Preventable Hospitalization Among Elderly Medicare Beneficiaries With Type 2 Diabetes

This study used a nationally representative cross-sectional sample of Medicare beneficiaries aged ≥65 years. Data were drawn from the 1999 Medicare Standard Analytic Files, a 5% nationally representative random sample of fee-for-service claims from Medicare beneficiaries (n = 2,055,561). Beneficiaries were excluded if they were aged <65 years, without both Medicare Part A and Part B coverage, or not residing in the U.S. Health maintenance organization enrollees, comprising 17.3% of Medicare beneficiaries nationwide in 1999 (25), were also excluded because they are unlikely to have complete claims information.

Subjects were defined as having type 2 diabetes if they had one claim in an inpatient setting or two claims in the physician/supplier file or outpatient file with a reported ICD-9-CM code of 250, 250.0, 250.00, 250.02, 250.4, 250.40, 250.42, 250.5, 250.50, 250.52, 250.6, 250.60, 250.62, 250.7, 250.70, 250.72, 250.8, 250.9, 250.90, or 250.92 (n = 193,556). Two outpatient claims were required to increase the specificity of diagnoses. We excluded 10,647 beneficiaries with type 1 diabetes (<1% of the population) from the analysis if there was not also a claim for type 2 diabetes during the year because 1) the diagnosis of type 1 diabetes in this age-group may often be inaccurate, as insulin-treated type 2 diabetes may be miscoded as type 1 diabetes, and 2) there are relatively few people in the Medicare population with type 1 diabetes (26). Chronic conditions were coded using the Clinical Classification System developed by the Agency for Healthcare Research and Quality (27). The Clinical Classification System aggregates all diagnosis codes into 259 unique, clinically homogeneous categories (e.g., essential hypertension, osteoarthritis, prostate cancer). We also merged data from the 1999 Area Resource File to obtain per capita community level indicators of income and hospital beds (28).

ACSCs were defined by panels of physicians as conditions that should not require hospitalization if ambulatory care treatment is provided in a timely and appropriate manner (24). The Medicare inpatient hospital claims data file was used to identify beneficiaries with all-cause hospitalization and preventable hospitalizations. The ICD-9-CM code reported as the primary diagnosis for each hospitalization was used to determine whether a hospitalization could be classified as potentially preventable (see online appendix at http://care.diabetesjournals.org). We adapted this definition of ACSC hospitalization from the study by Culler et al. (29) of Medicare beneficiaries aged ≥65 years.

Multiple logistic regression was used to test the hypothesis that specific comorbid conditions increased the likelihood of preventable hospitalization among Medicare beneficiaries aged ≥65 with type 2 diabetes. Multivariable analysis controlled for age; sex; race; mortality; number of unique physicians seen during the year; type of physician seen (at least one outpatient visit to a primary care physician, endocrinologist, or cardiologist); renal, ophthalmic, or neurological manifestations of diabetes (ICD-9-CM codes 250.4x, 250.5x, and 250.6x, respectively); number of comorbid conditions; per capita income at the county level; and hospital beds per capita. We chose these variables because each could influence the probability of preventable hospitalization (23,29,30). The chronic conditions chosen in the analysis were either cardiovascular related or among the 25 most prevalent chronic comorbid conditions of beneficiaries with type 2 diabetes. Cardiovascular comorbidities were of specific interest because of their high prevalence among beneficiaries with type 2 diabetes and the complications that result from these conditions (5,10–12). Analyses were conducted using SAS statistical software version 8.02 (31).

We identified a total of 193,556 Medicare beneficiaries aged ≥65 years with type 2 diabetes using the claims data (15.6% of the sample population). A disproportionate number of the Medicare population with type 2 diabetes were African American compared with those without type 2 diabetes (11 vs. 7%, P < 0.0001). Fifty-eight percent with type 2 diabetes were women, and 11% were aged ≥85 years. Almost all Medicare beneficiaries (96%) with type 2 diabetes had at least one other chronic condition. Almost half (46%) of Medicare beneficiaries aged ≥65 years with type 2 diabetes had five or more comorbid conditions (Fig. 1).

The ten most prevalent comorbidities accompanying type 2 diabetes are shown in Fig. 2; the top five were related to CVD. Comorbidity to type 2 diabetes increased health care utilization dramatically. A beneficiary aged ≥65 years with diabetes and no comorbid conditions saw 5.6 unique physicians a year on average, 16.4 physicians if they had five comorbidities, and 28.2 if they had ≥10 comorbidities.

A beneficiary aged ≥65 years with type 2 diabetes who was hospitalized during the year spent an average of 13.3 days in the hospital. Without comorbidity, this number decreased to 4.3 days per year. Five comorbidities almost doubled the average number of days spent in the hospital (to 8.0 days), and people with ≥10 comorbidities had an average of 23.7 hospital days.

