Impact of Multi-morbidity on Mortality in Diabetes
Impact of Multi-morbidity on Mortality in Diabetes
A national cohort of veterans with type 2 diabetes was created by linking multiple patient and administrative files from the VHA National Patient Care and Pharmacy Benefits Management (PBM) databases. We used a previously validated algorithm for identifying veterans with diabetes. Veterans were included in the cohort if they had: 1) type 2 diabetes defined by two or more International Classification of Diseases, Ninth Revision (ICD-9) codes for diabetes (250, 357.2, 362.0, and 366.41) in the previous 24 months (2000 and 2001) and during 2002 from inpatient stays and/or outpatient visits on separate days (excluding codes from lab tests and other non-clinician visits); or 2) prescriptions for insulin or oral hypoglycemic agents (VA classes HS501 or HS502, respectively; to capture those without a diabetes ICD-9 code) in 2002. PBM data were available during the entire period of analysis. When the data were merged based on the criteria above, the total sample included 832,000 veterans. We excluded those not taking prescription medications for diabetes (n = 201,255) and added those who had one ICD-9 code for diabetes and prescriptions filled in 2002 (n = 60,493); and 3,660 were excluded due to death prior to 2002, missing age or no service connection. The subset with complete data resulted in a final cohort of 625,903 veterans. The Department of Veterans Affairs maintains data through the VA Information Resource Center (VIReC). Data was requested from and approved for use by VIReC following data use agreement requirements. The study was approved by the Medical University of South Carolina Institutional Review Board (IRB) and the Ralph H. Johnson Veterans Affairs Medical Center Research and Development committee.
The main outcome measure was time to death. Veterans were followed from time of entry into the study until death, loss to follow-up, or through December 2006. A subject was considered censored if alive by December 2006.
The primary covariates were medical comorbidity and psychiatric comorbidity both defined as the count of diseases for each subject throughout the study period. All comorbidities were dichotomized as present or absent where presence was determined by ICD-9 codes at entry into the cohort based on a previously validated algorithm in veterans. Medical comorbidity variables included anemia, cancer, cardiovascular disease (CVD), cerebrovascular disease, congestive heart failure (CHF), fluid and electrolyte disorders, hypertension, hypothyroidism, liver disease, lung conditions (chronic pulmonary disease, pulmonary circulation disease), obesity, peripheral vascular disease, and other (acquired immunodeficiency syndrome–AIDS, rheumatoid arthritis, renal failure, peptic ulcer disease and bleeding, weight loss). Psychiatric comorbidities included psychoses, substance abuse (alcohol abuse, drug abuse) and depression.
We controlled for seven demographic variables. Age was treated as continuous and centered at a mean of 66 years. Race/ethnicity included four categories with non-Hispanic white (NHW) serving as the reference group. Race/ethnicity was retrieved from the 2002 outpatient and inpatient [Medical SAS] data sets. When missing or unknown, the variable was supplemented using the inpatient race1-race6 fields from the 2003 [Medical SAS] data sets, the outpatient race1-race7 fields from the 2004 [Medical SAS] data sets, and the VA Vital Status Centers for Medicare and Medicaid Services (CMS) field for race. Gender, marital status, and location of residence (urban versus rural or highly rural) were dichotomous. Highly rural was categorized as rural according to the VA definition of rurality. Percentage service-connectedness, representing the degree of disability due to illness or injury that was aggravated by or incurred in military service, was treated as dichotomous (1= > 50%, 0 = <50%). Region, which accounts for the five geographic regions of the country, was treated as a categorical variable: Northeast [VISNs 1, 2, 3], Mid-Atlantic [VISNs 4, 5, 6, 9, 10], South [VISNs 7, 8, 16, 17], Midwest [VISNs 11, 12, 15, 19, 23], and West [VISNs 18, 20, 21, 22].
In preliminary analyses, crude associations were examined between mortality and all measured covariates using chi-square tests for categorical variables and t-tests for continuous variables. Cox regression methods were used to model the association between time to death and medical and psychiatric comorbidity after adjusting for known covariates. Time to death was defined as the number of months from time of entry into the cohort to time of death or censoring (i.e., day last seen or May 2006). For the Cox model, appropriateness of the assumption of proportionality was determined by testing the coefficients of the interactions of time with the respective covariate in multivariate analyses. Initially, Cox models for each of the medical and psychiatric comorbidities were fitted adjusting for all covariates (race, socio-demographics). Then an interaction between medical and psychiatric comorbidity was tested to check whether the association between mortality and medical comorbidity was modified by the presence of psychiatric comorbidity. HR estimates of medical comorbidity for each level of psychiatric comorbidity and estimates for levels of psychiatric comorbidity for levels of medical comorbidity are reported since there was significant interaction (p = 0.003). The Kaplan-Meier method was used to plot the survival functions for both medical and psychiatric comorbidities separately. Residual analysis was used to assess goodness-of-fit of each of the models. All data analyses were conducted using SAS 9.3.
