BMI, Diabetes, and Complications Among US Older Adults
BMI, Diabetes, and Complications Among US Older Adults
Using a nationally representative survey of older adults from the Medicare Current Beneficiary Survey (MCBS), 1991–2010, this study examined the association between excess weight and time to first diagnosis of type 2 diabetes mellitus and its complications and used data from the MCBS merged by a unique identifier with Medicare claims data. The MCBS is a continuous longitudinal survey of Medicare beneficiaries sponsored by the Centers for Medicare and Medicaid Services. The survey is designed to be nationally representative of the Medicare beneficiary population as a whole. Data are collected on access to and utilization of health services, healthcare expenditures, sources of payment for healthcare-related services, health status and functioning, quality of life, and demographic and socioeconomic characteristics. The method for accounting for sample attrition changed slightly during the course of the survey. Before 1994, MCBS would supplement each panel once each year with newly chosen beneficiaries. In 1994, the MCBS switched from a purely longitudinal to a rotating panel design. Since then, each year of MCBS data contains four overlapping sample panels. Each year, one panel is rotated out of the MCBS and replaced with a newly selected one; therefore, a single panel can provide up to 4 years of information. In a typical year, the MCBS contains data on 14,000 to 17,000 beneficiaries.
Medicare claims data came from the inpatient, outpatient, Medicare Part B, and durable medical equipment claims provided by the Centers for Medicare and Medicaid Services. Claims data were available for 1991–2010. We used all of the claims data available for a respondent, even if the data fell outside the MCBS survey window. For example, if the respondent was first interviewed in 2001 and was 77 years old in that year, we used all of the available claims data from 1991 to 2000 for the look-back period to determine whether the person had been diagnosed as having diabetes mellitus before the first MCBS (baseline) interview and from 2001 to 2010 or from a departure from Medicare for any reason for the follow-up period. The mean number of years of the look-back period was 3.96 (standard deviation [SD] 3.56) and 6.68 years (SD 4.05) for the follow-up period. The hazard model accounted for right censoring of the data. A condition was counted as a diabetes mellitus complication only if the person had a prior diagnosis of diabetes mellitus recorded on a claim.
Beneficiaries in our analysis sample were born between 1926 and 1944, were not enrolled in a Medicare Advantage plan, and did not have a diabetes mellitus diagnosis at the time of their first (baseline) MCBS interview. Data on age, race, sex, educational attainment, smoking status, income strata, and BMI were drawn from the MCBS Cost and Use and Access to Care files. Beneficiaries with incomplete information were dropped. After exclusions, our sample size consisted of 14,657 beneficiaries. Coding of diabetes mellitus diagnosis and its complications was based on International Classification of Diseases, Ninth Revision Clinical Modification and Current Procedural Terminology, 4th Edition, codes.
We identified the time in 3-month intervals from an individual's first interview to the first diabetes mellitus diagnosis and to each major complication category (Table 1). The analysis focused on common cardiovascular, cerebrovascular, renal, ocular, and lower extremity complications. Risks were calculated both for the category as a whole (eg, any cardiovascular complication) and for specific conditions within each category (eg, AMI).
We calculated the BMI of individuals at the first MCBS interview (baseline). We classified them into six BMI categories based on criteria specified by the World Health Organization: BMI ≤18.50 (underweight); BMI between 18.50 and 24.99 (normal weight); BMI between 25.00 and 27.49 (overweight); BMI between 27.50 and 29.99 (preobese); BMI between 30.00 and 39.99 (obese); BMI ≥40.00 (morbidly obese). Normal-weight individuals were used as a reference group. BMI was calculated as weight divided by the square of height, based on the self-reported weight and height measures from MCBS. Other control variables included age, income, race, educational attainment (in years), smoking status, hypertension, and the year of the baseline interview to account for secular trends in practice and technology use. Results on explanatory variables other than BMI categories above normal are not reported to conserve space. We conducted separate analyses for men and women to measure sex-specific differences in time to diagnosis by sex. We estimated the association between BMI measured at baseline on time to first diagnosis of diabetes mellitus and to specific complications by calculating hazard ratios using Cox proportional hazards implemented by the stcox procedure of STATA 12.0 (StataCorp, College Station, TX). The hazard ratios represent the probability of being diagnosed as having diabetes mellitus and its complications relative to beneficiaries in the baseline group (individuals of normal weight). To assess the sensitivity of the results to the choice of statistical method and to compare hazard ratios with odds ratios, we also estimated logit models (results not shown).
