Characterizing Hand Infections in an Underserved Population: The Role of Diabetic Status in Antibiotic Choice and Infection Location

Abstract

Introduction Patients with diabetes mellitus (DM) in underserved communities are at greater risk for hand infections. We aimed to describe the features of hand infections presenting to an urban hospital via laboratories, microbiology, and antibiotic choice with respect to diabetic status.

Materials and Methods Patients presenting with any hand infection were reviewed and stratified by DM status and infection location. Labs, culture results, antibiotic regimens, and significant predictors of laboratories or infection location were analyzed.

Results Fifty-three patients were included: DM ( n = 24), no-DM ( n = 24), and unknown status ( n = 5). Culture rates were comparable between all groups. Mean erythrocyte sedimentation rate (ESR) was significantly higher in DM (76.19 vs. 51.33); mean white blood cell count (WBC) and C-reactive protein (CRP) were comparable. Diabetics had higher odds of increased ESR (odds ratio [OR] = 1.03). Diabetics received vancomycin/piperacillin/tazobactam (VAN/PTZ) significantly more often (52% vs. 8%). Providers treated DM with VAN/PTZ or any VAN-containing regimen more often than with any other regimen. Proximal infections had significantly higher mean CRP (136.9 vs. 50.5) and WBC (5.19 vs. 3.9) and higher CRP (OR = 1.02).

Conclusion This study highlights the need for systematic criteria to better risk- stratify patients for appropriate antibiotic treatment. It may not be appropriate to treat both groups differently, as overly aggressive antibiotic selection may contribute to drug-resistance development.

Introduction

Patients with diabetes mellitus (DM) from underserved communities are at greater risk for infections of the hand compared with the general population. 1 2 3 4 Hand abscesses can lead to significant morbidity, including stiffness, chronic pain, amputation, and sepsis without prompt and aggressive treatment. Diabetic patients are particularly at risk if improper antibiotic selection delays treatment; amputation rates of up to 63% have been reported in the past for diabetics with hand infections, and death from sepsis is possible. 5 6 7 Early identification and prompt, aggressive medical and surgical treatment have been shown to improve morbidity. 2 8 9 10

Objective triage data, such as diabetic status and laboratory values, may play a role in the development of risk stratification systems aimed at the selection of an appropriate empiric therapy. Whereas community-acquired hand infections in the nondiabetic patient are more likely monomicrobial and gram-positive in nature, it is well established that such hand infections in diabetic patients are more likely to be polymicrobial in nature, with variable combinations of gram-staining, drug resistance, and metabolic properties and/or fungal origin. 11 12 Diabetes is also a risk factor for infection with methicillin-resistant Staphylococcus aureus (MRSA). 12 13 Early and objective risk stratification of the diabetic with a polymicrobial/severe hand infection, as in the case of skin and soft tissue infections in adults, 14 may improve morbidity while avoiding the negative effects of overtreatment, including antibiotic side effects or promotion of development of drug- resistant organisms.

This study therefore sought to compare and characterize, by diabetic status, hand infections presenting to an urban hospital in a 2-year period regarding: (1) serum inflammatory markers, (2) infection site, (3) infectious agent, (4) provider’s choice of antibiotic therapy, and (5) 90-day adverse outcomes and length of stay (LOS). These characteristics were examined with the goal of determining which clinical features were suggestive of infection and could help guide ultimate antibiotic choice.

Materials and Methods

This was a retrospective review of a prospectively collected, single-center database. Patients who presented to an urban hospital between October 2014 and October 2016 with an infection of the hand were identified and then stratified into groups by the presence or absence of DM, as reported by the patient. Information regarding insulin dependency was not available, nor was it available about individual providers’ antibiotic selection criteria. Patients were also stratified by site of infection: proximal to the digit (proximal) or within the digit (distal). Infection types encountered included felons, other abscesses, paronychias, and hand infections not otherwise specified. Patients with recent history of surgery, comorbid infection proximal to the hand, history of osteomyelitis, or human or animal bite mechanisms were excluded.

