- Email: [email protected]
Cancer registry policies in the United States and geographic information systems applications in comprehensive cancer control
Available online at www.sciencedirect.com
Health Policy 87 (2008) 185–193
Cancer registry policies in the United States and geographic information systems applications in comprehensive cancer control Christie B. Ghetian a,∗ , Roxanne Parrott a , Julie E. Volkman a , Eugene J. Lengerich b a
Department of Communication Arts & Sciences, The Pennsylvania State University, University Park, PA, United States b Department of Public Health Sciences, The Pennsylvania State University, College of Medicine, and Penn State Cancer Institute, Hershey, PA, United States
Abstract Objectives: Through a long history of cancer policies, public health has a foundation for cancer research and data to use in applying technological advancements for U.S. cancer control efforts. Geographic information systems (GIS) are one technology enabling the visualization of cancer risk patterns associated with incidence, mortality, and accessibility to care. Methods: U.S. Comprehensive Cancer Control (CCC) program directors were interviewed from 49 of 50 states to assess use and function of GIS tools for mapping data related to cancer control policies and practices. Interviews were coded to obtain frequencies of response associated with content domains mapped using GIS tools and the perceived relative advantages. Results: Significant relationships were found between the mapping of behavioral risk factors, health care services, transportation access, and policy advantages identified by program managers. The mapping of cancer incidence, mortality, and staging, transportation access, and multiple layers of content were found to have significant associations with perceived research advantages. Conclusions: U.S. CCC program managers recognize several important advantages relating to health policy and research for use of GIS tools in cancer control efforts. The application of GIS in U.S. cancer control efforts is employed unevenly, suggesting the need for innovative policies to support accessibility. © 2007 Elsevier Ireland Ltd. All rights reserved. Keywords: GIS advantages; Cancer control; Policy development; Cancer mapping
The collection and use of cancer registry data for cancer control in the United States have long been ∗ Corresponding author at: 234 Sparks Building, University Park, PA, 16802, United States. Tel.: +1 814 865 4484; fax: +1 814 863 7986. E-mail address: [email protected] (C.B. Ghetian).
a national priority, as evidenced by numerous laws, rules and policies. Technologies, such as geographic information systems (GIS), provide an opportunity to use information provided through cancer registries to help guide and target cancer control efforts. The first statewide registry was begun in Connecticut in 1935 to facilitate follow-up treatments [1–3]. Congress’ pas-
0168-8510/$ – see front matter © 2007 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.healthpol.2007.12.007
186
C.B. Ghetian et al. / Health Policy 87 (2008) 185–193
sage of the National Cancer Institute Act in 1937 marked the beginning of the U.S. National Cancer Program [4] and an emphasis on population-based studies to relate human characteristics to the origins of cancer [5]. Throughout the 1940s, population-based cancer registries emerged [2,4], with the American College of Surgeons (ACoS) encouraging all hospitals to use cancer registries for the periodic review and evaluation of patients. U.S. cancer programs were asked to include a cancer registry component by 1956 [2], and the passage of the National Cancer Act in 1971 mandated that hospitals report all newly diagnosed cancers to state departments of public health [6], which contributed to data-based cancer research and innovations in diagnosis and treatment [7]. Two years later, the National Cancer Institute (NCI) instituted the Surveillance Evaluation and End Results (SEER) Program, which was a national effort to collect and evaluate cancer registry data from specific geographic regions and centrally locate it [8]. In 1989, the ACoS and the American Cancer Society (ACS) formed the Commission on Cancer (COC) and the National Cancer Data Base (NCDB) to formalize the cancer registration process. The NCDB includes more than 1500 U.S. hospital- and ambulatory care-based cancer registry programs which report cancer data, activities, and analyses on national cancer trends and disease patterns [2]. In the 1990s, research suggested that state health departments were best suited to address their state populations’ cancer needs [9], and the Cancer Registries Amendment Act of 1992 championed a state-based approach to cancer data management. It authorized the formation of the National Program of Cancer Registries (NPCR) within the Centers for Disease Control and Prevention [8], which in turn advanced state legislation and regulations to require hospitals and practitioners to make cancer diagnosis and treatment data available to their respective state registries in a standardized and timely manner [8]. By 1998, under the guidance of the NPCR, 13 new cancer registries were created and 36 improved their existing registries, accounting for 97% of the U.S. population [8]; in 2007, 45 states, the District of Columbia, Puerto Rico, the Republic of Palau, and the Virgin Islands maintain cancer registries under the NPCR [2]. The NPCR and SEER registries combined accounting for approximately 100% of the U.S. population [10]. With the adoption supporting policies and identification of funding, cancer registry data
have become the standard for surveillance of cancer incidence. Advances in computing technology now enable cancer registry data to be linked to other data, thereby providing additional insight into the distribution, determinants and treatment of cancer [2]. For example, cancer incidence data, in combination with vital statistics data from the National Center for Health Statistics, are used to estimate cancer survival rates [8]. Researchers have linked registry data with socioeconomic, environmental, and behavioral data to examine hypotheses on the etiology of cancer or the outcomes of cancer treatment [11,12]. Also, researchers have examined cancer incidence across time and geographic regions in relation to hypotheses about the changing patterns of cancer incidence. Today, state-of-the-art technology makes detailed insets, summary tables, figures, descriptive, and explanatory text accompanying computerized cancer maps not only possible, but accessible. For example, websites, such as http://statecancerprofiles.cancer.gov/ and http://cancercontrolplanet.cancer.gov/ offer interactive menus that draw upon cancer registry data to facilitate examination of states’ cancer profiles, intervention programs, and cancer data (e.g., incidence rates, death rates, cancer sites). The recent development and widespread distribution of geographic information systems technology facilitates the analysis of data in particularly innovative ways [13]. Specifically, GIS uses four general approaches for cancer research: (1) incidence mapping; (2) geographic correlation studies; (3) risk assessment in relation to a pre-specified point or line source; and (4) cluster detection and disease clustering [14]. For example, studies have used GIS technology to examine clusters of childhood leukemia and their possible association with environmental contamination [15]. The objective of the current study was to estimate the degree to which state program directors perceive policy or research advantages from the use of GIS for cancer control. In particular, we sought to determine what types of data sets are being used in state CCC, and what defining state characteristics are associated with these activities (as defined by their cancer control activities. The data from the current study provide a general indication of the research hypotheses being explored or components of the programs being developed. The results of this study may suggest areas for further devel-
C.B. Ghetian et al. / Health Policy 87 (2008) 185–193
opment of state cancer control policies and analytic capacity. 1. Method 1.1. Participants and procedures Individuals recruited to participate were the managers of CCC programs for the 50 states in the U.S. and were first contacted through use of email via use of a 2005 National CCC Program list; if an individual responded that she or he was not the CCC program manager for a given state, he/she was asked to identify the appropriate person who was contacted in a like fashion. Interviews were coordinated in batches of 10 to facilitate scheduling. At the agreed upon time, a female interviewer trained in the research protocol and phone interview script contacted the interviewee to talk about the use of CCC mapping in their state. Verbal consent to conduct and record the interview was obtained and participants were assured that no personally identifiable information would be included. Participants had to be 18 years or older to participate. At the beginning of the interview, the interviewer stated the purpose of the call was for the research entitled, “GIS Adoption Telephone Survey of Comprehensive Cancer Control and Prevention (CCCP) Directors.” The interviewer specifically stated that the interview was to assess the interviewee’s perspective about the use of mapping in their state’s CCC efforts. During the interview, respondents were asked about the use of mapping for CCC activities in their state and perceived advantages of using GIS tools and mapping to CCC activities. The specific content participants were asked about in relation to mapping activities included: behavioral risk factors; cancer incidence, mortality, and staging; demographics; environmental exposures; health care service availability; and transportation access. Interviews were transcribed verbatim with the advantages relating to cancer policy and research, summarized in Table 1. The definitions were derived post hoc based on interviewees’ responses and used in subsequent coding activities. 1.2. Coding training and coding Four coders were trained to understand the meaning of the seven policy and research advantages identi-
187
fied by participants’ responses. Then working in pairs, the coders identified the presence or absence of each of the policy and research advantages in each interview based on these definitions. After assessing five interviews, Cohen’s kappa, a measure of intercoder agreement for dichotomous data that takes chance agreement into account [16] was used to assess reliability. Cohen’s kappa ranged from .69 to 1.00 at the time of initial coding. Coders’ discrepancies were resolved through discussion until consensus was achieved, and the definitions were further refined. After another 10 interviews were coded, another assessment of intercoder agreement revealed that Cohen’s kappa ranged between .89 and 1.00, discrepancies were resolved through consensus, and the remaining interviews were coded. The final dataset combined coder responses to create a single response for each coded variable. The Statistical Package for Social Sciences 14.0 was used for analyses.
