Since 1993, Kaiser Permanente Northern California has maintained a registry of all its patients with diabetes defined by laboratory results, pharmacy use, and outpatient diagnoses. In 2005, a race-stratified random sample of registry members, aged 30 to 75 years, was invited to participate in the Diabetes Study of Northern California (DISTANCE) survey (N = 40,735). The response rate was 62% (n = 20,188) based on a standard algorithm from the Council of American Survey Research Organization (7). Details about the survey and its methods have been published previously (8).

The DISTANCE survey was administered in four different versions, three of which included a HRQL instrument. Respondents were excluded from this study if they completed a survey version that did not have a HRQL instrument (n = 6,421) or if they were aged <60 years (n = 7,125). There were 6,642 respondents who received a survey that included a HRQL instrument and were aged 60 to 75 years, and 6,317 (95%) completed the HRQL instrument. Respondents who completed the HRQL instrument were significantly younger, more physically active, and less depressed, and reported a higher income compared with respondents who did not complete the instrument.

HRQL was assessed using an instrument based on the Short Form 8-item (SF-8) health survey, which includes one question for each domain of the Medical Outcomes Short Form-36 (SF-36) health survey (physical functioning, role limitations due to physical health, bodily pain, general health perceptions, vitality, social functioning, role limitations due to emotional problems, and mental health). Validated methods were used to transform responses to each question to the SF-36 equivalent scores, and overall physical and mental HRQL scores were calculated (9,10). HRQL scores can range from 0 (worst) to 100 (best) and have a mean of 50 (SD, 10) points. Studies on the meaningful difference in HRQL scores have suggested that differences between 0.3 to 0.5 SDs could be considered a minimally important difference in HRQL (11). Previous studies using the SF-36 have recommended not using less than a 1-point difference when interpreting results (12). We chose a cutoff of 3 points a priori to be conservative in describing conditions that were associated with meaningful differences in HRQL.

Administrative records were used to ascertain the presence versus absence of geriatric syndromes, diabetes complications, and hypoglycemia for the 1-year period before each subject’s survey date. Multiple sources of administrative records were used, including diagnostic codes from outpatient and inpatient records and the Kaiser End-Stage Renal Disease (ESRD) database used for reporting to the Health Care Financing Administration. We included diagnostic codes for each condition according to the clinical location in which the exposure typically occurred (outpatient, inpatient, or outpatient/inpatient). Details about the diagnostic codes and clinical locations are available in the Supplementary Table A1.

The geriatric syndromes were determined by a review of the geriatrics literature (13–15) and geriatric diabetes guidelines (2) and included chronic pain, depression, urinary incontinence, being underweight, and falls. The diabetes complications included amputation, blindness, congestive heart failure (CHF), ESRD, foot ulcer, myocardial infarction (MI), peripheral neuropathy, and stroke. Outpatient diagnoses were used to identify depression, incontinence, blindness, and peripheral neuropathy. Chronic pain was defined by outpatient diagnoses of arthritis, back/neck/musculoskeletal pain, headache, fibromyalgia, or reflex sympathetic dystrophy (16,17). Inpatient diagnoses were used to identify CHF, MI, and stroke. Outpatient and inpatient diagnoses were used to identify fall history, foot ulcer, and hypoglycemia. Amputation was identified by outpatient diagnoses and procedure codes. Being underweight was defined by a BMI <18.5 kg/m² calculated from outpatient records or survey data. ESRD was defined by an active record in the Kaiser ESRD database. In addition, two pooled exposure variables were created to specify if any geriatric syndrome or if any diabetes complication was present.

We included a range of factors that may confound the relationship between HRQL and geriatric syndromes, diabetes complications, and hypoglycemia. We obtained self-reported data on 1) sociodemographics (age, sex, race/ethnicity, education, income, marital status); 2) health behaviors (smoking history, alcohol use, physical activity [18]); and 3) diabetes-related factors (duration, treatment, self-monitoring of blood glucose). We included obesity (BMI ≥ 30 kg/m²) calculated from outpatient records or survey data.

