- Email: [email protected]
Comorbidity and early diagnosis of head and neck cancer in a Medicare population
Comorbidity and Early Diagnosis of Head and Neck Cancer in a Medicare Population Britt C. Reid, DDS, PhD, Joan L. Warren, PhD, Gary Rozier, DDS, MPH Background: This study assessed potential opportunities for an early diagnosis of head and neck cancers (HNCA), and the role played by comorbidity, among contacts by Medicare patients with the healthcare system before their diagnosis. Comorbidity was hypothesized to affect the relationship between use of healthcare services and stage of disease. Methods:
The study database (n ⫽11,312) was constructed by linking files from the Surveillance, Epidemiology, and End Results Program 1991 through 1999 and those of the Center for Medicare and Medicaid Services Program. HNCA cases were comprised of the lip, oral cavity, pharynx, and larynx anatomic sites.
Results:
Among persons with no alcohol and tobacco–related comorbidities (ATC), increasing numbers of physician visits were independently associated with a reduced risk of advanced stage at diagnosis for all anatomic sites, but especially the pharynx and larynx. Among persons with one or more ATC, physician visits displayed no association with stage at diagnosis with the exception of laryngeal tumors, where physician visits were strongly associated with reduced risk of late stage at diagnosis. Over 46% of cases with advanced stage tumors, most of whom also had ATC disease, had ⱖ11 physician visits in the year preceding their diagnosis of HNCA.
Conclusions: Potential opportunities for an earlier diagnosis of HNCA were found among the numerous contacts with the healthcare system that patients had prior to their cancer diagnosis. Such opportunities were especially numerous and relevant for persons with ATC and form the basis for combining high-risk with opportunistic screening approaches. (Am J Prev Med 2004;27(5):373–378) © 2004 American Journal of Preventive Medicine
Introduction
T
he recent Surgeon General’s Report on Oral Health identified oral and pharyngeal cancers as imposing a significant burden on society, especially among minorities and the poor.1 The American Cancer Society (ACS) projects about 37,200 new cases of head and neck cancers (HNCA) with 11,000 associated deaths for the year 2003.2 More than half of these tumors are diagnosed at an advanced stage,3 despite most early cases being readily detectable by clinical examination.4 To reduce morbidity and mortality from head and neck cancer, Healthy People 2010 set goals to increase the number of adults with an annual oral cancer screening and the proportion of oral cancer diagnosed at the From the Health Services Research Program, Department of Health Promotion and Policy, School of Dentistry, University of Maryland (Reid), Baltimore, Maryland; Applied Research Program, National Cancer Institute (Warren), Bethesda, Maryland; and Department of Health Policy and Administration, School of Public Health, University of North Carolina at Chapel Hill (Rozier), Chapel Hill, North Carolina Address correspondence and reprint requests to: Britt C. Reid, DDS, PhD, School of Dentistry, University of Maryland, Department of Health Promotion and Policy, Room 3A-08, 666 Baltimore Street, Baltimore MD 21201. E-mail: [email protected].
local stage.5 Yet there is no consensus regarding who should undergo oral cancer screenings. The ACS recommends oral cancer screenings for all adults as part of a regular overall cancer screening.6 Both the Canadian Task Force on Clinical Preventive Services and U.S. Task Force on Clinical Preventive Services report that there is insufficient evidence to recommend routine population screening for oral cancer by healthcare professionals, but do recommend an annual examination by a physician or dentist for persons aged ⱖ60 years with heavy alcohol and tobacco use.7,8 Despite these recommendations, only 15% of adults aged ⱖ40 years report receipt of an oral cancer examination in their lifetime.9 Report of an examination varies little by smoking status,10 a highly relevant risk factor for the disease.11 General approaches to screening for head and neck cancer are sometimes described as (1) routine population screening (annual screening for all adults); (2) opportunistic screening (screening in conjunction with other scheduled visits); and (3) high-risk population screening (screening of tobacco and alcohol users). It has been argued that high-risk individuals should be targeted for aggressive regular screening
Am J Prev Med 2004;27(5) © 2004 American Journal of Preventive Medicine • Published by Elsevier Inc.
0749-3797/04/$–see front matter doi:10.1016/j.amepre.2004.07.018
373
because early head and neck symptoms provide inadequate information for an early diagnosis.12 It has also been found that considerable prior contact with primary healthcare providers existed among head and neck cancer patients, supporting an opportunistic strategy of integrating screening for head and neck cancers into existing healthcare services.13 Comorbidity is relevant for targeting high-risk individuals and integrating their head and neck cancer screening into existing healthcare services. Comorbid disease, in the present context, is a coexisting medical condition other than head and neck cancer. The major risk factors for head and neck cancer, tobacco and alcohol abuse, play the same role in many other chronic diseases including heart disease, lung disease, and cancers of other sites. As a result, persons most at risk for head and neck cancer can be expected to have considerable comorbid disease and related contacts with the healthcare system. Therefore, comorbidity potentially identifies high-risk individuals likely to have numerous contacts with the healthcare system, a fortuitous situation to exploit for early diagnosis. The purpose of this study is to assess opportunities for an early diagnosis of head and neck cancers by identifying the frequency of contacts with the healthcare system before the diagnosis. The hypothesis is that medical visits occur that are not exploited for their screening opportunities, and that they are particularly numerous among those who have comorbidities sharing a common etiology with head and neck cancers.
Methods Data Source The linked Surveillance, Epidemiology, and End Results (SEER)–Medicare data were used for this study. These data contain the Medicare claims for persons diagnosed with head and neck cancer, as reported by the National Cancer Institute’s SEER Program. SEER is a population-based system of cancer registries. The data for this analysis included 12 SEER registries that cover approximately 14% of the U.S. population. The Medicare data included in this analysis comprise the claims from all inpatient stays (Medicare Part A), as well as bills from physician and hospital outpatient departments (Medicare Part B). The hospital and physician claims include ICD-9-CM diagnoses for each hospitalization or billed service. Ninety-four percent of all SEER cancer cases diagnosed in persons aged ⱖ65 were found in the Medicare enrollment files. In addition to information about individuals, the SEER– Medicare data have also been matched by census track and ZIP code to data from the U.S. Census. These data have been used to obtain the median levels of income and education in the locales where patients reside. This study received approval from the Institutional Review Board of the University of Maryland, Baltimore, and the Centers for Medicare and Medicaid Services database review committee.