Statistical Methods

Analysis of variance (ANOVA) was used to compare the prescribed antibiotic regimens between and within disease groups. Ninety-day adverse outcomes (any complication, readmission, or subsequent surgery) were identified. Diabetes status, hemoglobin A _1C (HbA _1C ), blood glucose (mg/dL), white blood cell count (WBC), erythrocyte sedimentation rate (ESR), C-reactive protein (CRP), culture results, and LOS were analyzed via univariate analyses. ANOVA was used to compare antibiotic regiments between and within DM and no-DM cohorts. Multivariate analyses used regression modeling, controlling for age, sex, and diabetes status, to identify any significant independent predictors of laboratory values or infection location in diabetics with hand infections. An α level of 0.05 was used as the threshold for statistical significance.

Study Population

Of the 53 patients who met inclusion criteria, 45.3% ( n = 24) had a confirmed history of comorbid DM, 45.3% ( n = 24) had no recorded history of diabetes (no-DM), and 9.4% ( n = 5) were of unknown diabetic status. Mean interval between DM diagnosis and presentation was 4.5 years. Hypertension was the second most common systemic comorbidity, reported by 17% of patients (9 out of 53). Mean overall patient age was 46 years; DM patients were significantly older (54.0 vs. 40.9 years for no-DM, p = 0.009). Both DM and no-DM groups comprised 45.3% females ( n = 24). End-stage renal disease was reported by 4% of patients (2 out of 53), both of whom among no-DM patients. No patients had a known or documented history of peripheral vascular disease. The presence of neuropathy, without further specification, was reported by 29.2% (7 of 24) in patients in the DM group. Overall, six patients were current smokers (mean active smoking history, 33.6 years), three of whom were among the DM cohort and three among the no-DM cohort. Seven patients were former smokers, of whom six were in the DM cohort.

Results

Serum Marker Analysis

Mean HbA _1C was confirmed as lower in the no-DM group compared with the DM group (6.07 vs. 12.16, p = 0.003), as was glucose on admission (99.9 vs. 302.8 mg/dL) and highest random glucose reported (116 vs. 316 mg/dL) (both p < 0.001).

Mean ESR was significantly higher in DM patients compared with no-DM (76.19 vs. 51.33 mm/h, p = 0.015). Mean overall WBC was 9.6 and was not significantly different between the two groups (9.75 in DM vs. 9.4 in no-DM, p = 0.800). Overall mean CRP level was 64.7. Mean CRP was higher in diabetics, but it did not reach significance (87.2 in diabetics vs. 55.1 in nondiabetics, p = 0.250).

Regression modeling revealed that diabetic status was a predictor of increased ESR (odds ratio [OR] = 1.03, confidence interval [CI]: 1.01–1.05], p = 0.013]); this was not the case for CRP ( p = 0.250) or WBC ( p = 0.800).

Infection Site

Of patients with known DM status (DM or no-DM), 27.1% (13 out of 48) infections were proximal to the digits (proximal) and 72.9% (35 out of 48) were located on a digit (distal). Age (51 vs. 46.1 years, p = 0.401), gender distribution (38% vs. 49% female, p = 0.540), and blood glucose on admission (188 vs. 212, p = 0.636) comparable between patients with proximal and distal infections. Mean ESR did not significantly differ between proximal and distal infections (64.5 vs. 59 mm/h, p = 0.677). Infections proximal to the digits had significantly higher mean CRP (136.9 vs. 50.5 mg/L, p = 0.001) and WBC (5.19 vs. 3.9, p = 0.020) than distal infections.

Regression analysis revealed that proximal infections was a predictor of higher CRP (OR = 1.02 [CI: 1.01–1.03], p = 0.003); this was not the case for ESR or WBC (both p > 0.050).

Culture Results

In both DM and no-DM groups, Staphylococcus aureus was the most commonly identified pathogen: out of 22, 12 (54.5%) cultures were positive for S. aureus in the DM group and 14 (63.6%) cultures were positive for S. aureus among no-DM patients ( p = 0.760). Similarly, rates of MRSA identification in culture did not differ significantly between DM and no-DM (18% vs. 32%, p = 0.300). Overall rate of MRSA identification was 24% (12 of 49 total cultures).

