A total of 16,650,296 hospitalizations from NIS were included in the sample for primary analysis, and of these participants 12,080,290 had normoglycemia, 116,779 had prediabetes and 4,453,227 had diabetes. The characteristics of the study population are shown in Table 1. Among patients with prediabetes, the mean age was 61.4 years (SD 14.4), and 60,440 (51.8%) were female and 72,322 (64.2%) were white. Compared with participants with normoglycemia, those with prediabetes or diabetes were older and cost more (P < 0.05). Patients with diabetes had a longer length of stay than those with normoglycemia or prediabetes (P < 0.05).
Table 1
Baseline characteristics of the study sample.
| Normoglycemia (n = 12,080,290) | Prediabetes (n = 116,779) | Diabetes (n = 4,453,227) |
Age (years) * | 51.9 (19.5) | 61.4 (14.4) | 63.6 (13.7) |
Age group * | | | |
Youth (18 to < 45 years) | 4,694,236 (38.9%) | 15,551 (13.3%) | 432,048 (9.7%) |
Middle age (45 to < 65 years) | 3,499,388 (29.0%) | 46,456 (39.8%) | 1,669,873 (37.5%) |
Old age (65 to ≤ 85 years) | 3,886,666 (32.2%) | 54,772 (46.9%) | 2,351,306 (52.8%) |
Women * | 7,302,622 (60.5%) | 60,440 (51.8%) | 2,173,370 (48.8%) |
Race * | | | |
White | 7,843,881 (67.4%) | 72,322 (64.2%) | 2,700,876 (62.5%) |
Black | 1,720,532 (14.8%) | 17,109 (15.2%) | 812,117 (18.8%) |
Hispanic | 1,313,440 (11.3%) | 14,092 (12.5%) | 531,623 (12.3%) |
Asian or Pacific Islander | 322,615 (2.8%) | 5,298 (4.7%) | 115,567 (2.7%) |
Native American | 69,505 (0.6%) | 733 (0.7%) | 36,866 (0.9%) |
Other | 367,080 (3.2%) | 3,121 (2.8%) | 125,317 (2.9%) |
Smoking * | 2,210,713 (18.3%) | 32,055 (27.4%) | 1,099,697 (24.7%) |
Alcohol abuse/dependence * | 867,249 (7.2%) | 6,292 (5.4%) | 170,180 (3.8%) |
Primary expected payer * | | | |
Medicare | 4,421,944 (36.7%) | 55,699 (47.8%) | 2,721,469 (61.2%) |
Medicaid | 2,700,052 (22.4%) | 15,549 (13.3%) | 620,809 (14.0%) |
Private insurance | 3,936,256 (32.6%) | 38,487 (33.0%) | 842,877 (19.0%) |
Self-pay | 566,133 (4.7%) | 3,801 (3.3%) | 141,196 (3.2%) |
No charge | 49,947 (0.4%) | 348 (0.3%) | 12,977 (0.3%) |
Other | 386,428 (3.2%) | 2,721 (2.3%) | 108,121 (2.4%) |
Median household income for patient's ZIP Code * |
0-25th percentile | 3,502,111 (29.5%) | 26,448 (23.0%) | 1,526,313 (34.9%) |
26th to 50th percentile (median) | 3,102,730 (26.2%) | 28,264 (24.6%) | 1,191,160 (27.2%) |
51st to 75th percentile | 2,840,124 (24.0%) | 30,393 (26.5%) | 975,978 (22.3%) |
76th to 100th percentile | 2,412,055 (20.3%) | 29,766 (25.9%) | 681,405 (15.6%) |
Total charge * | 51,163 (90,793.9) | 61,427.9 (76,937.8) | 62,051.8 (91,491.5) |
Length of stay * | 4.5 (6.4) | 4.3 (5.9) | 5.4 (6.7) |
Patient Location: NCHS Urban-Rural Code* |
“Central” counties of metro areas of > = 1 million population | 3,554,978 (29.6%) | 42,336 (36.4%) | 1,312,583 (29.6%) |
“Fringe” counties of metro areas of > = 1 million population | 2,919,663 (24.3%) | 27,679 (23.8%) | 1,010,121 (22.8%) |
Counties in metro areas of 250,000-999,999 population | 2,496,754 (20.8%) | 22,921 (19.7%) | 916,088 (20.7%) |
Counties in metro areas of 50,000-249,999 population | 1,115,613 (9.3%) | 9,693 (8.3%) | 423,245 (9.5%) |
Micropolitan counties | 1,100,338 (9.2%) | 8,037 (6.9%) | 435,827 (9.8%) |
Not metropolitan or micropolitan counties | 819,195 (6.8%) | 5,580 (4.8%) | 338,209 (7.6%) |
Data are mean (SD) or n (%). * The three glycemic status groups have different distributions of these variables, with p < 0.05 using a χ² test.
NCHS National Center for Health Statistics, ZIP Zone Improvement Plan.
The results of multivariable regression models for the effects of prediabetes on an array of health outcomes among the whole study population are shown in Fig. 2. ORs were adjusted for age, sex, race, smoking, alcohol abuse/dependence, primary expected payer, median household income, total charge, LOS, and patient location. Compared with normoglycemia, prediabetes was significantly associated with an increased risk of a range of diseases, such as polycystic ovary syndrome (OR: 8.92 [95% CI, 8.30–9.59]), overweight and obesity (OR: 4.01 [95% CI, 3.95–4.06]), and arteriosclerosis (OR: 3.45 [95% CI, 3.35–3.56]). Furthermore, almost all cardiovascular diseases and three types of eye diseases in our predefined list were related to prediabetes (OR > 1). The unadjusted ORs for risk of 76 common diseases associated with prediabetes were shown in Supplementary Table 2.
