, 43100 Sussex County 50101, 50102, 50103, 50200, 50301, 50302, 50401, 50402, 50403, 50404, 50501, 50502, 50601, 50602, 50701, 50702, 50801, 50802, 50803, 50900, 51001, 51002, 51003, 51100, 51200, 51301, 51302, 51303, 51304, 51400, 51500, 51701, 51702, 51801, 51802, 51900 Source: Delaware Division of Public Health Delaware is comprised of 93 zip codes. As with census tract data, new variable "zcode" was created to represent the county in which a cancer case would have been diagnosed assuming the zip code of residence at time of diagnosis was coded accurately. Thus, the variable "zcode" answered the question, "If a person was diagnosed in this zip code, in what county would they have been diagnosed?" Table A4 shows the breakdown of counties by zip code. Table A4: Delaware Counties by Zip Code County Zip Codes Included New Castle County 19701, 19702, 19703, 19706, 19707, 19709, 19710, 19711, 19713, 19720, 19730, 19731, 19732, 19733, 19734, 19736, 19801, 19802, 19803, 19804, 19805, 19806, 19807, 19808, 19809, 19810 and 19938 if census tract = 16901 19977 if census tract = 16901 or 16902 Kent County 19901, 19902, 19903, 19904, 19905, 19934, 19936, 19943, 19946, 19952, 19953, 19954, 19955, 19961, 19962, 19964, 19979, 19980, and 19938 if census tract = 40100 or 40201 19963 if census tract = 42600 19977 if census tract = 40100, 40201, 40202, 40203, or 40404 Sussex County 19930, 19931, 19933, 19939, 19940, 19941, 19944, 19945, 19947, 19950, 19951, 19956, 19958, 19960, 19966, 19967, 19968, 19969, 19970, 19971, 19973, 19975, and 19963 if census tract = 50101, 50102, or 50103 Source: Delaware Division of Public Health Cross-tabulations ("ccode" x "zcode" x county) were performed within each dataset. Consistent results (i.e., "ccode" = "zcode" = county code) were assumed to be indicative of accurately coded geographic data. For example, consider a hypothetical record for Individual A, diagnosed with breast cancer. If the data indicated that Individual A resided in census tract 02700 (New Castle County), within zip code 19801 (New Castle County), and the county of residence for Individual A was coded as New Castle County, geographic data were assumed accurate. Conversely, geographic data were considered inconsistent if one of the following three scenarios existed: (a) "ccode" "zcode"; (b) "ccode" county; (c) "zcode" county). For example, if the data indicated that Individual B, a hypothetical prostate cancer patient, resided in census tract 50102 (Sussex County), within zip code 19901 (Kent County), and the county of residence for Individual B was coded as Sussex County, geographic data were inconsistent. Records with inconsistent geographic data were subsetted for further analyses; analyses revealed that geographic inconsistencies belonged to one of 12 categories. These categories are outlined in Table A5. Table A5: 12 Categories of Inconsistent Geographic Data 1 Zip code and census tract point to the same county of diagnosis; county code is non-matching 2 Zip code and county code point to the same county of diagnosis; census tract is non-matching 3 Census tract and county code point to the same county of diagnosis; zip code is non-matching 4 Zip code and census tract point to the same county of diagnosis; county code is missing 5 Zip code and county code point to the same county of diagnosis; census tract is missing 6 Census tract and county code point to the same county of diagnosis; zip code is missing 7 Zip code and census tract are non-matching; county code is missing 8 Census tract and county code are non-matching; zip code is missing 9 Zip code and county code are non-matching; census tract is missing 10 Zip code data available; census tract and county code data are both missing 11 Census tract data available; zip code and county code data are both missing 12 County code data available; zip code and census tract data are both missing Source: Delaware Division of Public Health Three categories of inconsistency among geographic variables warranted record elimination. First, records were eliminated if census tract data were missing. Second, records were eliminated if census tract pointed to a different county of diagnosis than did zip code and county code. For example, if the record for Individual C indicated that that the person was diagnosed in census tract 02700 (New Castle County), but zip code and county code data both indicated that the county of residence at time of diagnosis was Sussex County, the record was eliminated. Third, records were eliminated if census tract, zip code, and county code were