374
Selection Criteria People were included in the analysis if they were in the Medicare database as of 1991, with continuous enrollment in Part A and Part B, and had their first head and neck cancer between 1992 and 1999 as reported in the SEER data. This design ensured that all subjects had at least 1 year before diagnosis with HNCA to establish a pattern of health services usage. HNCA cases comprised the lip, oral cavity, pharynx, and larynx anatomic sites (International Classification of Disease [ICD-9-CM] codes 140.0 –148.9, 161.0 –161.9). Individuals whose cancer was diagnosed at autopsy, solely from death certificates and not microscopically confirmed, or with missing stage at diagnosis data were not included. Also excluded were individuals with cancers classified anatomically as “other oral cavity and pharynx” (n ⫽225) because the analysis required exclusive assignment to a specific anatomic site. Analyses excluded people enrolled in a health maintenance organization (HMO) (n ⫽3234) in the year before diagnosis with HNCR because the Medicare claims do not include any detailed health services data. Cases excluded for enrollment in an HMO did not differ from non-HMO enrolled cases in distributions of age, gender, or race/ethnicity. Cases with missing income and education (n ⫽51) were also excluded, resulting in a final sample of 11,312 cases.
Variable Descriptions Stage at diagnosis was classified as localized (historic stages in situ and local) or advanced (historic stages regional and distant). The summary system of “historic stage” was used in order for stage at diagnosis to remain consistent over the time span of the study. Historic stage at diagnosis is not completely comparable to the American Joint Committee on Cancer (AJCC) staging system because of AJCC changes in anatomic subsites and extent of disease definitions during the study time period. Histologic grade was the only variable available in this data set, other than cell type, which captured some of the biological attributes of the tumor; it was modeled as a dichotomous variable with the categories grade 1 or unknown, and ⱖgrade 2. Histologic grades “1” and “unknown” had similar bivariate relationships with stage at diagnosis, and were combined to simplify analysis and presentation. Highergrade scores indicate less histologic differentiation. Cases with a cancer diagnosis other than HNCA, occurring anytime prior to diagnosis with HNCA, were modeled as a dichotomous yes/no variable. Patient demographic variables gender and Hispanic ethnicity were modeled as dichotomous indicators. Race/ethnicity was modeled with white, black, and “other” categories. Age was modeled as a continuous variable. Marital status was modeled with the categories of currently married, divorced, never married, widowed, and unknown. Income and education were ecologic variables derived from an algorithm designed to minimize missing data. The algorithm searched for data from the following sources in priority order: 1990 census tract data, 2000 census tract data, 1990 ZIP code data, and lastly, 2000 ZIP code data. The 1990 census tract data was available for 97.6% of the study population. Income was assigned based on the median household income of the patient’s census tract/zip code; the resulting continuous variable was then divided into quartiles for analysis. Similarly,
American Journal of Preventive Medicine, Volume 27, Number 5
Table 1. Percentage of head and neck cancer patients by stage at diagnosis and selected characteristics (N ⫽ 11,312) Local stage (%) (n ⴝ 5585) Physician visits in year prior to cancer diagnosis None 06.8 1–5 18.3 6–10 20.6 ⱖ11 54.3 Anatomic site Lip 16.7 Oral cavity 36.9 Pharynx 05.3 Larynx 41.1 Number of ATC conditions 0 33.2 ⱖ1 66.8 Had prior cancer (other than head and neck) Yes 16.9 No 83.1 Histologic grade Grade 1 or unknown 52.2 Grade 2 or higher 47.8 Age at diagnosis (years) 65–74 52.1 ⱖ75 47.9 Ethnicity/race Hispanic 03.3 Non-Hispanic 96.7 White 91.1 Nonwhite 08.9 Gender Male 68.6 Female 31.4 Marital status Currently married 59.4 Not married/unknown 40.6 Incomea Above median 51.2 At or below median 48.8 Educationb Above median 52.6 At or below median 47.4
Advanced stage (%) (n ⴝ 5727) 11.7 22.3 19.2 46.8 03.0 44.7 25.7 26.6 37.3 62.7 14.2 85.8 29.7 70.2 54.8 45.2 04.2 95.8 86.5 13.5 66.2 33.8 50.5 49.5 48.9 51.1 47.8 52.2
Note: All column differences in proportions had p values ⱕ0.05. ATC, alcohol and tobacco–related comorbid conditions. a Determined by household income of a patient’s census/ZIP code area. b Determined by percentage of adults in a patient’s census/ZIP code area with 12 or more years of education.
education was defined by the percentage of people aged ⱖ25 years in a patient’s census tract/ZIP code with ⬍12 years of education. The resulting data were then divided into quartiles for analysis but collapsed to simplify presentation in Table 1. Physician visits in the year before a diagnosis with HNCA were modeled as counts. An alcohol and tobacco–related comorbidity (ATC) variable derived for the study is described in detail elsewhere.14 It is based on comorbid conditions associated with alcohol and tobacco use, the primary etiologic agents for HNCA. The ATC index consists of a simple count of 11 conditions present, and thus has a theoretical range of 0 to 11. Conditions were counted if listed on inpatient records during the 12 months before diagnosis with HNCA (excluding the
month of diagnosis), and at least once in physician claims, or in physician claims more than once in more than one 30-day period. This approach is similar to the one described in more detail by Klabunde et al.15 The ATC index was collapsed into a dichotomous variable to facilitate stratification of analyses and presentation of data. The sample distribution of ATC scores 0, 1, and ⱖ2 was 35%, 32%, and 33%, respectively. In statistical models of the relationship between stage at diagnosis and physician visits, the ATC variable functioned as an effect modifier with p values for interaction terms of ⬍0.01. All subsequent associations were therefore stratified by ATC status in order to display the role played by these comorbidities.
Statistics The Pearson chi-square test for proportions was used to develop p values where appropriate for comparing proportions. Logistic regression was used for adjusted odds ratios as measures of association. Odds ratios were determined to be statistically significant when 95% confidence intervals did not include the value 1.00. The STATA statistical package, version 8.0, was used for all analyses (STATA Corporation, College Station TX, 2003). All analyses were conducted in 2003 and 2004.
