Impact of Cognitive Impairment on Screening Mammography
Use in Older US Women
Kala M. Mehta, DSc, MPH, Kathy Z. Fung, MS, Christine E. Kistler, MD, Anna Chang, MD, and Louise C. Walter, MD
Screening mammography guidelines suggest that women with a life expectancy less than 4 to
5 years at the time of screening are unlikely to benefit from breast cancer screening and, thus, should not be screened.1–3 Although some cancer screening guidelines specify upper-age cutoffs for stopping screening as a surrogate for life expectancy (e.g., prostate-specific antigen screening guidelines suggest stopping at age 75 years),4 we do not know of any guidelines that specify the types of comorbidity that would preclude screening. This is despite the fact that certain comorbid conditions, such as dementia, are stronger predictors of life expectancy than age.5 Specifically, patients with dementia generally live less than 5 years6–12 and therefore are unlikely to benefit from screeningmammography.
In addition, having dementia or severe cognitive impairment increases the likelihood that elderly women will experience harm from screening mammography (e.g., more psychological distress from false-positive results because of the inability to understand screening procedures, and more complications from the treatment of clinically insignificant disease).13,14
Moreover, screeningmammography can distract care away from more pressing medical problems arising from either the cognitive impairment itself or from other comorbid conditions. However, it is unknown how often these women with severe cognitive impairment in the United States are undergoing screening mammography.
A few prior studies have examined screening mammography rates in women with cognitive impairment; however, they relied on self-report of screening mammography, which is likely to be inaccurate among women with cognitive impairment15–17 or older studies limited to a local geographic area.17,18 To our knowledge, there have not been any recent national studies that have used objective measures, such as
Medicare claims, to document the actual mammography rates in older women with severe cognitive impairment. Such data are needed to determine current practice patterns and to identify whether cognitive status appropriately factors into screening mammography decisions.
Therefore, we conducted a study to document the actual rates of screening mammography in a US-representative sample of older women stratified according to their cognitive status. We used Medicare claims data linked to the Health and Retirement Study (HRS) to define rates of screening mammography (based on claims data) for women with differing levels of cognitive impairment. We also calculated survival according to level of cognitive impairment to validate that women with severe cognitive impairment defined by a standardized instrument have a median survival less than 5 years and are therefore unlikely to benefit from screening mammography.
METHODS
The HRS is a US-representative longitudinal study of health and wealth of noninstitutionalized adults aged 50 years and older with biennial data collection starting in
1992.19 The primary sample was obtained through a multistage, clustered area probability frame screening of household units. For about half of the respondents born before 1913, samples were additionally drawn from Medicare enrollment lists. Data were collected primarily through telephone interviews, and the overall response rate was 81%.20 We merged survey data from the HRS and claims data from the
Center for Medicare Services to examine actual screening mammography rates of older women in the period 2 years prior to their 2002 HRS interview. Women eligible for our study included the
4312 women who were aged 70 years or older at the 2002 wave of theHRS study. We excluded 939 women (22%) enrolled in
Medicare managed care during the 2 years prior to their HRS interview because they lacked the opportunity to file Medicare
Objectives. We evaluated mammography rates for cognitively impaired women in the context of their life expectancies, given that guidelines do not recommend screening mammography in women with limited life expectancies because harms outweigh benefits.
Methods. We evaluated Medicare claims for women aged 70 years or older from the 2002 wave of the Health and Retirement Study to determine which women had screening mammography. We calculated population-based estimates of 2-year screening mammography prevalence and 4-year survival by cognitive status and age.
Results. Women with severe cognitive impairment had lower rates of mammography (18%) compared with women with normal cognition (45%).
Nationally, an estimated 120000 screening mammograms were performed among women with severe cognitive impairment despite this group’s median survival of 3.3 years (95% confidence interval=2.8, 3.7). Cognitively impaired women who had high net worth and were married had screening rates approaching 50%.