Methods
Using a nationally representative survey of older adults from the Medicare Current Beneficiary Survey (MCBS), 1991–2010, this study examined the association between excess weight and time to first diagnosis of type 2 diabetes mellitus and its complications and used data from the MCBS merged by a unique identifier with Medicare claims data. The MCBS is a continuous longitudinal survey of Medicare beneficiaries sponsored by the Centers for Medicare and Medicaid Services. The survey is designed to be nationally representative of the Medicare beneficiary population as a whole. Data are collected on access to and utilization of health services, healthcare expenditures, sources of payment for healthcare-related services, health status and functioning, quality of life, and demographic and socioeconomic characteristics. The method for accounting for sample attrition changed slightly during the course of the survey. Before 1994, MCBS would supplement each panel once each year with newly chosen beneficiaries. In 1994, the MCBS switched from a purely longitudinal to a rotating panel design. Since then, each year of MCBS data contains four overlapping sample panels. Each year, one panel is rotated out of the MCBS and replaced with a newly selected one; therefore, a single panel can provide up to 4 years of information. In a typical year, the MCBS contains data on 14,000 to 17,000 beneficiaries.
Medicare claims data came from the inpatient, outpatient, Medicare Part B, and durable medical equipment claims provided by the Centers for Medicare and Medicaid Services. Claims data were available for 1991–2010. We used all of the claims data available for a respondent, even if the data fell outside the MCBS survey window. For example, if the respondent was first interviewed in 2001 and was 77 years old in that year, we used all of the available claims data from 1991 to 2000 for the look-back period to determine whether the person had been diagnosed as having diabetes mellitus before the first MCBS (baseline) interview and from 2001 to 2010 or from a departure from Medicare for any reason for the follow-up period. The mean number of years of the look-back period was 3.96 (standard deviation [SD] 3.56) and 6.68 years (SD 4.05) for the follow-up period. The hazard model accounted for right censoring of the data. A condition was counted as a diabetes mellitus complication only if the person had a prior diagnosis of diabetes mellitus recorded on a claim.
Beneficiaries in our analysis sample were born between 1926 and 1944, were not enrolled in a Medicare Advantage plan, and did not have a diabetes mellitus diagnosis at the time of their first (baseline) MCBS interview. Data on age, race, sex, educational attainment, smoking status, income strata, and BMI were drawn from the MCBS Cost and Use and Access to Care files. Beneficiaries with incomplete information were dropped. After exclusions, our sample size consisted of 14,657 beneficiaries. Coding of diabetes mellitus diagnosis and its complications was based on International Classification of Diseases, Ninth Revision Clinical Modification and Current Procedural Terminology, 4th Edition, codes.
We identified the time in 3-month intervals from an individual's first interview to the first diabetes mellitus diagnosis and to each major complication category (Table 1). The analysis focused on common cardiovascular, cerebrovascular, renal, ocular, and lower extremity complications. Risks were calculated both for the category as a whole (eg, any cardiovascular complication) and for specific conditions within each category (eg, AMI).
We calculated the BMI of individuals at the first MCBS interview (baseline). We classified them into six BMI categories based on criteria specified by the World Health Organization: BMI ≤18.50 (underweight); BMI between 18.50 and 24.99 (normal weight); BMI between 25.00 and 27.49 (overweight); BMI between 27.50 and 29.99 (preobese); BMI between 30.00 and 39.99 (obese); BMI ≥40.00 (morbidly obese). Normal-weight individuals were used as a reference group. BMI was calculated as weight divided by the square of height, based on the self-reported weight and height measures from MCBS. Other control variables included age, income, race, educational attainment (in years), smoking status, hypertension, and the year of the baseline interview to account for secular trends in practice and technology use. Results on explanatory variables other than BMI categories above normal are not reported to conserve space. We conducted separate analyses for men and women to measure sex-specific differences in time to diagnosis by sex. We estimated the association between BMI measured at baseline on time to first diagnosis of diabetes mellitus and to specific complications by calculating hazard ratios using Cox proportional hazards implemented by the stcox procedure of STATA 12.0 (StataCorp, College Station, TX). The hazard ratios represent the probability of being diagnosed as having diabetes mellitus and its complications relative to beneficiaries in the baseline group (individuals of normal weight). To assess the sensitivity of the results to the choice of statistical method and to compare hazard ratios with odds ratios, we also estimated logit models (results not shown).
Source...