Results The distribution of cases by number of physician visits before diagnosis, anatomic site, and histologic grade varied greatly by stage at diagnosis. The distribution of cases by age, prior cancer, ATC, race/ethnicity, gender, marital status, income, and education varied only slightly to modestly by stage at diagnosis (Table 1). Among those with at least one physician visit in the year before HNCA diagnosis, cases with ATC had more physician visits than persons without ATC (data not shown). After stratifying by anatomic site, increasing numbers of physician visits were independently associated with reduced risk of advanced stage at diagnosis among persons with no ATC for all anatomic sites, but not statistically significant for the lip site (Table 2). The association was especially strong for the pharynx and larynx. Among cases with one or more ATC, physician visits displayed no association with stage at diagnosis with the exception of patients with laryngeal tumors where physician visits were associated with a larger reduction in risk of late stage at diagnosis compared to patients with no ATCs who had laryngeal tumors (Table 3). A moderate to poorly differentiated tumor increased the risk of advanced stage at diagnosis compared to well-differentiated and ungraded tumors among all anatomic sites, except for cases with both no ATC and pharyngeal cancers (Tables 2 and 3). A previous cancer diagnosis did not change risk of advanced stage at diagnosis. Am J Prev Med 2004;27(5)
375
Table 2. Adjusteda odds ratios for advanced stage at diagnosis by anatomic site among cases with no ATC Lip (n ⴝ 409) AOR
Oral cavity (n ⴝ 1669)
95% CI
AOR
95% CI
0.32 0.55 0.61 0.69 1.00
0.09–1.16 0.19–1.56 0.25–1.52 0.33–1.42 —
0.81 0.63* 0.65* 0.86 1.00
0.55–1.21 0.45–0.88* 0.47–0.90* 0.65–1.14 —
0.61 1.00
0.24–1.55 —
1.16 1.00
3.10* 1.00
1.76–5.48* —
1.58* 1.00
Pharynx (n ⴝ 640) AOR
Larynx (n ⴝ 1269)
95% CI
AOR
95% CI
0.29* 0.62 0.35* 0.31* 1.00
0.11–0.75* 0.25–1.52 0.15–0.81* 0.15–0.64* —
0.38* 0.41* 0.49* 0.81 1.00
0.22–0.66* 0.27–0.62* 0.34–0.70* 0.58–1.12 —
0.85–1.59 —
0.92 1.00
0.44–1.92 —
0.93 1.00
0.63–1.37 —
1.28–1.94* —
1.20 1.00
0.69–2.07 —
3.06* 1.00
2.38–3.95* —
b
Physician visits 20⫹ 11–20 6–10 1–5 0 Prior cancerc Yes No Histologic grade ⱖGrade 2 Grade 1/unknown
AOR, adjusted odds ratio; ATC, alcohol and tobacco–related comorbid condition; CI, confidence interval a All displayed odds ratios adjusted for gender, race/ethnicity, marital status, income, education, and age, in addition to the other covariates in the table. b Number of physician visits in year before diagnosis with head and neck cancer. c Had previous cancer (other than head and neck). *Significant at p ⱕ0.05 (bolded).
Overall, ⬎46% of persons with advanced stage tumors visited a physician ⱖ11 times in the year prior to their diagnosis with HNCA. The distribution of visits by ATC status differed, with 4.6 times more people with comorbidities having ⱖ11 visits compared to those without (Figure 1).
Discussion Potential missed opportunities for an earlier diagnosis of HNCA were found among the many contacts with the healthcare system that patients had prior to their diagnosis with HNCA. Over 93% of persons with localized tumors, and 88% of persons with advanced
tumors had at least one physician visit in the year prior to diagnosis with HNCA, and ⬎54% of persons with localized tumors and 46% of persons with advanced tumors had ⱖ11 (Table 1). Such contacts with the healthcare system are the basis for “opportunistic screening” as described by Prout et al.13 and others.16,17 Comorbidity was found to effect the association between physician visits and stage at diagnosis. Comorbid disease was found to be an independent predictor of survival among a similar population of HNCA patients, and some of the effects of comorbidity on survival were related to the association of comorbidity and stage at diagnosis.18
Table 3. Adjusteda odds ratios for advanced stage at diagnosis by anatomic site among cases with one or more ATC Oral cavity (n ⴝ 2951)
Lip (n ⴝ 691) AOR
95% CI
AOR
95% CI
Pharynx (n ⴝ 1127) AOR
Larynx (n ⴝ 2552)
95% CI
AOR
95% CI
b
Physician visits 20⫹ 11–20 6–10 1–5 0 Prior cancerc Yes No Histologic grade ⱖGrade 2 Grade 1/unknown
0.45 0.63 0.50 0.67 1.00
0.06–3.30 0.09–4.59 0.07–3.65 0.09–5.03 —
0.97 1.00 1.16 1.24 1.00
0.48–1.97 0.49–2.03 0.56–2.37 0.60–2.55 —
1.28 0.92 1.10 1.15 1.00
0.40–4.06 0.29–2.92 0.34–3.54 0.35–3.83 —
0.21* 0.23* 0.25* 0.41* 1.00*
0.09–0.46* 0.10–0.52* 0.11–0.57* 0.18–0.93* —
1.07 1.00
0.60–1.90 —
0.85 1.00
0.69–1.05 —
1.00 1.00
0.66–1.52 —
0.93 1.00
0.74–1.17 —
2.49* 1.00
1.59–3.90* —
2.02* 1.00
1.74–2.35* —
1.91* 1.00
1.36–2.69* —
2.57* 1.00
2.15–3.07* —
AOR, adjusted odds ratio; ATC, alcohol and tobacco–related comorbid condition; CI, 95% confidence interval. a All displayed odds ratios adjusted for gender, race/ethnicity, marital status, income, education, and age in addition to the other covariates in the table. b Number of physician visits in year before diagnosis with head and neck cancer. c Had previous cancer (other than head and neck). *Significant at p ⱕ 0.05 (bolded).
376
American Journal of Preventive Medicine, Volume 27, Number 5
Figure 1. Percentage distribution of persons with advanced stage at diagnosis by alcohol and tobacco–related comorbidity status and number of physician visits in the year prior to diagnosis (n ⫽5727).
