Frequency and Nature of Infectious Risk Moments During Acute Care Based on the INFORM Structured Classification Taxonomy

OBJECTIVE In this study, we sought to establish a comprehensive inventory of infectious risk moments (IRMs), defined as seemingly innocuous yet frequently occurring care manipulations potentially resulting in transfer of pathogens to patients. We also aimed to develop and employ an observational taxonomy to quantify the frequency and nature of IRMs in acute-care settings. DESIGN Prospective observational study and establishment of observational taxonomy. SETTING Intensive care unit, general medical ward, and emergency ward of a university-affiliated hospital. PARTICIPANTS Healthcare workers (HCWs) METHODS Exploratory observations were conducted to identify IRMs, which were coded based on the surfaces involved in the transmission pathway to establish a structured taxonomy. Structured observations were performed using this taxonomy to quantify IRMs in all 3 settings. RESULTS Following 129.17 hours of exploratory observations, identified IRMs involved HCW hands, gloves, care devices, mobile objects, and HCW clothing and accessories. A structured taxonomy called INFORM (INFectiOus Risk Moment) was established to classify each IRM according to the source, vector, and endpoint of potential pathogen transfer. We observed 1,138 IRMs during 53.77 hours of structured observations (31.25 active care hours) for an average foundation of 42.8 IRMs per active care hour overall, and average densities of 34.9, 36.8, and 56.3 IRMs in the intensive care, medical, and emergency wards, respectively. CONCLUSIONS Hands and gloves remain among the most important contributors to the transfer of pathogens within the healthcare setting, but medical devices, mobile objects, invasive devices, and HCW clothing and accessories may also contribute to patient colonization and/or infection. The INFORM observational taxonomy and IRM inventory presented may benefit clinical risk assessment, training and education, and future research. Infect Control Hosp Epidemiol 2018;39:272–279

Healthcare-associated infections (HAIs) remain a major threat to patient safety. A significant proportion of such infections are likely preventable through the application of infection prevention measures, [1][2][3][4] such as those aiming to reduce the transmission of pathogens that may lead to patient colonization or infection. 5 Hand hygiene, for example, is widely recognized as one of the most effective practices to reduce infection rates and patient colonization with multidrug-resistant bacteria by reducing the transmission of microorganisms. 6 Strong evidence also suggests that environmental contamination of surfaces and objects contribute to HAI, 7-12 yet the behavioral focus of such studies is often limited to hand hygiene and environmental cleaning. While the practice of hand hygiene has been increasingly studied over the last decade for its role in infection prevention, considerably less knowledge exists regarding other important infection-related behaviors.
A growing body of evidence suggests that practices beyond those addressed by hand hygiene may be relevant in the transmission of microorganisms that results in patient colonization and infection, such as handling of mobile objects, 13,14 healthcare worker (HCW) private 15 and professional attire, 16,17 and medical devices. 11,14,18 Therefore, we hypothesize that an important portion of infectious risks lie in infectious risk moments (IRM), defined as seemingly innocuous, yet frequently occurring care manipulations that potentially result in the transfer of pathogens. Such IRM include yet go beyond existing indications for hand hygiene. 13 The design of infection prevention strategies that consider a broad range of infectious risks must begin with systematic identification and classification of IRMs. In a 2-part project, we conducted (1) exploratory observations to establish a comprehensive inventory of potential IRMs, which served as a basis for developing a taxonomy for structured observations and (2) structured observations to quantify the frequency and nature of IRMs in 3 distinct typical healthcare settings.

Design
We conducted a prospective observational study in 2 parts. First, we conducted live exploratory observations to identify a wide range of potential IRM and to establish a structured taxonomy called INFORM (INFectiOus Risk Moment) for identifying and classifying IRMs. Second, we conducted live structured observations based on the INFORM taxonomy. Parts of this methodology have been pilot tested previously. 13 The observations reported in the current manuscript do not include the pilot observations.

Setting
An intensive care unit (ICU), general medical ward, and emergency ward, including trauma unit, located at a 900-bed, university-affiliated, tertiary-care hospital were purposefully sampled to represent a broad range of care activities and potential infectious risks. All healthcare workers (HCWs) from the participating wards were included in the study. The study hospital has a well-established infection prevention and control (IPC) group with extensive state-of-the-art, written IPC standard operating procedures, weekly IPC rounds, and a designated IPC nurse consultant for each hospital ward.