Mean WBC in patients whose cultures grew MRSA did not differ significantly from the mean WBC in patients whose cultures did not grow MRSA (8.9 vs. 9.8, p = 0.500). Gram-negative organisms were cultured in 24% (12 out of 49) of total cultures, and rates of identification of gram-negative organisms were comparable between DM and no-DM patients (27.2 vs. 22.7%, p = 0.730). Out of 22, 8 (36.3%) cultures in the DM group were polymicrobial, compared with 4 (18.2%) in the no-DM group ( p = 0.180). Thirteen (26.5%) of 49 cultures were polymicrobial overall. Supplementary Table S1 (available in the online version) provides a full list of culture results.

Antibiotic Therapy

Antibiotic regimens ( Table 1 ) differed between DM and no-DM patients: DM patients received a combination vancomycin (VAN)/piperacillin/tazobactam (PTZ) regimen more often than nondiabetics (52% vs. 8%, p < 0.001) ( Table 2 ). Providers were significantly more likely to treat diabetics with VAN/PTZ or any VAN-containing regimen than with any treatment containing clindamycin (both p < 0.010), ampicillin/sulbactam (both p < 0.050), or trimethoprim/sulfamethoxazole (TMP/SMX) (both p < 0.010).

Table 1. Initial antibiotic regimen by diabetic status.

Antibiotic Regimen	No-DM	DM	DM unknown
Abbreviation: DM, diabetes mellitus. Note: These data were not reported for 3 DM patients.
Vancomycin/Piperacillin/Tazobactam	2	11	–
Vancomycin/Ampicillin/Sulbactam	2	2	–
Vancomycin alone	3	1	1
Ampicillin/Sulbactam alone	3	1	–
Clindamycin	6	3	1
Trimethoprim/Sulfamethoxazole	–	1	1
Cefazolin/Ampicillin/Sulbactam	2	–	–
Ampicillin/Sulbactam/Clindamycin	1	1	–
Ampicillin/Sulbactam/Trimethoprim/Sulfamethoxazole	–	1	–
Vancomycin/Ceftriaxone	1	–	–
Vancomycin/Gentamicin	1	–	–
Ciprofloxacin alone	2	–	–
Clindamycin/Trimethoprim/Sulfamethoxazole	1	–	1
Ceftriaxone/Cefazolin	–	–	1
Total	24	21	5

Table 2. Comparing antibiotic regimens administered to nondiabetics (no-DM) and diabetics (DM).

Antibiotic regimen	No-DM (%)	DM (%)	p Value (%)
Abbreviations: AMS, ampicillin/sulbactam; CLIND, clindamycin; DM, diabetes mellitus; PTZ, piperacillin/tazobactam; TMP/SMX, trimethoprim/sulfamethoxazole; VAN, vancomycin. Note: Bold values indicate statistical significance, at an alpha level of 0.05.
VAN/PTZ	8.3	52.4	< 0.001
Any VAN	41.2	66.7	0.097
Any CLIND	33.3	19.0	0.289
Any AMS	33.3	23.8	0.490
Any TMP/SMX	4.2	9.5	0.480

Complications

Total six 90-day adverse outcomes were reported, four of which occurred among DM patients. Adverse outcomes among DM patients included a prolonged hospitalization (3 months), one readmission for osteomyelitis and amputation, one readmission for chronic osteomyelitis, and one case of lower extremity osteomyelitis. Adverse outcomes among the two no-DM patients included one readmission following a failed trial of oral antibiotics and one repeat incision and drainage procedure. Of note, one patient in the DM group was reported to have died 112 days following their initial hospital presentation due to causes unrelated to their hand infection. Mean LOS for DM patients was comparable to no-DM patients (5.4 vs. 4.4 days, p = 0.261).

Discussion

Hand infections in the urban population have been studied at length, 15 16 but this is the first study to our knowledge that examined the high-risk diabetic subset of this population. Our hospital is located in an area with a high prevalence of DM, but the rate of diabetes in hand infection patients was disproportionate to the rate of comorbid diabetes: approximately 15.7% of the population in the neighborhood surrounding the hospital is estimated to have diabetes, adjusted for age, 17 whereas 45.3% of the patients who presented with hand infections in our study were found to have comorbid DM. Seventeen percent were found to have comorbid hypertension, the second most prevalent comorbidity.