Among cases with no ATC, the number of physician visits displayed the largest reduction in risk of advanced stage for laryngeal and pharyngeal tumors, and had the least impact on lip cancers. No dose–response relationship between visits and stage was observed, and not all strata displayed statistically significant associations, suggesting that some opportunities exist for improving the proportion diagnosed at early stages among people with no comorbidity (Table 2). The relationship between physician visits and stage at diagnosis among cases with one or more ATC resembled that of cases without ATC for lip cancers but not the other anatomic sites. In the presence of ATC, oral cavity and pharyngeal cancers no longer displayed any association between physician visits and stage, while the association among laryngeal cancer patients strengthened. It is possible that in the majority of physician visits recorded for persons with one or more ATC and who eventually developed oral cavity or pharyngeal tumors, the provider focused solely on the comorbid conditions themselves, and thus offered little opportunity for early diagnosis. In these cases, comorbid conditions may so dominate the care of the patient that fewer chances for an early diagnosis of HNCA exist. This pattern of care was present despite the fact that all the comorbid conditions in this analysis were related to alcohol and tobacco use. The possible exclusive physician focus on one adverse effect of exposure to alcohol and tobacco products may result in lost opportunities for an earlier diagnosis of additional adverse effects, such as HNCA. Interestingly, in the case of laryngeal cancer, the presence of an ATC apparently did not “distract” physicians, but possibly alerted them to risk of these cancers among their patients. One way in which laryngeal cancers may differ from oral cavity and pharyngeal cancers is in early symptoms. Laryngeal tumors often display an early symptom, hoarseness, which a patient brings to the attention of a provider.12 It is possible that instead of a busy provider having to look for something
beyond the immediate scope of the visit as would be the case for oral cavity and pharyngeal tumors, the hoarseness becomes a chief complaint that, especially in the context of an ATC, a physician then couples with the need for evaluation. All histologic types were included in the analyses. It was reasoned that although squamous cell carcinomas (SCC) were the most prevalent histologic type, ultimately the only chance for HNCA to be picked up early may lie with providers; and even if providers were only looking for SCC this process may uncover many nonSCC lesions. To confirm this inclusion decision, analyses using only SCC lesions were run with the results supporting the inclusion of all histologic types. This study exhibited a number of strengths. The use of a Medicare population allowed us to minimize any impact that access to care may have had on the findings. Few large population groups in U.S. society have similar access to health care. The sample population has been shown to be generally representative of all U.S. citizens of this age group.19 The relatively large sample sizes also allowed for stable estimates of effect. In addition, a measure of comorbidity was available for all the analyses, which was important for this type of study. A limitation of this study includes not having any information on dental visits. These providers would be expected to diagnose a large number of oral cavity and lip tumors. However, the impact of dentist diagnoses would only affect the estimated associations between physician visits and stage at diagnosis to the extent that dentist screening behavior differs from that of physicians. Given the very low rates of oral cancer screening reported by patients, this study limitation may be small, but future studies would be improved by including these data. About half of all HNCA occurs before age 65 years, and the study findings may not be representative of patterns present among younger cancer patients. Physician visits may to some extent also be a proxy for health-seeking behavior on the part of patients. Therefore, any findings that potential opportunities for an early diagnosis of HNCA exists in the many contacts that patients have with the healthcare system will clearly not apply to persons who avoid the system. People with extremely high levels of alcohol or tobacco abuse, who are at high risk of HNCA, may also display little health-seeking behavior. Few biological variables were available for the study. The strong and consistent independent association of histologic grade with stage at diagnosis hints that some tumors may proceed to an advanced stage due to biological factors and may not ultimately be amenable to the types of health services interventions examined in this study. Other limitations include the lack of a variable measuring the pattern of provider contacts in the Am J Prev Med 2004;27(5)
377
What This Study Adds . . . It was previously reported that individuals at high risk for head and neck cancers should be targeted for screening because early symptoms provide inadequate information for early diagnosis, and that considerable prior contact with primary healthcare providers existed among head and neck cancer patients. This study found that comorbidity was relevant for targeting high-risk individuals and numerous opportunities may exist for integrating their head and neck cancer screening into existing healthcare services.
months prior to a diagnosis of HNCA, identification of the type of medical providers involved in the visits, and the reason for the visit. Although many aspects remain to be studied, elderly HNCA patients did display numerous contacts with healthcare providers in the year prior to their cancer diagnosis, and the impact of these visits on stage at diagnosis varied by comorbidity and anatomic site. At a minimum, the potential exists for opportunistic screening during provider contacts to achieve early diagnoses of HNCA, especially among people with ATC-related comorbid disease. We are grateful to Marie Topor and Angela Fahey of Information Management Systems (Silver Spring, MD) for programming and database assistance. BCR was funded by the National Institutes Health (grant 5-K22-DE014341-02).
References 1. National Institute of Dental and Craniofacial Research, National Institutes of Health. Oral health in America: a report of the Surgeon General. Washington DC: U.S. Department of Health and Human Services, 2000.
378
2. American Cancer Society. Cancer facts and figures 2003. Available at: www.cancer.org/docroot/STT/stt_0.asp. Accessed January 5, 2004. 3. National Cancer Institute. Surveillance, Epidemiology, and End Results Program 9 Registries Public-Use Database (1973–1995). Rockville MD: National Cancer Institute, 1998. 4. Smart CR. Screening for cancer of the aerodigestive tract. Cancer 1993; 72(suppl 3):1061–5. 5. U.S. Department of Health and Human Services. Healthy people 2010. Washington DC: U.S. Department of Health and Human Services, 2000. 6. American Cancer Society. Cancer prevention and early detection, 2002. Available at: http://www.cancer.org/docroot/PED/content/PED_2_3X_ Early_Detection.asp. Accessed January 5, 2004. 7. Hawkins RJ, Wang EE, Leake JL. Preventive health care, 1999 update: prevention of oral cancer mortality. The Canadian Task Force on Preventive Health Care. J Can Dent Assoc 1999;65:617. 8. U.S. Preventive Services Task Force. Screening oral cancer 1996. Available at: http://www.ahcpr.gov/clinic/uspstf/uspsoral.htm. Accessed January 5, 2004. 9. Horowitz AM, Nourjah PA. Factors associated with having oral cancer examinations among U.S. adults 40 years of age or older. J Public Health Dent 1996;56:331–5. 10. Macek MD, Reid BC, Yellowitz JA. Oral cancer examinations among adults at high risk: findings from the 1998 National Health Interview Survey. J Public Health Dent 2003;63:119 –25. 11. Rothman K, Keller A. The effect of joint exposure to alcohol and tobacco on risk of cancer of the mouth and pharynx. J Chronic Dis 1972;25:711– 6. 12. Dolan RW, Vaughan CW, Fuleihan N. Symptoms in early head and neck cancer: an inadequate indicator. Otolaryngol Head Neck Surg 1998;119: 463–7. 13. Prout MN, Heeren TC, Barber CE, et al. Use of health services before diagnosis of head and neck cancer among Boston residents. Am J Prev Med 1990;6:77– 83. 14. Reid BC, Alberg AJ, Klassen AC, et al. A comparison of three comorbidity indexes in a head and neck cancer population. Oral Oncol 2002;38:187–94. 15. Klabunde CN, Warren JL, Legler JM. Assessing comorbidity using claims data: an overview. Med Care 2002;40(suppl 8):26 –35. 16. Dombi C, Voros-Balog T, Czegledy A, Hermann P, Vincze N, Banoczy J. Risk group assessment of oral precancer attached to x-ray lung-screening examinations. Community Dent Oral Epidemiol 2001;29:9 –13. 17. Lim K, Moles DR, Downer MC, Speight PM. Opportunistic screening for oral cancer and precancer in general dental practice: results of a demonstration study. Br Dent J 2003;194:497–502. 18. Reid BC, Alberg AJ, Klassen AC, et al. Comorbidity and survival of elderly head and neck carcinoma patients. Cancer 2001;92:2109 –16. 19. Warren JL, Klabunde CN, Schrag D, Bach PB, Riley GF. Overview of the SEER-Medicare data: content, research applications, and generalizability to the United States elderly population. Med Care 2002;40(suppl 8):3–18.
