Association of Diabetes and Insulin Therapy With Risk of Hospitalization for Infection and 28-Day Mortality Risk

Summary

Among 30 239 community-dwelling adults in the United States, diabetes was associated with increased risk of hospitalization for infection but not 28-day mortality risk. Insulin therapy conferred a greater risk of hospitalization, with no increase in 28-day mortality.

Abstract

Background.

Epidemiologic and experimental evidence suggests that individuals with diabetes are at increased risk of infection. We sought to examine the association of diabetes and insulin therapy with hospitalization for infection and 28-day mortality.

Methods.

We performed a prospective cohort study using data from 30 239 community-dwelling participants aged ≥45 years enrolled in the REasons for Geographic and Racial Differences in Stroke (REGARDS) study. We defined diabetes as a fasting glucose level ≥126 mg/L (or ≥200 mg/L for those not fasting), the use of insulin or oral hypoglycemic agents, or self-reported history. We identified infection-related hospitalizations over the years 2003–2012. We fit Cox proportional hazards models to assess the association of diabetes with hazard rates of infection and logistic regression models for 28-day mortality.

Results.

Among 29 683 patients from the REGARDS study with complete follow-up, 7375 had diabetes. Over a median follow-up period of 6.5 years, we identified 2593 first and 3411 total infection hospitalizations. In adjusted analyses, participants with diabetes had an increased hazard of infection (hazard ratio, 1.50; 95% confidence interval [CI], 1.37–1.64) compared with those without diabetes. Participants with diabetes hospitalized for infection did not have an increased odds of death within 28 days (odds ratio, 0.94; 95% CI, .67–1.32). Participants receiving insulin therapy had greater hazard of infection (hazard ratio, 2.18; 95% CI, 1.90–2.51) but no increased odds of mortality (odd ratio, 1.07; 95% CI, .67–1.71).

Conclusions.

Diabetes is associated with increased risk of hospitalization for infection. However, we did not find an association with 28-day mortality. Insulin therapy conferred an even greater risk of hospitalization, without increased mortality.

Diabetes and infection are common, expensive, and often concurrent conditions [1–3]. Infection can lead to sepsis, which is organ dysfunction due to a dysregulated host response to infection. Although the pathophysiology and treatment of diabetes has been intensively studied, the association between diabetes, infection, and mortality risk has not been fully elucidated. Current research suggests that diabetes is a risk factor for infection through various host- and organism-specific mechanisms. For example, hyperglycemia is associated with deleterious effects on granulocyte chemotaxis, phagocytosis, and adherence [4–6]. In addition, individuals with diabetes may also experience increased risk of infection due to local tissue ischemia resulting from vascular insufficiency, peripheral or autonomic diabetic neuropathy, colonization by organisms such as methicillin-resistant Staphylococcus aureus, and molecular mechanisms (eg, increased pathogen expression of adhesion molecules in response to high concentrations of glucose) [7–10].

The notion that diabetes may be a risk factor for infection is supported by several population-based studies [11–14]. Whether infected or septic patients with diabetes are more likely to die than their counterparts without diabetes, however, is not clear. The literature is conflicted, with some studies finding harm [15–18], others no influence [19–22], and still others a possible protective influence of diabetes on outcomes [23–25]. One explanation for the diversity of findings is methodologic issues in the studies to date. For example, the diabetes-mortality relationship has generally been assessed in studies comprising patients presenting to the hospital for care. The role of chronic insulin therapy in the association between diabetes and infection has also not been examined in detail, although prior studies suggest an association of insulin use with infections of the skin or soft tissues [26, 27]. Furthermore, insulin use may function as a marker for advanced disease and could affect infection hospitalization risk as well as outcomes.

To address the limitations of prior investigations and to better define the relationship between diabetes, incidence of infection hospitalization, and mortality risk after infection, we conducted a longitudinal investigation using data from the population-based REasons for Geographic and Racial Differences in Stroke (REGARDS) cohort. We hypothesized that community-dwelling adults with diabetes would be more likely than those without diabetes to be hospitalized with infection and that, once hospitalized for an infection, individuals with diabetes would have increased mortality. We also hypothesized that participants with diabetes receiving insulin therapy would have additional increased risk of infection hospitalization and mortality risk.

