Bish et al
and a fault tree of the model was constructed. The ST-PRA model was refined and revised to create a model that had face validity with technical experts who understood the procedure and the processes of care. Once validated, the model was executed using a risk assessment software (Relex Reliability Studio 2007 v2; PTC Corporate Headquarters, Needham, MA) to determine the importance of individual measures and failure combinations. The fault tree contained 283 events (failure points) and 203 gates; the complete fault tree can be found in Slonim et al,6 with special attention to Figure 9, which depicts the part of the fault tree containing the top 5 minimal cut sets, as discussed subsequently in the “Results” section of this article. The specific sequence of risk points (i.e., failures) contributing most to an SSI were identified next, followed by a sensitivity analysis to ensure that model conclusions were robust and consistent with the literature. Finally, on the basis of the failure sequences, an intervention aimed at reducing the SSI risk was proposed.
Data Sources for Building the Fault Tree Model To ensure that the ST-PRA model captured all possible process factors, 4 different data sources were used: (1) the peerreviewed and gray literature, (2) national databases, (3) ASC site visits, and (4) technical expert panel (TEP) input. Each source informed the data collection effort for the other sources in an iterative fashion. That is, information gleaned during the literature review informed ways to analyze the databases; information collected during the site visits or during the TEP meetings informed additional data analyses and literature searches.
Literature Review An extensive peer-reviewed and gray literature review was conducted to identify potential risk factors associated with deep incisional SSIs from surgical procedures. The review was limited to publications in English published since 2000. However, some seminal papers before 2000 were included, particularly if they dictated current clinical practices. Search engines included PubMed, the Cochrane Collaborative, CINAHL, and others as appropriate. A series of key word search terms were used within the following categories of interest: (1) patient demographics (e.g., age, race, insurance status, income, socioeconomic status), (2) patient health status and comorbidity, (3) surgical risk factors (e.g., surgical risk factors, procedural complications, medical errors, surgical infections, SSIs, health care–associated [acquired] infections [HAIs], surgical procedures, arthroscopic procedures, outcomes), (4) organizational context (e.g., academic medical centers, community hospitals, for-profit surgical centers, freestanding ASCs), and (5) infection control practices. The literature review was also used for obtaining discrete probability estimates and their ranges, required for the sensitivity testing. An article was included when it studied ASC-specific studies on complications; complications in different contexts (e.g., academic); hospital- and care-related infections and their risks; risk estimates related to demographics, conditions, or procedures; support or counterevidence from the site visit estimates; and risk factor estimates related to the patients, providers, structures, or teams. Studies were excluded when these did not provide specific information on surgery complications, focused only on hospital and not surgical infections, did not address interventions, and were opinion and not empirically based. The extant peerreviewed literature was enhanced by a review of gray literature to provide important information for inclusion in the risk models.
National Databases Discharge databases have an important role in studying SSIs. Although ASCs provide an important setting for the performance
2
www.journalpatientsafety.com
J Patient Saf • Volume 00, Number 00, Month 2014
of ambulatory procedures, these are constrained in that complications, particularly infections, may receive follow-up care in the physician’s office; in the emergency department (ED); or, for severe cases, in the hospital. Therefore, we complemented our study with the data reported in publicly available databases. We analyzed the Agency for Healthcare Research and Quality’s State Ambulatory Surgery, Nationwide Inpatient Sample, and State Emergency Department databases for the states of California, Maryland, and New Jersey for the period of 2006 to 2008 to capture information about infection rates and demographic and institutional characteristics related to specific procedures.7
ASC Site Visits Another important source of process and sociotechnical information was 4 ASCs based in Maryland and Washington, DC, each representing a different care context: an academic hospital– associated ASC; a community hospital–associated ASC; a freestanding for-profit ASC; and a free-standing, hospital-associated, pediatric ASC. The different care contexts informed the ST-PRA model about the SSI risks in these settings and created the learning opportunities across these settings. Site visits included 3 major activities: • A review and comparison of policies and procedures related to SSIs, including preoperative, operative, and postoperative policies governing patient care, room cleaning, disinfection, and equipment disinfection and sterilization • Informal exploratory interviews with a selection of 5 to 8 staff members from the participating ASCs to learn about infection prevention policies and procedures in place • Process flow comparisons across sites, noting differences in policies, procedures, and facility characteristics
TEP Reviews A panel of technical experts was convened to guide the STPRA modeling. The TEP members were selected to ensure representation of the relevant issues, namely, surgeons, anesthesiologists, nurses, and a physician’s assistant, working in inpatient and/or outpatient settings, and representatives from the Centers for Disease Control and the Centers for Medicare and Medicaid Services. The TEP’s input helped to shape the fault tree’s design and the final selection of an intervention.