2. Results CCC directors in 49 of 50 states were interviewed; Louisiana’s director was not contacted due to Hurricane Katrina. Data were collected from July 14, 2005 to January 27, 2006 with all interviews conducted in English. Interviews lasted approximately 30 min. The interviewees included 5 males, 37 females, and 7 groups of 2 or more participants (58 individuals participated in 49 interviews). Approximately half had been in their current position for 2 years or less. The range of experience varied from one month to 12 years (M = 33 months; S.D. = 30.27 months). Half worked in public health for their states for more than 4 years, with an average of nearly 12 years. The level and areas of education of participants included: 6 Ph.D.s, 1 M.D., 2 M.D.s who also held M.S. degrees, 1 M.D. who also held a M.P.H. degree, 12 with M.P.H.s, 2 with M.S. degrees, 4 with M.P.A.s, 1 with a M.S.W., 1 with both a M.P.A. and M.S.W., 8 with Masters degrees in unspecified areas, 1 with a M.B.A., 4 with B.S. degrees, 3 with B.A.s, 1 with both a B.S. and R.N. credentials, and 1 with an Associates degree. One other interviewee was a radiation therapist. Descriptive statistics, including frequencies and percentages, were derived relating to program managers’ reports of GIS use in their states. Descriptive statistics
188
C.B. Ghetian et al. / Health Policy 87 (2008) 185–193
Table 1 Codes and definitions for content domains and relative advantages Codes Content domains Cancer incidence, mortality, and/or staging Health care service availability Demographics Behavioral Environmental exposures Transportation access Multi-layer Cancer policy advantages Identify at-risk populations Identify access issues
Identify service gaps Identify cancer staging Cancer research advantages Monitors/surveillance tool Multivariate modeling tool Generate etiological hypotheses
Definition Using mapping for any type of incidence or prevalence, mortality, early or late-stage screening, survivorship or cancer diagnosis Use of maps to identify health service availability (e.g. screening facilities) and health insurance status Use of mapping for information on race/ethnicity, gender and age Use of mapping to assess lifestyle choices including alcohol use, physical inactivity, diet tobacco use Use of mapping to identify radon, watershed locations, toxic release inventory data and pesticide exposure Use of mapping for transportation routes Use of mapping to overlay content areas to determine risk relationships Includes mapping to identify high risk populations in cancer control and prevention planning Assesses the use of maps to identify health service availability (e.g. screening facilities), transportation availability, travel distance and health insurance status Examines discussion about mapping to identify health service gaps, needs and disparities Refers to mapping that compares areas of late-versus early-stage cancer screening or diagnoses Examines mapping used to monitor the status of interventions and use of resources Represents a specific statement of how mapping is used to overlay data or multivariate modeling Assesses any discussion about mapping providing a relative advantage or generation of hypotheses for research purposes
were also computed to identify the number of managers who identified GIS mapping as affording cancer policy or research advantages for CCC. Chi-square analyses were performed to identify associations between states’ CCC-related map use and the naming of policy and research advantages for mapping specific content domains. However, because chi-square tests rely on large samples for approximation and, in some cases, there were zeros in the cells, Fisher’s exact test was also employed. This measure of statistical significance for categorical data is appropriate for small sample sizes, more specifically for two-by-two contingency tables. Although Fisher’s exact test does not operate through formal test statistics or critical values, its calculated p-value is useful for indicating a possible relationship between variables. Significance suggests that the probability of obtaining the observed value due to chance is unlikely [17].
2.1. Cancer policy and research advantages identified for GIS use Within the cancer policy domain, the frequency of response for the four issues identified by participants as advantages for using GIS mapping were: (a) identify at-risk populations (n = 7), (b) identify access issues (n = 20), (c) identify service gaps (n = 33), and (d) identify cancer staging (n = 13). Fig. 1 illustrates the relationship between content currently mapped and likelihood of mentioning any of the cancer policy advantages. The frequency of response for the three cancer research issues identified in participants’ responses regarding the advantages of using GIS for CCC activities were: (a) monitoring and surveillance tool (n = 16), (b) multivariate modeling tool (n = 24), and (c) generate etiologic hypotheses tool (n = 13). Fig. 2
C.B. Ghetian et al. / Health Policy 87 (2008) 185–193
Fig. 1. Relative advantages related to cancer policy by mapped content domain. Bars represent the within-group percentages of states that reported the relative advantage (i.e., identify at-risk populations, identify access issues, identify service gaps, and identify cancer staging) of utilizing GIS in CCC activities and mapped a specific content domain (i.e., access, behavioral risk factors, cancer incidence/mortality/staging, etc.).
illustrates the findings related to the cancer research advantages. 2.2. Relationships between content mapped and perceived advantages The following section outlines each of the content domains mapped through state CCC efforts. Within each content domain, significant relationships between it and both the cancer policy and the cancer research advantages are discussed (see Table 2). 2.2.1. Behavioral risk factors Seventeen (35%) program managers stated that behavioral risk factors were a content domain being mapped via application of GIS in their CCC activities. Mapping behavioral risk factors was found to have a significant relationship to identifying cancer policy issues as advantages. Six of the seven (86%) participants who stated the ability to identify at-risk populations as an advantage mapped behavioral risk
189
Fig. 2. Relative advantages related to cancer policy by mapped content domain. Bars represent the within-group percentages of states that reported the relative advantage (i.e., monitoring and surveillance tool, multivariate modeling tool, and generate etiological hypothesis tool) of utilizing GIS in CCC activities and mapped a specific content domain (i.e., access, behavioral risk factors, cancer incidence/mortality/staging, etc.).