American Journal of Preventive Medicine, Volume 27, Number 5
The study database (n ⫽11,312) was constructed by linking files from the Surveillance, Epidemiology, and End Results Program 1991 through 1999 and those of the Center for Medicare and Medicaid Services Program. HNCA cases were comprised of the lip, oral cavity, pharynx, and larynx anatomic sites.
Results:
Among persons with no alcohol and tobacco–related comorbidities (ATC), increasing numbers of physician visits were independently associated with a reduced risk of advanced stage at diagnosis for all anatomic sites, but especially the pharynx and larynx. Among persons with one or more ATC, physician visits displayed no association with stage at diagnosis with the exception of laryngeal tumors, where physician visits were strongly associated with reduced risk of late stage at diagnosis. Over 46% of cases with advanced stage tumors, most of whom also had ATC disease, had ⱖ11 physician visits in the year preceding their diagnosis of HNCA.
Conclusions: Potential opportunities for an earlier diagnosis of HNCA were found among the numerous contacts with the healthcare system that patients had prior to their cancer diagnosis. Such opportunities were especially numerous and relevant for persons with ATC and form the basis for combining high-risk with opportunistic screening approaches. (Am J Prev Med 2004;27(5):373–378) © 2004 American Journal of Preventive Medicine
Introduction
T
he recent Surgeon General’s Report on Oral Health identified oral and pharyngeal cancers as imposing a significant burden on society, especially among minorities and the poor.1 The American Cancer Society (ACS) projects about 37,200 new cases of head and neck cancers (HNCA) with 11,000 associated deaths for the year 2003.2 More than half of these tumors are diagnosed at an advanced stage,3 despite most early cases being readily detectable by clinical examination.4 To reduce morbidity and mortality from head and neck cancer, Healthy People 2010 set goals to increase the number of adults with an annual oral cancer screening and the proportion of oral cancer diagnosed at the From the Health Services Research Program, Department of Health Promotion and Policy, School of Dentistry, University of Maryland (Reid), Baltimore, Maryland; Applied Research Program, National Cancer Institute (Warren), Bethesda, Maryland; and Department of Health Policy and Administration, School of Public Health, University of North Carolina at Chapel Hill (Rozier), Chapel Hill, North Carolina Address correspondence and reprint requests to: Britt C. Reid, DDS, PhD, School of Dentistry, University of Maryland, Department of Health Promotion and Policy, Room 3A-08, 666 Baltimore Street, Baltimore MD 21201. E-mail: [email protected].
local stage.5 Yet there is no consensus regarding who should undergo oral cancer screenings. The ACS recommends oral cancer screenings for all adults as part of a regular overall cancer screening.6 Both the Canadian Task Force on Clinical Preventive Services and U.S. Task Force on Clinical Preventive Services report that there is insufficient evidence to recommend routine population screening for oral cancer by healthcare professionals, but do recommend an annual examination by a physician or dentist for persons aged ⱖ60 years with heavy alcohol and tobacco use.7,8 Despite these recommendations, only 15% of adults aged ⱖ40 years report receipt of an oral cancer examination in their lifetime.9 Report of an examination varies little by smoking status,10 a highly relevant risk factor for the disease.11 General approaches to screening for head and neck cancer are sometimes described as (1) routine population screening (annual screening for all adults); (2) opportunistic screening (screening in conjunction with other scheduled visits); and (3) high-risk population screening (screening of tobacco and alcohol users). It has been argued that high-risk individuals should be targeted for aggressive regular screening
Am J Prev Med 2004;27(5) © 2004 American Journal of Preventive Medicine • Published by Elsevier Inc.
0749-3797/04/$–see front matter doi:10.1016/j.amepre.2004.07.018
373
because early head and neck symptoms provide inadequate information for an early diagnosis.12 It has also been found that considerable prior contact with primary healthcare providers existed among head and neck cancer patients, supporting an opportunistic strategy of integrating screening for head and neck cancers into existing healthcare services.13 Comorbidity is relevant for targeting high-risk individuals and integrating their head and neck cancer screening into existing healthcare services. Comorbid disease, in the present context, is a coexisting medical condition other than head and neck cancer. The major risk factors for head and neck cancer, tobacco and alcohol abuse, play the same role in many other chronic diseases including heart disease, lung disease, and cancers of other sites. As a result, persons most at risk for head and neck cancer can be expected to have considerable comorbid disease and related contacts with the healthcare system. Therefore, comorbidity potentially identifies high-risk individuals likely to have numerous contacts with the healthcare system, a fortuitous situation to exploit for early diagnosis. The purpose of this study is to assess opportunities for an early diagnosis of head and neck cancers by identifying the frequency of contacts with the healthcare system before the diagnosis. The hypothesis is that medical visits occur that are not exploited for their screening opportunities, and that they are particularly numerous among those who have comorbidities sharing a common etiology with head and neck cancers.
Methods Data Source The linked Surveillance, Epidemiology, and End Results (SEER)–Medicare data were used for this study. These data contain the Medicare claims for persons diagnosed with head and neck cancer, as reported by the National Cancer Institute’s SEER Program. SEER is a population-based system of cancer registries. The data for this analysis included 12 SEER registries that cover approximately 14% of the U.S. population. The Medicare data included in this analysis comprise the claims from all inpatient stays (Medicare Part A), as well as bills from physician and hospital outpatient departments (Medicare Part B). The hospital and physician claims include ICD-9-CM diagnoses for each hospitalization or billed service. Ninety-four percent of all SEER cancer cases diagnosed in persons aged ⱖ65 were found in the Medicare enrollment files. In addition to information about individuals, the SEER– Medicare data have also been matched by census track and ZIP code to data from the U.S. Census. These data have been used to obtain the median levels of income and education in the locales where patients reside. This study received approval from the Institutional Review Board of the University of Maryland, Baltimore, and the Centers for Medicare and Medicaid Services database review committee.
