Disabilities are Risk Factors for Late Stage or Poor Prognosis Cancers

Presenter Name: Donald Austin, MD, MPH

Institution: Department of Public Health and Preventive Medicine, Oregon Health & Science University

Primary Research Interests: Cancer epidemiology, quality of care

INTRODUCTION

Anecdotal stories suggest that people with disabilities receive different screening levels for cancer than do the general population. In addition, the similar stories suggest that smoking prevalence may be elevated among groups of people with certain disabilities. If these suspicions are accurate, it might require a policy change in how screening and prevention programs are planned and implemented, so as to better serve the needs of this population.

Because a major portion of people with marked disabilities are included in Oregon’s Medicaid population, we used that population to determine if people with disabilities in certain domains experienced cancer outcomes that were suggestive of being screened less or of having higher smoking prevalence. In cooperation with officials in the Oregon Medical Assistance Programs (OMAP), the state Medicaid office, we sought to determine whether impartial solid evidence existed that would validate the anecdotal suggestions.

RESEARCH OBJECTIVE/RESEARCH QUESTION

To research the possibility of less cancer screening and higher smoking prevalence among groups with disabilities of certain types, we addressed the following research questions:

  1. Do Medicaid clients with certain disabilities have higher rates of smoking related cancer than Medicaid clients without known disabilities?
  2. Are Medicaid clients with certain disabilities and who develop cancer of the female breast, cervix, or colorectum, diagnosed at a more advanced stage of disease than Medicaid clients without known disabilities?

If our analysis of the data obtained to answer these questions resulted in significant and convincing evidence that the answer to either or both of these research questions was "Yes," we then proposed to develop suggested policy changes that would address the issues. Because there is a spectrum of possible reasons why an affirmative answer to the research questions might occur, we planned a review of a sample of medical records of representatives of the groups experiencing the undesirable outcomes, to obtain information about the possible etiology and resolution of the problem.

METHODS

We obtained an unduplicated file of all Oregon Medicaid clients for the years 1994-98. There were over a million individuals on the file, with an identifying number. For all OMAP data, each individual is identified by a unique number termed the "prime" number. In addition, we obtained a separate name and address file for these individuals, with initial and all subsequent addresses, and with information on exact dates starting and stopping coverage. This file had over 15 million entries. A third eligibility file with about seven million entries contained the eligibility basis for each member. By linking these three files on the prime number, we were able to create a file with named individuals and their demographic characteristics, with a usual address, and with the program for which they qualified for eligibility. In addition, for a subset of the individuals in the eligibility file, some were receiving aid based on a disability from a sister agency, then named the Senior and Disabled Services Division (SDSD). Again, using the prime number, we linked the file of SDSD recipients to the OMAP file appending information on the evaluation of the functional disability done by SDSD caseworkers, and creating our initial research file.

Lastly, using the names, addresses, dates of birth, sex, and if available, the Social Security Numbers for individuals on the initial research file, we linked to the Oregon State Cancer Registry (OSCaR) for the years 1996-1998, obtaining any diagnosed cancers, along with date of diagnosis, site (type) of cancer, and stage at the time of diagnosis. This file containing individuals, eligibility codes and dates, caseworker disability assessments, and cancer data constituted our final research file. The final research file included 805,608 individuals for the years 1996-98. Mental retardation information was also available on a separate sister agency file, linked by prime number to the final research file, but for most analyses the numbers were too small for meaningful analysis.

Even though we had cancer data for three years, we first analyzed only two years, since we were searching for hypothesized evidence and were conducting multiple comparisons. We considered that statistically significant or nearly significant results would be specific data hypotheses to be independently tested in the third year of data. Thus the possibility of chance associations is minimized in the analysis and resultant findings may be considered as having been tested in an independent body of data using the exact procedures as produced the findings in the 2-year body of data.

We used two different measures of disability. The most crude was the Program Eligibility Recording Code (PERC), which generally allowed for blindness, and “disabilities” to be distinguished from old age or individuals and families below federal poverty level.. Those designated as having a disability by this code (PERC-disabled) constituted 8.9% individuals on the final research file.

