Skip to main content

Table 3 Description of people with diabetes identified through cluster analysis

From: Evaluating the relationship between clinical and demographic characteristics of insulin-using people with diabetes and their health outcomes: a cluster analysis application

  CLUSTER NUMBER
  1 2 3 4 5 6 7
Cluster Label Low ER Utilizers with High Comorbid Burden Endocrinology Utilizers with Concordant Comorbidities Low ER Utilizers with High Diabetes-related Claims without Discordant Comorbidities High Inpatient Utilizers with High Comorbid Burden High ER Utilizers Endocrinology Utilizers without Concordant Comorbidities Lowest Overall Utilizers
Number of People (n, %) 22,508 (22.4%) 13,986 (13.9%) 22,880 (22.7%) 5629 (5.6%) 11,351 (11.3%) 10,049 (10.0%) 14,247 (14.2%)
Age (mean, SD) (years) 69.80 (9.85) 65.16 (10.83) 60.71 (10.83) 69.54 (11.98) 64.92 (12.54) 38.34 (12.16) 61.31 (12.51)
Male (%) 48.0% 50.8% 45.4% 48.9% 54.8% 43.0% 49.5%
Household Income (US dollars)1 Median (Q1 - Q3) $51,665 ($42,344–$65,516) $59,733 ($47,149–$77,449) $51,879 ($42,139–$66,124) $53,612 ($43,017–$68,301) $49,154 ($40,163–$61,588) $65,972 ($51,625–$84,930) $54,107 ($43,152–$69,429)
Medicare Advantage (%) 80.3% 62.2% 42.6% 81.5% 72.5% 3.8% 51.7%
Diabetes Type T1DM (%) 3.3% 13.6% 2.2% 2.1% 3.6% 69.6% 8.1%
Number of HbA1c Tests
 One or more HbA1c tests performed (%) 66.4% 75.8% 71.3% 54.3% 63.8% 63.0% 15.2%
 Mean (SD) 1.01 (0.91) 1.29 (1.00) 1.02 (0.83) 0.89 (1.21) 1.00 (0.99) 0.87 (0.81) 0.17 (0.41)
Latest HbA1c Value
 N 7007 6215 7494 1471 3260 3045 2918
 Mean (SD) 8.56 (1.79) 8.60 (1.69) 9.35 (2.07) 8.25 (1.81) 8.95 (2.05) 8.82 (2.02) 8.87 (1.97)
Utilization of Diabetes Technology
 Pump (%) 2.8% 10.6% 2.1% 4.2% 4.3% 29.2% 1.7%
 BGM (%) 33.3% 38.6% 31.8% 29.6% 35.3% 39.6% 26.2%
 CGM (%) 1.0% 6.8% 0.8% 1.0% 1.2% 14.5% 0.3%
Type of Insulin (%)
 Basal only 45.0% 27.9% 55.3% 32.5% 40.8% 17.3% 47.4%
 Bolus only 6.2% 15.6% 6.0% 10.9% 8.2% 36.7% 7.2%
 Both 48.7% 56.5% 38.8% 56.6% 51.0% 46.0% 45.3%
Months on Insulin (mean, SD) 4.45 (3.58) 5.02 (3.72) 3.34 (3.15) 2.68 (3.18) 3.70 (3.47) 4.29 (3.25) 3.87 (3.42)
Clinical Characteristics
Number of Concordant Comorbidities (mean, SD) 5.56 (2.17) 4.56 (2.31) 3.29 (1.61) 6.72 (2.34) 5.53 (2.34) 0.52 (0.76) 3.67 (2.19)
Number of Discordant Comorbidities (mean, SD) 1.40 (1.07) 1.46 (1.26) 0.50 (0.74) 1.44 (1.16) 1.31 (1.14) 0.89 (1.10) 0.88 (0.99)
Charlson Comorbidity Index (CCI) (mean, SD) 2.70 (1.44) 2.47 (1.50) 1.15 (0.78) 3.92 (2.03) 2.47 (1.61) 1.09 (0.74) 0.30 (0.55)
Diabetes Complication Severity Index (DCSI) (mean, SD) 1.79 (1.39) 1.33 (1.39) 0.26 (0.58) 3.48 (1.95) 1.66 (1.53) 0.22 (0.56) 0.15 (0.49)
Median (Q1 - Q3) Number of Medical Claims 11.00 (7.00–18.00) 11.00 (7.00–17.00) 4.00 (2.00–6.00) 43.00 (30.00–64.00) 14.00 (9.00–23.00) 4.00 (2.00–7.00) 3.00 (0.00–6.00)
Proportion of Medical Claims 2,3 That are Diabetes Related (mean, SD) 0.46 (0.24) 0.53 (0.24) 0.78 (0.23) 0.53 (0.25) 0.51 (0.24) 0.69 (0.30) 0.19 (0.22)
Healthcare Resource Utilization4
All-cause Utilization
  All-cause ER visits (%) 0.0% 10.7% 6.2% 45.2% 100.0% 12.0% 6.5%
  All-cause ER visits (mean, SD) 0.00 (0.00) 0.12 (0.36) 0.07 (0.26) 0.82 (1.37) 1.77 (1.50) 0.15 (0.44) 0.07 (0.28)