There were 32,518 preventable hospitalizations among Medicare beneficiaries aged ≥65 years with type 2 diabetes, constituting 7% of all inpatient stays. The chronic conditions most commonly associated with preventable hospitalizations were CHF, pneumonia, chronic obstructive pulmonary disease (COPD), dehydration, and diabetes. Forty percent of all preventable hospitalizations were attributed to CHF, 21% to pneumonia, 13% to COPD, and 2% to diabetes. The comorbidities associated with preventable hospitalizations in diabetes were similar to those in the generic population (P > 0.05). Hospitalization for CHF disproportionately affected beneficiaries with type 2 diabetes compared with Medicare beneficiaries without type 2 diabetes (P < 0.0001).

Expenditures for preventable hospitalizations were high. The average per capita cost of a preventable hospitalization was $7,935 among type 2 diabetes beneficiaries. Total expenditures per capita for preventable hospitalizations were far higher among type 2 diabetes beneficiaries with CVD comorbidities than for those without CVD comorbidities ($26,733 vs. $13,150). Among those with a preventable hospitalization, nearly three-fourths were hospitalized once during the year with a preventable condition and more than one-quarter experienced multiple hospital admissions for an ACSC, further raising Medicare expenditures.

Several specific cardiovascular comorbidities increased the likelihood of preventable hospitalization after controlling for potential confounders (Table 1). Beneficiaries with type 2 diabetes and CHF were over three times more likely to have a preventable hospitalization than those without CHF. Cardiomyopathy, coronary atherosclerosis, hypertension, and cardiac dysrythmias also raised the odds of preventable hospitalization among beneficiaries with type 2 diabetes, controlling for other factors.

Certain noncardiovascular comorbidities were also associated with a greater likelihood of preventable hospitalization. Beneficiaries with type 2 diabetes and COPD were 2.4 times more likely to experience preventable hospitalization than type 2 diabetes beneficiaries without COPD. The likelihood of preventable hospitalization also increased for beneficiaries with type 2 diabetes if they had comorbid asthma or lower respiratory disorders, Alzheimer’s disease/dementia, personality/anxiety disorder, depression, and osteoporosis.

A few comorbidities appeared to be associated with a lower probability of preventable hospitalization among those with type 2 diabetes. Those with lipid disorders were 34% less likely to have a preventable hospitalization. This occurs because of the low prevalence of beneficiaries with lipid disorders without a comorbid claim for conduction disorders. Once conduction disorders are controlled for in the analysis, the added effect of lipid disorders on preventable hospitalizations is negative. Claims for a diagnosis of cataract, glaucoma, hyperplasia of the prostate, cancer of the prostate, spondylosis/intervertebral disc disorders, other connective tissue disease, or other eye disorder also lowered the odds of preventable hospitalization.

Other factors that significantly increased the likelihood of preventable hospitalization (P < 0.05) included age ≥75 years, nonwhite race, female sex, and mortality during the year. The odds of preventable hospitalization increased by 17% for each added comorbid condition, when all other factors were held constant.

With one exception, seeing any particular type of specialist at any point during the year was not a significant predictor of hospitalization (P > 0.05). Beneficiaries with diabetes who saw a cardiologist did have a 12% lower odds of experiencing a preventable hospitalization compared with beneficiaries who did not see a cardiologist (P < 0.001). None of the odds ratios corresponding to the interaction terms between the types of physicians were significant, nor were the odds ratios associated with the number of different physicians seen, the per capita income at the county level, or the number of hospital beds per capita.

We performed a sensitivity analysis of the data by eliminating people from the model who died during the year. There were no significant changes to the comorbid conditions, the demographic variables, or the community level indicators. The “type of doctor seen” variables did become significant predictors of preventable hospitalizations. Patients who were seen solely by their primary care physician, endocrinologist, or cardiologist were significantly more likely to experience preventable hospitalization during the year. Patients who were seen by primary care physicians and endocrinologists, cardiologists and endocrinologists, or primary care physicians and cardiologists were significantly less likely to experience a preventable hospitalization.

Comorbid conditions increase the likelihood of preventable hospitalizations among patients aged ≥65 years with type 2 diabetes. Using a nationally representative sample, our data suggest that nearly 7% of all hospitalizations in Medicare beneficiaries with type 2 diabetes could be avoided. Specific conditions associated with increased rates of preventable hospitalization included CHF, other cardiovascular-related conditions, COPD, asthma, and mental disorders.