374
Selection Criteria People were included in the analysis if they were in the Medicare database as of 1991, with continuous enrollment in Part A and Part B, and had their first head and neck cancer between 1992 and 1999 as reported in the SEER data. This design ensured that all subjects had at least 1 year before diagnosis with HNCA to establish a pattern of health services usage. HNCA cases comprised the lip, oral cavity, pharynx, and larynx anatomic sites (International Classification of Disease [ICD-9-CM] codes 140.0 –148.9, 161.0 –161.9). Individuals whose cancer was diagnosed at autopsy, solely from death certificates and not microscopically confirmed, or with missing stage at diagnosis data were not included. Also excluded were individuals with cancers classified anatomically as “other oral cavity and pharynx” (n ⫽225) because the analysis required exclusive assignment to a specific anatomic site. Analyses excluded people enrolled in a health maintenance organization (HMO) (n ⫽3234) in the year before diagnosis with HNCR because the Medicare claims do not include any detailed health services data. Cases excluded for enrollment in an HMO did not differ from non-HMO enrolled cases in distributions of age, gender, or race/ethnicity. Cases with missing income and education (n ⫽51) were also excluded, resulting in a final sample of 11,312 cases.
Variable Descriptions Stage at diagnosis was classified as localized (historic stages in situ and local) or advanced (historic stages regional and distant). The summary system of “historic stage” was used in order for stage at diagnosis to remain consistent over the time span of the study. Historic stage at diagnosis is not completely comparable to the American Joint Committee on Cancer (AJCC) staging system because of AJCC changes in anatomic subsites and extent of disease definitions during the study time period. Histologic grade was the only variable available in this data set, other than cell type, which captured some of the biological attributes of the tumor; it was modeled as a dichotomous variable with the categories grade 1 or unknown, and ⱖgrade 2. Histologic grades “1” and “unknown” had similar bivariate relationships with stage at diagnosis, and were combined to simplify analysis and presentation. Highergrade scores indicate less histologic differentiation. Cases with a cancer diagnosis other than HNCA, occurring anytime prior to diagnosis with HNCA, were modeled as a dichotomous yes/no variable. Patient demographic variables gender and Hispanic ethnicity were modeled as dichotomous indicators. Race/ethnicity was modeled with white, black, and “other” categories. Age was modeled as a continuous variable. Marital status was modeled with the categories of currently married, divorced, never married, widowed, and unknown. Income and education were ecologic variables derived from an algorithm designed to minimize missing data. The algorithm searched for data from the following sources in priority order: 1990 census tract data, 2000 census tract data, 1990 ZIP code data, and lastly, 2000 ZIP code data. The 1990 census tract data was available for 97.6% of the study population. Income was assigned based on the median household income of the patient’s census tract/zip code; the resulting continuous variable was then divided into quartiles for analysis. Similarly,
American Journal of Preventive Medicine, Volume 27, Number 5
Table 1. Percentage of head and neck cancer patients by stage at diagnosis and selected characteristics (N ⫽ 11,312) Local stage (%) (n ⴝ 5585) Physician visits in year prior to cancer diagnosis None 06.8 1–5 18.3 6–10 20.6 ⱖ11 54.3 Anatomic site Lip 16.7 Oral cavity 36.9 Pharynx 05.3 Larynx 41.1 Number of ATC conditions 0 33.2 ⱖ1 66.8 Had prior cancer (other than head and neck) Yes 16.9 No 83.1 Histologic grade Grade 1 or unknown 52.2 Grade 2 or higher 47.8 Age at diagnosis (years) 65–74 52.1 ⱖ75 47.9 Ethnicity/race Hispanic 03.3 Non-Hispanic 96.7 White 91.1 Nonwhite 08.9 Gender Male 68.6 Female 31.4 Marital status Currently married 59.4 Not married/unknown 40.6 Incomea Above median 51.2 At or below median 48.8 Educationb Above median 52.6 At or below median 47.4
Advanced stage (%) (n ⴝ 5727) 11.7 22.3 19.2 46.8 03.0 44.7 25.7 26.6 37.3 62.7 14.2 85.8 29.7 70.2 54.8 45.2 04.2 95.8 86.5 13.5 66.2 33.8 50.5 49.5 48.9 51.1 47.8 52.2
Note: All column differences in proportions had p values ⱕ0.05. ATC, alcohol and tobacco–related comorbid conditions. a Determined by household income of a patient’s census/ZIP code area. b Determined by percentage of adults in a patient’s census/ZIP code area with 12 or more years of education.
education was defined by the percentage of people aged ⱖ25 years in a patient’s census tract/ZIP code with ⬍12 years of education. The resulting data were then divided into quartiles for analysis but collapsed to simplify presentation in Table 1. Physician visits in the year before a diagnosis with HNCA were modeled as counts. An alcohol and tobacco–related comorbidity (ATC) variable derived for the study is described in detail elsewhere.14 It is based on comorbid conditions associated with alcohol and tobacco use, the primary etiologic agents for HNCA. The ATC index consists of a simple count of 11 conditions present, and thus has a theoretical range of 0 to 11. Conditions were counted if listed on inpatient records during the 12 months before diagnosis with HNCA (excluding the
month of diagnosis), and at least once in physician claims, or in physician claims more than once in more than one 30-day period. This approach is similar to the one described in more detail by Klabunde et al.15 The ATC index was collapsed into a dichotomous variable to facilitate stratification of analyses and presentation of data. The sample distribution of ATC scores 0, 1, and ⱖ2 was 35%, 32%, and 33%, respectively. In statistical models of the relationship between stage at diagnosis and physician visits, the ATC variable functioned as an effect modifier with p values for interaction terms of ⬍0.01. All subsequent associations were therefore stratified by ATC status in order to display the role played by these comorbidities.
Statistics The Pearson chi-square test for proportions was used to develop p values where appropriate for comparing proportions. Logistic regression was used for adjusted odds ratios as measures of association. Odds ratios were determined to be statistically significant when 95% confidence intervals did not include the value 1.00. The STATA statistical package, version 8.0, was used for all analyses (STATA Corporation, College Station TX, 2003). All analyses were conducted in 2003 and 2004.