The mean HbA _1C of the no-DM group was found to be in the prediabetic range, although several patients within the no-DM group had HbA _1C values available. Other metabolic data (random blood glucose and blood glucose on admission) were expectedly lower for no-DM as compared with DM group, and mean values of these categories were not in the diabetic range.

Consistent with data reported by other centers, S. aureus was the most commonly identified pathogen in culture overall. 18 19 20 21 22 23 The etiologic agents of hand infection were comparable between the DM and no-DM groups: rates of identification of all S. aureus , MRSA, gram-negative organisms, and polymicrobial cultures did not differ significantly between DM and no-DM groups. The literature describes the rates of culture of gram-negative organisms from hand infections as ranging from 51.1 to 73% in diabetics and 8 to 11.3% overall. 19 21 24 25 Our population overall grew gram-negative organisms in culture at a higher rate than expected at 24%, but rate of gram-negative cultures among diabetics was lower than expected. Similar variation in incidence has been reported for polymicrobial infections of the hand: sources report values between 41 and 100% in diabetics and 11.7 to 19% overall. 5 7 9 19 21 24 25 26 27 Our overall polymicrobial culture rate of 24% was again higher than expected, whereas the polymicrobial culture rate in diabetics was slightly lower than reported in the literature. There were no predominant second bacteria cultured.

Diabetic status has been found to be associated with higher likelihood of polymicrobial culture, and some studies note an increased risk of MRSA. 7 9 11 12 13 24 26 However, diabetic patients were actually less likely to have a methicillin-resistant S. aureus infection versus the nondiabetic cohort with an MRSA rate of 18% compared with 30%. No patient with an HbA _1C greater than 10 had a positive MRSA culture. Our data showed no association between culture outcome and diabetic status. Based on our data, a known medical history of DM should not be a factor in selecting empiric antibiotic coverage for MRSA. The Centers for Disease Control and Prevention (CDC) recommends empiric coverage for MRSA with community prevalence greater than 10 to 15%. 28

Despite no significant difference in microbiology, diabetic patients were more likely to have received a combination VAN/PTZ regimen than nondiabetic patients. This regimen was selected in diabetics over clindamycin-containing regimens, which are known to have a high MRSA-resistance rate in our community, ampicillin-sulbactam–containing regimens, and TMP/SMX-containing regimens. Early hand infection studies noted higher polymicrobial infection rates in diabetics, suggesting broad empiric coverage, 5 29 30 but Fowler et al 31 more recently found MRSA to be the most commonly cultured organism from infected hands; despite the incidence of polymicrobial cultures increasing, they recommended against the standard use of empiric treatment with broad-spectrum antibiotics, with gram-negative coverage recommended to be at the discretion of the provider. 31 This approach was also supported by our data, with the rates of polymicrobial and gram-negative cultures found to be similar in all patients. Our high rates of polymicrobial and gram-negative infections suggest that all patients, both diabetic and nondiabetic, should be receiving empiric coverage for these organisms as well. However, a notable consequence of enforcing this standard would be the possible risk of development of resistance to PTZ. Additionally, treatment of MRSA presents an ongoing challenge despite the use of vancomycin: both community-acquired MRSA (ca-MRSA) and multidrug- resistant MRSA (MDR-MRSA) rates have increased as well. 16 32 In 2006, vancomycin and/or clindamycin were considered first-line treatment for all nonabscess infections of the hand, but clindamycin resistance has been increasing since 2013. 11 33 Tosti et al 32 found that 20% of MRSA isolates in an urban hospital were clindamycin-resistant in 2014. Though their study found 100% sensitivity to vancomycin in their patient population, other studies have observed development of vancomycin resistance in MRSA cultures. 20 34 35 It is imperative to halt the development of such resistance while introducing minimal harm to patients. Tools such as rigorous, focused patient education and risk stratification schemes for hand infections at presentation can help accomplish this goal while alternative antibiotics are tested and developed. The available data have consistently suggested that early presentation to care is a key factor for improving outcomes of hand infections. 1 2 9 10 29 36 Patient education about hand hygiene as well as warning signs and symptoms of infection can help encourage early presentation. Development of risk-stratification criteria for hand infections may help avoid overtreatment with vancomycin. Interestingly, one study found that empiric coverage for MRSA did not affect outcomes in patients with hand infections: a randomized controlled trial at a level one county hospital showed no difference between empiric treatment with cefazolin versus vancomycin. 35