factors in their states (2 (1, N = 49) = 9.38, p < .01; one-tailed Fisher’s exact test, p < .01). No other identified policy advantages revealed significant associations to the mapping of behavioral risk factors. In addition, no significant relationships were revealed among the seventeen state program managers who reported the current mapping of behavioral risk factors and the identification of GIS as advantageous for cancer research efforts. 2.2.2. Cancer incidence, mortality, and staging Sixty-nine percent of participants (n = 34) reported that cancer incidence, mortality, and staging constituted a content domain being mapped via GIS technologies. No significant relationships were found between the mapping of this content domain and any of the cancer policy advantages identified. Among the 34 program managers who indicated that they currently utilized GIS technologies to map cancer incidence, mortality, and staging, however, a relationship was found between mapping this content and a greater likelihood of reporting the use of
190
C.B. Ghetian et al. / Health Policy 87 (2008) 185–193
Table 2 Current content domains by relative advantages related to cancer policy (N = 49) Identify at-risk populations (n = 7)
Behavioral risk factors (n = 17) Cancer incidence/morality/staging (n = 34) Demographics (n = 25) Environmental exposures (n = 10) Health care services (n = 22) Transportation access (n = 2) Multi-layer (n = 13)
n (%)
χ2 (1,
6 (86) 6 (86) 3 (43) 2 (29) 3 (43) 0 (0) 3 (43)
9.38 1.03 .22 .34 .01 .35 1.12
N = 49)
Identify access issues (n = 20)
p
Fisher’s exact text, p
<.01 .31 .64 .56 .91 .56 .29
<.01 .30 .48 .44 .62 .73 .27
Identify service gaps (n = 33)
Behavioral risk factors (n = 17) Cancer incidence/morality/staging (n = 34) Demographics (n = 25) Environmental exposures (n = 10) Health care services (n = 22) Transportation access (n = 2) Multi-layer (n = 13)
n (%)
7 (35) 14 (70) 8 (40) 4 (20) 5 (25) 2 (10) 6 (30)
χ2 (1, N = 49)
p
Fisher’s exact text, p
.00 .01 1.64 .00 5.41 3.02 .21
.97 .94 .20 .95 .02 .08 .65
.60 .60 .16 .62 .02 .16 .45
Identify cancer staging (n = 13)
n (%)
χ2 (1, N = 49)
p
Fisher’s exact text, p
n (%)
χ2 (1, N = 49)
p
Fisher’s exact text, p
12 (36) 23 (70) 16 (48) 6 (18) 13 (39) 2 (6) 10 (30)
.49 .01 .26 .31 1.24 1.01 .74
.12 .95 .61 .58 .27 .32 .39
.49 .60 .42 .42 .21 .45 .31
4 (31) 7 (54) 7 (54) 4 (31) 5 (38) 0 (0) 3 (23)
.12 2.01 .06 1.17 .30 .75 .11
.73 .16 .81 .28 .59 .39 .74
.50 .14 .53 .24 .42 .54 .53
Note: Percentages are within-group for the relative advantage category (i.e., relative advantage n is used as the denominator).
GIS for generating etiological hypotheses, a cancer research advantage. Twelve of 13 (92%) managers that reported the generation of etiological hypotheses as an advantage were currently mapping cancer incidence, mortality, and staging in their respective state (χ2 (1, N = 49) = 4.38, p = .04; one-tailed Fisher’s exact test, p = .04). No other perceived research advantages were significantly associated with this domain. 2.2.3. Demographics Twenty-five (51%) participants declared GIS mapping of demographics as a current content domain for CCC activities. The mapping of demographic information had no significant effects on the likelihood of naming any cancer policy advantages of GIS applications. Likewise, GIS mapping of demographic data did not significantly affect the likelihood of directors naming any of the three cancer research advantages. 2.2.4. Environmental exposures Twenty percent (n = 10) of program managers revealed current utilization of GIS technologies for
the mapping of environmental exposures as a content area. Analyses indicated no significant relationships between the mapping of environmental exposures and identification of either cancer policy or research advantages. 2.2.5. Health care services Of the 49 participants, 45% (n = 22) reported current use of GIS applications to map health care services as a content domain. Analyses revealed that mapping health care services significantly affected the likelihood that state program managers would specify the identification of access issues as a policy advantage of use of GIS tools for mapping. Five of the 20 (25%) program managers that stated the identification of access issues as an advantage, currently mapped health care services (χ2 (1, N = 49) = 5.41, p = .02; one-tailed Fisher’s exact test, p = .02). No other cancer policy advantages were significantly associated with health care service mapping. No significant associations related to any research advantages were found in relation to the 22 state pro-
C.B. Ghetian et al. / Health Policy 87 (2008) 185–193
gram managers that indicated current use of GIS to map health care services as a content domain. 2.2.6. Transportation access Two state program managers (4%) indicated that their state CCC efforts included the mapping of transportation-related access issues as a content area in their GIS use. Although analyses revealed that the mapping of transportation routes did not significantly affect the likelihood that program managers would report any of the policy advantages related to GIS, the policy advantage, “identifies access issues,” did approach significance (χ2 (1, N = 49) = 3.02, p = .08; one-tailed Fisher’s exact test, p = .16). In regard to the research advantages afforded by GIS, the mapping of transportation-related access as a content area was found to significantly affect the likelihood that program managers would name the generation of etiologic hypotheses as an advantage of GIS mapping. Two of 13 (15%) of program managers who named the generation of etiologic hypotheses as an advantage also indicated their use of GIS to map access routes (χ2 (1, N = 49) = 5.77, p = .02; one-tailed Fisher’s exact test, p = .07). Noted in Table 3, no other research advantages identified by program managers revealed significant relationships with mapping transportation access. 2.2.7. Multi-layer content Thirteen (27%) program directors reported the use of GIS to map multiple content areas. In other words, these states mapped two or more content domains (e.g., access, behavioral risk factors, demographics) in a single map. The mapping of multiple content areas had no significant effect on the prospect of program directors naming any of the policy advantages of GIS mapping. Among the 13 state program directors that reported the use of GIS to map multiple content areas in their CCC activities, however, was a greater likelihood for naming cancer monitoring and surveillance as a relative cancer research advantage of GIS utilization. One of 16 (6%) directors that identified cancer monitoring and surveillance as a GIS advantage was currently mapping multiple layers of content (χ2 (1, N = 49) = 5.01, p = .03; one-tailed Fisher’s exact test, p = .02) (see Table 3). Neither of the other cancer research advantages was significantly related to the mapping of multiple content areas.