Our work confirms and extends the findings of others by analyzing a disease-specific population, identifying multiple comorbid conditions, and including patient access measures at the community level. Culler et al. (29) reported that diabetes increased the odds of a preventable hospitalization among Medicare beneficiaries aged ≥65 years by 23% compared with beneficiaries without diabetes (29). Wolff et al. (30) reported that preventable hospitalization rates increased as the number of chronic conditions a beneficiary had increased. Our data extend these studies by identifying specific comorbid conditions that are associated with preventable hospitalizations in a disease-specific population.

While our study is the first to address preventable hospitalizations in elderly type 2 diabetic patients with comorbidities, several reports have identified inadequacies in primary care for older adults with diabetes and comorbid illnesses. Asch et al. (32) estimated that only 36% of patients with diabetes receive an HbA_1c or fructosamine assessment every 6 months and only 43% receive eye exams every year. In addition, Weiner et al. (33) found that 45% of elderly Medicare beneficiaries received no cholesterol screening.

The presence of comorbidities in our analysis was related to a remarkable proliferation in the number of different physicians seeing individual Medicare beneficiaries. Thus, while the average beneficiary without comorbidity saw 5.6 physicians, this number increased fivefold, to 28.2 doctors, if the individual had 10 comorbidities. It may be speculated that communication among caregivers deteriorates as the number of unique physicians involved in the care of these elderly beneficiaries increases, which could complicate outpatient care management.

There are several mechanisms by which comorbid conditions can increase preventable hospitalizations in diabetes. Ability to adhere to prescribed treatment regimens may be more difficult because of complex, hard to understand, or possibly opposing treatment recommendations by different physicians (34,35). Other data show that individuals with diabetes especially benefit from interventions aimed at improving patient adherence to therapeutic recommendations (36). Comorbidity may also discourage adherence to multiple complex treatment regimens due to depressed mood about high burden of disease—individuals with diabetes and depression are less likely than those without depression to achieve adequate levels of glycemic control (37,38).

Our data strongly support the need for improving the current system of chronic disease management in Medicare. According to the Institute of Medicine, “the most effective approach to chronic illness includes access to the entire medical record for each patient, coordination of treatment between various health care providers, and simplicity of system use for both the practitioner and the patient” (39). A second approach advocated by the Institute of Medicine is multidisciplinary disease management. Multidisciplinary disease management has been found to lower HbA_1c levels, improve self-care practices, and reduce hospital utilization among individuals with diabetes in a health maintenance organization setting compared with similar patients receiving care from their primary care physician (40).

Several limitations in our data are worth mentioning. First, because we used administrative claims data, we were unable to measure several factors that may confound the relation between comorbidity and preventable hospitalizations, such as availability of primary care physicians and specialists, individual indicators of socioeconomic status, satisfaction with care, and disease severity. Despite this limitation, however, the strength of our data lies in its large size, national representativeness, and extensive listing of multiple chronic conditions.

The second limitation is that we relied on the claims-coding process to identify comorbid conditions and preventable hospitalizations. Claims-based selection and misclassification bias would likely bias the prevalence rates of our chronic conditions downward. Despite any underestimation of the prevalence of chronic conditions, it is unlikely that the relative impact of any specific comorbidity would change dramatically. In addition, while preventable hospitalizations have been linked to poorer outpatient access, chart reviews have not been performed to validate whether patients with preventable hospitalizations are more likely than similar patients without preventable hospitalization to fail to receive timely and appropriate services. This was not possible given the framework of our study.

The third limitation is the cross-sectional nature of our data. This precluded us from establishing causal relationships between comorbidities and preventable hospitalizations. Finally, because we only studied fee-for-service Medicare beneficiaries aged ≥65 years with type 2 diabetes with both Part A and Part B Medicare, our findings may not be generalizable to younger individuals, Medicare beneficiaries enrolled in managed care plans, or Medicare beneficiaries without supplemental insurance.

In conclusion, this study demonstrates that comorbidities complicate care among elderly Medicare beneficiaries with type 2 diabetes. Cardiovascular conditions, in particular, resulted in a greatly increased rate of hospitalization that may be prevented through improvements in timely and effective outpatient care. These preventable hospitalizations are not only a burden to individuals, but also a considerable cost to the Medicare program. The enormous number of unique physicians involved in caring for individuals with diabetes and comorbidities highlights the challenges of delivering coordinated care to elderly beneficiaries. Given the aging U.S. population and the burden of type 2 diabetes, improved outpatient strategies for the management of high-risk elderly patients with type 2 diabetes are worthy of attention by policy makers, health care administrators, and clinicians.

This research has been sponsored by Partnership for Solutions, Robert Wood Johnson Foundation, Princeton, NJ.

The authors are very grateful to Robert Herbert for his extensive programming support.