MATERIALS AND METHODS

Study Design

We used data from the REGARDS study, one of the largest population-based cohorts of community-dwelling adults in the United States [28]. REGARDS includes 30 239 black or white adults aged ≥45 years from the 48 contiguous US states and the District of Columbia [28]. The study oversampled blacks and individuals living in the Southeastern United States, with 21% of the cohort originating from the coastal plains of North Carolina, South Carolina, and Georgia (the “stroke buckle”), 35% from the remainder of North Carolina, South Carolina, and Georgia plus Tennessee, Mississippi, Alabama, Louisiana, and Arkansas (the “stroke belt”), and 44% from other states (“non–stroke belt”). REGARDS participants are 42% black and 45% male, with 69% aged >60 years at the time of enrollment.

The REGARDS study enrolled participants between 2003 and 2007, obtaining baseline data for each participant using phone interview and an in-person evaluation. REGARDS contacted study participants at 6-month intervals by telephone, identifying the date, location, and attributed reason for all hospitalizations. If the participant died, the study team reviewed death certificates, autopsy reports, and medical records and interviewed proxies to ascertain the circumstances of the participant’s death. The REGARDS study was approved by the institutional review boards of participating institutions, and all participants provided verbal consent before the telephone interview and written informed consent before the in-home study visit.

Identification of Serious Infection and Sepsis Events

We identified participants experiencing hospitalizations for serious infections, based on criteria introduced by Angus et al [29] (including admissions to the emergency department or hospital for infection). We searched all reported hospitalizations, and 2 trained abstractors independently reviewed all relevant medical records to confirm infection as a reason for hospitalization [29]. Sepsis events were defined as presentation with an infection plus ≥2 systemic inflammatory response syndrome criteria [30], based the worst vital signs and laboratory test values observed during the initial 28 hours of hospitalization. We defined severe sepsis as sepsis with dysfunction of ≥1 organ and septic shock as sepsis with hypotension refractory to fluid administration and/or the use of vasopressors, both according to previously published criteria for the Sepsis-related Organ Failure Assessment (SOFA) score [31]. Initial review of 1349 hospital records indicated excellent interrater agreement for presence of a serious infection (κ = 0.92). We included hospitalization events during the follow-up period, from 5 February 2003 through 31 December 2012. We also determined all-cause mortality risk within 28 days of infection, with deaths through 28 January 2013 included in this analysis.

Primary Exposure and Participant Characteristics

Participant characteristics were determined at the time of REGARDS enrollment. The primary exposure for this study was history or presence of diabetes. Diabetes was defined as a fasting glucose level ≥126 mg/L (or a glucose level ≥200 mg/L for those not fasting), the use of insulin or oral hypoglycemic agents, or self-reported history of diabetes. REGARDS study procedures did not permit differentiation between types 1 and 2 diabetes. We also defined participants with diabetes as users or nonusers of insulin therapy.

Demographic characteristics included age, sex, race, region, and self-reported annual household income and education. Region was defined as stroke buckle, stroke belt, or nonbelt. Health behaviors included smoking status and alcohol use. Smoking status was classified as current, former, or never. Alcohol use categories included none, moderate (1 drink per day for women or 2 drinks per day for men), and heavy (>1 drink per day for women and >2 drinks per day for men) [32]. We also identified body mass index and classified participants into standard categories: normal/underweight (<25 kg/m²), overweight (25–30 kg/m²), or obese (>30 kg/m²) [33].

Chronic conditions and biomarkers included lung disease, statin use, dyslipidemia, hypertension, myocardial infarction (MI), stroke, chronic kidney disease, and elevated high-sensitivity C-reactive protein (CRP). Based on the initial medication inventory, we defined use of β-agonists, leukotriene inhibitors, inhaled corticosteroids, combination inhalers, ipratropium, cromolyn, aminophylline, or theophylline as a surrogate for chronic lung disease. Statin use was defined in a similar manner, with specific statins reported by subjects including atorvastatin, fluvastatin, lovastatin, pravastatin, rosuvastatin, and simvastatin. Dyslipidemia was defined as low-density lipoprotein cholesterol levels >130 mg/dL or the use of lipid-lowering medications. Criteria for hypertension included systolic blood pressure ≥140 mm Hg, diastolic blood pressure ≥90 mm Hg, or use of antihypertensives. MI was defined as a self-reported history or baseline electrocardiographic evidence of MI. Participants self-reported history of prior stroke or transient ischemic attacks. Chronic kidney disease was defined as an estimated glomerular filtration rate <60 mL/min/1.73m². Serum creatinine was assayed using colorimetric reflectance spectrophotometry (Ortho Vitros Clinical Chemistry System 950IRC; Johnson & Johnson Clinical Diagnostics), and the estimated glomerular filtration rate was based on the Chronic Kidney Disease Epidemiology Collaboration equation [34]. Elevated CRP levels were defined as those >3.0 mg/dL; CRP was assayed using particle-enhanced immunonephelometry (N High-sensitivity CRP; Siemens).

