Risk Factors of Clostridium Difficile Infection After Spinal Surgery: National Health Insurance Database

The purpose of this study was to evaluate risk factors of Clostridium Difficile infection (CDI) after spinal surgery using the Health Insurance Review and Assessment Service (HIRA) data. The incidence of postoperative CDI was investigated using HIRA data from 2012 to 2016. Cases involving CDI that occurred within a 30-day postoperative period were identified. Risk factors, including age, sex, comorbidities, postoperative infection, spinal surgery procedure, type of antibiotic, and duration of antibiotic use, were evaluated. Duration of hospital stay, medical cost, and mortality were also evaluated. In total, 71,322 patients were included. Presumed cases of CDI were identified in 57 patients, with CDI rate of 0.54 per 10,000 patient days. Advanced age, staged operation, postoperative infection, and the use of multiple antibiotics were significant risk factors. First-generation cephalosporins were shown to be associated with a lower incidence of CDI. CDI was also associated with longer hospital stays and increased medical cost, and it was an independent risk factor for increased mortality. Extra attention should be paid to patients at high risk for the development of postoperative CDI, and unnecessary use of multiple antibiotics should be avoided. Level of Evidence: Level III, retrospective cohort study


Results
Of 71,322 patients, 57 were diagnosed with CDI within 30 days after surgery (incidence 0.08%). The CDI rate per total number of patient days was 0.54 per 10,000 patient days. The average time from surgery to CDI diagnosis was 18.16 ± 7.81 days. Lumbar surgery (80.68%) was the most common procedure, and the incidence of CDI therein was 0.07%. The percentage of other procedures are detailed in Table 1.
Patients who acquired postoperative CDI after spinal surgery were significantly older (67.12 ± 10.94 vs. 57.27 ± 14.78); there was no difference in sex ( Table 2). The proportion of patients over 65 years of age was greater in the CDI group than in the non-CDI group (P < 0.001). CDI patients also had a higher prevalence of chronic obstructive pulmonary disease (P = 0.030) and a higher incidence of postoperative infection, such as urinary tract infection (P < 0.001), sepsis (P < 0.001), or pneumonia (P < 0.001), compared to patients without CDI. Patients with CDI were more likely to have been prescribed proton pump inhibitors postoperatively (P = 0.002); however, there was no significant difference in the use of H2 blockers between the CDI and non-CDI groups. As for the use of antibiotics, broad spectrum antibiotics were more often prescribed to CDI patients, including third generation cephalosporins (P < 0.001), fourth generation cephalosporins (P = 0.013), penicillins (P < 0.001), glycopeptides (P < 0.001), carbapenems (P = 0.009), and ketolides (P = 0.043). On the other hand, CDI patients were less likely to be given first generation cephalosporins (P = 0.004).
The effect of the duration of antibiotics use and the number of co-administered antibiotics on the incidence of CDI was investigated. (Table 3) The higher the number of administered antibiotics, the higher the incidence of CDI (P < 0.001). There was no significant difference according to the duration of antibiotics used.
Postoperative CDI after spinal surgery resulted in remarkably negative outcomes. Mean hospitalization period increased more than three-fold for patients diagnosed with CDI versus. those not diagnosed with CDI (64.91 ± 74.87 days vs. 18.85 ± 25.37 days, P < 0.001). When multiple regression analysis was performed to investigate risk factors for increased hospitalization period, CDI increased the hospitalization period increased by 30.12 days (P < 0.001) after adjusting for other variables. The total medical cost increased by more than four-fold for those patients with CDI diagnosis ($14,704 ± 15,590 vs. $3,580 ± 4,190, P < 0.001).
Multivariate analyses, identifying independent risk factors for mortality within 30 days postoperatively, identified CDI as a significant risk factor (OR = 7.717, 95% CI = 1.670-35.662, P = 0.009) ( Table 5). In addition, the use of antibiotics, such as fourth-generation cephalosporins, penicillins, glycopeptides, and carbapenems, were associated with significantly increased odds of death, whereas the first-and second-generation cephalosporins were associated with lower risk. Advanced age and the use of three or more antibiotics were also associated with an increased odds of death.

