evaluation and implementation of state comprehensive cancer control plans: evolving lessons apha...

28
Evaluation and Implementation of State Comprehensive Cancer Control Plans: Evolving Lessons APHA 2005 Annual Meeting Epidemiology Section Session 3187.0 12:30–2:00 PM Monday, December 12, 2005

Upload: dwight-hall

Post on 13-Dec-2015

219 views

Category:

Documents


0 download

TRANSCRIPT

Evaluation and Implementation of State Comprehensive Cancer Control Plans: Evolving Lessons

APHA 2005 Annual MeetingEpidemiology Section

Session 3187.0

12:30–2:00 PMMonday, December 12, 2005

Utilizing research and data: Use of epidemiologic data in community assessments

Jung Y. Kim, MPH Department of Preventive Medicine,UMDNJ-New Jersey Medical School

3

Co-authors of this presentation include:

Margaret L. Knight, RN, MEdDaniel M. Rosenblum, PhDJudith B. Klotz, DrPHStanley H. Weiss, MD

4

Standardized Evaluation Goal: To identify greatest cancer burden and

health disparities, in order to propose local and statewide priorities and to assess progress toward reducing cancer burden

Standardized methods and time periods for cancer data are critical to establish common baseline data and enable valid comparisons

To be discussed: Data utilized for this assessment and data sources Common errors in data use and interpretation

5

Data and Sources Demographics and health status indicators Cancer incidence and stage at diagnosis Cancer mortality Healthy New Jersey 2010 objectives Staging of cancer Prevalence Estimates of medically underserved

populations

6

Demographics Source: U.S. Census Bureau Data files generated using

American FactFinder, www.factfinder.census.gov/

Characterize population’s age, gender, race, ethnicity, languages spoken at home, ability to speak English

Identify municipalities with populations of low income, low educational attainment, high poverty, high unemployment

7

DemographicsRace and ethnicity Race and ethnicity are separate categories, and are not

mutually exclusive.Hispanics/Latinos may be of any race.

While Census data allows multiple races and combined race/ethnicity, other data sources may not use similar categories.

Thus, the ability to compare race/ethnicity categories depends upon the data, i.e., whether data exists on

white Hispanics, black Hispanics, as well as white non-Hispanics, black non-Hispanics, etc.

– enabling direct comparisons of either race or Hispanic ethnicity

8

Demographics

Health Status Indicators Source:

Center for Health Statistics,NJ Department of Health and Senior Services, www.state.nj.us/health/chs/index.html

CDC’s BRFSS, www.cdc.gov/brfss/index.htm

Birth rate, death rate, percentage of low birth weight babies, infant mortality rate, estimated obese and overweight populations, estimated population who smoke

9

Cancer Incidence Source: New Jersey State Cancer Registry,

Cancer Epidemiology Services Division of the NJ Department of Health & Senior Services, by special request

Counts, age-adjusted rates, and stage at diagnosis provided by Gender Age group (15-39, 40-49, 50-64, 65-74, 75+) Race (black, white) and Hispanic ethnicity

10

Cancer MortalitySource: Cancer P.L.A.N.E.T.,

State Cancer Profiles, http://statecancerprofiles.cancer.gov/

User specifies data parameters Geographic region (state or county level) Type of cancer Race (White, Black, American Indian/

Alaskan Native, Asian/Pacific Islander, Hispanic)

Gender

11

Cancer Mortality

Special request to NCI for counts and age-adjusted rates by county Gender Age group (0-49, 50+ and 0-64, 65+) Race (black, white) and Hispanic ethnicity

12

Age Adjustment Rates of cancer incidence and mortality among

different age groups are vastly different Age adjustment allows comparison among

various groups that may have different age structures

Eliminates the effect of the underlying age distribution of the population

Rates age-adjusted to a standard population by 5-year age groups (19 groups)

13

Age Adjustment Since 1940, the age distribution of the U.S.

population has dramatically changed

1940 1970 2000

Median age (years) 29.0 28.1 35.3% of Total population, selected groups

< 5 years 8.0 8.4 6.835 to 44 years 13.9 11.4 16.065 years + 6.8 9.9 12.4

Age group with highest % of population 15-24 5-14 35-44

Source: U.S. Census Bureau, Census 2000 Special Reports, Series CNSR-4; Demographic Trends in the 20th Century, 2002; Statistical Abstract of the United States: 2003.

14

Age Adjustment Prior to data year 1999, reports of incidence

and mortality rates were age-adjusted to earlier populations, commonly 1970 U.S. standard population for incidence 1940 U.S. standard population for mortality

Beginning with the 1999 reporting year, the U.S. DHHS required health data to be age-adjusted using the 2000 U.S. standard population.

Source: Martin, RM. Age standardization of death rates in New Jersey: Implications of a change in the standard population. NJDHSS, Center for Health Statistics, 2003.

15

Disease rates that vary by age can be affected enormously by a change in standardizationAge-Adjusted Death Pop. Standard PercentRate in NJ (per 100,000) 1940 2000 Change

All causes 460.3 861.4 87%Cancer, all sites 127.9 212.5 66%

These differences in rates are purely statistical and are due to the aging of the U.S. population, since the elderly are given greater weight when using the 2000 population standard.