Among participants hospitalized for infection, we also obtained hospitalization characteristics. We determined admission destination (intensive care unit or floor) and length of stay for events resulting in hospital admission. Infection type was also identified, based on available clinical documentation. Participants were additionally classified based on admission from or discharge to a nursing, assisted care, or rehabilitation facility.

Data Analysis

We compared participant characteristics between participants with or without diabetes, using t tests or nonparametric Wilcoxon rank-sum tests for continuous variables and Pearson χ² tests for categorical variables. Among participants hospitalized for infection, we also compared baseline and discharge characteristics using exact and nonparametric tests, as appropriate. We fit Cox proportional hazards models to assess the association of diabetes with rates of infection, including time from enrollment to first hospitalization as the outcome. Participants were censored at loss to follow-up, death, or 31 December 2012. Sequential models were constructed, with adjustment for demographics and health behaviors, chronic conditions, and body mass index class. We verified the proportional hazards assumption using Schoenfeld residuals and tests of the interaction with the logarithm of time. We present a complete case analysis but also performed the analysis using multiple imputation with chained equations to account for missing participant characteristic values (data not shown).

To assess the association of diabetes with mortality risk during hospitalization or within 28 days of hospitalization for infection or sepsis, we fit logistic regression models. Models were adjusted for factors associated with both diabetes and mortality, in addition to SOFA score. Participants with missing follow-up information after hospitalization who did not die during the infection event were excluded from mortality analyses. We performed a sensitivity analysis in which we used logistic regression models fit with generalized estimating equations to assess mortality among all infection hospitalizations (including those beyond the first event), adjusting for the same characteristics as outlined above. We performed an additional sensitivity analysis in which we limited mortality analyses to participants with sepsis or severe sepsis. We also constructed models for infection hospitalization risk and mortality with the primary exposure classified into 3 groups: no diabetes, diabetes with no insulin, and diabetes with insulin. We conducted all analyses using Stata 13.1 software (StataCorp).

RESULTS

A total of 30 239 patients were enrolled in REGARDS, of whom 29 683 had complete follow-up information. Among participants with complete follow-up, 7375 (24.9%) were identified as having diabetes at baseline (Figure 1). Compared with participants without diabetes, those with diabetes were older and more likely to be female, black, from the stroke belt, have a lower level of education and income, be overweight, and have a higher comorbid burden (Table 1). In addition, 1685 participants with diabetes (22.9%) were receiving insulin at the time of baseline enrollment. Participants with diabetes receiving insulin differed from those who were not in that they were more likely to be black, have lower education and income, and have an increased comorbid burden (eTable 1).

Figure 1.

Study population flowchart for the REasons for Geographic and Racial Differences in Stroke (REGARDS) study. Participants were excluded if their follow-up information was incomplete.

Table 1.