191
3. Discussion In the United States, we are fortunate that cancer registration polices enacted years ago now enable federal and state governments to accurately monitor temporal trends in cancer burden and to compare cancer burden in selected demographic groups. With the advent of GIS technology, governments have the potential to further use cancer registry data for cancer prevention and control in small geographic areas within their jurisdiction. This new technology affords great potential to develop policies that can apply limited cancer prevention and control resources to high-burden areas. Our research found, however, that the adoption of this technology for policy purposes was far from consistent among the states, and that there were wide variances in the perceived advantages related to use of GIS tools for mapping in performance of CCC activities. States were using data on behavioral risks and health care services in GIS technology for the policy domains of identifying at-risk populations and in identifying access issues, respectively. However, states were not consistently using demographic, environmental exposure, transportation access and even cancer incidence/mortality and staging data in the policy domain. These unused data sets may represent missed opportunities for states to target the reduction of cancer burden. States were consistently using only multi-layer data and cancer incidence/mortality/staging data in GIS technology in the research domain for monitoring and surveillance and generating etiologic research, respectively. Again, states may be missing opportunities afforded by data on behavioral risks, demographics, environmental exposures, health care services, and transportation access in cancer research. Since cancer registry data is already being collected by almost every state, the future use of GIS technology for cancer policy and research purposes will depend upon training of personnel and the diffusion of the GIS technology. Programs that train policymakers to utilize results from GIS analysis of multiple data sets should be developed and evaluated. Training materials and processes should highlight the positive attributes of GIS technology that will enhance its diffusion, including relative advantage and trialability of GIS technology. This study is limited by two issues. Though the director of Comprehensive Cancer Control in the state
192
C.B. Ghetian et al. / Health Policy 87 (2008) 185–193
Table 3 Current content domains by relative advantages related to cancer research (N = 49) Monitoring and surveillance (n = 16) n (%) Behavioral risk factors (n = 17) 1 (6) Cancer incidence/morality/staging (n = 34) 11 (69) Demographics (n = 25) 7 (44) Environmental exposures (n = 10) 3 (19) Health care services (n = 22) 7 (44) Transportation access (n = 2) 1 (6) Multi-layer (n = 13) 1 (6)
χ2 (1,
N = 49)
.08 .01 .50 .04 .01 .29 5.01
p
Fisher’s exact text, p
.77 .95 .48 .84 .91 .59 .03
.51 .60 .34 .58 .58 .55 .02
Multivariate modeling (n = 24) n (%) 10 (42) 18 (75) 13 (54) 3 (13) 9 (38) 1 (4) 8 (33)
χ2 (1, N = 49)
p
Fisher’s exact text, p
1.01 .70 .19 1.81 1.04 .00 1.12
.32 .40 .67 .18 .31 .98 .29
.24 .30 .44 .16 .23 .75 .23
Generate etiologic hypotheses (n = 13) n (%) Behavioral risk factors (n = 17) 6 (46) Cancer incidence/morality/staging (n = 34) 12 (92) Demographics (n = 25) 9 (69) Environmental exposures (n = 10) 3 (23) Health care services (n = 22) 6 (46) Transportation access (n = 2) 2 (15) Multi-layer (n = 13) 5 (38)
χ2 (1, N = 49)
p
Fisher’s exact text, p
1.03 4.38 2.35 .08 .01 5.77 1.29
.31 .04 .13 .78 .92 .02 .26
.25 .04 .11 .53 .58 .07 .22
Note: Percentages are within-group for the relative advantage category (i.e., relative advantage n is used as the denominator).
was encouraged to seek additional input, the responses largely reflect the knowledge and perspective of one person. Second, policy constructs for the reduction of cancer burden have not been developed and disseminated. However, the strengths of the study are substantial. First, this is the only study to have examined the utilization of data sets relevant to cancer policy and research at the state level. Second, the number and rate of respondents was high, with a 98% participation rate. Third, respondents had the opportunity to review transcripts of their interview to assure that their intended meaning was present, increasing the internal validity associated with the seven advantages extracted to represent the advantages perceived to relate to policy and research in fulfillment of CCC activities. The coding of data involved multiple investigators, increasing the external validity associated with the conclusions drawn from this project. 4. Conclusion As noted at the outset, policies, especially those to collect and use cancer registry information, have long been a national priority. The advent of GIS technology
enables states to utilize additional data sets for cancer policy and research. However, states must now adopt policies that encourage the development and utilization of cancer registry data in small geographic areas, thus illustrating the maxim that health research is integrally related to political environments. Although significant relationships were identified, such that participants’ naming policy advantages for GIS and cancer mapping were more likely to display behavioral risk factors to identify at-risk populations and health care services to identify access issues, it is difficult to separate the direction of the relationship. Mapping a content domain may influence the likelihood of perceiving particular policy advantages, or vice versa. Acknowledgements This research was supported by a Cooperative Agreement between the Centers for Disease Control and Prevention (CDC) and the Association of American Medical Colleges (AAMC), award number MM-0718. Its contents are the responsibility of the authors and do not necessarily reflect the official views of the CDC or AAMC.
C.B. Ghetian et al. / Health Policy 87 (2008) 185–193
References [1] Gagnon F, Turgeon J, Dallaire C. Health public policy: a conceptual cognitive framework. Health Policy 2007;81:42–55. [2] National Cancer Institute. Surveillance Epidemiology and End Results. Brief History of Cancer Registration, http://training.seer.cancer.gov/module cancer registration/unit1 history.html; 2007. [3] Heston JF, Kelly JB, Meigs JW, Flannery JT. Forty-five years of cancer incidence in Connecticut: 1935–79. National Cancer Institute (NCI) Monograph No. 70. NIH Publication No. 862652. NCI, Bethesda, MD; 1986. [4] Hubbard SM, Mayer DK, DeVita Jr VT. Evolution of the National Cancer Program: an American investment for conquering cancer. Seminars in Oncology Nursing 2002;18:252–64. [5] Schottenfeld D, Beebe-Dimmer JL. Advances in cancer epidemiology: understanding causal mechanisms and the evidence for implementing interventions. Annual Reviews of Public Health 2005;26:37–60. [6] Polednak AP. Trends in cancer incidence in Connecticut, 1935–1991. Cancer 1994;74:2863–72. [7] Flannery JT, Boice Jr JD, Devesa SS, Kleinerman RA, Curtis RE, Fraumeni Jr JF. Cancer Registration in Connecticut and the study of multiple primary cancers, 1935–82. Journal of the National Cancer Institute Monographs 1985;68:13–24. [8] National Cancer Institute. Cancer Surveillance Implementation Plan, March 1999. http://cancercontrol.cancer.gov/sig/ index.htm; 2007.
193
[9] National Cancer Legislation Advisory Committee. National Cancer Policy Legislative History, http://www.cancersource .com/nclac/leghistory.htm; 2007. [10] Centers for Disease Control and Prevention. National Program of Cancer Registries, http://www.cdc.gov/cancer/npcr/; 2007. [11] Boss LP, Suarez L. Uses of data to plan cancer prevention and control programs. Public Health Reports 1990;105:354– 60. [12] Pommerenke FA, Miller RW, Sirvastava S, Ackerman SP. Targeting cancer control: the state cancer control map and data program. American Journal of Public Health 1994;84:1479– 82. [13] Goodchild MF, Anselin L, Appelbaum R, Harthorn B. Toward spatially integrated social science. International Regional Science Review 2000;23:139–59. [14] Wakefield J, Elliott P. Issues in the statistical analysis of small area health data. Statistics in Medicine 1999;18:2377– 99. [15] Openshaw S, Charlto M, Wymer C, Craft A. Developing a mark 1 geographical analysis machine for the automated analysis of point data sets. International Journal of Geographical Information Systems 1987;1:335–58. [16] Cohen J. Coefficient of agreement for nominal scales. Educational and Psychological Measurement 1960;20:37– 46. [17] Fisher RA. Statistical method for research workers. 14th ed. New York: Hafner; 1970.