Results The distribution of cases by number of physician visits before diagnosis, anatomic site, and histologic grade varied greatly by stage at diagnosis. The distribution of cases by age, prior cancer, ATC, race/ethnicity, gender, marital status, income, and education varied only slightly to modestly by stage at diagnosis (Table 1). Among those with at least one physician visit in the year before HNCA diagnosis, cases with ATC had more physician visits than persons without ATC (data not shown). After stratifying by anatomic site, increasing numbers of physician visits were independently associated with reduced risk of advanced stage at diagnosis among persons with no ATC for all anatomic sites, but not statistically significant for the lip site (Table 2). The association was especially strong for the pharynx and larynx. Among cases with one or more ATC, physician visits displayed no association with stage at diagnosis with the exception of patients with laryngeal tumors where physician visits were associated with a larger reduction in risk of late stage at diagnosis compared to patients with no ATCs who had laryngeal tumors (Table 3). A moderate to poorly differentiated tumor increased the risk of advanced stage at diagnosis compared to well-differentiated and ungraded tumors among all anatomic sites, except for cases with both no ATC and pharyngeal cancers (Tables 2 and 3). A previous cancer diagnosis did not change risk of advanced stage at diagnosis. Am J Prev Med 2004;27(5)
375
Table 2. Adjusteda odds ratios for advanced stage at diagnosis by anatomic site among cases with no ATC Lip (n ⴝ 409) AOR
Oral cavity (n ⴝ 1669)
95% CI
AOR
95% CI
0.32 0.55 0.61 0.69 1.00
0.09–1.16 0.19–1.56 0.25–1.52 0.33–1.42 —
0.81 0.63* 0.65* 0.86 1.00
0.55–1.21 0.45–0.88* 0.47–0.90* 0.65–1.14 —
0.61 1.00
0.24–1.55 —
1.16 1.00
3.10* 1.00
1.76–5.48* —
1.58* 1.00
Pharynx (n ⴝ 640) AOR
Larynx (n ⴝ 1269)
95% CI
AOR
95% CI
0.29* 0.62 0.35* 0.31* 1.00
0.11–0.75* 0.25–1.52 0.15–0.81* 0.15–0.64* —
0.38* 0.41* 0.49* 0.81 1.00
0.22–0.66* 0.27–0.62* 0.34–0.70* 0.58–1.12 —
0.85–1.59 —
0.92 1.00
0.44–1.92 —
0.93 1.00
0.63–1.37 —
1.28–1.94* —
1.20 1.00
0.69–2.07 —
3.06* 1.00
2.38–3.95* —
b
Physician visits 20⫹ 11–20 6–10 1–5 0 Prior cancerc Yes No Histologic grade ⱖGrade 2 Grade 1/unknown
AOR, adjusted odds ratio; ATC, alcohol and tobacco–related comorbid condition; CI, confidence interval a All displayed odds ratios adjusted for gender, race/ethnicity, marital status, income, education, and age, in addition to the other covariates in the table. b Number of physician visits in year before diagnosis with head and neck cancer. c Had previous cancer (other than head and neck). *Significant at p ⱕ0.05 (bolded).
Overall, ⬎46% of persons with advanced stage tumors visited a physician ⱖ11 times in the year prior to their diagnosis with HNCA. The distribution of visits by ATC status differed, with 4.6 times more people with comorbidities having ⱖ11 visits compared to those without (Figure 1).
Discussion Potential missed opportunities for an earlier diagnosis of HNCA were found among the many contacts with the healthcare system that patients had prior to their diagnosis with HNCA. Over 93% of persons with localized tumors, and 88% of persons with advanced
tumors had at least one physician visit in the year prior to diagnosis with HNCA, and ⬎54% of persons with localized tumors and 46% of persons with advanced tumors had ⱖ11 (Table 1). Such contacts with the healthcare system are the basis for “opportunistic screening” as described by Prout et al.13 and others.16,17 Comorbidity was found to effect the association between physician visits and stage at diagnosis. Comorbid disease was found to be an independent predictor of survival among a similar population of HNCA patients, and some of the effects of comorbidity on survival were related to the association of comorbidity and stage at diagnosis.18
Table 3. Adjusteda odds ratios for advanced stage at diagnosis by anatomic site among cases with one or more ATC Oral cavity (n ⴝ 2951)
Lip (n ⴝ 691) AOR
95% CI
AOR
95% CI
Pharynx (n ⴝ 1127) AOR
Larynx (n ⴝ 2552)
95% CI
AOR
95% CI
b
Physician visits 20⫹ 11–20 6–10 1–5 0 Prior cancerc Yes No Histologic grade ⱖGrade 2 Grade 1/unknown
0.45 0.63 0.50 0.67 1.00
0.06–3.30 0.09–4.59 0.07–3.65 0.09–5.03 —
0.97 1.00 1.16 1.24 1.00
0.48–1.97 0.49–2.03 0.56–2.37 0.60–2.55 —
1.28 0.92 1.10 1.15 1.00
0.40–4.06 0.29–2.92 0.34–3.54 0.35–3.83 —
0.21* 0.23* 0.25* 0.41* 1.00*
0.09–0.46* 0.10–0.52* 0.11–0.57* 0.18–0.93* —
1.07 1.00
0.60–1.90 —
0.85 1.00
0.69–1.05 —
1.00 1.00
0.66–1.52 —
0.93 1.00
0.74–1.17 —
2.49* 1.00
1.59–3.90* —
2.02* 1.00
1.74–2.35* —
1.91* 1.00
1.36–2.69* —
2.57* 1.00
2.15–3.07* —
AOR, adjusted odds ratio; ATC, alcohol and tobacco–related comorbid condition; CI, 95% confidence interval. a All displayed odds ratios adjusted for gender, race/ethnicity, marital status, income, education, and age in addition to the other covariates in the table. b Number of physician visits in year before diagnosis with head and neck cancer. c Had previous cancer (other than head and neck). *Significant at p ⱕ 0.05 (bolded).
376
American Journal of Preventive Medicine, Volume 27, Number 5
Figure 1. Percentage distribution of persons with advanced stage at diagnosis by alcohol and tobacco–related comorbidity status and number of physician visits in the year prior to diagnosis (n ⫽5727).