Ignoring lesser effects (adjusted OR < 1.5) and associations with a P value of 6.6×10− 4 or greater, prediabetes was strongly associated with 23 of the 76 diseases studied (Fig. 2). Further, multimorbidity analyses were based on 22 non-overlapping prediabetes-related diseases, including cardiovascular (hypertension, angina pectoris, myocardial infarction, cerebral infarction, arteriosclerosis), nervous system (headaches, transient ischemic attack, sleep disorders - mostly sleep apnea syndrome), musculoskeletal (gout, osteoarthritis, sciatica, back pain), endocrine (overweight and obesity, hyperlipidemia, polycystic ovary syndrome), eye (choroidal and retinal diseases, cataract and other lens disorders, glaucoma), ear (hearing loss), respiratory (asthma), cancer (endometrial cancer), and genitourinary (infertility) diseases. The number of diseases significantly related to prediabetes is more than that of diabetes (Supplementary Table 3). Meanwhile, the results of sensitivity analysis were similar to that of the overall study population (Supplementary Table 4 and Supplementary Table 5).
The relative risks of co-existence of the one, two, three, and four or more prediabetes-related diseases by glycemic status are shown in Table 2. The adjusted OR for prediabetes versus normoglycemia was 2.53 (95% CI 2.47–2.59) for accompanying at least one prediabetes-related disease, 4.74 (4.63–4.85) for accompanying two diseases (simple multimorbidity), 7.35 (7.17–7.54) for accompanying three diseases, and 11.74 (11.43–12.05) for accompanying four or more diseases (complex multimorbidity). Participants with prediabetes were at a higher risk of accompanying prediabetes-related multimorbidity compared with patients with normoglycemia. Further sex-specific analysis showed similar increases in the risk of prediabetes-related multimorbidity (Supplementary Table 6).
Table 2
Association of prediabetes with co-existing prediabetes-related disease and multimorbidity
Number of prediabetes-related diseases | Cases, n (%) | OR (95% CI) | Adjusted OR (95% CI) * |
Prediabetes (N = 116,779) | | |
One | 25,174 (21.6) | 3.46 (3.38–3.54) | 2.53 (2.47–2.59) |
Two | 31,842 (27.3) | 6.65 (6.51–6.79) | 4.74 (4.63–4.85) |
Three | 25,853 (22.1) | 10.6 (10.4–10.8) | 7.35 (7.17–7.54) |
Four or more | 22,368 (19.2) | 17.9 (17.5–18.3) | 11.74 (11.43–12.05) |
IFG (N = 12,469) | | | |
One | 2,230 (17.9) | 3.15 (2.91–3.40) | 2.11 (1.94–2.30) |
Two | 3,187 (25.6) | 5.92 (5.50–6.37) | 3.64 (3.35–3.96) |
Three | 3,080 (24.7) | 10.2 (9.52–11.03) | 6.12 (5.62–6.65) |
Four or more | 3,055 (24.5) | 17.7 (16.43–19.04) | 10.12 (9.30-11.02) |
IGT (N = 7,201) | | | |
One | 1,581 (22.0) | 1.69 (1.57–1.83) | 1.70 (1.56–1.85) |
Two | 1,754 (24.4) | 2.48 (2.30–2.66) | 2.38 (2.18–2.60) |
Three | 1,425 (19.8) | 3.60 (3.33–3.89) | 3.35 (3.05–3.68) |
Four or more | 1,234 (17.1) | 5.42 (5.01–5.87) | 4.96 (4.50–5.47) |
*OR for prediabetes versus normoglycemia, adjusted for age, sex, race, smoking, and alcohol abuse/dependence, primary expected payer, median household income, total charge, length of stay, patient location; Four disease outcomes: co-existing of one, two, three, and four or more prediabetes-related diseases (in this case, complex multimorbidity).
OR odds ratio, CI confidence interval, IFG impaired fasting glucose, IGT impaired glucose tolerance.
In the prediabetes population, only 12,469 patients were diagnosed with IFG, and 7,201 patients were diagnosed with IGT. Compared with normoglycemia, IGT was significantly associated with an increased risk of co-existing complex multimorbidity (OR: 4.96 [95% CI 4.50–5.47]), while those with IFG had a higher risk of co-existing complex multimorbidity (OR: 10.12 [95% CI 9.30-11.02]). In the age-specific analysis (Supplementary Table 7), we found the prevalence of each disease outcome was lower in the youth group compared to the other two groups. Patients aged 18 to < 45 years had a higher risk of co-existing complex multimorbidity (OR: 42.32 [95% CI 39.65–45.16]) compared to the other two groups.
Figure 3 demonstrates the proportions of the 23 prediabetes-related diseases among the prediabetes patients with complex multimorbidity. The prediabetes-related diseases that occurred most frequently in patients with prediabetes and complex multimorbidity were hyperlipidemia (17,271 [77.2%] of 22,368 patients), hypertension (17,198 [76.9%]), overweight and obesity (14,927 [66.7%]), sleep disorders (11,783 [52.7%]), and osteoarthritis (11,356 [50.8%]). The relevant proportions in patients with different numbers of prediabetes-related diseases are shown in Supplementary Table 8. Nearly half of people with simple multimorbidity (i.e., two comorbid diseases) suffer from hypertension (51.7%) or/and hyperlipidemia (48.8%).
The proportion of three age segments (18 to ≤ 45, 45 to ≤ 65, and > 65 years) among the prediabetes patients with different numbers of prediabetes-related diseases is shown in Fig. 4. In the four disease outcomes subpopulations, the proportion of the old age group was the highest among all groups. Meanwhile, the proportion of old age people increased from 42.7% in the group with one prediabetes-related disease to 52.8% in the group with complex multimorbidity.