Multiple Hospitalizations for Patients With Diabetes

We used 1999 data from the Healthcare Cost and Utilization Project (HCUP) State Inpatient Databases. Sponsored by the Agency for Healthcare Research and Quality, HCUP is a federal/state/industry partnership to build a multistate health care data system. The core of HCUP is hospital discharge abstract data obtained from state data organizations or hospital associations. The State Inpatient Databases include data on inpatient stays from virtually all community hospitals in participating states. Community hospitals are short-term, nonfederal, general, and specialty hospitals, including academic medical centers but excluding long-term care and psychiatric hospitals. We selected five states (California, Missouri, New York, Tennessee, and Virginia) that provide patient numbers (PNUMs) encrypted from the patient’s name and/or social security number. PNUMs are used to identify multiple admissions by distinct patients across hospitals within the same state. These five states also provide a relatively high representation of racial/ethnic minorities.

We identified 1,129,506 nonneonatal, nonmaternal discharges with the principal or secondary diagnosis indicating diabetes (ICD-9-CM code of 250.xx). Other researchers have used the same method to identify hospital discharges by patients with diabetes (3,4,7). We excluded discharges that had missing PNUMs, the same PNUM but inconsistent age (different by >1 year) or sex, or missing admission dates. If a patient had 2 consecutive stays in which the discharge date of the first stay overlapped with the admission date of the second by >2 days, both discharges were excluded. This left 1,076,562 discharges (95.3%) remaining in the study. To avoid the counting of transfers, if the readmission occurred on the same day of or 1 day before discharge of the previous stay, the readmission was not considered as a separate hospitalization. The final study population consisted of 648,748 patients with a total of 993,074 hospitalizations.

The purpose of our analysis is to examine the resource implications and opportunities for improving outcomes of multiple hospitalizations by patients with diabetes. We classified a patient as having multiple hospitalizations if the patient had two or more hospital stays during 1999, regardless of whether the readmission was planned or unplanned, preventable or unpreventable, or related or unrelated to prior stay. This definition has been used in a study by researchers of the Veterans Administration (8). We examined multiple hospitalizations using three indicators: the percent of patients with multiple hospitalizations, the percent of total stays represented by multiple hospitalizations, and the average number of stays per patient. We also examined total hospital costs that were derived from total hospital charges. We applied the cost-to-charge ratios developed from national data compiled by Centers for Medicare and Medicaid Services and tested against accounting systems from four states (9). We used hospital wage indices to adjust for geographic differences in labor cost. We calculated total hospital costs separately for patients with single hospitalization and patients with multiple stays.

Patient demographic and socioeconomic characteristics include age, sex, race/ethnicity, expected payer, location of residence, and median income of the patient’s community. We used the patient’s county or zip code to determine whether the patient lived in a rural or urban area. Rural areas are defined as areas outside a Metropolitan Statistical Area. We estimated income using the median household income of the patient’s zip code and categorized income into three levels using quartiles specific to each state. Low income refers to zip codes in the lowest quartile of median income within that state, while high income refers to zip codes in the highest quartile. Medium income includes those zip codes with median incomes falling in the two middle quartiles.

We performed stratified analyses by patient demographic and socioeconomic characteristics to identify differences across patient subgroups in the extent of multiple hospitalizations and resource implications. Student’s t test and χ² test were used to determine statistical significance of comparisons across age-groups (nonelderly adults being the reference), sex, race (non-Hispanic whites being the reference), payer (private insurance being the reference), location, and income. Differences in the likelihood of having multiple stays across racial/ethnic and socioeconomic groups were further assessed by logistic regression to control for patient age, sex, and clinical characteristics at the first stay, which include the type of diabetes complications and chronic conditions (see online appendix at http://care.diabetesjournals.org), length of stay (in log function), and disposition (e.g., transfer to nursing home or other health care facility, home health care). Admission month of the first stay was included to adjust for different follow-up period lengths. Dummy variables indicating individual states were used to control for state effects.

We also compared the presence of different clinical conditions among patients with multiple hospitalizations by age, sex, race/ethnicity, and payer. Three types of conditions were identified: 1) acute complications of diabetes (e.g., ketoacidosis, hyperosmolarity, diabetic coma), 2) chronic complications of diabetes (e.g., ophthalmic disease, lower extremity disease, renal disease, cardiovascular disease), and 3) general chronic conditions. The online appendix provides the ICD-9-CM codes used to define these conditions. Each condition was considered present if the code corresponding to the condition was listed as a principal or secondary diagnosis during any hospital stay.