Sensitivity Analysis After fault tree development, a sensitivity analysis was conducted for events in which the literature reported large variations in the probabilities or for which failure probabilities were highly site dependent. We varied these probabilities in the base case (i.e., original model) within the ranges suggested in the literature. When a probability estimate was unavailable, an anchor estimate was obtained from technical experts. For example, hand washing is a common approach for preventing bacterial spread and has a positive impact on preventing SSIs. Compliance rates for hand washing range between 40% and 90%. In the sensitivity analysis, we varied the probability for hand washing compliance from 40% to 90% to better understand its impact on mitigating SSIs. The sensitivity analyses ensured that the conclusions were appropriate even if the basic-level event probabilities were somewhat uncertain at the beginning of the modeling exercise.
Minimal Cut Sets and Importance Measures A cut set corresponds to a set of basic events whose occurrence causes the top event (i.e., SSI).8 The minimal cut set identifies the major critical paths, composed of multiple failure points. Cut set analysis is complemented by importance measures © 2014 Lippincott Williams & Wilkins
J Patient Saf • Volume 00, Number 00, Month 2014
that rank the most significant individual risks based upon their contribution to the top-level event (i.e., SSI). These measures assess the criticality of the risk in the model by assessing the absolute risk, the risk’s relative importance, or the risk’s frequency in the model. Commonly used relative importance measures in risk assessment modeling include the criticality, Birnbaum, and Fussell-Vesely measures.9 Both the Birnbaum and Fussell-Vesely measures anchor an individual risk estimate within the context of the model’s other risks, whereas the criticality measure is a measure of absolute risk, which identifies the independent risk contribution of a basic event. The importance measure selected depends, in part, on the type of model created and the purpose of the modeling exercise. For this study, we focused on the criticality measure because this measure permitted the rank ordering of the most critical contributors to the relatively rare event of SSIs and facilitated the identification of interventions that are most likely to improve system performance.
RESULTS Data Sources Literature Review The literature demonstrated mixed results as to whether infection rates are lower for ASCs compared with hospital settings.2,10 However, recently, HAIs originating in ASCs and other outpatient settings have reportedly increased.11–13 In 2008, the largest HAI outbreak was reported and traced to unsafe injection practices.14 This event represented the largest hepatitis C virus outbreak in U.S. history, exceeding 63,000 contacts and 115 infections possibly linked to this event. A literature survey suggested the following general risk factors for the development of SSIs, which were incorporated into the ST-PRA2: • Patient factors (e.g., comorbidities, malnutrition, obesity, tobacco use, age, ASA score, positive methicillin-resistant Staphylococcus aureus [MRSA] nasal culture) • Surgery-related factors Procedural factors (e.g., surgery type, antibiotic prophylaxis, surgical duration, hypothermia, blood loss requiring transfusion, spillage of luminal contents) Surgeon-related factors (e.g., S. aureus carriers among orthopedic team members, wristwatch users) • Operating room (OR) environmental factors (e.g., people in the OR; frequent OR door opening; movement of OR personnel, which disrupts the positive-pressure OR environment) ○
Risk Assessment of Surgical Site Infections
of facilities. Some ASCs were equipped with isolation bays reserved for MRSA patients, whereas others were not. The patient flow mapping and structural factors served as the foundation for building the process component of the fault tree and the riskinformed intervention.
TEP Reviews Once constructed, the fault tree was validated by the TEP for content, structure, and accuracy. The fault tree’s logic was revised based on TEP comments. The TEP also provided a framework for face validity for the probability estimates. When large variability in probability estimates remained either because of large reported variations in the literature or because the probabilities were highly site dependent, a sensitivity analysis was performed by running the model for the base case and for each variation. The literature review, ASC site visits, and discussion with the TEP identified similar lapses and variability in infection control practices among health care professionals and among the different types of ASCs. Often, these lapses resulted from inconsistency in guidelines or personnel training. There is considerable variability in the knowledge of infection control practices among health care professionals.15 Information on how the system fails and how frequently it fails was embedded into the model through Boolean logic and probability estimators.