  All-cause IP visits (%) 11.6% 7.5% 5.4% 99.1% 22.2% 6.0% 3.2%
  All-cause IP visits (mean, SD) 0.13 (0.36) 0.08 (0.31) 0.06 (0.25) 2.11 (1.29) 0.26 (0.53) 0.07 (0.28) 0.03 (0.20)
  All-cause IP days for patients with ≥1 admit (mean, SD) 3.84 (2.04) 3.88 (2.69) 4.07 (2.81) 25.34 (24.57) 4.13 (2.75) 3.61 (2.96) 5.03 (4.14)
  All-cause PCP visits (%) 83.7% 71.4% 79.8% 81.3% 84.1% 55.8% 41.8%
  All-cause PCP visits (mean, SD) 2.60 (2.32) 1.73 (1.88) 1.87 (1.58) 3.44 (3.69) 3.31 (2.97) 1.10 (1.41) 0.86 (1.39)
  All-cause Endocrinologist visits (%) 0.7% 100.0% 4.0% 13.5% 12.1% 39.2% 2.1%
  All-cause Endocrinologist visits (mean, SD) 0.01 (0.08) 1.83 (0.94) 0.04 (0.23) 0.22 (0.69) 0.16 (0.47) 0.58 (0.85) 0.02 (0.15)
  All-cause Cardiologist visits (%) 29.0% 27.5% 5.1% 36.8% 28.8% 2.0% 6.1%
  All-cause Cardiologist visits (mean, SD) 0.46 (0.98) 0.43 (0.90) 0.07 (0.33) 0.79 (1.48) 0.53 (1.13) 0.03 (0.24) 0.08 (0.37)
  No visit to PCP, Endocrinologist or Cardiologist (%) 12.4% 0.0% 17.4% 12.7% 11.2% 21.4% 54.6%
Diabetes-related Utilization3
  Diabetes-related ER visits (%) 0.0% 8.5% 5.2% 35.6% 80.8% 9.3% 4.0%
  Diabetes-related IP visits (%) 11.2% 7.3% 5.4% 95.7% 21.3% 5.8% 2.7%
  Diabetes-related IP days for patients with ≥1 admit (mean, SD) 3.81 (2.02) 3.87 (2.67) 4.05 (2.79) 23.64 (22.99) 4.07 (2.71) 3.52 (2.84) 4.88 (4.07)
Cluster Characteristics • No patients with all-cause or diabetes-related ER visits
• Second highest mean CCI and DCSI scores
• 100% had an Endocrinologist visit
• Highest proportion with 1+ HbA1c tests performed (76%)
• Highest average HbA1c lab value (mean = 9.35)
• Highest proportion of medical claims that were diabetes-related (mean = 0.78)
• Highest proportion of new insulin users (mean = 2.68 months)
• Highest mean CCI and DCSI scores
• Highest average number of medical claims (median = 43)
• Nearly 100% had an Inpatient Hospitalization
• 100% had an ER visit
• 81% had a diabetes-related ER visit
• Higher observed HbA1c (mean = 8.95)
• Highest proportion of type 1 diabetics (70%)
• Youngest (mean age = 38 years)
• Relatively high proportion with an Endocrinology visit (39%)
• Highest diabetes technology use (insulin pump, CGM, and/or BGM)
• Healthiest, according to CCI and DCSI scores
• Largest proportion without PCP, Endocrinologist, or Cardiologist visits (55%)
• Lowest total number of medical claims (median = 3)
  1. Abbreviations: BGM = blood glucose monitor; CCI = Charlson Comorbidity Index; CGM = continuous glucose monitor; DCSI = diabetes complications severity index; ER = emergency room; HbA1c = hemoglobin A1c; IP = inpatient; n, N = number of people with diabetes; PCP = primary care physician; Q1 = first quartile; Q3 = third quartile; SD = standard deviation; T1DM = type 1 diabetes mellitus
  2. 1 Missing values for median household income were imputed for 454 patients for the purpose of cluster analysis
  3. 2 Proportion is set to zero for those members who had zero medical claims overall during the baseline period
  4. 3 Diabetes–related determined by diagnosis in any diagnostic position on a medical claim
  5. 4 Statistical comparisons for average utilization were conducted using Kruskal-Wallis tests