Baseline Participant Characteristics by Diabetes Status^a

Characteristic	Participants, %b		P Valuec
	No Diabetes (n = 22 308)	Diabetes (n = 7375)
Demographics
Age, mean (SD), y	64.7 (9.6)	65.6 (8.8)	< .001
Age group			< .001
45–49 y	5.5	3.3
50–59 y	27.2	23.1
60–69 y	36.5	41.1
70–79 y	23.5	26.0
≥80 y	7.3	6.6
Sex			.004
Male	44.4	46.4
Female	55.6	53.6
Race			< .001
White	63.6	44.5
Black	36.4	55.5
BMI class			< .001
Normal/underweight	28.8	12.3
Overweight	32.4	56.5
Obese	38.7	31.2
Missing	0.5	1.3
Educational level			< .001
Less than high school	10.3	19.1
High school graduate	25.0	28.3
Some college	26.9	26.5
College or higher	37.7	26.2
Missing	0.1	0.2
Annual income			< .001
<$20 000	15.7	25.1
$20 000–$34 000	23.3	26.8
$35 000–$74 000	31.0	25.8
≥$75 000	17.9	9.6
Declined to report	12.2	12.9
Geographic regiond			<.001
Stroke buckle	20.5	22.3
Stroke belt	34.4	35.5
Non–belt/buckle	45.1	42.3
Health behaviors
Smoking status			< .001
Current	14.6	14.2
Past	39.3	43.4
Never	46.2	42.4
Missing	0.4	0.3
Alcohol use			<.001
Heavy	4.7	2.0
Moderate	36.3	24.2
None	59.0	73.7
Missing	1.8	2.4
Chronic medical conditions and biomarkers
Chronic lung disease	8.9	10.2	.001
Chronic kidney disease	8.8	17.3	< .001
Elevated CRP level	35.8	48.4	< .001
History of stroke	5.1	10.3	< .001
History of MI	10.7	18.6	< .001
Hypertension	53.2	77.1	< .001
Dyslipidemia	53.3	71.2	< .001
Statin use	26.7	46.2	< .001

Abbreviations: BMI, body mass index; CRP, C-reactive protein; MI, myocardial infarction; SD, standard deviation.

^aA total of 29 683 REasons for Geographic and Racial Differences in Stroke (REGARDS) study participants were included in the analysis.

^bData represent percentage of participants unless otherwise specified.

^cP values based on Pearson χ² test of association for categorical variables and t test of equal means for continuous variables comparing participants with and without diabetes. Categories listed as missing were not included in tests of association.

^dSee Study Design for explanation of regions.

During the study period, there were 2593 first infection hospitalization events (8.7%) and 3411 total events. The median follow-up time was 6.6 years (interquartile range, 5.1–8.1 years) for censored participants and 3.6 years (1.8–5.5 years) for those hospitalized with infection. Among participants hospitalized for infection, those with diabetes were younger and more likely to be black, have skin or soft-tissue infections, meet criteria for severe sepsis, be discharged to a nursing home, have a higher SOFA score, and have longer stay when admitted (Table 2). For participants with diabetes hospitalized for infection, those using insulin were more likely to be black, have infections classified as skin or soft-tissue infection, sepsis, or catheter related, meet criteria for severe sepsis, be discharged to a nursing home, and have a higher SOFA score with a longer hospital stay (eTable 2).

Table 2.

Characteristics of First Hospitalization for Infection by Diabetes Status^a

Characteristic	Participants, %b		P Valuec
	No Diabetes (n = 1701)	Diabetes (n = 892)
Age, mean (SD), y	68.1 (9.4)	67.0 (8.8)	.005d
Sex			.65
Male	47.4	48.3
Female	52.6	51.7
Race			<.001
White	71.3	56.2
Black	28.7	43.8
Chronic medical conditions and biomarkers
Chronic kidney disease	15.5	24.1	<.001
Elevated CRP level	44.1	51.7	<.001
History of stroke	7.6	14.5	<.001
History of MI	16.4	25.1	<.001
Hypertension	61.5	78.3	<.001
Infection type
Pneumonia	30.9	30.5	.85
Kidney and urinary tract	18.1	19.1	.55
Abdominal	20.2	14.0	<.001
Skin and soft tissue	11.6	14.8	.02
Bronchitis, influenza, and other lung infections	10.8	9.8	.43
Sepsis	3.7	5.8	.01
Fever of unknown origin	1.7	1.4	.49
Surgical wound	1.0	1.2	.58
Catheter (intravenous, central, or dialysis)	0.1	0.6	.02e
Meningitis	0.3	0.1	.67e
Unknown/other	1.7	2.8	.06
Sepsis syndrome within 28 h	54.2	52.4	.37
Severe sepsis among participants with sepsis	69.3	80.3	<.001
Septic shock among participants with sepsis	15.1	14.6	.80
Worst syndrome within 28 h			<.001
Infection without sepsis	45.8	47.7
Sepsis only	16.6	10.3
Severe sepsis only	29.5	34.4
Septic shock	8.2	7.6
Admitted to ICU vs floor	7.1	8.5	.20
Admitted from nursing home	5.4	5.7	.70
Discharged to nursing home	6.8	10.7	.001
SOFA score, median (IQR)	1 (0–2)	1 (0–3)	<.001f
SOFA score category			<.001
0	42.7	30.0
1	23.6	26.7
2	14.8	16.6
3–4	13.1	17.5
5	5.9	9.2
Length of hospital stay, median (IQR), d	4 (3–7)	5 (3–8)	<.001f