Risk‐Informed Intervention We determined the SSI rate and the top 5 minimal cut sets for each variation of the base model considered in the sensitivity analysis to understand how and whether these had changed over the base case. The following events were included in the top 5 minimal cut sets of the base case and all of its variations: • Event 173 Staff fails to provide patient with instructions for weight reduction • Event 660 Patient fails to notice infection during home care • Event 642 Staff fails to protect patient effectively • Event 543 SSI risk for obese patient, weight not reduced and nutrition not improved • Event 450 Obese but not diabetic patient (30 ≤ body mass index < 40) • Event 182 Failure to administer indicated antibiotics
○
National Databases Database analysis demonstrated infection rates that were significantly lower than those reported for similar procedures, possibly because of underreporting or missing data. Consequently, these databases were mainly used to provide estimates on volumes, institutional characteristics, and patient demographics for the most common outpatient procedures and as inputs in the fault tree model as probability estimates.
ASC Site Visits Despite the different care contexts, many policies and procedures were common across the 4 sites. Some process variability in the preoperative, operative, and postoperative processes was also evident. For example, although each ASC monitored the blood glucose levels of patients with diabetes, only the pediatric ASC engaged an endocrinologist to ensure patient optimization before surgery. Other differences included the capacity and layout © 2014 Lippincott Williams & Wilkins
The events overlapping these minimal cut sets provided the most significant opportunity for preventing risk. As a result, these 5 events became the focus of our next steps, determining the importance measures and designing a risk-informed intervention. Table 1 presents the events from the ST-PRA model ranked in order of criticality. For example, Event 642 Failure to protect the patient effectively ranked as the most critical unique event with the highest independent contribution to the occurrence of SSIs (i.e., 51.87% of all SSIs occur because of this failure). Figure 1 depicts the fault tree section that contains this event. This event is followed in criticality by events corresponding to obese patients whose weight is not reduced and whose nutrition is not improved before the surgery (Events 450 and 543 in Table 1) because of failures such as the failure of staff to provide the patient with instructions for weight reduction . Because of its high criticality and the fact that it appeared in 4 of the 5 top cut sets, Event 642 Failure to protect the patient effectively and its component events were recommended as the focus for intervention development. The most important components that compose this failure, as indicated by their contribution to the risk for developing an SSI, include the following: • Failure to prepare the skin appropriately preoperatively • Antibiotic-related failures related to timing or administration www.journalpatientsafety.com
3
J Patient Saf • Volume 00, Number 00, Month 2014
Bish et al
TABLE 1. Importance Measures on Events in Rank Order of Criticality Event Event 642 Failure to protect the patient effectively (failure related to skin preparation, antibiotics administration, shaving surgical site, etc) Event 450 Obese but not diabetic patient (30 ≤ BMI < 40) Event 543 SSI risk for obese patient, weight not reduced and nutrition not improved Event 173 Staff fails to provide patient with instructions for weight reduction Event 142 Staff not well trained in infection control Event 182 Failure to administer indicated antibiotics Event 677 Patient is a smoker but has no COPD Event 456 Patient is a smoker and has COPD Event 659 Glove puncture Event 30 Failure to prepare skin appropriately Event 138 Failure to remove watch, jewelry, artificial nails Event 644 Failure to readminister antibiotics for longer surgery Event 419 Patient fails to come for postoperative visit Event 136 Failure to wash hands properly (OR staff) Event 687 Patient does not provide accurate information on smoking status Event 241 Tourniquet time > 60 min Event 660 Patient fails to notice infection during home care
Criticality
Birnbaum
Fussell-Vesely
0.5187
0.0113
0.5187
0.3147 0.3043 0.3042 0.2583 0.2331 0.2003 0.1649 0.1550 0.1457 0.1291 0.1166 0.1162 0.10332 0.0995 0.0956 0.0929
0.0047 0.0409 0.0015 0.0038 0.0051 0.0070 0.0135 0.0038 0.0051 0.0038 0.0051 0.0034 0.0038 0.0012 0.0051 0.0025
0.3147 0.3043 0.3042 0.2583 0.2331 0.2003 0.1649 0.1550 0.1457 0.1291 0.1166 0.1162 0.1033 0.0995 0.0956 0.0929
BMI, body mass index; COPD, chronic obstructive pulmonary disease.