Health Policy 87 (2008) 185–193
Cancer registry policies in the United States and geographic information systems applications in comprehensive cancer control Christie B. Ghetian a,∗ , Roxanne Parrott a , Julie E. Volkman a , Eugene J. Lengerich b a
Department of Communication Arts & Sciences, The Pennsylvania State University, University Park, PA, United States b Department of Public Health Sciences, The Pennsylvania State University, College of Medicine, and Penn State Cancer Institute, Hershey, PA, United States
Abstract Objectives: Through a long history of cancer policies, public health has a foundation for cancer research and data to use in applying technological advancements for U.S. cancer control efforts. Geographic information systems (GIS) are one technology enabling the visualization of cancer risk patterns associated with incidence, mortality, and accessibility to care. Methods: U.S. Comprehensive Cancer Control (CCC) program directors were interviewed from 49 of 50 states to assess use and function of GIS tools for mapping data related to cancer control policies and practices. Interviews were coded to obtain frequencies of response associated with content domains mapped using GIS tools and the perceived relative advantages. Results: Significant relationships were found between the mapping of behavioral risk factors, health care services, transportation access, and policy advantages identified by program managers. The mapping of cancer incidence, mortality, and staging, transportation access, and multiple layers of content were found to have significant associations with perceived research advantages. Conclusions: U.S. CCC program managers recognize several important advantages relating to health policy and research for use of GIS tools in cancer control efforts. The application of GIS in U.S. cancer control efforts is employed unevenly, suggesting the need for innovative policies to support accessibility. © 2007 Elsevier Ireland Ltd. All rights reserved. Keywords: GIS advantages; Cancer control; Policy development; Cancer mapping
The collection and use of cancer registry data for cancer control in the United States have long been ∗ Corresponding author at: 234 Sparks Building, University Park, PA, 16802, United States. Tel.: +1 814 865 4484; fax: +1 814 863 7986. E-mail address: [email protected] (C.B. Ghetian).
a national priority, as evidenced by numerous laws, rules and policies. Technologies, such as geographic information systems (GIS), provide an opportunity to use information provided through cancer registries to help guide and target cancer control efforts. The first statewide registry was begun in Connecticut in 1935 to facilitate follow-up treatments [1–3]. Congress’ pas-
0168-8510/$ – see front matter © 2007 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.healthpol.2007.12.007
186
C.B. Ghetian et al. / Health Policy 87 (2008) 185–193
sage of the National Cancer Institute Act in 1937 marked the beginning of the U.S. National Cancer Program [4] and an emphasis on population-based studies to relate human characteristics to the origins of cancer [5]. Throughout the 1940s, population-based cancer registries emerged [2,4], with the American College of Surgeons (ACoS) encouraging all hospitals to use cancer registries for the periodic review and evaluation of patients. U.S. cancer programs were asked to include a cancer registry component by 1956 [2], and the passage of the National Cancer Act in 1971 mandated that hospitals report all newly diagnosed cancers to state departments of public health [6], which contributed to data-based cancer research and innovations in diagnosis and treatment [7]. Two years later, the National Cancer Institute (NCI) instituted the Surveillance Evaluation and End Results (SEER) Program, which was a national effort to collect and evaluate cancer registry data from specific geographic regions and centrally locate it [8]. In 1989, the ACoS and the American Cancer Society (ACS) formed the Commission on Cancer (COC) and the National Cancer Data Base (NCDB) to formalize the cancer registration process. The NCDB includes more than 1500 U.S. hospital- and ambulatory care-based cancer registry programs which report cancer data, activities, and analyses on national cancer trends and disease patterns [2]. In the 1990s, research suggested that state health departments were best suited to address their state populations’ cancer needs [9], and the Cancer Registries Amendment Act of 1992 championed a state-based approach to cancer data management. It authorized the formation of the National Program of Cancer Registries (NPCR) within the Centers for Disease Control and Prevention [8], which in turn advanced state legislation and regulations to require hospitals and practitioners to make cancer diagnosis and treatment data available to their respective state registries in a standardized and timely manner [8]. By 1998, under the guidance of the NPCR, 13 new cancer registries were created and 36 improved their existing registries, accounting for 97% of the U.S. population [8]; in 2007, 45 states, the District of Columbia, Puerto Rico, the Republic of Palau, and the Virgin Islands maintain cancer registries under the NPCR [2]. The NPCR and SEER registries combined accounting for approximately 100% of the U.S. population [10]. With the adoption supporting policies and identification of funding, cancer registry data
have become the standard for surveillance of cancer incidence. Advances in computing technology now enable cancer registry data to be linked to other data, thereby providing additional insight into the distribution, determinants and treatment of cancer [2]. For example, cancer incidence data, in combination with vital statistics data from the National Center for Health Statistics, are used to estimate cancer survival rates [8]. Researchers have linked registry data with socioeconomic, environmental, and behavioral data to examine hypotheses on the etiology of cancer or the outcomes of cancer treatment [11,12]. Also, researchers have examined cancer incidence across time and geographic regions in relation to hypotheses about the changing patterns of cancer incidence. Today, state-of-the-art technology makes detailed insets, summary tables, figures, descriptive, and explanatory text accompanying computerized cancer maps not only possible, but accessible. For example, websites, such as http://statecancerprofiles.cancer.gov/ and http://cancercontrolplanet.cancer.gov/ offer interactive menus that draw upon cancer registry data to facilitate examination of states’ cancer profiles, intervention programs, and cancer data (e.g., incidence rates, death rates, cancer sites). The recent development and widespread distribution of geographic information systems technology facilitates the analysis of data in particularly innovative ways [13]. Specifically, GIS uses four general approaches for cancer research: (1) incidence mapping; (2) geographic correlation studies; (3) risk assessment in relation to a pre-specified point or line source; and (4) cluster detection and disease clustering [14]. For example, studies have used GIS technology to examine clusters of childhood leukemia and their possible association with environmental contamination [15]. The objective of the current study was to estimate the degree to which state program directors perceive policy or research advantages from the use of GIS for cancer control. In particular, we sought to determine what types of data sets are being used in state CCC, and what defining state characteristics are associated with these activities (as defined by their cancer control activities. The data from the current study provide a general indication of the research hypotheses being explored or components of the programs being developed. The results of this study may suggest areas for further devel-
C.B. Ghetian et al. / Health Policy 87 (2008) 185–193
opment of state cancer control policies and analytic capacity. 1. Method 1.1. Participants and procedures Individuals recruited to participate were the managers of CCC programs for the 50 states in the U.S. and were first contacted through use of email via use of a 2005 National CCC Program list; if an individual responded that she or he was not the CCC program manager for a given state, he/she was asked to identify the appropriate person who was contacted in a like fashion. Interviews were coordinated in batches of 10 to facilitate scheduling. At the agreed upon time, a female interviewer trained in the research protocol and phone interview script contacted the interviewee to talk about the use of CCC mapping in their state. Verbal consent to conduct and record the interview was obtained and participants were assured that no personally identifiable information would be included. Participants had to be 18 years or older to participate. At the beginning of the interview, the interviewer stated the purpose of the call was for the research entitled, “GIS Adoption Telephone Survey of Comprehensive Cancer Control and Prevention (CCCP) Directors.” The interviewer specifically stated that the interview was to assess the interviewee’s perspective about the use of mapping in their state’s CCC efforts. During the interview, respondents were asked about the use of mapping for CCC activities in their state and perceived advantages of using GIS tools and mapping to CCC activities. The specific content participants were asked about in relation to mapping activities included: behavioral risk factors; cancer incidence, mortality, and staging; demographics; environmental exposures; health care service availability; and transportation access. Interviews were transcribed verbatim with the advantages relating to cancer policy and research, summarized in Table 1. The definitions were derived post hoc based on interviewees’ responses and used in subsequent coding activities. 1.2. Coding training and coding Four coders were trained to understand the meaning of the seven policy and research advantages identi-
187
fied by participants’ responses. Then working in pairs, the coders identified the presence or absence of each of the policy and research advantages in each interview based on these definitions. After assessing five interviews, Cohen’s kappa, a measure of intercoder agreement for dichotomous data that takes chance agreement into account [16] was used to assess reliability. Cohen’s kappa ranged from .69 to 1.00 at the time of initial coding. Coders’ discrepancies were resolved through discussion until consensus was achieved, and the definitions were further refined. After another 10 interviews were coded, another assessment of intercoder agreement revealed that Cohen’s kappa ranged between .89 and 1.00, discrepancies were resolved through consensus, and the remaining interviews were coded. The final dataset combined coder responses to create a single response for each coded variable. The Statistical Package for Social Sciences 14.0 was used for analyses.
