Pediatric patients with NMS are considered a high-risk population with surgical site infection rates that are much higher than for idiopathic scoliosis. In 2013, the Best Practice Guideline described strategies to prevent surgical site infection in high-risk pediatric spine surgery patients. We found a significant decrease in deep surgical site infections, from 16.1% to 4.4%, in the 4 years before (2008–2012) and after (2014–2018) the strategies were adopted at our institution.
Our infection rate in Group 2 is on the lower end of reported rates in the literature. Ramo et al demonstrated deep infection in 10.3% of patients, 73% occurring within 12 months. Mackenzie et al demonstrated a 13.1% rate of surgical site infection in this population in a multicenter analysis. A meta-analysis by Zhou et al found a rate of 13% in NMS patients, the highest incidence in their study. Our Group 1 surgical site infection rate of 16.1% is consistent with these previous publications.
Although we found no difference in patient demographics and characteristics preoperatively between the two groups and believe them to be comparable, there was a significant change in implant material from stainless steel in Group 1 to cobalt-chrome in Group 2. 80% of the infections in Group 1 were in patients with stainless steel implantation and one (16.7%) in Group 2. Four (66.7%) infections in Group 2 included cobalt-chrome instrumentation, with one (10%) in Group 1. Although the change in instrumentation between groups was significant and could potentially be a confounding variable, we do not believe this variable likely changed the significant results of this study. This is supported by the largest and only multicenter study by Wright et al, who reported on 1156 PSF procedures and concluded no difference in infection risk between stainless steel, cobalt chromium, or titanium.
The microbiology associated with these infections has also been reported extensively in the literature. Sponseller et al identified primarily polymicrobial infections in 210 patients, with an almost equal presence of gram-positive and gram-negative organisms. Contrary, Master et al reported a majority of gram-positive organisms in their review of 151 patients. In their study, 75% of infections were unimicrobial and 25% polymicrobial. Several other studies have reported elevated rates of coagulase-positive and coagulase-negative staphylococcus, polymicrobial, and gram-negative infections.[4–6]
In our study, we report a higher number of gram-negative organisms identified, almost all in Group 1. Additionally, we saw half of the infections in Group 1 result as polymicrobial. At our institution, intermittent use of vancomycin in bone graft and wound closure began in mid-2011. Prior to April 2011, no patients received local vancomycin powder and 80% of infections in Group 1 fell into this time frame. Following the institution of the BPG, use of vancomycin powder in the graft and wound closure became essentially universal and was used in 97% of patients in Group 2. Though multiple protocol modifications were made concurrently, the fact that no patients in Group 2 had a MRSA infection and that the percentage of isolated gram-negative bacteria dropped to 28.5%, adds credibility to our suspicion that the local addition of vancomycin powder played an important role in lowering our infection rate. This is consistent with multiple studies appearing in the literature in 2011 recommending use of local vancomycin to reduce infection rates.[25–27]
The difference in mean time to infection was markedly different between the groups. This did not achieve statistical significance, likely due to the small number of infections in Group 2. Our postoperative protocol consists of visits at the following time points: 2-weeks, 4-weeks, 3-months, 6-months, 12-months, and annually after that. Consequently, these patients are closely monitored in the early postoperative period. As the presenting symptoms of deep infections are rarely subtle and unlikely to be ignored by patients, parents, or providers, we attribute this difference to a delay in infection rather than a delay in detection. Although it is impossible to ascertain the reason for this difference, one may hypothesize that the additional precautions taken led to less wound contamination in the perioperative period, and yet this vulnerable population remained susceptible to later infections possibly seeding the implants.
We investigated the groups with regards to the dates of their primary procedures to search for any clusters of infections that might have been a confounder. We found one pair of patients in Group 1 and two pairs of patients in Group 2 that had primary procedure dates within 1 month of each other. However, the time over which they developed infections and the organisms involved varied and thus were not consistent with any clustering of infection.
There are limitations to this study. There is the inherent limitation of being a retrospectively designed study. This high-risk population is inherently heterogeneous, but we found no significantly different patient characteristics, such as BMI, age, and lab values, between the two groups. Also, we excluded superficial surgical site infections from our chart review. They typically do not require trips to the operating room, and may be mentioned briefly in clinic note or handled over a phone call with nursing, we were concerned that accurately gathering this information retrospectively would not be possible. As the cost to the patient, family and healthcare system is far greater with a deep infection we focused our efforts on those alone. Deep infections are also more accurately captured retrospectively as they undergo operative treatment.
There are many possible confounding variables in a study, such as this operative time and surgeon expertise. However, we found that operative times were nearly identical (P = 0.8457) as well as percentage of cases that had dual attendings (P = 0.855). As Group 2 actually had a higher percentage of cases done by junior attendings than Group 1, it is unlikely that surgeon experience contributed to the improvement observed.
Perhaps the biggest limitation is that there were many components to the Best Practice Guideline, and that it is impossible to say from this data which of the policies truly impacted the change in infection rate we observed. Faced with a high infection rate, our institution and others were motivated to use multiple approaches simultaneously to combat these infections, which resulted in the establishment of the Best Practice Guideline. Compliance with some of these guidelines was easily monitored (addition of gram-negative antibiotic coverage, nutrition consults, and vancomycin use), while others were much harder to monitor (such as limiting operating room access and chlorhexidine wipes the night prior). This data shows that the BPG strategy was effective at rapidly reducing the infection rate to 1/4 of that seen in the years prior. However, this move, which was made in the interest of patient safety, provides less clarity of true causation as would be established with a single change or a randomized study. To the patients that were spared an infection through the addition of several low cost modifications with little morbidity, this was no doubt a worthy trade off. Nevertheless, it is possible, even probable, that some of the components in this guideline are superfluous and continued study is needed to streamline our approach. For example, Group 2 had more nutrition consults than Group 1 (P = 0.002), however, we have separately analyzed the effect of nutrition consults and found there was no significant weight gain preoperatively in the patients who received a nutrition consult. Continued review of these guidelines is needed so that they can continue to evolve to optimize care.
Another inherent limitation of this study is the impact of heightened awareness that instituting a protocol such as this brings. This is certainly a potential confounder and we acknowledge that attention to infection may play an equally important role in prevention as many of the components of this protocol. However, as the study period spanned more than 4 years following the institution of this protocol, it was likely not the only factor, as the "Hawthorn effect" typically diminishes over time.
In conclusion, this study represents a dramatic 72% decrease in the incidence of deep surgical site infections in patients with NMS undergoing primary PSF after the implementation of the strategies taken from the 2013 Best Practice Guideline. Further prospective studies are required to elucidate the efficacy of these perioperative strategies and decrease the incidence of surgical site infections in high-risk NMS patients.
Spine. 2021;46(17):1147-1153. © 2021 Lippincott Williams & Wilkins