Abstract
Background:
Sepsis is the leading cause of maternal death in sub-Saharan Africa (SSA), a region that sees some of the highest rates of maternal death and morbidity in the world. As one of the most commonly performed surgical procedures in SSA and a proved risk factor for surgical site infection (SSI), cesarean section (CS) is an important operation to target because of its massive impact on maternal and neonatal health. There is currently insufficient published data available on the patient and facility-based context around SSI after CS to establish a true and clear understanding of this infectious category. The objective of this study was to collect accurate and valid data on the incidence of SSI after CS and the circumstances around SSI in two Kenyan hospitals.
Hypothesis:
Our primary analysis focused on the consequences of timing of peri-operative antimicrobial prophylaxis. We hypothesized that patients who were given antibiotics pre-operatively would show lower SSI rates than those given the agents post-operatively.
Methods:
This was an Institutional Review Board-approved observational study of 609 women who had CS operations at two Kenyan hospitals from September to December 2015. Thika provided antimicrobial prophylaxis prior to incision for all patients, and Kiambu provided only post-operative prophylaxis. It should be noted that this difference was the result of a previous intervention at Thika and not a part of this observational study.
Results:
Patients at the two hospitals had similar pre-operative characteristics indicating a relatively healthy population. The median age was 26 ± 6 years (range 18–43) at Thika and 26 ± 5 (18–44) at Kiambu. Median parity was 1 ± 1 (range 0–7) at Thika and 1 ± 1 (0–10). Patients also went through a comparable number of antenatal care visits (median 4 ± 1 at both hospitals). The number of patients with prolonged rupture of the membranes was 103 (34.4%) at Thika and 99 (32.9%) at Kiambu. There was a slightly higher number of patients with meconium-stained liquor at Kiambu Hospital (115) than at Thika (74). The SSI rate was 4.0% (12/299; 11 superficial, 1 deep) at Thika and 9.3% (28/301; 18 superficial, 7 deep, 3 organ/space) at Kiambu.
Conclusions:
The data show a striking difference between SSI rates in patients who were given properly timed pre-operative antibiotics and patients who were given only post-operative antibiotics. Administration of post-operative antibiotics is currently the norm in much of SSA, and there is strong evidence that many of the infectious problems encountered in this population would be reduced by the provision of antibiotic prophylaxis prior to the incision.
Hospital-acquired infections (HAI) and surgical site infections (SSI) constitute a large social and economic burden for healthcare institutions in developed countries, where they affect up to 15% of hospitalized patients in regular wards and as many as 50% or more of patients in intensive care units [1,2]. In low- and middle-income countries (LMIC), the magnitude of the problem is almost certainly greater. There is a particularly egregious gap in SSI rates between high-income and LMIC; a World Health Organization review stating that the risks of SSI in LMIC were “strikingly higher than in equivalent surgical procedures in high-income countries” [3]. However, good estimates of the rates of SSI remain underestimated or unknown in most LMIC, likely because of non-adherence to infection surveillance guidelines as well as inadequate resource allocation for surveillance activities.
Surgical site infection and sepsis have a heavy impact on maternal and neonatal health in sub-Saharan Africa (SSA), a region that sees some of the highest rates of maternal death and morbidity in the world [4]. A global review on maternal deaths found that sepsis accounted for as many as 15.0% of all maternal deaths in SSA in comparison with a maximum of 2.1% in developed countries [5]. As one of the most commonly performed surgical procedures in SSA, cesarean section (CS) is an important operation to target for the prevention and treatment of SSI. The relative youth and good health of patients undergoing this procedure can help provide a clearer map of the pathways and processes leading to infection [6-8].
One proved means of reducing the risk of SSI is the administration of pre-operative antibiotic prophylaxis [9], the effectiveness of which has been confirmed in both developed and developing countries [10]. However, post-operative rather than pre-operative antibiotics still are widely used despite the lack of evidence supporting their effectiveness for the prevention of SSI [11]. The aim of this study was to collect accurate and valid data on the incidence of SSI after CS in two Kenyan government sector hospitals. Specifically, we aimed to investigate the effects of antibiotic timing on the primary outcome of SSI, collect data on risk factors for SSI to inform future points for intervention, and demonstrate that it is possible to collect high-quality data on SSI after CS in a developing country without dedicating large resource reserves.
Patients and Methods
This was an observational study carried out from September 1 to December 31, 2015, at two government-sector district hospitals in Kiambu County, Kenya. At Thika, pre-operative prophylaxis had become routine four years prior as part of an infection control quality-improvement project. Kiambu had no history of external interventions or research with regard to the timing of antibiotics and administered only post-operative antibiotics to patients. All patients undergoing CS in both facilities were recruited into the study. These sites were selected as they conduct a similarly large number of CS per month; approximately 200 are conducted per month, from a pool of more than 1,400 laboring women.
The co-principal investigator, project manager, obstetricians, and theatre and surveillance nurses from both study sites attended two one-day training sessions that included an overview of the study protocol and Good Clinical Practice, a practical demonstration of specimen collection and handling, and a site initiation visit.