2. Results CCC directors in 49 of 50 states were interviewed; Louisiana’s director was not contacted due to Hurricane Katrina. Data were collected from July 14, 2005 to January 27, 2006 with all interviews conducted in English. Interviews lasted approximately 30 min. The interviewees included 5 males, 37 females, and 7 groups of 2 or more participants (58 individuals participated in 49 interviews). Approximately half had been in their current position for 2 years or less. The range of experience varied from one month to 12 years (M = 33 months; S.D. = 30.27 months). Half worked in public health for their states for more than 4 years, with an average of nearly 12 years. The level and areas of education of participants included: 6 Ph.D.s, 1 M.D., 2 M.D.s who also held M.S. degrees, 1 M.D. who also held a M.P.H. degree, 12 with M.P.H.s, 2 with M.S. degrees, 4 with M.P.A.s, 1 with a M.S.W., 1 with both a M.P.A. and M.S.W., 8 with Masters degrees in unspecified areas, 1 with a M.B.A., 4 with B.S. degrees, 3 with B.A.s, 1 with both a B.S. and R.N. credentials, and 1 with an Associates degree. One other interviewee was a radiation therapist. Descriptive statistics, including frequencies and percentages, were derived relating to program managers’ reports of GIS use in their states. Descriptive statistics
188
C.B. Ghetian et al. / Health Policy 87 (2008) 185–193
Table 1 Codes and definitions for content domains and relative advantages Codes Content domains Cancer incidence, mortality, and/or staging Health care service availability Demographics Behavioral Environmental exposures Transportation access Multi-layer Cancer policy advantages Identify at-risk populations Identify access issues
Identify service gaps Identify cancer staging Cancer research advantages Monitors/surveillance tool Multivariate modeling tool Generate etiological hypotheses
Definition Using mapping for any type of incidence or prevalence, mortality, early or late-stage screening, survivorship or cancer diagnosis Use of maps to identify health service availability (e.g. screening facilities) and health insurance status Use of mapping for information on race/ethnicity, gender and age Use of mapping to assess lifestyle choices including alcohol use, physical inactivity, diet tobacco use Use of mapping to identify radon, watershed locations, toxic release inventory data and pesticide exposure Use of mapping for transportation routes Use of mapping to overlay content areas to determine risk relationships Includes mapping to identify high risk populations in cancer control and prevention planning Assesses the use of maps to identify health service availability (e.g. screening facilities), transportation availability, travel distance and health insurance status Examines discussion about mapping to identify health service gaps, needs and disparities Refers to mapping that compares areas of late-versus early-stage cancer screening or diagnoses Examines mapping used to monitor the status of interventions and use of resources Represents a specific statement of how mapping is used to overlay data or multivariate modeling Assesses any discussion about mapping providing a relative advantage or generation of hypotheses for research purposes
were also computed to identify the number of managers who identified GIS mapping as affording cancer policy or research advantages for CCC. Chi-square analyses were performed to identify associations between states’ CCC-related map use and the naming of policy and research advantages for mapping specific content domains. However, because chi-square tests rely on large samples for approximation and, in some cases, there were zeros in the cells, Fisher’s exact test was also employed. This measure of statistical significance for categorical data is appropriate for small sample sizes, more specifically for two-by-two contingency tables. Although Fisher’s exact test does not operate through formal test statistics or critical values, its calculated p-value is useful for indicating a possible relationship between variables. Significance suggests that the probability of obtaining the observed value due to chance is unlikely [17].
2.1. Cancer policy and research advantages identified for GIS use Within the cancer policy domain, the frequency of response for the four issues identified by participants as advantages for using GIS mapping were: (a) identify at-risk populations (n = 7), (b) identify access issues (n = 20), (c) identify service gaps (n = 33), and (d) identify cancer staging (n = 13). Fig. 1 illustrates the relationship between content currently mapped and likelihood of mentioning any of the cancer policy advantages. The frequency of response for the three cancer research issues identified in participants’ responses regarding the advantages of using GIS for CCC activities were: (a) monitoring and surveillance tool (n = 16), (b) multivariate modeling tool (n = 24), and (c) generate etiologic hypotheses tool (n = 13). Fig. 2
C.B. Ghetian et al. / Health Policy 87 (2008) 185–193
Fig. 1. Relative advantages related to cancer policy by mapped content domain. Bars represent the within-group percentages of states that reported the relative advantage (i.e., identify at-risk populations, identify access issues, identify service gaps, and identify cancer staging) of utilizing GIS in CCC activities and mapped a specific content domain (i.e., access, behavioral risk factors, cancer incidence/mortality/staging, etc.).
illustrates the findings related to the cancer research advantages. 2.2. Relationships between content mapped and perceived advantages The following section outlines each of the content domains mapped through state CCC efforts. Within each content domain, significant relationships between it and both the cancer policy and the cancer research advantages are discussed (see Table 2). 2.2.1. Behavioral risk factors Seventeen (35%) program managers stated that behavioral risk factors were a content domain being mapped via application of GIS in their CCC activities. Mapping behavioral risk factors was found to have a significant relationship to identifying cancer policy issues as advantages. Six of the seven (86%) participants who stated the ability to identify at-risk populations as an advantage mapped behavioral risk
189
Fig. 2. Relative advantages related to cancer policy by mapped content domain. Bars represent the within-group percentages of states that reported the relative advantage (i.e., monitoring and surveillance tool, multivariate modeling tool, and generate etiological hypothesis tool) of utilizing GIS in CCC activities and mapped a specific content domain (i.e., access, behavioral risk factors, cancer incidence/mortality/staging, etc.).