Conclusions. Although severe cognitive impairment is associated with lower screening mammography rates, certain subgroups with cognitive impairment are often screened despite lack of probable benefit. Given the limited life expectancy of women with severe cognitive impairment, guidelines should explicitly recommend against screening these women. (Am J Public Health.
2010;100:1917–1923. doi:10.2105/AJPH.2008.158485)
RESEARCH AND PRACTICE
October 2010, Vol 100, No. 10 | American Journal of Public Health Mehta et al. | Peer Reviewed | Research and Practice | 1917 claims. In addition, women had to be eligible for mammography screening. Therefore, we excluded 706 women because they had a history of breast cancer (n=223; 5%) or a breast neoplasm (n=483; 11%). The identification of a history of breast cancer or breast neoplasm was based on Medicare claims during 1998 or 1999, or if the women reported a history of breast cancer during their 2000 HRS interview. We also used claims data to exclude 287 women (6%) whose first mammogram during the study interval was performed for nonscreening purposes (e.g., bilateral or unilateral diagnostic mammogram or breast mass within 1 year prior to the index mammogram). Last, we excluded 249 women (6%) who were missing cognitive impairment information.
This left a final screen-eligible cohort for analysis of 2131 women.
Primary Predictor
An overall cognitive score was calculated for each participant based on a 35-point cognitive instrument developed for HRS.21,22 The scale includes memory (20 points), calculation and attention (7 points), and orientation and naming questions (8 points). Memory was tested by asking the women to recall 10 nouns immediately and after a 5-minute delay, with1point per recalled word for a total of 20 points. Calculation and attention were assessed with the serial 7s test, in which participants start with 100 and consecutively subtract 7 five times, and by counting backward from 20 to 10 and 86 to 76.
Orientation and naming included the following items: naming the month, day, year, day of the week; the object used to cut paper (scissors); the plant that lives in the desert (cactus); and the president and vice president of the United States.
An overall score of 20 to 35 was considered normal cognitive function, 11 to 19 was defined as mild-to-moderate cognitive impairment, and a score of 10 or below was defined as severe cognitive impairment, based on prior work with this measure.23,24 In addition, 273 participants used a proxy informant to complete the HRS interview. For these participants, we defined severe cognitive impairment on the basis of the proxy’s report that the respondent’s memory or ability to make judgments and decisions was ‘‘fair or poor.’’ These definitions for severe cognitive impairment have been used to define dementia in prior studies that have used HRS data.25–27
Outcome Variables
Because self-report of mammography is unreliable in women with cognitive impairment, our primary outcome was receipt of screening mammography within the previous 2 years based on Medicare claims (Current Procedure
Terminology28 code 76092; International Classification of Diseases, Ninth Revision [ICD-9]29 codes V76.11, V76.12; Healthcare Common
Procedure Coding System30,31 codes G0202
G0203; or revenue center code32 403).33 To validate that our cognitive score identified women who had limited survival and would be unlikely to benefit from mammography, we calculated 4-year survival estimates and life expectancy estimates for age and cognitive status groups by using vital status information from the
National Death Index.34
TABLE 1—Characteristics of US Women Aged 70 Years or Older, Health and Retirement
Study, 2002
Characteristic
Cognitive Statusa
Normal Cognitive
Status (n = 1167), %
Mild-to-Moderate
Cognitive Impairment
(n = 609), %
Severe Cognitive
Impairment
(n = 355), %
Total
(N = 2131), %
Age, y
70–74 39 19 15 31
75–79 30 25 25 28
80–84 22 28 26 24
‡85 9 28 38 17
Race/ethnicity
Non-Hispanic White 93 79 66 86
Non-Hispanic Black 4 12 19 8
Latino 2 7 11 4
Others 1 2 4 2
Education, > high school 81 54 36 69
Married 38 25 24 33
Total net worth, $
> 100 000 66 44 31 56
10 000–100 000 24 33 24 27
< 10 000 10 24 45 17
Comorbid illnesses
High blood pressure 61 61 63 61
Diabetes 14 19 20 16
Lung disease 9 8 8 9
Any psychiatric illness 12 15 28 15
Arthritis 72 71 72 72
Stroke 8 13 29 12
Heart disease 27 32 41 30
IADL dependenceb 21 39 79 32
ADL dependencec 4 11 51 11
Notes. ADL = activities of daily living; IADL = instrumental activities of daily living. Women with normal cognition were
62% of the total sample; women with mild-to-moderate cognitive impairment, 29%; women with severe cognitive impairment, 9%. ac2 tests and t tests were used to compare the 3 cognitive status groups. All characteristics were different between women with normal cognitive status and those with cognitive impairment. bNeeding help from another person to perform 1 or more of 6 IADL: using a map, managing money, taking medication, grocery shopping, preparing hot meals, and using the telephone. cNeeding help from another person to perform 1 or more of 5 ADL: bathing or showering, getting in and out of bed, dressing, eating, and toileting.