We specified two types of linear regression models to examine associations between HRQL and geriatric syndromes, diabetes complications, and hypoglycemia: 1) exposure-specific models and 2) combined exposure models. Because conditions may co-occur and interact, each exposure was modeled individually as a predictor of HRQL in exposure-specific models. Also, we modeled pooled exposure variables and hypoglycemia simultaneously in combined exposure models to compare associations between HRQL and geriatric syndromes, diabetes complications, and hypoglycemia. All exposures were specified a priori as binary variables to indicate if the condition was present or absent. All covariates were modeled categorically, except for age and diabetes duration, and were included in all models. Expansion weights were applied to all models to account for the complex sampling design (race-stratified random sampling). We report β-values, which indicate differences in HRQL scores based on the presence versus absence of each exposure. P values were based on two-sided tests and considered significant at P < 0.05. Analyses were performed using SAS 9.0 software (SAS Institute Inc, Cary, NC).

Symptom-based conditions may be under-reported in administrative data; therefore, sensitivity analyses were conducted that respecified models using survey data instead of administrative data. Chronic pain was assessed by asking if bodily pain was present most of the time during the last 6 months (19). Depression was assessed using the 8-item Personal Health Questionnaire depression scale, and scores ≥15 were considered diagnostic (20). Peripheral neuropathy was defined by symptoms occurring ≥3 days/week. Hypoglycemia was defined by a history of “a severe low blood sugar reaction” in the past year. Urinary incontinence was asked only in women and included if it affected daily activities “quite a bit” or “extremely.”

In exposure-specific models, any geriatric syndrome (−5.9 [95% CI −6.5 to −5.4], P < 0.001), any diabetes complication (−4.5 [−5.1 to −4.0], P < 0.001), and hypoglycemia (−4.7 [−8.2 to −1.3], P < 0.001) were associated with lower physical HRQL (Table 2). The lowest physical HRQL was associated with amputation (−9.8 [−13.9 to −5.8], P < 0.001), CHF (−7.2 [−8.7 to −5.8], P < 0.001), falls (−6.3 [−8.6 to −4.1], P < 0.001), chronic pain (−5.6 [−6.1 to −5.0], P < 0.001), and MI (−5.3 [−7.0 to −3.5], P < 0.001). In combined exposure models, the association between physical HRQL and geriatric syndromes (−5.3 [−5.8 to −4.8], P < 0.001) and diabetes complications (−3.5 [−4.0 to −2.9], P < 0.001) were still clinically significant but attenuated. The association with hypoglycemia was no longer significant (Table 3).

Overall, geriatric syndromes and diabetes complications were not associated with clinically significant lower mental HRQL in exposure-specific models (Table 2). However, lower mental HRQL was associated with several individual geriatric syndromes and diabetes complications: depression (−7.9 [95% CI −8.7 to −7.2], P < 0.001), underweight (−5.6 [−9.7 to −1.4], P = 0.008), amputation (−5.5 [−9.0 to −2.0], P = 0.002), hypoglycemia (−5.0 [−8.0 to −2.0], P = 0.001), CHF (−4.0 [−5.3 to −2.7], P < 0.001), and falls (−3.6 [−5.5 to −1.6], P < 0.001). In combined exposure models, only hypoglycemia (−4.0 [−7.0 to −1.1], P = 0.008) remained associated with lower mental HRQL (Table 3).

The use of survey data instead of administrative data showed that fewer respondents had geriatric syndromes and diabetes complications (Table 4). Chronic pain, depression, and peripheral neuropathy occurred less frequently, whereas urinary incontinence and hypoglycemia occurred more frequently. In exposure-specific models, associations were larger between physical HRQL and any geriatric syndrome (survey vs. administrative data, −10.1 vs. −5.9 points) and any diabetes complication (−5.8 vs. −4.5 points). Specifically, larger differences in physical HRQL were associated with chronic pain (survey vs. administrative data, −10.3 vs. −5.6 points), depression (−10.3 vs. −4.9 points), and urinary incontinence (−7.3 vs. −3.1 points). Also, larger differences in mental HRQL were associated with depression (survey vs. administrative data, −17.3 vs. −7.9 points) and incontinence (−4.6 vs. −0.7 points). Associations with HRQL were slightly smaller for hypoglycemia.

In the combined exposure models, the association between physical HRQL and any geriatric syndromes was larger (survey vs. administrative data, −9.4 vs. −5.3 points), whereas associations with diabetes complications and hypoglycemia changed minimally. Associations between mental HRQL and any geriatric syndrome, any diabetes complication, and hypoglycemia did not change substantially.