Discussion
In this study, we found CDI after spinal surgery to be associated with advanced age, staged operation, postoperative infection, and administration of three or more antibiotics. The third-generation cephalosporins, clindamycins, penicillins, and fluoroquinolones were previously reported as risk factors for CDI 10,11 . In this study, first-generation cephalosporins were associated with a low incidence of CDI in the multivariate analysis, whereas the third-and fourth-generation cephaslosporins, penicillins, glycopeptides, carbapenems, and ketolides were statistically significant risk factors in the univariate analysis alone. This may be due to the fact that, clinically, many patients are treated with more than one type of antibiotic perioperatively for broad coverage.
A previous study reported that more than 10 days of antibiotics were identified as a significant risk factor for CDI 12 ; however, although higher CDI incidence was noted when antibiotics were administered for a longer period of time, the trend was not statistically significant in our study. The period of antibiotics use was calculated as the number of days the antibiotics administered regardless of the number of antibiotics used. However, antibiotic  www.nature.com/scientificreports www.nature.com/scientificreports/ days was defined as the summation of the number of days of administration of each antibiotic in the previous study 12 . The difference in the definition of the duration of antibiotic administration may have resulted in such disparity.
Multiple studies have reported on the risk of perioperative antibiotic prophylaxis and CDI. Prophylactic monotherapy with cefoxitin (second generation cephalosporin) was associated with a higher risk of CDI, and the addition of another antibacterial to cefoxitin further increased the risk 13 . The use of multiple antibiotics rather than the duration of antibiotic treatment was associated with CDI 14 . In the present study, the increased number www.nature.com/scientificreports www.nature.com/scientificreports/ of antibiotics was associated with an increased risk of CDI after spinal surgery. The Surgical Infection Society Guidelines on surgical prophylaxis recommend the use of antimicrobial prophylaxis for spinal procedures with or without instrumentation 15 . They also recommend monotherapy with cefazolin, either used as a single dose or administered up to 24 hours postoperatively 15 . Clindamycin and vancomycin could be used instead for patients with beta-lactam allergies. None of these antibiotics was associated with an increased risk of CDI in the present study, and therefore, these antibiotics could safely be used for prevention of both surgical site infection and CDI.
Although the surgical antibiotic prophylaxis recommends monotherapy with first-generation cephalosporin, other types of antibiotics or multiple drugs were prescribed perioperatively. We identified that prophylactic antibiotics were prescribed in combination in 21% of patients on the day of surgery. Among patients without any postoperative infection, only 30.5% was administered first-generation cephalosporins alone for surgical antibiotic prophylaxis. The second-generation cephalosporins were used as prophylactic antibiotics in many patients (15.3%), and many hospitals prescribed second-generation cephalosporins as discharge medication after intravenous administration of first-generation cephalosporins (18.2%). Similar pattern of such misuse of antibiotics was previously addressed in 2012 HIRA report on the adherence of surgical prophylaxis protocol. Combination of antibiotics was given to 19.8% of surgical patients, and 16.4% was given antibiotics as discharge medication 16 .
Similar to the results of the previously published papers on CDI after spinal surgery, we identified advanced age and postoperative infection, including urinary tract infection, as significant risk factors [6][7][8] . The difference from the previous studies is that we were able to identify antibiotic regimens and other perioperative medications. We demonstrated that the use of multiple antibiotics was a more significant risk factor than type of antibiotics and duration of use. There is a controversy regarding whether antacids increase the risk of CDI 17 . Although proton pump inhibitors were not shown to be associated with increased risk in the multivariate analysis, they were more commonly prescribed in the CDI patients. Therefore, caution should be taken when prescribing proton pump inhibitors to patients with an increased risk of CDI.
A novel finding in this study was that staged operations were shown to be an independent risk factor for CDI. Staged operation accounted for 0.2% of the cases, and it was associated with 5.336 times higher risk of CDI. Recently, the number of adult spinal deformity surgeries is rapidly increasing, and circumferential spine fusion is performed frequently 18,19 . Because these procedures are complex and the prolonged operation time poses significant burden on the patients, the surgeries are often performed in a staged manner 20 . Therefore, patients who receive staged operations may need more meticulous patient care and monitoring for development of postoperative complications, including CDI. Also, unlike in previous reports on CDI after spinal surgery, we subdivided postoperative infection into any infection, urinary tract infection, sepsis and pneumonia, and confirmed that all types of postoperative infection were risk factors for CDI.