Exploratory Observations
Observers with backgrounds in nursing (C.D.A. and V.G.) and human factors/psychology (L.C.) and extensive experience conducting observations for patient safety research carried out exploratory observations in all 3 settings. Field notes documented the care processes observed and any potential IRMs, which were operationally defined as behaviors potentially resulting in the transmission of pathogens that may result in patient colonization or infection. The observers discussed all identified potential IRMs regularly throughout the exploratory observation period together with a senior infection prevention physician (H.S.) and all potential IRM were collected in a database.
Based on the definition of IRMs and following the hand hygiene literature, IRMs were limited to moments resulting in potential transfer of pathogens to patients and their immediate surroundings (eg, bedding), rather than the larger translocation of microorganisms throughout the healthcare environment. For example, an HCW entering a patient room then, without doing hand hygiene, touching the patient's bedside monitor to silence an alarm (a behavior that occurs often and may introduce nonpatient flora to the patient environment) was not considered an IRM. Only behaviors that resulted in potential transfer of pathogens directly to the patient were considered.
We distinguished between noncritical patient sites (eg, intact skin, intact dressings, patient clothing), critical patient sites, defined as "body sites or medical devices that have to be protected against microorganisms potentially leading to HAI" 19 (eg, mucous membranes, catheter insertion sites, or open wounds), and patient bedding. Exploratory observations were conducted until saturation was achieved in each setting, that is, until no new IRMs were observed.

Structured Observation Taxonomy and Mobile Observation Tool Validation
Following exploratory observations, all IRMs were extracted from field notes and were systematically coded according to the source, vector, and endpoint from, through, and to which pathogens were transferred, respectively. This structure was used to establish the INFORM classification taxonomy, on which structured observations were based ( Figure 1). A mobile observation tool based on the INFORM taxonomy was programmed with Filemaker 14 (FileMaker, Santa Clara, CA).
To ensure the quality of observations, 2 observers (L.C. and S. P.) validated the mobile observation tool during a 1-month test period. The percentage of agreement between the 2 observers was calculated to measure sensitivity (detection of the same IRM) and Cohen's κ was calculated to determine interobserver agreement (ie, consistent classification of IRM) using STATA version 14 software (StataCorp, College Station, TX).

Structured Observations
Structured observations were carried out in the same 3 clinical settings using the mobile observation tool. Two observers (L.C. and S.P.) conducted live, structured observations in parallel to ensure systematic documentation of all IRMs. Structured observations targeted periods of active patient care, and both observers focused on the same HCW at once. Observation sessions of 30-60 minutes were deliberately conducted at different times throughout the workday to include many different HCWs who performed a diverse range of care tasks for multiple patients during each session. During live observations, both observers independently noted the source, vector, and endpoint of pathogens for each IRM according to the observational taxonomy as well as demographic information about the HCW being observed (ie, gender and professional category) and contextual information (ie, date, time, ward name, and patient isolation status) using the mobile observation tool (Appendix 1). No identifying patient or HCW data were collected during observations. For each observation period, we recorded the total amount of observation time, as well as the amount of active patient care time to calculate the density of IRMs per setting. Following each structured observation session, all observed IRMs were compared between the 2 observers, and any discrepancies were discussed until a consensus agreement was achieved. Frequent discussion among researchers to achieve consensus after each observation period was maintained throughout the study to ensure quality and to avoid drift between observers.

Ethics
The Cantonal Ethics Committee of Zurich formally waived the ethics requirement for this study (KEK-StV-Nr.73/14). Participation in observations was voluntary, and HCWs were free to opt out or stop observations at any time without providing justification.

Exploratory Observations
A total of 129.17 hours of exploratory observations resulted in the identification of 292 unique IRMs. Identified IRMs included moments of potential direct contact transmission (potentially infected or colonized HCW to patient) as well as potential indirect contact transmission via vectors such as care devices, mobile objects, and HCW clothing and accessories. Following exploratory observations, IRMs were systematically coded according to the source, vector, and endpoint of potential pathogen transfer, and these codes formed the basis of the INFORM structured taxonomy ( Figure 1).

Structured Observation Taxonomy and Mobile Observation Tool Validation
The 3-level taxonomy begins with classification of surfaces (loci) involved in the observed IRM according to source, vector, or endpoint of potential pathogen transfer (level 1: locus), then assigns each source, vector, and endpoint to a main category (level 2: surface), and specifies the exact nature (level 3: surface detail). Each observed IRM is then represented as a transmission chain composed of 3 loci (source, vector, and endpoint), with each locus having 2 levels of detail (surface and surface detail). Table 1 lists examples of archetypal observed and classified IRMs for each of the observed vectors.
During the 1-month test of the taxonomy using the mobile observation tool (5.5 hours of active patient care), observers 1 and 2 detected 123 (78.9%) and 118 (75.6%) of all observed IRMs, respectively. Based on this detection rate, the decision was made to have 2 observers present for all structured observations to ensure the highest possible sensitivity. For moments identified by both observers during the pilot test, the Cohen's κ measure of interobserver agreement was 0.75, indicating substantial agreement between individual observers. 20 Bedside  Table 2.
The vectors in the identified IRMs included hands (n = 596; 44.54%), gloves (n = 457; 34.16%), medical devices (n = 115; 8.59%), mobile objects (n = 102; 7.62%), invasive devices (n = 53; 3.96%), and HCW clothing and accessories (n = 15; 1.12%). Overall, 25.8% of IRM concerned moments of potential transmission of pathogens to a critical site, described in detail in Table 2A. Among the 217 IRMs dealing with medical devices and mobile objects as vectors, 143 IRMs (65.90%) involved the lack of disinfection of a device or object prior to patient contact. The 3 most frequently occurring IRMs per clinical setting are described in detail in Table 3.