Methods
Study Population
A national cohort of veterans with type 2 diabetes was created by linking multiple patient and administrative files from the VHA National Patient Care and Pharmacy Benefits Management (PBM) databases. We used a previously validated algorithm for identifying veterans with diabetes. Veterans were included in the cohort if they had: 1) type 2 diabetes defined by two or more International Classification of Diseases, Ninth Revision (ICD-9) codes for diabetes (250, 357.2, 362.0, and 366.41) in the previous 24 months (2000 and 2001) and during 2002 from inpatient stays and/or outpatient visits on separate days (excluding codes from lab tests and other non-clinician visits); or 2) prescriptions for insulin or oral hypoglycemic agents (VA classes HS501 or HS502, respectively; to capture those without a diabetes ICD-9 code) in 2002. PBM data were available during the entire period of analysis. When the data were merged based on the criteria above, the total sample included 832,000 veterans. We excluded those not taking prescription medications for diabetes (n = 201,255) and added those who had one ICD-9 code for diabetes and prescriptions filled in 2002 (n = 60,493); and 3,660 were excluded due to death prior to 2002, missing age or no service connection. The subset with complete data resulted in a final cohort of 625,903 veterans. The Department of Veterans Affairs maintains data through the VA Information Resource Center (VIReC). Data was requested from and approved for use by VIReC following data use agreement requirements. The study was approved by the Medical University of South Carolina Institutional Review Board (IRB) and the Ralph H. Johnson Veterans Affairs Medical Center Research and Development committee.
Outcome Measure
The main outcome measure was time to death. Veterans were followed from time of entry into the study until death, loss to follow-up, or through December 2006. A subject was considered censored if alive by December 2006.
Primary Covariates
The primary covariates were medical comorbidity and psychiatric comorbidity both defined as the count of diseases for each subject throughout the study period. All comorbidities were dichotomized as present or absent where presence was determined by ICD-9 codes at entry into the cohort based on a previously validated algorithm in veterans. Medical comorbidity variables included anemia, cancer, cardiovascular disease (CVD), cerebrovascular disease, congestive heart failure (CHF), fluid and electrolyte disorders, hypertension, hypothyroidism, liver disease, lung conditions (chronic pulmonary disease, pulmonary circulation disease), obesity, peripheral vascular disease, and other (acquired immunodeficiency syndrome–AIDS, rheumatoid arthritis, renal failure, peptic ulcer disease and bleeding, weight loss). Psychiatric comorbidities included psychoses, substance abuse (alcohol abuse, drug abuse) and depression.
Demographic Variables
We controlled for seven demographic variables. Age was treated as continuous and centered at a mean of 66 years. Race/ethnicity included four categories with non-Hispanic white (NHW) serving as the reference group. Race/ethnicity was retrieved from the 2002 outpatient and inpatient [Medical SAS] data sets. When missing or unknown, the variable was supplemented using the inpatient race1-race6 fields from the 2003 [Medical SAS] data sets, the outpatient race1-race7 fields from the 2004 [Medical SAS] data sets, and the VA Vital Status Centers for Medicare and Medicaid Services (CMS) field for race. Gender, marital status, and location of residence (urban versus rural or highly rural) were dichotomous. Highly rural was categorized as rural according to the VA definition of rurality. Percentage service-connectedness, representing the degree of disability due to illness or injury that was aggravated by or incurred in military service, was treated as dichotomous (1= > 50%, 0 = <50%). Region, which accounts for the five geographic regions of the country, was treated as a categorical variable: Northeast [VISNs 1, 2, 3], Mid-Atlantic [VISNs 4, 5, 6, 9, 10], South [VISNs 7, 8, 16, 17], Midwest [VISNs 11, 12, 15, 19, 23], and West [VISNs 18, 20, 21, 22].
Statistical Analysis
In preliminary analyses, crude associations were examined between mortality and all measured covariates using chi-square tests for categorical variables and t-tests for continuous variables. Cox regression methods were used to model the association between time to death and medical and psychiatric comorbidity after adjusting for known covariates. Time to death was defined as the number of months from time of entry into the cohort to time of death or censoring (i.e., day last seen or May 2006). For the Cox model, appropriateness of the assumption of proportionality was determined by testing the coefficients of the interactions of time with the respective covariate in multivariate analyses. Initially, Cox models for each of the medical and psychiatric comorbidities were fitted adjusting for all covariates (race, socio-demographics). Then an interaction between medical and psychiatric comorbidity was tested to check whether the association between mortality and medical comorbidity was modified by the presence of psychiatric comorbidity. HR estimates of medical comorbidity for each level of psychiatric comorbidity and estimates for levels of psychiatric comorbidity for levels of medical comorbidity are reported since there was significant interaction (p = 0.003). The Kaplan-Meier method was used to plot the survival functions for both medical and psychiatric comorbidities separately. Residual analysis was used to assess goodness-of-fit of each of the models. All data analyses were conducted using SAS 9.3.