completely non-matching (i.e., census tract indicated New Castle County as the county of residence at time of diagnosis, while zip code and county code indicated Kent and Sussex Counties, respectively, as the county of residence at the time of diagnosis). Provided they did not meet any of the above three scenarios, records were retained for incidence rate analyses. Table A6 shows the total number of records, the number of records with inconsistent geographic data, and the number of records ultimately eliminated from further analyses among the five cancer data sets. Table A6: Breakdown of Geographically Inconsistent Records, by Data Set No. Cases with No. Cases No. Cases Retained Total No. Cases Inconsistent Ultimately for Incidence Rate (from DCR) Geographic Data Eliminated Analysis All Site, 2000-04 21,191 860 570 20,721 Breast, 2000-04 2,886 98 59 2,827 Colorectal, 2000-04 2,312 88 49 2,263 Lung, 2000-04 3,327 126 86 3,241 Prostate, 2000-04 3,336 123 83 3,253 Source: Delaware Division of Public Health Calculating Age-Adjusted Incidence Rates CCD population denominators were calculated by the Delaware Health Statistics Center using estimates from the Delaware Population Consortium and the 2000 Census. CCD population denominators were stratified by gender (male, female) and age group (0-4 yrs, 5-14 yrs, 15-24 yrs, 25-34 yrs, 35-44 yrs, 45-54 yrs, 55-64 yrs, 65-74 yrs, 75-84 yrs, and 85+ years). Population denominators were available for the 6-year time period encompassing 2000-2005. Denominators for years 2000 through 2004 were summed to obtain the 2000-2004 population for each CCD. Five-year (2000-2004) population estimates ranged in size from 27,924 for Kenton to 414,251 for New Castle. Please refer to Appendix B for total and gender-specific populations by CCD. Within these tables, CCDs are arranged in order of descending size by 2000-2004 population; below each 5-year population estimate is a one-year population estimate (2005 only) for that CCD. Within each of the five cancer data files, cross-tabulations (age group x CCD) were performed to determine the number of cancer cases by CCD and the age groups in which they were diagnosed. These frequencies were used to calculate crude and age-adjusted incidence rates. Crude incidence rates represent the total number of new cancer diagnoses over the total population at risk, without consideration of any demographic characteristics of the population. Age-adjusted incidence rates take into account the age distribution of the population at risk; age-adjusted incidence rates are useful for comparing rates between two populations that differ in age composition. To calculate crude incidence rates, the number of cancer cases diagnosed in a particular age group in a particular CCD was divided by the population size for that specific cohort; these values were then multiplied by 100,000 (see Equation A1). To determine the 2000-2004 crude incidence rate for an entire CCD, the number of cancer cases diagnosed in a CCD over the 5-year period was divided by the total population of the CCD for the same 5-year period, and this value was multiplied by 100,000. Equation A1: 2000-2004 Crude All Site Incidence Rate, 25-34 year olds, Milford South (not shown) To calculate age-adjusted incidence rates, crude incidence rates for each age group were multiplied by the appropriate 2000 U.S. Standard Million Population weight . Table A7 displays the U.S. Standard Million population weights, by age group. Age-adjusted incidence rates for each of the 10 age groups were summed to yield the age-adjusted incidence rate for an entire population. Table A7: U.S. Standard Million Population Weights, by Age Group Age Group U.S. Standard Million Population Weight 0-4 yrs 0.069135 5-14 yrs 0.145565 15-24 yrs 0.138646 25-34 yrs 0.135573 35-44 yrs 0.162613 45-54 yrs 0.134834 55-64 yrs 0.087247 65-74 yrs 0.066037 75-84 yrs 0.044842 85+ yrs 0.015508 Source: Centers for Disease Control and Prevention, National Center for Health Statistics Upper and lower (95%) confidence intervals were calculated around age-adjusted incidence rates for each CCD. The following formulas were used to generate upper and lower confidence limits: (not shown) Appendix B: Population Census Data (not shown) For questions or comments related to this report, please contact the Division of Public Health at the following address: Delaware Health and Social Services Division of Public Health 540 S. DuPont Highway Dover, DE 19901 Phone: 302-744-1040 Fax: 302-739-2545 http://www.dhss.delaware.gov/dhss/dph/index.html April, 2008