• Staff not well trained in infection control practices • Glove puncture • Failure to remove watch, jewelry, or artificial nails
Because improvement efforts can never be 100% successful at mitigating risk, we examined the variable impact of an intervention using values of 25%, 50%, and 75% reduction in noncompliance rates, as presented in Table 2. For example, if an ASC chooses to focus on improving skin preparation practices, the intervention would reduce the likelihood of that risk factor (i.e., Event 30 Failure to prepare the skin appropriately) from 0.125 to 0.0625 if the noncompliance rate was reduced by 50%. The original risk for developing an SSI is estimated by the model as 0.0044 (i.e., 44/10,000 cases will develop an SSI after surgery). The combined impact on the risk for developing an SSI for each of these combinations at different impact levels is presented in Table 3. For example, if the ASC selected interventions that targeted both the failure to prepare the skin appropriately (i.e., Event 30) and training for staff in infection control practices (i.e., Event 142, staff not well trained in infection control) and expected only a 25% reduction in noncompliance rate for each, the probability of an SSI would actually be further reduced to 0.0039. On the basis of results presented in Tables 1 to 3, we proposed an intervention aimed at Event 642 Failure to protect the patient effectively. Specifically, the intervention was designed to target skin preparation practices; proper timing and administration of antibiotics; staff training in infection control practices; preventing glove punctures; and procedures to ensure removal of watches, jewelry, and artificial nails. The proposed intervention targeted 2 important processes of patient care: (1) infection control practices and (2) communications between health care providers.
Infection Control Practices The intervention included developing and instituting guidelines for infection control practices at ASCs, modeled after the guidelines provided to hospitals and including policies and procedures for the prevention of preoperative infection transmission
4
www.journalpatientsafety.com
(e.g., hygiene, infectious waste, personal protective equipment, infectious patients, prevention of patient-to-patient transmission, assessment of risk factors of SSIs, risk procedures), prevention of intraoperative infection transmission (e.g., surgical disinfection and antisepsis, skin preparation, disinfection in the surgical environment), and prevention of postoperative infection transmission (e.g., dressing the postoperative wound). The intervention should also include opportunities for staff training on the proper technique to ensure consistency across providers and personnel and observation of staff technique in the ASC environment. We also recommended that a bundle addressing the following issues be identified: • Identification of high-risk patients (e.g., diabetes, MRSA, obesity) • Procedures to ensure removal of jewelry, watches, and artificial nails • Guidance on routine double gloving and proper response to glove punctures • Skin preparation practices including antisepsis and draping • Proper administration and timing of antibacterial agents
Communication Between Health Care Providers Improving communications across providers was an important component of the intervention. A nursing checklist could be used during the preoperative screening telephone call to ensure that major risk factors such as obesity and poor blood glucose control are properly identified. A referral back to the surgeon should be performed for any identified patient risk for developing an SSI (e.g., smoking, diabetes, and obesity) to ensure optimal preoperative control. This communication should also be shared with the anesthesiologist and the patient's primary care physician. A preoperative briefing should occur to determine whether the patient can proceed with surgery safely within the ASC environment, if inpatient surgery is the more appropriate option, or if the procedure should be done at all. In addition, ASCs should establish stopgap measures that prevent surgery from occurring for patients with multiple known risk factors who present out of control on © 2014 Lippincott Williams & Wilkins
the morning of surgery (e.g., blood glucose levels out of control, uncompensated congestive heart failure). Finally, a case debrief is an important mechanism for improving future care for infectious and noninfectious complications.
DISCUSSION The ST-PRA is a valuable tool for assessing the risks associated with patient safety events in several health care contexts. The ST-PRA is also in alignment with recent studies that suggest that interventions should be driven by outcomes and not by performance measures that may have little or no association with outcomes.5 In this study, we used the ST-PRA model to identify the risks of SSIs in the ASC environment. The ST-PRA considers both individual and unique risk combinations contributing to adverse outcomes. By including both quantitative and qualitative data, the tool’s scientific integrity can be maintained while reallife experiences are integrated into the models. The ST-PRA model allowed us to rank failure points (i.e., events) based on their contribution to an SSI. We found that Event 642 Failure to protect the patient effectively was the most critical unique event with the
highest independent contribution to the occurrence of SSIs: 51.87% of SSIs were caused by this failure. Consequently, we proposed an intervention aimed at this failure. Each of these component failures has received considerable attention in the literature. The patient’s skin is a major source of contamination in SSIs, and several studies have demonstrated the inadequacy of skin preparation practices.16 Hence, skin preparation practices need to be standardized and staff needs to be adequately trained in these procedures. Studies demonstrate varying practices among health care providers for the proper administration of antibiotics in orthopedic and nonorthopedic surgical cases.9–11 Prophylactic antibiotic consistency is a primary prevention strategy for SSIs. Glove perforation rates approach 35% and double the risk for an SSI. They increase with the surgery duration and often go unnoticed.17–19 Double gloving reduces the glove puncture risk and was found in the ST-PRA model as an effective strategy for mitigating the risk for SSIs in the ASCs.19 However, it should also be noted that double gloving reduces surgical dexterity.19–21 More studies are needed to understand the tradeoff between the reduction in glove puncture risk and reduction
and a fault tree of the model was constructed. The ST-PRA model was refined and revised to create a model that had face validity with technical experts who understood the procedure and the processes of care. Once validated, the model was executed using a risk assessment software (Relex Reliability Studio 2007 v2; PTC Corporate Headquarters, Needham, MA) to determine the importance of individual measures and failure combinations. The fault tree contained 283 events (failure points) and 203 gates; the complete fault tree can be found in Slonim et al,6 with special attention to Figure 9, which depicts the part of the fault tree containing the top 5 minimal cut sets, as discussed subsequently in the “Results” section of this article. The specific sequence of risk points (i.e., failures) contributing most to an SSI were identified next, followed by a sensitivity analysis to ensure that model conclusions were robust and consistent with the literature. Finally, on the basis of the failure sequences, an intervention aimed at reducing the SSI risk was proposed.