Many studies have reported that CDI increases the length of hospitalization, medical cost, and mortality 6,21 . Similarly, we found that CDI increased mortality rate after spinal surgery by more than seven-fold, and CDI was associated with increased hospital stay and escalated treatment costs. The global surge in the incidence of hypervirulent C. difficile 027/BI/NAP1 subtype is likely to produce a significant burden on the health care system in the near future 13,22 .
The incidence of CDI after spinal surgery has been reported as 0.08-0.11% [6][7][8] . Our study found that the CDI incidence within 30 days after spinal surgery was 0.08%, which was lower than previous results. This disparity may be due to the difference in the definition of CDI. We identified CDI cases as those that satisfied both criteria of a specific diagnostic code and specific antibiotic administration, whereas other studies identified patients with the diagnoses alone 6,8 , or selected patients based on their medical record 7 .
There are several limitations in our study. We used administrative data and were unable to review the medical records for confirmation of the diagnosis or the treatment. The insurance claim data did not include individual health and nutritional status, such as body mass index, functional status, or smoking status, and it also did not provide laboratory data, such as hypoalbuminemia. Therefore, the severity or the stage of the conditions could not be investigated. Another inherent limitation of HIRA data is that there could be discordance between the diagnoses entered in the system and the actual health status of the patient, especially for mild conditions. Therefore, we could not calculate the comorbidity index and analyze its association with the CDI incidence. HIRA-NIS data includes insurance claim data of 13% of randomly selected inpatients, and each sample is labeled with an allocation number. The sample is renewed annually, and since the data is anonymous, a single patient who received operations more than once in different years may be classified as multiple different samples. Although we excluded surgery due to infection, we did not distinguish between trauma, degenerative disease, and congenital deformity. We also could not identify emergency surgeries. Furthermore, even though this study investigated 71,322 patients undergoing spinal surgery, the absolute number of patients (n = 57) with CDI was still low. Finally, our study restricted CDIs to those that occurred within 30 days after spinal surgery. Although many previous studies also used the same definition 6-8 , CDI has also been reported to occur after more than 30 days after antibiotics administration 23 .
In conclusion, we found that advanced age, staged operation, postoperative infection, and the use of multiple antibiotics were independent risk factors for CDI after spinal surgery. Although the incidence was low (0.08%), CDI was associated with negative outcomes of increased hospital stay, increased medical cost, and increased mortality. Therefore, modifiable risk factors should be minimized in order to prevent CDI and to optimize surgical outcome after spinal surgery. In particular, the unnecessary use of multiple antibiotics should be avoided.

Methods
Database and data collection. The National Health Insurance (NIH) covers about 98% of the South Korean population 24 . The data of the Health Insurance Review and Assessment Service (HIRA) includes information on patient diagnoses, past medical or surgical history, treatment procedures, and prescription dispensing information. We used the National Inpatient Sample of the HIRA (HIRA-NIS), which contains claims data of 13% Statistical Analysis. Statistical analysis was performed by independent two-sample t-test and chi-square (or Fisher's exact) test. For continuous variables, the data are expressed as a mean ± standard deviation. For categorical variables, the data are expressed as counts and percentages. To assess potential risk factors for CDI after spinal surgery, a multivariable logistic regression model was created to assess for multivariable predictors of CDI after spinal surgery. A multivariable logistic regression model was also constructed for mortality. Results are reported as odds ratio (OR) and 95% CIs. Risk factors associated with CDI diagnosis were assessed using multivariable logistic regression including all variables. The Hosmer-Lemeshow and model Χ2 goodness of fit tests were used for the model. All statistical analyses were performed with IBM SPSS Statistics for Windows/Macintosh, Version 23.0 (IBM Corp., Armonk, NY, USA). P values < 0.05 were considered significant.

Data availability
The data will be available upon reasonable request.