Abbreviations: CRP, C-reactive protein; ICU, intensive care unit; IQR, interquartile range; MI, myocardial infarction; SD, standard deviation; SOFA, Sepsis-related Organ Failure Assessment.

^aBased on a total of 2593 initial infection events.

^bData represent percentage of participants unless otherwise specified.

^c P values based on Pearson χ² test of association for categorical variables with 75% of cell counts ≥5 for comparison of participants with or without diabetes.

^dBased on Fisher exact test for categorical variables with 25% of cell counts <5.

^eBased on t test of equal means for normally distributed continuous variables.

^fBased on nonparametric Wilcoxon rank sum test for nonsymmetric continuous variables.

Participants with diabetes had an increased hazard of hospitalization for infection than those without diabetes (Table 3; Figure 2A). After adjustment for demographic characteristics, health behaviors, and chronic conditions, the association between diabetes and infection hazard remained (Table 3). However, the odds of death within 28 days after an infection hospitalization did not differ between those with and those without diabetes (Table 4). In a sensitivity analysis including infection events beyond the first, we observed similar results when comparing the odds of death within 28 days between participants with diabetes and those without (unadjusted odds ratio [OR], 1.21 [95% confidence interval (CI), .95–1.54; adjusted OR, 0.85 [.64–1.13]). We also examined mortality risk for first infection events, restricting the population to those meeting the criteria for sepsis or severe sepsis, and we obtained similar results (eTable 3). Finally, performing the analysis using multiple imputation to account for missing values produced similar associations (data not shown).

Table 3.

Association of Baseline Diabetes and Insulin Status With Risk of First Infection^a

Exposure	Participants, No.	Infection Events, No. (%)	IR per 1000 Person-Years (95% CI)	HR (95% CI)
				Crude	Add Demographics and Health Behaviors	Add Chronic Medical Conditions	Add BMI Class
Diabetes status
No Diabetes	22 308	1701 (7.6)	12.2 (11.6–12.8)	Reference	Reference	Reference	Reference
Diabetes	7375	892 (12.1)	21.2 (19.9–22.7)	1.76 (1.62–1.91)	1.72 (1.58–1.87)	1.55 (1.42–1.69)	1.50 (1.37–1.64)
Diabetes and insulin status
No Diabetes	22 308	1701 (7.6)	12.2 (11.6–12.8)	Reference	Reference	Reference	Reference
Diabetes, no insulin therapy	5690	608 (10.7)	18.3 (16.9–19.9)	1.52 (1.38–1.67)	1.48 (1.34–1.63)	1.36 (1.24–1.50)	1.33 (1.21–1.47)
Diabetes, with insulin therapy	1685	284 (16.9)	32.1 (28.6–36.0)	2.68 (2.37–3.04)	2.74 (2.40–3.12)	2.29 (2.00–2.62)	2.18 (1.90–2.51)

Abbreviations: BMI, body mass index; CI, confidence interval; HR, hazard ratio; IR, incidence rate.

^aA total of 29 683 participants in the REasons for Geographic and Racial Differences in Stroke (REGARDS) study were included in the crude analysis, and 28 763 in the fully adjusted analysis. HRs estimated using Cox proportional hazard regression.

Demographics and health behaviors included age, sex, race, education, income, region, smoking status, and alcohol use. Chronic medical conditions included chronic lung disease, chronic kidney disease, elevated C-reactive protein level, stroke, myocardial infarction, hypertension, dyslipidemia, and statin use.

Figure 2.

Kaplan-Meier failure curves for infection hospitalization by diabetes and insulin status. A, Infection hospitalization by diabetes status. B, Infection hospitalization by diabetes and insulin status. A total of 29 683 REasons for Geographic and Racial Differences in Stroke (REGARDS) study participants were included in the infection risk analysis.