Among cases with no ATC, the number of physician visits displayed the largest reduction in risk of advanced stage for laryngeal and pharyngeal tumors, and had the least impact on lip cancers. No dose–response relationship between visits and stage was observed, and not all strata displayed statistically significant associations, suggesting that some opportunities exist for improving the proportion diagnosed at early stages among people with no comorbidity (Table 2). The relationship between physician visits and stage at diagnosis among cases with one or more ATC resembled that of cases without ATC for lip cancers but not the other anatomic sites. In the presence of ATC, oral cavity and pharyngeal cancers no longer displayed any association between physician visits and stage, while the association among laryngeal cancer patients strengthened. It is possible that in the majority of physician visits recorded for persons with one or more ATC and who eventually developed oral cavity or pharyngeal tumors, the provider focused solely on the comorbid conditions themselves, and thus offered little opportunity for early diagnosis. In these cases, comorbid conditions may so dominate the care of the patient that fewer chances for an early diagnosis of HNCA exist. This pattern of care was present despite the fact that all the comorbid conditions in this analysis were related to alcohol and tobacco use. The possible exclusive physician focus on one adverse effect of exposure to alcohol and tobacco products may result in lost opportunities for an earlier diagnosis of additional adverse effects, such as HNCA. Interestingly, in the case of laryngeal cancer, the presence of an ATC apparently did not “distract” physicians, but possibly alerted them to risk of these cancers among their patients. One way in which laryngeal cancers may differ from oral cavity and pharyngeal cancers is in early symptoms. Laryngeal tumors often display an early symptom, hoarseness, which a patient brings to the attention of a provider.12 It is possible that instead of a busy provider having to look for something
beyond the immediate scope of the visit as would be the case for oral cavity and pharyngeal tumors, the hoarseness becomes a chief complaint that, especially in the context of an ATC, a physician then couples with the need for evaluation. All histologic types were included in the analyses. It was reasoned that although squamous cell carcinomas (SCC) were the most prevalent histologic type, ultimately the only chance for HNCA to be picked up early may lie with providers; and even if providers were only looking for SCC this process may uncover many nonSCC lesions. To confirm this inclusion decision, analyses using only SCC lesions were run with the results supporting the inclusion of all histologic types. This study exhibited a number of strengths. The use of a Medicare population allowed us to minimize any impact that access to care may have had on the findings. Few large population groups in U.S. society have similar access to health care. The sample population has been shown to be generally representative of all U.S. citizens of this age group.19 The relatively large sample sizes also allowed for stable estimates of effect. In addition, a measure of comorbidity was available for all the analyses, which was important for this type of study. A limitation of this study includes not having any information on dental visits. These providers would be expected to diagnose a large number of oral cavity and lip tumors. However, the impact of dentist diagnoses would only affect the estimated associations between physician visits and stage at diagnosis to the extent that dentist screening behavior differs from that of physicians. Given the very low rates of oral cancer screening reported by patients, this study limitation may be small, but future studies would be improved by including these data. About half of all HNCA occurs before age 65 years, and the study findings may not be representative of patterns present among younger cancer patients. Physician visits may to some extent also be a proxy for health-seeking behavior on the part of patients. Therefore, any findings that potential opportunities for an early diagnosis of HNCA exists in the many contacts that patients have with the healthcare system will clearly not apply to persons who avoid the system. People with extremely high levels of alcohol or tobacco abuse, who are at high risk of HNCA, may also display little health-seeking behavior. Few biological variables were available for the study. The strong and consistent independent association of histologic grade with stage at diagnosis hints that some tumors may proceed to an advanced stage due to biological factors and may not ultimately be amenable to the types of health services interventions examined in this study. Other limitations include the lack of a variable measuring the pattern of provider contacts in the Am J Prev Med 2004;27(5)
377
What This Study Adds . . . It was previously reported that individuals at high risk for head and neck cancers should be targeted for screening because early symptoms provide inadequate information for early diagnosis, and that considerable prior contact with primary healthcare providers existed among head and neck cancer patients. This study found that comorbidity was relevant for targeting high-risk individuals and numerous opportunities may exist for integrating their head and neck cancer screening into existing healthcare services.
months prior to a diagnosis of HNCA, identification of the type of medical providers involved in the visits, and the reason for the visit. Although many aspects remain to be studied, elderly HNCA patients did display numerous contacts with healthcare providers in the year prior to their cancer diagnosis, and the impact of these visits on stage at diagnosis varied by comorbidity and anatomic site. At a minimum, the potential exists for opportunistic screening during provider contacts to achieve early diagnoses of HNCA, especially among people with ATC-related comorbid disease. We are grateful to Marie Topor and Angela Fahey of Information Management Systems (Silver Spring, MD) for programming and database assistance. BCR was funded by the National Institutes Health (grant 5-K22-DE014341-02).
References 1. National Institute of Dental and Craniofacial Research, National Institutes of Health. Oral health in America: a report of the Surgeon General. Washington DC: U.S. Department of Health and Human Services, 2000.
378
2. American Cancer Society. Cancer facts and figures 2003. Available at: www.cancer.org/docroot/STT/stt_0.asp. Accessed January 5, 2004. 3. National Cancer Institute. Surveillance, Epidemiology, and End Results Program 9 Registries Public-Use Database (1973–1995). Rockville MD: National Cancer Institute, 1998. 4. Smart CR. Screening for cancer of the aerodigestive tract. Cancer 1993; 72(suppl 3):1061–5. 5. U.S. Department of Health and Human Services. Healthy people 2010. Washington DC: U.S. Department of Health and Human Services, 2000. 6. American Cancer Society. Cancer prevention and early detection, 2002. Available at: http://www.cancer.org/docroot/PED/content/PED_2_3X_ Early_Detection.asp. Accessed January 5, 2004. 7. Hawkins RJ, Wang EE, Leake JL. Preventive health care, 1999 update: prevention of oral cancer mortality. The Canadian Task Force on Preventive Health Care. J Can Dent Assoc 1999;65:617. 8. U.S. Preventive Services Task Force. Screening oral cancer 1996. Available at: http://www.ahcpr.gov/clinic/uspstf/uspsoral.htm. Accessed January 5, 2004. 9. Horowitz AM, Nourjah PA. Factors associated with having oral cancer examinations among U.S. adults 40 years of age or older. J Public Health Dent 1996;56:331–5. 10. Macek MD, Reid BC, Yellowitz JA. Oral cancer examinations among adults at high risk: findings from the 1998 National Health Interview Survey. J Public Health Dent 2003;63:119 –25. 11. Rothman K, Keller A. The effect of joint exposure to alcohol and tobacco on risk of cancer of the mouth and pharynx. J Chronic Dis 1972;25:711– 6. 12. Dolan RW, Vaughan CW, Fuleihan N. Symptoms in early head and neck cancer: an inadequate indicator. Otolaryngol Head Neck Surg 1998;119: 463–7. 13. Prout MN, Heeren TC, Barber CE, et al. Use of health services before diagnosis of head and neck cancer among Boston residents. Am J Prev Med 1990;6:77– 83. 14. Reid BC, Alberg AJ, Klassen AC, et al. A comparison of three comorbidity indexes in a head and neck cancer population. Oral Oncol 2002;38:187–94. 15. Klabunde CN, Warren JL, Legler JM. Assessing comorbidity using claims data: an overview. Med Care 2002;40(suppl 8):26 –35. 16. Dombi C, Voros-Balog T, Czegledy A, Hermann P, Vincze N, Banoczy J. Risk group assessment of oral precancer attached to x-ray lung-screening examinations. Community Dent Oral Epidemiol 2001;29:9 –13. 17. Lim K, Moles DR, Downer MC, Speight PM. Opportunistic screening for oral cancer and precancer in general dental practice: results of a demonstration study. Br Dent J 2003;194:497–502. 18. Reid BC, Alberg AJ, Klassen AC, et al. Comorbidity and survival of elderly head and neck carcinoma patients. Cancer 2001;92:2109 –16. 19. Warren JL, Klabunde CN, Schrag D, Bach PB, Riley GF. Overview of the SEER-Medicare data: content, research applications, and generalizability to the United States elderly population. Med Care 2002;40(suppl 8):3–18.
American Journal of Preventive Medicine, Volume 27, Number 5