factors in their states (2 (1, N = 49) = 9.38, p < .01; one-tailed Fisher’s exact test, p < .01). No other identified policy advantages revealed significant associations to the mapping of behavioral risk factors. In addition, no significant relationships were revealed among the seventeen state program managers who reported the current mapping of behavioral risk factors and the identification of GIS as advantageous for cancer research efforts. 2.2.2. Cancer incidence, mortality, and staging Sixty-nine percent of participants (n = 34) reported that cancer incidence, mortality, and staging constituted a content domain being mapped via GIS technologies. No significant relationships were found between the mapping of this content domain and any of the cancer policy advantages identified. Among the 34 program managers who indicated that they currently utilized GIS technologies to map cancer incidence, mortality, and staging, however, a relationship was found between mapping this content and a greater likelihood of reporting the use of
190
C.B. Ghetian et al. / Health Policy 87 (2008) 185–193
Table 2 Current content domains by relative advantages related to cancer policy (N = 49) Identify at-risk populations (n = 7)
Behavioral risk factors (n = 17) Cancer incidence/morality/staging (n = 34) Demographics (n = 25) Environmental exposures (n = 10) Health care services (n = 22) Transportation access (n = 2) Multi-layer (n = 13)
n (%)
χ2 (1,
6 (86) 6 (86) 3 (43) 2 (29) 3 (43) 0 (0) 3 (43)
9.38 1.03 .22 .34 .01 .35 1.12
N = 49)
Identify access issues (n = 20)
p
Fisher’s exact text, p
<.01 .31 .64 .56 .91 .56 .29
<.01 .30 .48 .44 .62 .73 .27
Identify service gaps (n = 33)
Behavioral risk factors (n = 17) Cancer incidence/morality/staging (n = 34) Demographics (n = 25) Environmental exposures (n = 10) Health care services (n = 22) Transportation access (n = 2) Multi-layer (n = 13)
n (%)
7 (35) 14 (70) 8 (40) 4 (20) 5 (25) 2 (10) 6 (30)
χ2 (1, N = 49)
p
Fisher’s exact text, p
.00 .01 1.64 .00 5.41 3.02 .21
.97 .94 .20 .95 .02 .08 .65
.60 .60 .16 .62 .02 .16 .45
Identify cancer staging (n = 13)
n (%)
χ2 (1, N = 49)
p
Fisher’s exact text, p
n (%)
χ2 (1, N = 49)
p
Fisher’s exact text, p
12 (36) 23 (70) 16 (48) 6 (18) 13 (39) 2 (6) 10 (30)
.49 .01 .26 .31 1.24 1.01 .74
.12 .95 .61 .58 .27 .32 .39
.49 .60 .42 .42 .21 .45 .31
4 (31) 7 (54) 7 (54) 4 (31) 5 (38) 0 (0) 3 (23)
.12 2.01 .06 1.17 .30 .75 .11
.73 .16 .81 .28 .59 .39 .74
.50 .14 .53 .24 .42 .54 .53
Note: Percentages are within-group for the relative advantage category (i.e., relative advantage n is used as the denominator).
GIS for generating etiological hypotheses, a cancer research advantage. Twelve of 13 (92%) managers that reported the generation of etiological hypotheses as an advantage were currently mapping cancer incidence, mortality, and staging in their respective state (χ2 (1, N = 49) = 4.38, p = .04; one-tailed Fisher’s exact test, p = .04). No other perceived research advantages were significantly associated with this domain. 2.2.3. Demographics Twenty-five (51%) participants declared GIS mapping of demographics as a current content domain for CCC activities. The mapping of demographic information had no significant effects on the likelihood of naming any cancer policy advantages of GIS applications. Likewise, GIS mapping of demographic data did not significantly affect the likelihood of directors naming any of the three cancer research advantages. 2.2.4. Environmental exposures Twenty percent (n = 10) of program managers revealed current utilization of GIS technologies for
the mapping of environmental exposures as a content area. Analyses indicated no significant relationships between the mapping of environmental exposures and identification of either cancer policy or research advantages. 2.2.5. Health care services Of the 49 participants, 45% (n = 22) reported current use of GIS applications to map health care services as a content domain. Analyses revealed that mapping health care services significantly affected the likelihood that state program managers would specify the identification of access issues as a policy advantage of use of GIS tools for mapping. Five of the 20 (25%) program managers that stated the identification of access issues as an advantage, currently mapped health care services (χ2 (1, N = 49) = 5.41, p = .02; one-tailed Fisher’s exact test, p = .02). No other cancer policy advantages were significantly associated with health care service mapping. No significant associations related to any research advantages were found in relation to the 22 state pro-
C.B. Ghetian et al. / Health Policy 87 (2008) 185–193
gram managers that indicated current use of GIS to map health care services as a content domain. 2.2.6. Transportation access Two state program managers (4%) indicated that their state CCC efforts included the mapping of transportation-related access issues as a content area in their GIS use. Although analyses revealed that the mapping of transportation routes did not significantly affect the likelihood that program managers would report any of the policy advantages related to GIS, the policy advantage, “identifies access issues,” did approach significance (χ2 (1, N = 49) = 3.02, p = .08; one-tailed Fisher’s exact test, p = .16). In regard to the research advantages afforded by GIS, the mapping of transportation-related access as a content area was found to significantly affect the likelihood that program managers would name the generation of etiologic hypotheses as an advantage of GIS mapping. Two of 13 (15%) of program managers who named the generation of etiologic hypotheses as an advantage also indicated their use of GIS to map access routes (χ2 (1, N = 49) = 5.77, p = .02; one-tailed Fisher’s exact test, p = .07). Noted in Table 3, no other research advantages identified by program managers revealed significant relationships with mapping transportation access. 2.2.7. Multi-layer content Thirteen (27%) program directors reported the use of GIS to map multiple content areas. In other words, these states mapped two or more content domains (e.g., access, behavioral risk factors, demographics) in a single map. The mapping of multiple content areas had no significant effect on the prospect of program directors naming any of the policy advantages of GIS mapping. Among the 13 state program directors that reported the use of GIS to map multiple content areas in their CCC activities, however, was a greater likelihood for naming cancer monitoring and surveillance as a relative cancer research advantage of GIS utilization. One of 16 (6%) directors that identified cancer monitoring and surveillance as a GIS advantage was currently mapping multiple layers of content (χ2 (1, N = 49) = 5.01, p = .03; one-tailed Fisher’s exact test, p = .02) (see Table 3). Neither of the other cancer research advantages was significantly related to the mapping of multiple content areas.