Table 4.

Association of Diabetes and Insulin Status With 28-Day Mortality Risk After First Infection^a

Exposure	Infection Events, No.	Deaths, No. (%)	OR (95% CI)
			Crude	Add Baseline Factorsb	Add SOFA Score
Diabetes status
No diabetes	1652	110 (6.7)	Reference	Reference	Reference
Diabetes	863	73 (8.5)	1.30 (.95–1.76)	1.07 (.78–1.47)	0.94 (.67–1.32)
Diabetes and insulin status
No diabetes	1652	110 (6.7)	Reference	Reference	Reference
Diabetes, no insulin therapy	589	46 (7.8)	1.19 (.83–1.70)	1.02 (.71–1.47)	0.87 (.59–1.30)
Diabetes with insulin therapy	274	27 (9.9)	1.53 (.99–1.38)	1.18 (.75–1.86)	1.07 (.67–1.71)

Abbreviations: CI, confidence interval; OR, odds ratio; SOFA, Sepsis-related Organ Failure Assessment.

^aA total of 2515 initial infection events were included in this analysis, excluding 78 events without follow-up. The ORs were estimated using logistic regression, and the 28-day mortality included deaths during hospitalization or within 28 days after admission for infection.

^bAdjusted for chronic kidney disease, elevated C-reactive protein level, stroke, myocardial infarction, and hypertension.

When we tested our secondary hypothesis about the differential effects among participants with diabetes on insulin therapy, we found that participants with diabetes receiving insulin had a >2-fold higher hazard of hospitalization for infection (hazard ratio, 2.18; 95% CI, 1.90–2.51) than those without diabetes, an effect present but attenuated among those with diabetes not receiving insulin therapy (1.33; 1.21–1.47) (Table 3; Figure 2B) However, once again, there was no increase in the odds of death within 28 days for participants receiving insulin (OR, 1.07; 95% CI, .67–1.71) (Table 3; eTable 3).

DISCUSSION

Our observations affirm that diabetes is associated with increased hazard of infection hospitalization. We also found that participants receiving insulin therapy experienced a more pronounced infection risk. However, somewhat surprisingly, neither diabetes nor insulin use was associated with increased 28-day mortality after infection. The current study builds on prior efforts by linking robust baseline data with infection incidence and outcomes in a large cohort of community-dwelling adults. Our results suggest a need for improved control of risk factors for infection among individuals with diabetes, as well as continuing research on differential infection mechanisms during diabetic conditions.

There are a number of plausible explanations for increased infection among participants with diabetes. Diabetic complications, such as neuropathy, microvascular damage, and glycosuria, are recognizable contributors to the propensity for skin, soft-tissue, and urinary tract infections [8, 9]. Impaired immunologic function (reduced neutrophil and macrophage activity or T-lymphocyte subset alteration) in the presence of hyperglycemia may also be responsible for increased risk of pneumonia [15, 18]. Complications aside, diabetic infections occur in an altered physiologic milieu, where the intrinsic properties of the host have been altered by repeated or continual hyperglycemia and/or metabolic syndrome. Concentrations and effects of metabolic mediators other than insulin with in vivo effects on immunologic function (eg, adipokines such as resistin and leptin) may also be altered in these individuals [35, 36].

We also showed that participants with diabetes receiving insulin therapy experience an even higher risk of infection than those not receiving insulin. This finding could be explained by increased severity of disease, a mechanism directly related to insulin, or increased event detection. The mechanisms outlined above that might explain why diabetes is generally associated with increased risk of infection may be more pronounced in the population receiving insulin therapy. Injecting insulin regularly could also have increased the occurrence or severity of skin or soft-tissue infections by allowing pathogenic organisms to more easily cross the dermal barrier [26, 27]. However, it is important to consider alternative explanations not directly related to the disease. For instance, the population receiving insulin may have been more closely monitored for infectious complications, thus reducing the threshold for admission by a care provider. Further research should examine the association between insulin use and infection in greater detail to better understand potential mechanisms.