Among patients with diabetes who had been hospitalized during 1999 in five selected states (California, Missouri, New York, Tennessee, and Virginia), 30% had two or more stays that contributed to 55% of total hospitalizations and 54% of total hospital costs. The average length of stay is slightly longer for patients with multiple stays than for patients with single hospitalization (7.4 vs. 6.8 days) (Table 1). The total hospital cost per patient is nearly three times as high for patients with multiple stays as for patients with a single stay ($23,119 vs. $8,508). For patients with multiple hospitalizations, the total hospital cost covers all patient stays.

Table 1 also shows proportions of multiple hospitalizations, average length of stay, and total hospital cost per patient by age and sex. The percent of patients with multiple hospitalizations (two or more stays) increased by age, ranging from 16.0% for patients age 1–17 years to 31.4% for patients age ≥65 years. Multiple stays accounted for 35% of total hospitalizations for pediatric patients compared with 55% for adults. There was no substantial difference in the extent of multiple hospitalizations between male and female patients except for children. Girls had higher proportions of multiple hospitalizations than boys (percent of distinct patients: 18.2 vs. 13.4%; percent of total stays: 39.2 vs. 29.6%; average number of stays per patient: 1.35 vs. 1.23).

Table 2 shows differences in the percent of patients with multiple hospitalizations by race/ethnicity, payer, location of residence, and income of community. Among the elderly, Hispanic patients had the highest percent of multiple hospitalizations (37.2%), followed by non-Hispanic blacks (34.0%) and non-Hispanic whites (30.9%). For patients <65 years of age, the percents of multiple hospitalizations were similar for Hispanics and non-Hispanic blacks but higher for both of these groups than for non-Hispanic whites. The difference was greatest among children. The percents of multiple hospitalizations for Asian-Pacific Islanders were comparable to those of non-Hispanic whites.

Both Medicare and Medicaid patients had higher percents of multiple hospitalizations than privately insured patients. The difference was greater among the nonelderly: compared with privately insured, the percent of multiple hospitalizations for Medicaid patients was 80% higher among pediatric patients and 55% higher among nonelderly adults. The percents of multiple hospitalizations were similar between urban and rural patients except for pediatric patients. Children living in urban areas had a higher percent of multiple hospitalizations than children residing in rural areas (16.6 vs. 13.6%). Patients from low-income zip codes had higher percents of multiple hospitalizations than patients in high-income zip codes. The percent of children with multiple stays was 18.6% for the low-income group compared with 12.6% for the high-income group.

Results of the full logistic regression confirm the patterns of disparities already reported. Controlling for patient age, sex, and clinical characteristics at the first hospital stay, significantly higher odds of having multiple stays were found for Hispanics and non-Hispanic blacks relative to non-Hispanic whites (P < 0.0001), Medicare or Medicaid patients compared with privately insured (P < 0.0001), and patients in low-income areas (P < 0.05).

Table 3 presents average total hospital costs for patients with multiple stays by race/ethnicity, payer, location, and income. Among adults who had two or more hospitalizations, the average total hospital cost per patient was significantly higher for racial/ethnic minority groups, Medicare or Medicaid patients, and patients from low-income areas. Similar patterns were found for pediatric patients but not reaching statistical significance due to relatively small patient volumes.

Over 60% of pediatric diabetes patients who had multiple hospitalizations had experienced acute complications of diabetes as a primary or coexisting condition for the stay. This compared with only 10% for nonelderly adults and <5% for the elderly. In contrast, chronic complications of diabetes were more common among adults. About 40% of adult diabetes patients with multiple hospitalizations had developed lower extremity disease. Hypertension was the most common cardiovascular disease among adult patients, followed by ischemic heart disease and congestive heart failure. (Detailed results by age/sex are available from authors.)

Table 4 shows differences in presence of diabetes complications and comorbidities by race/ethnicity and payer among patients with multiple stays. Only adult patients (age ≥18 years) were included in the analysis. The acute complication rate was much higher for non-Hispanic blacks than for other racial/ethnic groups. All three racial/ethnic minority groups had higher renal disease rates than non-Hispanic whites. Asian/Pacific Islanders had a higher rate for stroke than other racial/ethnic groups (one in four versus one in five) but lowest rates for several comorbidities, including obesity, depression, and substance abuse. Across payers, the acute complication rate was more than twice as high for the uninsured as for patients with insurance coverage. The uninsured also had the highest rate for substance abuse. Compared with privately insured patients, Medicaid patients had higher rates for a number of conditions, including renal disease, congestive heart failure, stroke, chronic obstructive pulmonary disease, depression, and substance abuse.

