Diabetes Comorbidity and Age Influence Rehabilitation Outcomes After Hip Fracture

Data were from the Uniform Data System for Medical Rehabilitation (UDSMR), which contains patient- and facility-level information, a measure of function, length of stay (LOS), and discharge setting (7–9). Complete information on the variables and protocol for data collection is available in the Inpatient Rehabilitation Facility-Patient Assessment Instrument training manual (10) through the CMS and at http://www.udsmr.org.

The sample included 79,526 patients with hip fracture (ICD-9 codes 820.0–820.9 and 821.0–821.39) from 915 rehabilitation facilities across all 10 CMS regions. Criteria developed by CMS for prospective payment were used to exclude patients with an atypical course of rehabilitation (11,12).

Three patient groups were created using the following CMS criteria: 1) patients with no diabetes, 2) patients with non-tier diabetes (ICD-9 250.0–250.3), and 3) patients with tier 3 comorbid diabetes (ICD-9 250.4–250.9, 357.2, and 785.4) (4).

Three outcomes were examined: LOS was calculated as the total number of days spent in the rehabilitation unit or hospital. Functional status was determined using the Inpatient Rehabilitation Facility-Patient Assessment Instrument developed by CMS (10,11). The instrument includes six subscales: self-care, sphincter control, transfers, mobility, communication, and social cognition with total values ranging from 18 to 126 (13,14). Discharge setting was dichotomized into patients who returned home versus patients who were discharged to alternate destinations after rehabilitation.

Sociodemographic variables included age in years, sex, race/ethnicity (black, Hispanic, white, other), and marital status (married versus not married). A sum of the remaining nondiabetes comorbidities was calculated as an indicator of combined diseases or chronic conditions.

One-way ANOVA with post hoc and χ² tests were used to assess differences among variables across the diabetes groups on the basis of comorbidity status. Age-centered multiple linear regression models were computed to estimate the effects of diabetes status on LOS and functional status ratings. The odds of discharge to home based on diabetes status were computed using an age-centered logistic regression model. The regression models contained relevant covariates. Age by diabetes interaction terms was included in each model.

The mean age was 79.4 years (SD 9.6) with an average LOS of 13.2 days (SD 5.3). The prevalence of diabetes in the sample was 23.0%, with 18.3% of patients diagnosed with non-tier diabetes and 4.7% with tier-level diabetes. Statistically significant differences (P < 0.05) existed across the three groups for sociodemographic variables (age, sex, and race/ethnicity) and for the outcome and clinical variables (LOS, admission and discharge functional status, discharge home status, and sum of comorbidities).

Linear and logistic regression analyses yielded significant diabetes-by-age interactions for the outcomes of LOS, functional status ratings, and discharge setting. The linear regression models demonstrated significant (P < 0.05) group-by-age interactions between tier diabetes and no diabetes for both LOS and discharge functional status ratings. There were also significant age-by-diabetes group interactions between the non-tier (OR = 0.98, P < 0.01) and no diabetes (OR = 0.98, P < 0.01) groups compared with the tier-level diabetes group for the discharge home versus not home variable.

Increasing age was associated with decreased difference for each of the outcome measures (LOS, functional status, and discharge home) across the three diabetes status groups. This moderating effect is demonstrated in Figure 1 for the discharge home variable. Figure 1 shows the expected values for discharge home by age for each of the disability status groups and demonstrates a statistically significant difference (P < 0.01) that decreases as age increases. The pattern was similar for LOS and functional status.

Individuals with diabetes had longer LOS in rehabilitation facilities than individuals without diabetes. Patients with CMS-defined tier comorbid diabetes stayed longer than patients with non-tier diabetes. Similar associations were found for functional status ratings at discharge and for the percent of patients who were discharged home. The results were moderated by age. In each case, the difference among the diabetes groups (no diabetes versus non-tier diabetes versus tier diabetes) was larger in younger subjects and smaller in older subjects.

Our dataset did not have information about the length of time someone had diabetes or diabetes-related treatment(s). Our sample included only subjects living at home before hip fracture, and the findings are not generalizable to individuals living in institutional settings. We also lacked information on functional status before the hip fracture.

In summary, diabetes was prevalent in 23% of patients in our sample with hip fracture and was associated with poorer outcomes, particularly in younger individuals. Our findings support the use of the CMS tier comorbidity system for patients with diabetes receiving rehabilitation after hip fracture. The tier diabetes group displayed longer LOS and lower levels of functional status than patients with no diabetes or nontier diabetes. Our findings also suggest that diabetes tier status has an impact that is more pronounced in younger patients. Further research is necessary to better understand the mechanism underlying the moderating effects of age on the recovery of individuals with diabetes after hip fracture.

This work was funded in part by the National Institute on Disability and Rehabilitation Research, U.S. Department of Education (H133G080163 and H133F090030) and the National Institutes of Health, National Institute of Child Health and Development (K12-HD-055929).

No potential conflicts of interest relevant to this article were reported.

T.A.R. analyzed data, developed the first draft, and contributed to revisions. J.E.G. analyzed data and contributed to the data interpretation and revisions. A.D. assisted with data interpretation and contributed to revisions. S.J.M. and C.V.G. provided data, assisted with interpretation, and contributed to revisions. K.J.O. was responsible for funding, assisted with data interpretation, and contributed to revisions.