Our finding of similar infection- and sepsis-associated mortality risk, regardless of diabetes status, fits with findings of prior studies showing that diabetes is not related to in-hospital infection or sepsis-related mortality [19, 21, 22]. Interesting arguments have been advanced regarding the effect of diabetes during an infection, which may contribute to the observations of Schuetz et al [37] and Whitcomb et al [38], in that these studies found diabetes to be protective. Authors of a large epidemiologic study using the National Hospital Discharge Survey found that patients with diabetes were less likely to develop acute respiratory distress syndrome and postulated that this difference could be due to a blunted inflammatory response [39]. In a similar manner, the low-level of chronic inflammation associated with obesity and hyperglycemia may prime the body for oxidant stress associated with sepsis [40]. The same metabolic syndrome and obesity that is commensurate with rising rates of type 2 diabetes in the United States may then provide a caloric reserve that exerts a protective effect through an intensive care unit stay [41, 42].

As findings of an observational, population-based study, these results in and of themselves do not yield insight into the underlying mechanisms linking diabetes with infection and sepsis. However, our study improves on prior efforts, which were unable to combine longitudinal infection incidence with granular acute infection event data. These results offer further grounds for investigating why diabetes may not contribute to increased infection or sepsis mortality, such as through diminution of the immunologic or inflammatory cascade or through greater caloric reserve to survive critical illness. They also reinforce the need for proactive community-based interventions to reduce the risk of infection, such as regular preventive diabetic foot and skin checks, with adequate wound care.

Our study has several notable strengths, including a large sample size, standardized infection hospitalization detection and adjudication, comprehensive baseline participant data, and a follow-up period of nearly 10 years. However, our results must be interpreted in light of several important limitations. The REGARDS cohort enrolled black and white community-dwelling adults aged ≥45 years, which limits the generalizability of our results. For instance, our findings may not generalize to younger individuals or those of Hispanic ethnicity. Baseline diabetes status was ascertained, but we were not able to identify incident diabetes during the course of the follow-up period. However, it is unclear whether or not this misclassification would be differential with respect to infection hazard. We also were not able to obtain data on glycemic control or diabetes duration. Despite a comprehensive approach to identification of infection events, we may have failed to capture all hospitalizations. Further efforts should examine this association and collect more granular details pertaining to diabetes, as well as factors affecting hospitalization and postdischarge outcomes.

In conclusion, using data from one of the largest population-based cohort studies in the United States, we found that diabetes is associated with a higher risk of hospitalization for infection, but those with diabetes are no more likely to die from these conditions within 28 days of hospitalization. Furthermore, individuals with diabetes on insulin therapy are at higher risk of acquiring infection but are similarly not at increased risk of death. Our findings highlight the need for improved control of infection risk factors among individuals with diabetes and further research to explain why diabetes may not be associated with poor infection outcomes.

Notes

Acknowledgments. The authors thank the other investigators, the staff, and the participants of the REasons for Geographic and Racial Differences in Stroke (REGARDS) study for their valuable contributions. A full list of participating REGARDS investigators and institutions can be found at http://www.regardsstudy.org and http://www.regardssepsis.org.

Author contributions. H. E. W. and N. I. S. conceived the study. M. M. S. and H.E.W. organized and oversaw data collection. J. P. D., H. E. W., and N. I. S. conducted the analysis, and all authors contributed to review of results. J. P. D. and S. N. produced an initial draft of the manuscript, and all authors contributed to its editorial review and revision. J. P. D., H. E. W., and N. I. S. assume responsibility for the work as a whole.

Disclaimer. The content is solely the responsibility of the authors and does not necessarily represent the official views of the funding agencies. Representatives of the funding agencies have been involved in the review of the manuscript but not directly involved in the collection, management, analysis or interpretation of the data.

Financial support. This study was supported by the National Institute for Nursing Research (grant R01-NR012726), the National Center for Research Resources (grant UL1-RR025777), the Center for Clinical and Translational Science, the Lister Hill Center for Health Policy of the University of Alabama at Birmingham, and the Agency for Healthcare Research and Quality (grant T32-HS013852 to J. P. D.)The parent REGARDS study was supported by the National Institute of Neurological Disorders and Stroke, National Institutes of Health (cooperative agreement U01-NS041588).

Potential conflicts of interest. M. M. S. reports investigator-initiated research from Amgen.N. I. S. reports grants from Thermo Fisher Scientific and Rapid Pathogen Screening. All others have no disclosures. The authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.