191
3. Discussion In the United States, we are fortunate that cancer registration polices enacted years ago now enable federal and state governments to accurately monitor temporal trends in cancer burden and to compare cancer burden in selected demographic groups. With the advent of GIS technology, governments have the potential to further use cancer registry data for cancer prevention and control in small geographic areas within their jurisdiction. This new technology affords great potential to develop policies that can apply limited cancer prevention and control resources to high-burden areas. Our research found, however, that the adoption of this technology for policy purposes was far from consistent among the states, and that there were wide variances in the perceived advantages related to use of GIS tools for mapping in performance of CCC activities. States were using data on behavioral risks and health care services in GIS technology for the policy domains of identifying at-risk populations and in identifying access issues, respectively. However, states were not consistently using demographic, environmental exposure, transportation access and even cancer incidence/mortality and staging data in the policy domain. These unused data sets may represent missed opportunities for states to target the reduction of cancer burden. States were consistently using only multi-layer data and cancer incidence/mortality/staging data in GIS technology in the research domain for monitoring and surveillance and generating etiologic research, respectively. Again, states may be missing opportunities afforded by data on behavioral risks, demographics, environmental exposures, health care services, and transportation access in cancer research. Since cancer registry data is already being collected by almost every state, the future use of GIS technology for cancer policy and research purposes will depend upon training of personnel and the diffusion of the GIS technology. Programs that train policymakers to utilize results from GIS analysis of multiple data sets should be developed and evaluated. Training materials and processes should highlight the positive attributes of GIS technology that will enhance its diffusion, including relative advantage and trialability of GIS technology. This study is limited by two issues. Though the director of Comprehensive Cancer Control in the state
192
C.B. Ghetian et al. / Health Policy 87 (2008) 185–193
Table 3 Current content domains by relative advantages related to cancer research (N = 49) Monitoring and surveillance (n = 16) n (%) Behavioral risk factors (n = 17) 1 (6) Cancer incidence/morality/staging (n = 34) 11 (69) Demographics (n = 25) 7 (44) Environmental exposures (n = 10) 3 (19) Health care services (n = 22) 7 (44) Transportation access (n = 2) 1 (6) Multi-layer (n = 13) 1 (6)
χ2 (1,
N = 49)
.08 .01 .50 .04 .01 .29 5.01
p
Fisher’s exact text, p
.77 .95 .48 .84 .91 .59 .03
.51 .60 .34 .58 .58 .55 .02
Multivariate modeling (n = 24) n (%) 10 (42) 18 (75) 13 (54) 3 (13) 9 (38) 1 (4) 8 (33)
χ2 (1, N = 49)
p
Fisher’s exact text, p
1.01 .70 .19 1.81 1.04 .00 1.12
.32 .40 .67 .18 .31 .98 .29
.24 .30 .44 .16 .23 .75 .23
Generate etiologic hypotheses (n = 13) n (%) Behavioral risk factors (n = 17) 6 (46) Cancer incidence/morality/staging (n = 34) 12 (92) Demographics (n = 25) 9 (69) Environmental exposures (n = 10) 3 (23) Health care services (n = 22) 6 (46) Transportation access (n = 2) 2 (15) Multi-layer (n = 13) 5 (38)
χ2 (1, N = 49)
p
Fisher’s exact text, p
1.03 4.38 2.35 .08 .01 5.77 1.29
.31 .04 .13 .78 .92 .02 .26
.25 .04 .11 .53 .58 .07 .22
Note: Percentages are within-group for the relative advantage category (i.e., relative advantage n is used as the denominator).
was encouraged to seek additional input, the responses largely reflect the knowledge and perspective of one person. Second, policy constructs for the reduction of cancer burden have not been developed and disseminated. However, the strengths of the study are substantial. First, this is the only study to have examined the utilization of data sets relevant to cancer policy and research at the state level. Second, the number and rate of respondents was high, with a 98% participation rate. Third, respondents had the opportunity to review transcripts of their interview to assure that their intended meaning was present, increasing the internal validity associated with the seven advantages extracted to represent the advantages perceived to relate to policy and research in fulfillment of CCC activities. The coding of data involved multiple investigators, increasing the external validity associated with the conclusions drawn from this project. 4. Conclusion As noted at the outset, policies, especially those to collect and use cancer registry information, have long been a national priority. The advent of GIS technology
enables states to utilize additional data sets for cancer policy and research. However, states must now adopt policies that encourage the development and utilization of cancer registry data in small geographic areas, thus illustrating the maxim that health research is integrally related to political environments. Although significant relationships were identified, such that participants’ naming policy advantages for GIS and cancer mapping were more likely to display behavioral risk factors to identify at-risk populations and health care services to identify access issues, it is difficult to separate the direction of the relationship. Mapping a content domain may influence the likelihood of perceiving particular policy advantages, or vice versa. Acknowledgements This research was supported by a Cooperative Agreement between the Centers for Disease Control and Prevention (CDC) and the Association of American Medical Colleges (AAMC), award number MM-0718. Its contents are the responsibility of the authors and do not necessarily reflect the official views of the CDC or AAMC.
C.B. Ghetian et al. / Health Policy 87 (2008) 185–193
References [1] Gagnon F, Turgeon J, Dallaire C. Health public policy: a conceptual cognitive framework. Health Policy 2007;81:42–55. [2] National Cancer Institute. Surveillance Epidemiology and End Results. Brief History of Cancer Registration, http://training.seer.cancer.gov/module cancer registration/unit1 history.html; 2007. [3] Heston JF, Kelly JB, Meigs JW, Flannery JT. Forty-five years of cancer incidence in Connecticut: 1935–79. National Cancer Institute (NCI) Monograph No. 70. NIH Publication No. 862652. NCI, Bethesda, MD; 1986. [4] Hubbard SM, Mayer DK, DeVita Jr VT. Evolution of the National Cancer Program: an American investment for conquering cancer. Seminars in Oncology Nursing 2002;18:252–64. [5] Schottenfeld D, Beebe-Dimmer JL. Advances in cancer epidemiology: understanding causal mechanisms and the evidence for implementing interventions. Annual Reviews of Public Health 2005;26:37–60. [6] Polednak AP. Trends in cancer incidence in Connecticut, 1935–1991. Cancer 1994;74:2863–72. [7] Flannery JT, Boice Jr JD, Devesa SS, Kleinerman RA, Curtis RE, Fraumeni Jr JF. Cancer Registration in Connecticut and the study of multiple primary cancers, 1935–82. Journal of the National Cancer Institute Monographs 1985;68:13–24. [8] National Cancer Institute. Cancer Surveillance Implementation Plan, March 1999. http://cancercontrol.cancer.gov/sig/ index.htm; 2007.
193
[9] National Cancer Legislation Advisory Committee. National Cancer Policy Legislative History, http://www.cancersource .com/nclac/leghistory.htm; 2007. [10] Centers for Disease Control and Prevention. National Program of Cancer Registries, http://www.cdc.gov/cancer/npcr/; 2007. [11] Boss LP, Suarez L. Uses of data to plan cancer prevention and control programs. Public Health Reports 1990;105:354– 60. [12] Pommerenke FA, Miller RW, Sirvastava S, Ackerman SP. Targeting cancer control: the state cancer control map and data program. American Journal of Public Health 1994;84:1479– 82. [13] Goodchild MF, Anselin L, Appelbaum R, Harthorn B. Toward spatially integrated social science. International Regional Science Review 2000;23:139–59. [14] Wakefield J, Elliott P. Issues in the statistical analysis of small area health data. Statistics in Medicine 1999;18:2377– 99. [15] Openshaw S, Charlto M, Wymer C, Craft A. Developing a mark 1 geographical analysis machine for the automated analysis of point data sets. International Journal of Geographical Information Systems 1987;1:335–58. [16] Cohen J. Coefficient of agreement for nominal scales. Educational and Psychological Measurement 1960;20:37– 46. [17] Fisher RA. Statistical method for research workers. 14th ed. New York: Hafner; 1970.