Diagnosis of SSI
The primary outcome of interest was the development of SSI, which was diagnosed using the U.S. Centers for Disease Control (CDC) criteria [12], as closely as was possible within the constraints of available resources and services. Incision dressings were removed on Day 3 and inspected for any signs of SSI. If signs were present, the infection was classified per the CDC criteria and a swab taken as per protocol for CS before starting broad-spectrum antibiotics and awaiting culture results. If no signs were present, the patient was discharged and contacted by a study staff member on Days 10 and 30 using the phone number provided by the patient to inquire about the status of the incision according to standard operating procedure (SOP). If the patient indicated any sign of SSI, she was asked to report to the hospital, where the study clinician examined the surgical site, confirmed SSI, classified it according to the CDC criteria, took a swab for CS, and gave appropriate treatment. The participants were free to call or visit the hospitals if they developed or suspected any problem in the surgical site before or in between the scheduled calls. Cultures were conducted at Aga Khan University Hospital laboratories. Patients were observed from admission to discharge with a 30-day follow-up period.
Data analysis
Logistic regression analyses were done to determine predictors of SSI. A univariable screen of the different variables was performed at p < 0.25. Predictors meeting this criterion were included in stepwise backward analysis in which p < 0.05 served as the criterion for retention in the final model. All analyses were done using Stata 14.0 (College Station, TX).
Ethics
Ethical approval was sought and given from the Kenyatta University Ethical Review committee. Permission to conduct the study was given by the hospitals. Informed consent was obtained from the study participants.
Results
The study enrolled all women undergoing CS operations during the observation period; 299 women were treated at Thika and 301 at Kiambu. Patients at the two hospitals showed broadly similar baseline characteristics, as shown in Table 1. The ages of the patients were comparable at Thika (26 ± 5.7 years) and Kiambu (26 ± 5.2), as were the ratios of patients who had received antenatal care (99.0% and 98.3%). However, there were some differences. In general, patients at Thika had greater parity than those at Kiambu, which saw 41.5% nulliparous patients to Thika's 22.7%. Patients at Thika also were more likely to have had previous CS operations (42.8%) than were patients at Kiambu (27.8%). Of significance, more patients (37.9%) at Kiambu underwent more than four vaginal examinations than did those at Thika (21.1%), and more patients at Kiambu (38.2%) exhibited meconium-stained liquor compared with patients at Thika (24.8%).
Baseline Characteristics
Table 2 displays the surgical elements for the study population. In line with the findings of previous studies and systematic reviews, the majority of the operations at Thika (91.6%) and Kiambu (96.7%) were emergencies. The type of pre-operative skin preparation agent differed between the centers: Thika favored the use of iodine (55.5%) or an alcohol and iodine solution (42.1%), whereas Kiambu favored an iodine and chlorhexidine solution (72.0%), alcohol and iodine (13.0%), or alcohol and chlorhexidine (14.6%). Both hospitals used spinal anesthesia on the majority of patients, 70.2% at Thika and 98.0% at Kiambu. The Pfannenstiel incision was heavily favored at Kiambu (88.4%), whereas Thika was almost evenly split between Pfannenstiel (52.2%) and vertical (47.8%).
Surgical Elements
ASA = American Society of Anesthesiologists; NNIS = National Nosocomial Infection Survey.
Table 3 displays the post-operative factors. Of note, 12 SSIs, as defined by the CDC, were observed at Thika: 11 superficial, no deep, and one each organ/space unknown. However, 28 SSIs were observed at Kiambu: 17 superficial, seven deep, three organ/space, and one unknown.
Operative Outcome
SD = standard deviation.
The main divergence between the patients at the two hospitals was in the administration of pre-operative antibiotics, as shown in Table 4. All patients at Thika received antibiotic prophylaxis no more than 120 minutes before the incision time. In contrast, patients at Kiambu received oral and intravenous post-operative antibiotics and no prophylaxis. From the univariable analysis, number of vaginal examinations, post-partum hemorrhage, and administration of pre-operative antibiotic prophylaxis emerged as factors of possible significance. These candidates were then fit in a model using stepwise backward selection with a level of significance of 5% that yielded only the administration of pre-operative antibiotic prophylaxis as a statistically significant factor for the primary outcome of SSI (odds ratio 0.41; 95% confidence interval 0.20–0.82; p = 0.01).
Antibiotic Administration
IV = intravenous; PO = oral.
Discussion
Research findings
The collected data show a striking difference between SSI rates in the two populations. Potential reasons for the observed different in infection rates, as suggested by published studies and systematic reviews of SSI after CS [13-17], included previous CS operations, choice of skin preparation agents, type of anesthesia, number of vaginal examinations, and presence of meconium-stained liquor. Therefore, we believe that the timing of antibiotic therapy was likely the primary reason for the lower infection rate seen at Thika and that these data present strong evidence that many of the infectious problems encountered in this population would be reduced by the provision of antibiotic prophylaxis prior to the incision.