Data Sources for Building the Fault Tree Model To ensure that the ST-PRA model captured all possible process factors, 4 different data sources were used: (1) the peerreviewed and gray literature, (2) national databases, (3) ASC site visits, and (4) technical expert panel (TEP) input. Each source informed the data collection effort for the other sources in an iterative fashion. That is, information gleaned during the literature review informed ways to analyze the databases; information collected during the site visits or during the TEP meetings informed additional data analyses and literature searches.
Literature Review An extensive peer-reviewed and gray literature review was conducted to identify potential risk factors associated with deep incisional SSIs from surgical procedures. The review was limited to publications in English published since 2000. However, some seminal papers before 2000 were included, particularly if they dictated current clinical practices. Search engines included PubMed, the Cochrane Collaborative, CINAHL, and others as appropriate. A series of key word search terms were used within the following categories of interest: (1) patient demographics (e.g., age, race, insurance status, income, socioeconomic status), (2) patient health status and comorbidity, (3) surgical risk factors (e.g., surgical risk factors, procedural complications, medical errors, surgical infections, SSIs, health care–associated [acquired] infections [HAIs], surgical procedures, arthroscopic procedures, outcomes), (4) organizational context (e.g., academic medical centers, community hospitals, for-profit surgical centers, freestanding ASCs), and (5) infection control practices. The literature review was also used for obtaining discrete probability estimates and their ranges, required for the sensitivity testing. An article was included when it studied ASC-specific studies on complications; complications in different contexts (e.g., academic); hospital- and care-related infections and their risks; risk estimates related to demographics, conditions, or procedures; support or counterevidence from the site visit estimates; and risk factor estimates related to the patients, providers, structures, or teams. Studies were excluded when these did not provide specific information on surgery complications, focused only on hospital and not surgical infections, did not address interventions, and were opinion and not empirically based. The extant peerreviewed literature was enhanced by a review of gray literature to provide important information for inclusion in the risk models.
National Databases Discharge databases have an important role in studying SSIs. Although ASCs provide an important setting for the performance
2
www.journalpatientsafety.com
J Patient Saf • Volume 00, Number 00, Month 2014
of ambulatory procedures, these are constrained in that complications, particularly infections, may receive follow-up care in the physician’s office; in the emergency department (ED); or, for severe cases, in the hospital. Therefore, we complemented our study with the data reported in publicly available databases. We analyzed the Agency for Healthcare Research and Quality’s State Ambulatory Surgery, Nationwide Inpatient Sample, and State Emergency Department databases for the states of California, Maryland, and New Jersey for the period of 2006 to 2008 to capture information about infection rates and demographic and institutional characteristics related to specific procedures.7
ASC Site Visits Another important source of process and sociotechnical information was 4 ASCs based in Maryland and Washington, DC, each representing a different care context: an academic hospital– associated ASC; a community hospital–associated ASC; a freestanding for-profit ASC; and a free-standing, hospital-associated, pediatric ASC. The different care contexts informed the ST-PRA model about the SSI risks in these settings and created the learning opportunities across these settings. Site visits included 3 major activities: • A review and comparison of policies and procedures related to SSIs, including preoperative, operative, and postoperative policies governing patient care, room cleaning, disinfection, and equipment disinfection and sterilization • Informal exploratory interviews with a selection of 5 to 8 staff members from the participating ASCs to learn about infection prevention policies and procedures in place • Process flow comparisons across sites, noting differences in policies, procedures, and facility characteristics
TEP Reviews A panel of technical experts was convened to guide the STPRA modeling. The TEP members were selected to ensure representation of the relevant issues, namely, surgeons, anesthesiologists, nurses, and a physician’s assistant, working in inpatient and/or outpatient settings, and representatives from the Centers for Disease Control and the Centers for Medicare and Medicaid Services. The TEP’s input helped to shape the fault tree’s design and the final selection of an intervention.
