BACKGROUND: Diabetes mellitus requires continuous medical care and patient self-management in order to prevent short-term complications and decrease the risk of long-term complications, which can result in substantial increases in the total economic burden of the disease. Findings from randomized clinical trials have shown that improved glycemic control may reduce the risk of long-term complications as long as a target for hemoglobin A1c is not set below 7% for intensive glycemic control. However, limited data from clinical practice are available regarding the relationship between glycemic control and medical costs associated with diabetes care OBJECTIVE. To assess the potential relationships between glycemic levels, diabetes-related hospitalizations, and hospital costs among adult patients with either type 1 or type 2 diabetes mellitus who were assigned to a primary care provider (PCP) in a clinic that was affiliated with a managed care organization (MCO). METHODS: A retrospective cohort analysis was conducted using data from approximately 200,000 members of the Fallon Clinic Health Plan who were assigned to a clinic PCP at any time during a 5-year study period beginning January 1, 2002, and ending December 31, 2006. Patients aged 30 years or older with at least 2 medical claims with any listed diagnosis of diabetes mellitus (ICD-9-CM code 250.xx) during the study period and 2 or more A1c values within 1 year of each other during the study period (mean 7.6 tests over 39 months; median=6.8), were identified and stratified into 1 of 5 groups defined by 1% increments of A1c, based on their mean A1c values during the entire study period. A1c data were available only for tests ordered by a clinic provider, tests ordered by other specialists in the MCO's network were absent from the database. The study follow-up period started with each patient's first A1c test (index date) and continued until plan disenrollment, death, or December 31, 2006, whichever was earlier (end date), regardless of when the diagnosis of diabetes mellitus was made. Study measures included the proportion of patients with 1 or more diabetes-related hospitalizations, number of diabetes-related inpatient stays, and the associated estimated hospitalization costs over the follow-up period Diabetes-related hospitalizations were identified based on a diagnosis, in any of 10 diagnosis fields, for 1 of 16 selected complications of diabetes identified by the authors. Hospital costs were estimated using discharge data (diagnoses and costs calculated from cost-to-charge ratios) contained in the 2004 Healthcare Cost and Utilization Project (HCUP) database and inflated to 2007 dollars using the medical care component of the Consumer Price Index. Multivariate models controlled for age, sex, number of A1c tests, diagnosis of cancer, and follow-up time A multivariate logistic regression analysis was conducted with the occurrence of at least 1 diabetes-related hospital admission as the dependent variable In the logistic regression analysis, follow-up time was defined as time from the index date to the date of the first diabetes-related hospitalization, plan disenrollment, death, or the study end date, whichever occurred first. A generalized linear model with a Poisson distribution and a log link was employed to estimate the rate of hospital admissions In the Poisson regression analysis, follow-up time was defined as duration of the entire study follow-up period and was an offset variable. Costs were estimated using a 2-part model. first, we calculated the probability of having a hospitalization, as determined by the logistic regression above; second, a generalized linear model with a negative binomial distribution and a log link was used to predict the mean cost of diabetes-related hospitalizations only for patients with an inpatient stay, with the duration of the entire study follow-up period as an offset variable. We calculated the mean per patient cost of diabetes-related hospitalizations by multiplying the probability of having a hospitalization (as determined by the first part of the model) by the mean costs for patients who had such admissions (as determined by the second part of the model). RESULTS: 9,887 patients met study selection criteria Mean A1c level was <7% for 5,649 (57.1%) patients, 7% to <8% for 2,747 (27 8%), 8% to <9% for 1,002 (10.1%), 9% to <10% for 312 (3.2%), and 10% or more for 177 (1 8%). Over a mean (median) 40 (40) months of follow-up (interquartile range= 30-50 months), 287% (n=2,838) of patients had 1 or more diabetes-related hospital admissions In the logistic regression analysis, odds of having at least 1 diabetes-related hospital stay did not significantly differ for patients with mean A1c of <7% compared with patients in most higher mean A1c categories (7% to <8%, 8% to <9%, or 9% to <10%), however, odds of having a diabetes-related hospitalization were significantly higher for patients with mean A1c of 10% or more compared with patients with mean A1c of <7% (odds ratio=2.13, 95% confidence interval =1.36-3.33). In the negative binomial regression analysis of those with at least 1 hospital admission, estimated costs per hospitalized patient increased by mean A1c level. In the Poisson regression analysis, the rate of diabetes-related hospitalizations significantly increased by A1c level (13 per 100 patient-years for patients with mean A1c of <7% vs. 30 per 100 patient-years for mean A1c of 10% or more when covariates were held at mean levels, P<0.001) In the 2-part model results, adjusted mean estimated costs of diabetes-related hospitalizations per study patient were $2,792 among those with mean A1c of <7% and $6,759 among those with mean A1c of 10% or more. CONCLUSIONS. In this managed-care plan, the odds of having at least 1 diabetes-related hospitalization were not significantly associated with higher mean A1c except for patients with mean A1c of at least 10%. However, higher mean A1c levels were associated with significantly higher estimated hospitalization costs among those with at least 1 hospitalization and with higher rates of diabetes-related hospital utilization per 100 patient-years.