Biomarkers of inflammation may have a role to play in the development of reliable risk stratification criteria. WBC and CRP levels have been suggested to aid clinicians when deciding whether to add broad-spectrum antibiotics to a regimen. 31 In one study, WBC more than 8,700 (8.7) have been found to be a risk factor predicting MRSA as the etiologic agent of hand infection, with each 1,000 cell increase in WBC conferring an additional 25% risk of MRSA. 16 However, this is not consistent with our data: mean WBC in patients whose cultures grew MRSA was 8.9, and mean WBC in patients whose cultures did not grow MRSA was 9.8. Additionally, whereas CRP and ESR were elevated in both the DM and no-DM groups, mean WBC was not in either group. Moreover, relative utility of CRP, ESR, and WBC is controversial for infections not involving the bone: Houshian et al 19 found CRP to have inferior sensitivity for upper extremity infections than ESR, whereas Bishop et al 37 reported CRP to detect hand infection with higher sensitivity than either WBC or ESR. CRP has been found to be nonspecific for bacterial infection, however, and higher levels have been reported in association with gram-negative compared with gram-positive bacteremia. 38 39 Additionally, our data have evidenced the relative insensitivity of WBC, with only 13 of 49 patients demonstrating leukocytosis. Despite no difference in microbiologic agents, diabetic patients in this study had higher odds of having an increased ESR. This phenomenon has been described in the literature: diabetics have been described to have higher ESR than nondiabetics at baseline, and both diabetics and prediabetics are known to have higher CRP levels than normoglycemic individuals. 40 41 Additionally, increases in CRP levels and WBC may be associated with worse glycemic control. 40 42

Location of infection within the hand may also affect these inflammatory markers: infections proximal to the digits had significantly higher mean CRP levels and WBC count compared with infections of the digits, though mean WBC was within normal limits for both groups. Proximal infections had higher odds of an elevated CRP level, a finding echoed in the literature. 43 ESR did not vary with infection site in this study; this consistency of measurement may be useful in a risk stratification scheme, provided ESR levels are interpreted carefully based on diabetic status. Finally, as has been previously reported, we found the most common site of infection to be the finger. 3 This may help target patient education interventions by encouraging focus on careful fingernail hygiene and vigilance after minor trauma. Hand hygiene has also been described to effectively reduce infection risks in diabetic patients who use fingerstick testing. 44

There are several limitations in this study: namely, more comprehensive patient data were not available regarding final diagnoses, diabetic status, and detailed outcomes. Thus, it is conceivable that some patients with hand infections may have been missed. Additionally, more rigorous trending of blood sugar levels, acknowledgment of changes in antibiotic regimens during the infection course, and determination of patients’ diabetic status would have improved our ability to analyze the effects of diabetes on risk and differences in treatment by providers. More complete data on outcomes would allow a determination to be made about whether diabetics in our underserved population are more likely to suffer worse fates with hand infections than nondiabetics, and whether providers’ antibiotic choices were appropriate. 19 43

Conclusion

This study underlies the need for systematic criteria to aid in risk-stratifying patients for appropriate antibiotic use. Future assessment of the utility of physical examination and laboratory findings propose more predictive criteria for HI severity. Regarding laboratory markers, our findings suggest that if CRP is to be incorporated into predictive HI criteria, infection site must also be accounted for. Our findings also highlight that the microbiology of hand infections in diabetics may not differ compared with nondiabetics as much as was previously thought. In our underserved population, diabetics were more often prescribed combination VAN/PTZ compared with nondiabetics, despite no difference in rates of identification of MRSA or non-MRSA. It may not be appropriate to treat both the groups differently in urban populations similar to our own; antibiotic selection in this population may be overly aggressive and potentially contributes to evolving drug resistance. Future studies should focus on the outcomes of hand infections by diabetic status and antibiotic regimen to establish more patient-specific guidelines on drug selection to maximize outcomes and minimize resistance.