Sensitivity Analysis After fault tree development, a sensitivity analysis was conducted for events in which the literature reported large variations in the probabilities or for which failure probabilities were highly site dependent. We varied these probabilities in the base case (i.e., original model) within the ranges suggested in the literature. When a probability estimate was unavailable, an anchor estimate was obtained from technical experts. For example, hand washing is a common approach for preventing bacterial spread and has a positive impact on preventing SSIs. Compliance rates for hand washing range between 40% and 90%. In the sensitivity analysis, we varied the probability for hand washing compliance from 40% to 90% to better understand its impact on mitigating SSIs. The sensitivity analyses ensured that the conclusions were appropriate even if the basic-level event probabilities were somewhat uncertain at the beginning of the modeling exercise.
Minimal Cut Sets and Importance Measures A cut set corresponds to a set of basic events whose occurrence causes the top event (i.e., SSI).8 The minimal cut set identifies the major critical paths, composed of multiple failure points. Cut set analysis is complemented by importance measures © 2014 Lippincott Williams & Wilkins
J Patient Saf • Volume 00, Number 00, Month 2014
that rank the most significant individual risks based upon their contribution to the top-level event (i.e., SSI). These measures assess the criticality of the risk in the model by assessing the absolute risk, the risk’s relative importance, or the risk’s frequency in the model. Commonly used relative importance measures in risk assessment modeling include the criticality, Birnbaum, and Fussell-Vesely measures.9 Both the Birnbaum and Fussell-Vesely measures anchor an individual risk estimate within the context of the model’s other risks, whereas the criticality measure is a measure of absolute risk, which identifies the independent risk contribution of a basic event. The importance measure selected depends, in part, on the type of model created and the purpose of the modeling exercise. For this study, we focused on the criticality measure because this measure permitted the rank ordering of the most critical contributors to the relatively rare event of SSIs and facilitated the identification of interventions that are most likely to improve system performance.
RESULTS Data Sources Literature Review The literature demonstrated mixed results as to whether infection rates are lower for ASCs compared with hospital settings.2,10 However, recently, HAIs originating in ASCs and other outpatient settings have reportedly increased.11–13 In 2008, the largest HAI outbreak was reported and traced to unsafe injection practices.14 This event represented the largest hepatitis C virus outbreak in U.S. history, exceeding 63,000 contacts and 115 infections possibly linked to this event. A literature survey suggested the following general risk factors for the development of SSIs, which were incorporated into the ST-PRA2: • Patient factors (e.g., comorbidities, malnutrition, obesity, tobacco use, age, ASA score, positive methicillin-resistant Staphylococcus aureus [MRSA] nasal culture) • Surgery-related factors Procedural factors (e.g., surgery type, antibiotic prophylaxis, surgical duration, hypothermia, blood loss requiring transfusion, spillage of luminal contents) Surgeon-related factors (e.g., S. aureus carriers among orthopedic team members, wristwatch users) • Operating room (OR) environmental factors (e.g., people in the OR; frequent OR door opening; movement of OR personnel, which disrupts the positive-pressure OR environment) ○
Risk Assessment of Surgical Site Infections
of facilities. Some ASCs were equipped with isolation bays reserved for MRSA patients, whereas others were not. The patient flow mapping and structural factors served as the foundation for building the process component of the fault tree and the riskinformed intervention.
TEP Reviews Once constructed, the fault tree was validated by the TEP for content, structure, and accuracy. The fault tree’s logic was revised based on TEP comments. The TEP also provided a framework for face validity for the probability estimates. When large variability in probability estimates remained either because of large reported variations in the literature or because the probabilities were highly site dependent, a sensitivity analysis was performed by running the model for the base case and for each variation. The literature review, ASC site visits, and discussion with the TEP identified similar lapses and variability in infection control practices among health care professionals and among the different types of ASCs. Often, these lapses resulted from inconsistency in guidelines or personnel training. There is considerable variability in the knowledge of infection control practices among health care professionals.15 Information on how the system fails and how frequently it fails was embedded into the model through Boolean logic and probability estimators.