discussion
Hands and gloves continue to be among the most important contributors to the transfer of pathogens in the healthcare setting. Nonetheless, we identified moments dealing with other vectors such as medical devices, mobile objects, invasive devices, and HCW clothing and accessories, which may also contribute to patient colonization and/or infection. While previous studies have shown that indications for hand hygiene occur between 8 per hour in pediatric wards and 30 per hour in ICUs, 21,22 we found that IRMs occurred with a frequency of 42.8 IRM per active care hour overall and up to 56.3 IRM per active care hour in emergency settings. Similar to opportunities for hand hygiene, the high frequency with which IRMs occur suggests that the cumulative risk of negative patient outcomes due to IRMs may be significant, although the risk of patient infection or colonization with multiresistant pathogens at any single IRM may be low. The fact that 25.8% of IRMs concerned moments of potential pathogen transfer to critical patient sites further highlights the clinical relevance of IRM for infection prevention.
The structured observations in this study were targeted to moments resulting in potential pathogen transfer to the patient, as opposed to movement of pathogens around the larger healthcare environment. Our exploratory observations nonetheless revealed that pathogen transfer from outside to inside the patient zone likely occurred, for example when coming from one patient to silence an alarm on another patient's monitor without hand hygiene, or when transporting mobile objects that come into contact with multiple consecutive patients during clinical rounds. These findings are consistent with other studies demonstrating that HCW hand hygiene compliance prior to initial contact with the patient or the patient environment is suboptimal. 23 Our results also challenge the "patient zone" concept, which defines the patient and his/her immediate surroundings (eg, bed rails, bedside table, and medical equipment) and frequently touched surfaces (eg, monitors, knobs, and buttons) as the patient zone and assumes that surfaces within the patient zone are colonized by patient flora. 19 When disinfection is omitted prior to contact with the patient or patient environment, 23 it is likely that pathogens from the healthcare environment are introduced to these surfaces. Such ambiguity is a major challenge to safe behavior. 24 For this reason, during observations, we considered that environmental surfaces could potentially harbor pathogenic bacteria regardless of their location inside or outside of the patient zone. Similarly, our findings are consistent with multiple systematic reviews demonstrating that the frequent movement of healthcare equipment 25 and care items 14 between patients, together with suboptimal or missing disinfection of such items, result in the transfer of pathogens between patients. Potential contamination or missing disinfection of medical devices and mobile objects (classified as source = "unknown status" and source detail = "no disinfection observed") accounted for 16.2% of IRMs observed in this study ( Table 2).
The transmission-based observational approach employed in this study, which sought to identify all behaviors potentially resulting in transmission pathogen, differs from traditional rule-based observations that measure compliance with existing local or national guidelines. Observations using the INFORM taxonomy could hence be employed in additional settings, regardless of local guidelines, to identify the most frequently occurring IRMs and to establish local infection prevention priorities.
This study has several limitations. It is possible that being observed influenced HCW behavior during this study. 26 It is unlikely, however, that this resulted in systematic bias because HCWs were not aware of exactly what was being observed. Observations were limited to contact transmission (ie, the most common mode of transmission 5 ) and did not consider airborne and droplet transmission. Furthermore, our observations did not consider other behaviors that may also impact infectious risks, such as those interfering with the patient's defense system against infectious risks (eg, immune status, skin integrity, cough reflex, etc) because the associated HCW behavior rarely occurs at the bedside. Moreover, these observations were conducted in a single university hospital located in a high-income setting, which limits the generalizability of our findings. Further exploration of the nature and frequency of IRMs using the INFORM structured observational taxonomy is warranted to assess local priorities for infection prevention efforts in additional care settings. Finally, the risk of transmission during each type of IRM remains unknown. We aimed to bridge this gap through a modified Delphi survey with an international panel of experts in infectious diseases, infection prevention and control, and microbiology, in which experts rated the likelihood of infectious outcomes (eg, colonization, infection) following archetypical IRM. 27 Despite these limitations, the combination of methods employed in this study was well suited to identify a wide range of potential IRMs and to systematically observe their frequency and nature in multiple healthcare settings. The resulting mobile observation tool featuring the INFORM taxonomy of source, vector, and endpoint of pathogens was useful for the systematic documentation and categorization of IRMs. Further observations based on the INFORM taxonomy may prove useful in other settings to identify the most frequently occurring IRMs, to establish educational content, and to prioritize targeted infection prevention strategies.

acknowledgments
We would like to wholeheartedly thank all healthcare workers for their participation in this study. We also gratefully acknowledge Claudia Dell Apollonia and Verena Gabler for their contributions to data collection during exploratory observations. We also thank Jasmina Bogdanovic and Sabine Greschek for their critical feedback on an advanced draft of this paper.