The second measure of disability was generated by us, based on the SDSD caseworker functional assessment results that were initially and periodically done to determine the level of assistance needed in several different domains. The recorded functional assessment was an 18 point graduated measure on a number of different combined criteria, with 18 signifying needing no assistance and 1 signifying being totally dependent for all assessed functions. Functions assessed included mobility (mobility and transferring), cognition (orientation, adaptation, memory, awareness, judgment). The scoring allowed separation of most levels of these functions, and for this research we focused on mobility and cognition. We used the scale to create a three-category scale for mobility (no assistance needed, some assistance needed, dependent) for both assessed aspects of mobility. This allowed two three-point scales to be combined into a five-point score ranging from 2 to 6. We considered those with a score of 2 to be severely disabled, those with scores of 3-5 to be moderately disabled, and those with a score of 6 to be independent. Similarly, for cognition we assigned three levels of impairment to each of the five measures (no evidence of impairment, occasional to frequent impairment, consistent impairment). Assigning one to three points to these three levels, and combining the five measures (unweighted) created a 10-point scale ranging from 5 to 15. Those scoring 13-15 were considered without significant impairment, 10-12 were considered moderately impaired, and those scoring 9-5 points were classified as severely cognitively impaired. Those individuals designated as having a moderate or severe disability by this system constituted 7% of the final research file.

For our first analysis, we computed cancer incidence rates. We compiled the number of days that each individual was eligible for Medicaid in the period being analyzed (initially 1996-97) providing that the eligible days preceded the cancer diagnosis. These were summed and categorized by age group, creating an exact denominator of person-time. For the numerator, we assigned each diagnosis to the appropriate age and sex group and computed the age-specific, sex-specific, average annual cancer rates. We could also allocate individuals and their person-time contribution according to disability subgroups that we created and compute incidence rates for subgroups of interest. We computed age-adjusted incidence rates for "smoking-related cancers" which were comprised of the sites of lung, bronchus, trachea, larynx, hypo pharynx, and oral cavity including tongue. For groups categorized as disabled by our different criteria, we calculated the rate ratios (RR) for these cancers, as compared to the Medicaid group not known to have any disability. For our second analysis, we identified breast, cervix and colorectal cancers for which screening and early diagnosis is the control strategy. We determined the proportion of cases of these "screen able" cancers diagnosed with metastases at the time of diagnosis. For males, only colorectal cancer data were examined. We computed the odds ratios (OR) for the different disability groups, again as compared to the group not known to have any disabilities by any administrative data. We also conducted a multivariate analysis, in which we adjusted the OR for each disability group by the other disability and for age.

Lastly, for those specific analyses demonstrating an association between a disability measure and increased risk of the cancer outcome of interest, we conducted an identical analysis of the 1998 data, and if both results were close to the same (even if the smaller number was not statistically significant) we considered the result confirmed, and to obtain a better estimate of the risk, we combined all three years of observation.

RESULTS

The risk of smoking-related cancers was elevated in both disability groups by both the PERC (eligibility program) measure and the SDSD measure (based on the functional assessments), though the SDSD measure provided finer discrimination and higher risk estimates. Generally, women had higher risks than men and those with either cognitive or mobility impairments had double the risk of those with no known disabilities.

Generally, the risk of being diagnosed with advanced (ie, metastatic) disease was greater for people with disabilities, and the SDSD measure was superior to the PERC measure for identifying elevated risk. For colorectal cancer, the risk of a late diagnosis was significantly greater for women with cognitive disabilities (OR = 3.3; p < 0.05) when controlling for mobility disabilities. When limiting our analysis to individuals who were eligible for at least 95% of the 36 month of the study and comparing the two means of measuring disability using administrative data, the SDSD measures gave higher age-adjusted risk estimates for all screen able cancers combined in females, and the two disability groups had equal risk estimates, though the two groups had many members who were the same.

CONCLUSIONS

People in Oregon’s Medicaid population with disabilities in mobility or cognition have higher risk of developing smoking related cancers and of being diagnosed at a later stage for cancers for which screening and early diagnosis is the control strategy.

It is possible to use administrative data to classify individuals into functional disability groups and those measures provide sufficient discriminatory power to research other morbidities and outcomes that may be associated with the functional disability categories.

IMPLICATIONS

Research:

  1. Administrative data can provide disability categorization adequate for research.
  2. People with disabilities are at risk for late cancer diagnoses, with poor prognoses.
  3. People with certain disabilities are at elevated risk for smoking-related cancers.

What implications do your research or training project findings have for changes in concepts of health and disability?

New strategies for cancer prevention and risk factor intervention, and for early diagnosis and treatment of cancer may necessary for protecting people with disabilities.