Risk‐Informed Intervention We determined the SSI rate and the top 5 minimal cut sets for each variation of the base model considered in the sensitivity analysis to understand how and whether these had changed over the base case. The following events were included in the top 5 minimal cut sets of the base case and all of its variations: • Event 173 Staff fails to provide patient with instructions for weight reduction • Event 660 Patient fails to notice infection during home care • Event 642 Staff fails to protect patient effectively • Event 543 SSI risk for obese patient, weight not reduced and nutrition not improved • Event 450 Obese but not diabetic patient (30 ≤ body mass index < 40) • Event 182 Failure to administer indicated antibiotics
○
National Databases Database analysis demonstrated infection rates that were significantly lower than those reported for similar procedures, possibly because of underreporting or missing data. Consequently, these databases were mainly used to provide estimates on volumes, institutional characteristics, and patient demographics for the most common outpatient procedures and as inputs in the fault tree model as probability estimates.
ASC Site Visits Despite the different care contexts, many policies and procedures were common across the 4 sites. Some process variability in the preoperative, operative, and postoperative processes was also evident. For example, although each ASC monitored the blood glucose levels of patients with diabetes, only the pediatric ASC engaged an endocrinologist to ensure patient optimization before surgery. Other differences included the capacity and layout © 2014 Lippincott Williams & Wilkins
The events overlapping these minimal cut sets provided the most significant opportunity for preventing risk. As a result, these 5 events became the focus of our next steps, determining the importance measures and designing a risk-informed intervention. Table 1 presents the events from the ST-PRA model ranked in order of criticality. For example, Event 642 Failure to protect the patient effectively ranked as the most critical unique event with the highest independent contribution to the occurrence of SSIs (i.e., 51.87% of all SSIs occur because of this failure). Figure 1 depicts the fault tree section that contains this event. This event is followed in criticality by events corresponding to obese patients whose weight is not reduced and whose nutrition is not improved before the surgery (Events 450 and 543 in Table 1) because of failures such as the failure of staff to provide the patient with instructions for weight reduction . Because of its high criticality and the fact that it appeared in 4 of the 5 top cut sets, Event 642 Failure to protect the patient effectively and its component events were recommended as the focus for intervention development. The most important components that compose this failure, as indicated by their contribution to the risk for developing an SSI, include the following: • Failure to prepare the skin appropriately preoperatively • Antibiotic-related failures related to timing or administration www.journalpatientsafety.com
3
J Patient Saf • Volume 00, Number 00, Month 2014
Bish et al
TABLE 1. Importance Measures on Events in Rank Order of Criticality Event Event 642 Failure to protect the patient effectively (failure related to skin preparation, antibiotics administration, shaving surgical site, etc) Event 450 Obese but not diabetic patient (30 ≤ BMI < 40) Event 543 SSI risk for obese patient, weight not reduced and nutrition not improved Event 173 Staff fails to provide patient with instructions for weight reduction Event 142 Staff not well trained in infection control Event 182 Failure to administer indicated antibiotics Event 677 Patient is a smoker but has no COPD Event 456 Patient is a smoker and has COPD Event 659 Glove puncture Event 30 Failure to prepare skin appropriately Event 138 Failure to remove watch, jewelry, artificial nails Event 644 Failure to readminister antibiotics for longer surgery Event 419 Patient fails to come for postoperative visit Event 136 Failure to wash hands properly (OR staff) Event 687 Patient does not provide accurate information on smoking status Event 241 Tourniquet time > 60 min Event 660 Patient fails to notice infection during home care
Criticality
Birnbaum
Fussell-Vesely
0.5187
0.0113
0.5187
0.3147 0.3043 0.3042 0.2583 0.2331 0.2003 0.1649 0.1550 0.1457 0.1291 0.1166 0.1162 0.10332 0.0995 0.0956 0.0929
0.0047 0.0409 0.0015 0.0038 0.0051 0.0070 0.0135 0.0038 0.0051 0.0038 0.0051 0.0034 0.0038 0.0012 0.0051 0.0025
0.3147 0.3043 0.3042 0.2583 0.2331 0.2003 0.1649 0.1550 0.1457 0.1291 0.1166 0.1162 0.1033 0.0995 0.0956 0.0929
BMI, body mass index; COPD, chronic obstructive pulmonary disease.