The results of this study indicate that multiple hospitalizations are common among patients with diabetes but vary by age, race/ethnicity, payer, and income. Although only 30% of patients were hospitalized more than once, these patients accounted for a majority of the inpatient costs for patients with diabetes, suggesting that efforts aimed at reducing the need for multiple hospitalizations could pay major dividends. Prior studies have shown that focused efforts using disease or case management can improve outcomes (10) and reduce utilization and costs (11). Future evaluation of disease management programs could include multiple hospitalizations as an outcome measure.

While we generally consider the elderly to be sicker and more likely to use inpatient care, there was no substantial difference in the percent of patients with multiple hospitalizations between nonelderly adults and the elderly (28.2 vs. 31.4%). Our data do not permit separate analysis by type 1 versus type 2 diabetes, but patients with type 1 diabetes most likely accounted for a larger proportion of the nonelderly group than the elderly group. Some of these patients may be in an advanced stage of diabetes with multiple complications that could increase their likelihood of having multiple hospitalizations.

The proportions of multiple hospitalizations were higher for some groups that are considered vulnerable, including racial/ethnic minorities, those covered under Medicare or Medicaid, and those living in lower-income communities. Compared with non-Hispanic whites, Hispanics and non-Hispanic blacks had higher likelihoods of having multiple hospitalizations. Among patients with multiple stays, Hispanics and non-Hispanic blacks had higher total hospital costs per patient and higher prevalence rates for acute complications and end-stage renal disease. These results are consistent with other studies that found Hispanics and non-Hispanic blacks were less likely to receive HbA_1c test and eye exam (12), less likely to do self-monitoring of blood glucose level, and more likely to have poor glycemic control (13). Better glycemic control can help reduce hospitalization (14) and achieve significant cost savings (15). Of note, lack of insurance was not associated with higher rates of multiple hospitalizations. While they did not have a higher proportion of multiple stays, the uninsured adults had a clinically distinct pattern compared with other groups: higher rates of acute complications, lower rates of chronic complications, and higher rates of substance abuse. This lack of a consistent association between the socioeconomic vulnerable subgroups and multiple hospitalizations suggests that the latter may be determined by an interplay of factors, including eligibility for insurance, access to primary and specialty care, demographics such as age, prevalence of complications, and patient preferences and bias (16).

It is worth noting that compared with adults, greater differences were found for pediatric patients in multiple hospitalizations by race, payer, and income of community. This may be related to the higher prevalence of acute diabetes complications among pediatric patients, shown by our data, many of which are preventable for the more vulnerable subgroups. A relatively higher admission rate for acute complications has been reported by other studies for Medicaid patients (17) and patients living in low-income areas (18).

Our study relies on hospital discharge data that have several limitations. First, we were unable to distinguish betweein type 1 and type 2 diabetes because the fifth digit in the ICD-9-CM codes that specify whether the patient was insulin dependent is inaccurate in identifying the type of diabetes. Patients with different types of diabetes could have different patterns of hospital utilization and outcomes. Second, the data do not provide information on duration of diabetes. The likelihood of developing complications and being hospitalized could increase by how long the patient has had diabetes. Related to this, the data do not provide a measure of disease severity beyond that captured by complications and comorbidities. Third, the data capture only inpatient stays that would underreport those conditions, such as eye disorders, and that are more likely to be treated in the outpatient setting. Fourth, this study examines a 1-year period of time. It is likely that some patients who had only one hospitalization during 1999 were also hospitalized within 1 year of their inpatient stay, either in 1998 or 2000. However, the approach we used provides an annual picture of utilization patterns similar to time frames in many national surveys of health care utilization, such as the National Health Interview Survey or the Medical Expenditure Panel Survey. Fifth, we only captured hospital stays with diabetes listed as one of the discharge diagnoses. It is possible that some diabetic patients may not have diabetes reported in their discharge records. Thus, the study likely underestimates the percentage of diabetic patients with multiple hospitalizations and the associated costs.

In summary, the results of this study reveal multiple hospitalizations to be an important contributor to hospital utilization and costs for patients with diabetes. Among patients with multiple stays, >60% of children had experienced acute complications; ∼90% of adults had cardiovascular disease, 25% renal disease, and 40% lower extremity disease. Some of the multiple hospitalizations that resulted from these complications, as well as some of the complications themselves, are preventable with quality outpatient care, including patient self-management of the condition. Future research should explore what type of patients would have a higher probability of preventable readmissions and identify cost-effective interventions to target this specific group of patients.

The authors express special thanks to Zhengyi Fang at Social and Scientific Systems for his super programming skills and to Eva Weinstein, an intern at Agency for Healthcare Research and Quality, for assistance in literature search.