Source: Martin, RM. Age standardization of death rates in New Jersey: Implications of a change in the standard population. NJDHSS, Center for Health Statistics, 2003.

Age Adjustment

16

Healthy New Jersey 2010 Incidence and mortality target rates in Healthy

New Jersey 2010 (HNJ2010) were generated prior to the release of the NJ-CCCP prior to release of the 2000 population standard*

These HNJ2010 rates should not be compared to rates age-adjusted using the 2000 standard population

* In May 2005, NJDHSS published updated baselines, targets, and preferred endpoints for cancer incidence and mortality objectives using the 2000 standard population.

17

The 1996–2000 breast cancer mortality rate for all females in New Jersey was 31.3 per 100,000

Recalculated As published using 2000 in HNJ2010 standard pop.*

1998 Baseline 24.7 31.2Target rate 17.0 21.5Percent reduction 31% 31%

* Target recalculated to achieve equivalent percent reduction

(All rates per 100,000)

Healthy New Jersey 2010

18

Staging of Cancer Stage is determined just once for cancer

registry data – at the time of diagnosis Categories in major staging scheme: in situ

(non-invasive, only reported for some cancers), localized, regional, distant

Unstaged cancer – insufficient information to classify; no conclusion can be made about severity of cancer

Comparisons of the stage distributions between populations are important, but potentially problematic due to the unstaged cases

19

Staging of Cancer Proportion of unstaged cancers varies by

region and cancer site, typically between 10%–20%

Although elimination of unstaged cases and recomputation of percentages might appear straightforward, a complete analysis should include the reasons for variation in the unstaged proportion, which are not typically available

20

PrevalenceOne measure of the burden of a

diseaseAt a given point in time, how many

people have the diseaseTotal cancer prevalence at the county

level was not availableGiven its importance, a method for

estimating prevalence was developed

21

Estimates of MUAs

Medically Underserved Areas (MUAs) Source: NCI’s Cancer Information Service Customized Consumer Health Profile maps

and data provided for each county, identifying geographic areas of potentially medically underserved populations

Consumer marketing profiles developed using demographics, information on health behavior from various health and consumer surveys

22

Some Common Errors in Data Use and Interpretation

1. Differences in rates are often over-emphasized or sensationalized Example: Breast cancer incidence rate County A = 144.2 per 100,000 State = 138.5 per 100,000

Although the rate for County A was 5th highest in the state, it was only 4% higher than the overall rate for the state

This minor difference, in our opinion, should not be emphasized because it is insufficient to justify a major change in policy or funding for that county

23

2. Differences in gender-specific rates overlooked

Example: Total cancer incidence rateCounty B had a combined rate (male+female) that ranked 7th highest (worst) rate of 21 counties in the state, but Rate among females: 2nd highest Rate among males: 20th highest (2nd lowest)

Some Common Errors in Data Use and Interpretation

24

3. Neglecting sample size when comparing distribution at stage of diagnosis and failing to perform appropriate statistical calculations

Example: Oral cancer cases (for which staging information was available) diagnosed at the distant stage In County C, 13% among black females

7% among white females

13% represents 2 out of just 15 staged cases Too few cases to base conclusion on percentages alone Using Fisher’s Exact Test, this difference was not

statistically significant (p=0.605; 95% CI, 0.34-13.43)

Some Common Errors in Data Use and Interpretation

25

4. Use of single-year data at the county levelExample: Total cancer mortalityCounty D reported that it had the highest combined mortality rate in the state. The cited rate was valid for that year (2000), but One of the smallest counties in the state Since small numbers can vary substantially year to

year, trends for many chronic diseases are best based on statistics such as five-year averages

Five-year averages showed no significant disparity between that county and the rest of the state (for both 1996–2000 and 1998–2002 periods)

Some Common Errors in Data Use and Interpretation

26

5. Comparing rates for race to Hispanic ethnicity

Example: Cervical cancer incidence

County E reported the rate for Hispanic women was higher than that for white women. New Jersey data for Hispanic women include some

women who are white, since race and ethnicity are not mutually exclusive

The rate for Hispanic women could be compared to the rate for non-Hispanic women

If the available rates are age-adjusted, comparison to all women is only valid when the proportion of Hispanics in the relevant age strata of the total population is low

Some Common Errors in Data Use and Interpretation

27

6. Comparison of rates based on different time periods

7. Use of hospital discharge data as source for a county’s cancer burden Not a valid approach – counts for these data are

based on discharges, not patients Thus, patients with multiple hospital stays are

counted multiple times Many cancer patients do undergo multiple

hospitalizations

Some Common Errors in Data Use and Interpretation

28

Summary Accurate, up-to-date assessment of community

resources and identification of the community’s specific cancer burden and needs are needed to influence public support, funding, and policy

Understanding common mistakes such as those in the examples above may help guide the oversight of capacity and needs assessments, particularly the appropriate use and interpretation of epidemiologic data, which are essential to successful implementation at local and state levels