RESEARCH AND PRACTICE
1918 | Research and Practice | Peer Reviewed | Mehta et al. American Journal of Public Health | October 2010, Vol 100, No. 10
Covariates
We also assessed other factors known to influence the use of screening mammography, including demographic characteristics such as age, education, marital status, and total net worth. Total net worth was calculated as the sum of all assets and income, as this is a better indicator of socioeconomic status in the elderly compared with income alone. We categorized net worth into low ($100000) based on previous work.35 Race/ ethnicity was assessed with standard questions derived from the US Census. We categorized individuals into the following 4 mutually exclusive groups: White (non-Hispanic White), African
American (non-Hispanic African American),
Hispanic (Latino), and persons of other race/ ethnicity. Self-report of medical conditions included hypertension, diabetes, lung disease, psychiatric illness, arthritis, stroke, and heart disease.
We also assessed dependence in 6 instrumental activities of daily living (IADL) by calculating a summary score ranging from 0 to 6 points. One point each was given if a woman needed help from another person to perform the following: using a map, managing money, taking medications, grocery shopping, preparing hot meals, or using a telephone. Similarly, we assessed dependence in 5 basic activities of daily living (ADL) and calculated a summary score ranging from 0 to 5 points. One point each was given for needing help with the following: bathing, bed transfers, dressing, eating, or toileting.
Statistical Analysis
We summarized baseline characteristics for our cohort according to their cognitive status.
We then calculated the rate of screening mammography according to cognitive status.
We stratified the cohort on the basis of age, level of education, race/ethnicity, and net worth. To assess the effect of these covariates on the rates of screening mammography, we calculated logistic regression models with screening mammography as the outcome and cognitive status as the primary predictor. Adjustments to this logistic model were added, including demographics and comorbidity.
We then calculated 4-year survival in all screen-eligible respondents according to their cognitive status by using the method of Kaplan
Meier. We also calculated survival for 5-year age and cognitive status groups. We calculated a Cox proportional hazard regression model of the survival estimates to estimate life expectancy for each of the age and cognitive status groups. For the severe cognitive impairment group, median life expectancy was reported from actual data, not from the model. We also determined screening mammography rates and mortality according to cognitive impairment and ADL dependence. For all analyses, we applied study weights to account for the complex study design and we used these weights to calculate population estimates of screening mammography use overall and in specific strata. We conducted all analyses with SAS version 9.1.3 (SAS Institute Inc, Cary, NC).
RESULTS
Characteristics of the 2131 women in our cohort are presented in Table 1. The median age was 77 years (interquartile range=73–
80 years). The majority of women were White and unmarried. Seventy-two percent of the sample had normal cognitive status, 29% had mild-to-moderate cognitive impairment, and
9% had severe cognitive impairment.
Women with cognitive impairment were more likely to be older, non-White race/ethnicity
(African American or Latino), had lower levels of formal education, and had lower net worth. Women with cognitive impairment also had higher levels of comorbid conditions (e.g., hypertension, any psychiatric illness, diabetes, stroke, and heart disease) and higher levels of functional impairment. For example, 51% of women with severe cognitive impairment were dependent in 1 or more ADL compared with
4% of women with normal cognitive status
(P