• Staff not well trained in infection control practices • Glove puncture • Failure to remove watch, jewelry, or artificial nails
Because improvement efforts can never be 100% successful at mitigating risk, we examined the variable impact of an intervention using values of 25%, 50%, and 75% reduction in noncompliance rates, as presented in Table 2. For example, if an ASC chooses to focus on improving skin preparation practices, the intervention would reduce the likelihood of that risk factor (i.e., Event 30 Failure to prepare the skin appropriately) from 0.125 to 0.0625 if the noncompliance rate was reduced by 50%. The original risk for developing an SSI is estimated by the model as 0.0044 (i.e., 44/10,000 cases will develop an SSI after surgery). The combined impact on the risk for developing an SSI for each of these combinations at different impact levels is presented in Table 3. For example, if the ASC selected interventions that targeted both the failure to prepare the skin appropriately (i.e., Event 30) and training for staff in infection control practices (i.e., Event 142, staff not well trained in infection control) and expected only a 25% reduction in noncompliance rate for each, the probability of an SSI would actually be further reduced to 0.0039. On the basis of results presented in Tables 1 to 3, we proposed an intervention aimed at Event 642 Failure to protect the patient effectively. Specifically, the intervention was designed to target skin preparation practices; proper timing and administration of antibiotics; staff training in infection control practices; preventing glove punctures; and procedures to ensure removal of watches, jewelry, and artificial nails. The proposed intervention targeted 2 important processes of patient care: (1) infection control practices and (2) communications between health care providers.
Infection Control Practices The intervention included developing and instituting guidelines for infection control practices at ASCs, modeled after the guidelines provided to hospitals and including policies and procedures for the prevention of preoperative infection transmission
4
www.journalpatientsafety.com
(e.g., hygiene, infectious waste, personal protective equipment, infectious patients, prevention of patient-to-patient transmission, assessment of risk factors of SSIs, risk procedures), prevention of intraoperative infection transmission (e.g., surgical disinfection and antisepsis, skin preparation, disinfection in the surgical environment), and prevention of postoperative infection transmission (e.g., dressing the postoperative wound). The intervention should also include opportunities for staff training on the proper technique to ensure consistency across providers and personnel and observation of staff technique in the ASC environment. We also recommended that a bundle addressing the following issues be identified: • Identification of high-risk patients (e.g., diabetes, MRSA, obesity) • Procedures to ensure removal of jewelry, watches, and artificial nails • Guidance on routine double gloving and proper response to glove punctures • Skin preparation practices including antisepsis and draping • Proper administration and timing of antibacterial agents
Communication Between Health Care Providers Improving communications across providers was an important component of the intervention. A nursing checklist could be used during the preoperative screening telephone call to ensure that major risk factors such as obesity and poor blood glucose control are properly identified. A referral back to the surgeon should be performed for any identified patient risk for developing an SSI (e.g., smoking, diabetes, and obesity) to ensure optimal preoperative control. This communication should also be shared with the anesthesiologist and the patient's primary care physician. A preoperative briefing should occur to determine whether the patient can proceed with surgery safely within the ASC environment, if inpatient surgery is the more appropriate option, or if the procedure should be done at all. In addition, ASCs should establish stopgap measures that prevent surgery from occurring for patients with multiple known risk factors who present out of control on © 2014 Lippincott Williams & Wilkins
the morning of surgery (e.g., blood glucose levels out of control, uncompensated congestive heart failure). Finally, a case debrief is an important mechanism for improving future care for infectious and noninfectious complications.
DISCUSSION The ST-PRA is a valuable tool for assessing the risks associated with patient safety events in several health care contexts. The ST-PRA is also in alignment with recent studies that suggest that interventions should be driven by outcomes and not by performance measures that may have little or no association with outcomes.5 In this study, we used the ST-PRA model to identify the risks of SSIs in the ASC environment. The ST-PRA considers both individual and unique risk combinations contributing to adverse outcomes. By including both quantitative and qualitative data, the tool’s scientific integrity can be maintained while reallife experiences are integrated into the models. The ST-PRA model allowed us to rank failure points (i.e., events) based on their contribution to an SSI. We found that Event 642 Failure to protect the patient effectively was the most critical unique event with the
highest independent contribution to the occurrence of SSIs: 51.87% of SSIs were caused by this failure. Consequently, we proposed an intervention aimed at this failure. Each of these component failures has received considerable attention in the literature. The patient’s skin is a major source of contamination in SSIs, and several studies have demonstrated the inadequacy of skin preparation practices.16 Hence, skin preparation practices need to be standardized and staff needs to be adequately trained in these procedures. Studies demonstrate varying practices among health care providers for the proper administration of antibiotics in orthopedic and nonorthopedic surgical cases.9–11 Prophylactic antibiotic consistency is a primary prevention strategy for SSIs. Glove perforation rates approach 35% and double the risk for an SSI. They increase with the surgery duration and often go unnoticed.17–19 Double gloving reduces the glove puncture risk and was found in the ST-PRA model as an effective strategy for mitigating the risk for SSIs in the ASCs.19 However, it should also be noted that double gloving reduces surgical dexterity.19–21 More studies are needed to understand the tradeoff between the reduction in glove puncture risk and reduction