Some other interesting and important points were demonstrated. First, this study follows up on an intervention project on the timing of antibiotic prophylaxis and examines the effects of this intervention at a hospital that is similar in most aspects save for the intervention. Second, patient follow-up to 30 days was successful using mobile phones, showing that this technology, which has been used broadly in other areas of healthcare, also can be used for SSI surveillance. Lastly, the data categories collected were more comprehensive than in most other published studies of a similar type and used globally standardized definitions and terminology.
Limitations
One of limitation of this study as a piece of observational work is that the use of pre-operative antibiotic prophylaxis is perfectly associated with hospital site; it was administered to all patients at Thika and no patients at Kiambu. This raises the possibility that pre-operative prophylaxis is confounded by some other (unmeasured or hard-to-measure) aspect of the care delivered at these institutions (e.g., quality of sterilization services, theatre hygiene practices) and the use of pre-operative prophylaxis merely acts as a geographic surrogate for where the surgery took place. This is indeed a limitation of this work, and we hope to conduct interventional research to demonstrate a causal association better.
The staff and infrastructure divergences between the two hospitals are another possible confounding factor. Although similar in size and patient volume, Thika Hospital is a Level 5 center that offers a slightly larger menu of services than Kiambu District Hospital; Level 5 hospitals in Kenya are secondary regional referral centers that provide specialized care. It also is noted that the successful and sustained effect of the effort to implement universal pre-incision antibiotics was likely in large part attributable to the buy-in of staff and facility “champions.”
Although the above-listed limitations may be said to take away from the clarity of the findings, we strongly believe that the significance of the impact of properly administered pre-operative antibiotics on post-cesarean infection rates is undeniable, and that their effect is powerful enough to compensate for many of the resource and infrastructure challenges encountered in district-level hospitals in lower-resource regions. We concede, however, that further research is needed to generate the evidence that would provide conclusive support for this conjecture.
Models for surveillance
Despite the considerable burden of SSI in SSA, relatively few interventional studies targeting these infections have been conducted in this region [10,18]; and much of the available observational data use non-standardized definitions, making attempts to compare or synthesize the data challenging [10]. This lack of interventional data may be attributable in part to the scarcity of published surveillance data on the baseline SSI rates; without the descriptive data necessary to guide effective interventions, it can be difficult to locate the most pressing action points. It also has been found that in the observational studies that have been published, there is a marked lack of consistent and standardized terminology. We suggest this study as a model for surveillance within LMIC because it meets all of the data requirements necessary for a full analysis of the risk factors and context for SSI after CS without a large commitment of resources. Furthermore, the use of mobile phone technology should be considered for use as an infection surveillance tool, particularly in areas where patients may not have the resources to attend clinics, where it would be an important way of reducing the time, costs, and labor needed for a surveillance project.
Models for intervention
The prevailing implementation and intervention methods currently favored by the infection control community have shifted away from a focus on process control (SCIP) toward more holistic hospital culture-based programs. The Comprehensive Unit-Based Safety Program (CUSP) or Surgical Unit-Based Safety Program (SUSP) models are based on forming and empowering an interdisciplinary team of surgeons, nurses, anesthesiologists, and infection control workers at the target facility to identify areas of concern related to SSI and implement an intervention to target these areas. This team meets regularly to review progress and identify potential new areas of note. Although CUSP and SUSP programs have been effective in reducing SSI rates in high-risk populations, most testing has been done at hospitals in developed countries [19-21]. These programs require skilled personnel on site to educate and guide the local healthcare staff in making the required changes and infection control staff who can use proper metrics to assess true infection rates. The only analysis of this type in sub-Saharan Africa utilized highly regarded teaching hospitals with such staff in place, and it is likely that implementation of such programs would not be feasible for many of the more resource-constrained healthcare centers in SSA and other developing regions.
The comprehensive and exclusive use of pre-operative antibiotics prophylaxis at Thika is the result of a targeted intervention implemented in 2011 [18]. This process involved SSI surveillance of all operations at the hospital over a period of 16 months, the development of an antibiotic prophylaxis policy that synthesized relevant research papers and policy documents, the endorsement of all consultant surgeons and anesthetists at the hospital, and extensive training of all medical, nursing, and theater staff. Evaluation of the data showed reduction in the risk of infection, particularly superficial SSI, after the policy was introduced. This intervention also was associated with a reduction in costs and hours of labor.
Given the marked and persistent effects of this intervention on SSI rates at Thika, as shown in this study, as well as the economic and labor-reduction benefits, we suggest that this model for the implementation of a single intervention that shifts hospital policy from post- to pre-operative antibiotics is preferable to culture-based implementation and intervention strategies in LMIC.
Footnotes
Acknowledgment
Presented in part at the Surgical Infection Society-NA 2018 Congress.
Funding Information
No funding was received.
Author Disclosure Statement
No competing financial interests exist for any of the authors.