Implementing an electronic hand hygiene system improved compliance in the intensive care unit

Abstract

Background:

Hand hygiene (HH) compliance is low and difficult to improve among health care workers. We aim to validate an electronic HH system and assess the impact of this system on HH compliance and quality changes over time at both group and individual levels.

Methods:

An automated electronic HH system was installed in a 10-bed surgical intensive care unit.

Results:

The full HH compliance rate increased significantly from 8.4% in week 1 to 20.5% in week 16 with week 10 being the highest (27.4%). The partial compliance rate maintained relative consistency between 13.2% and 20.0%. The combined compliance rate (full compliance rate + partial compliance rate) increased from 23.5% in week 1 to 34.6% in week 16 with week 10 being the highest (41.4%).

Discussion:

We found significant variations among providers in terms of HH opportunities per shift, full compliance, partial compliance and combined compliance rates. The average duration of hand rubbing over time in partial compliance occurrences did not change significantly over time.

Conclusions:

A sensor-based platform with automated HH compliance and quality monitoring, real time feedback and comprehensive individual level analysis, improved providers’ HH compliance in an intensive care unit. There were significant variations among individual providers.

BACKGROUND

Despite extensive focus, hospital-acquired infections are still one of the most common adverse events in health care delivery, affecting 7%−10% of all patients.¹ Hospital-acquired infections and spread of antimicrobial resistance can be reduced by improving hand hygiene (HH) compliance among health care workers.² However, it remains a challenge to achieve and sustain high HH compliance consistently. A systematic review from 96 studies found the median overall HH compliance rate of health care workers was only 40% with large variations in rate.³ In addition, HH compliance has been difficult to measure and there are currently several different monitoring methods being used.⁴ Research has shown that HH compliance declines significantly when the health care workers do not feel they are being observed.⁵ Manual observation of HH by a trained observer is the most commonly used method, but only captures a small fraction of the total HH events while being time-consuming, subject to bias (observer, observation, and selection biases) and rarely considerate of the HH technique.⁶

Technological advancements have made automated and continuous HH compliance monitoring possible.^6,7 Automated systems operate 24 hours a day and collect significantly more HH events than manual observations. They also enable real-time data to be provided without requiring unnecessary expenditure of staff hours.^6,7 Studies examining the use of automated HH systems in health care settings have yielded varying results and highlight the importance of device implementation strategies for sustained improvement in HH performance by health care personnel.⁸ Differentiation in situational HH opportunities is an important factor for an effective implementation of automated HH compliance monitoring systems,⁹ but most monitoring systems are only able to measure patient room entries and exits, contributing to a limited picture of the true HH behavior.⁷ In addition, many systems do not measure HH at room levels and the change of HH compliance over time on an individual level (intraindividual variability). Such insights could be used to personalize HH intervention and help providers perform HH at the right time and in the right areas where it is needed. Of note, HH compliance of individual providers can’t be assessed by the measurement of hand sanitizer consumption or automated activation counts of dispensers, which only provide surrogate measures of HH compliance.

Furthermore, relatively few studies investigate the quality of HH in health care settings.¹⁰ These studies show low adherence (8%−15%) with the World Health Organization (WHO) 6-step technique for HH. Various strategies have been used to improve HH quality, and frequent short training sessions have reduced methicillin-resistant staphylococcus aureus infection rates.¹¹ Remote video auditing has been used to improve the quantity of HH in appropriate clinical opportunities from 10% to 85%.¹² A recent study measured the quality and quantity of handwashing by providers using an automatic video auditing system with real-time feedback at handwash sinks.¹³ However, patient and provider privacy is a serious concern for any video observation systems.¹³

The Sanibit system is a sensor-based platform with automated HH compliance and quality monitoring, real-time feedback, and comprehensive HH compliance analysis. In this project, we aim to validate this electronic HH system against manual observation, and assess the impact of this system on HH compliance and quality changes over time at both group and individual levels in an intensive care unit (ICU) setting.

METHODS

Study design and settings

This prospective, observational study was conducted from October 12, 2020 to January 31, 2021 at a Level 1, 348-bed trauma hospital (University of Louisville Hospital, Louisville, KY, USA). The study was approved by the University of Louisville Institutional Review Board (IRB #19.0174, University of Louisville Human Subjects Protection Program Office, MedCenter One, Suite 200, 501 E. Broadway, Louisville, KY 40202, USA) and adheres to enhancing the quality and transparency of health research (EQUATOR) guidelines. Written informed consents were obtained from all participants. The clinical trial was registered at www.clinicaltrials.org prior to the implementation of this study (NCT03948672).

An automated electronic HH system (the Sanibit system, Microsensor Labs, Chicago, IL, USA) was installed in a 10-bed surgical ICU. Fifteen health care workers (12 nurses, 2 patient care assistants, and 1 secretary consisting of 75% of ICU staff) signed informed consents to participate in this study. Participation was incentivized by daily raffle drawings of $20 gift cards during the first month followed by performance-based gift cards thereafter. Participation was voluntary and participants could withdraw their consent at any point during the study without repercussion.

All participants were educated on the functionalities of the Sanibit system with real-time reminders turned on. Health care workers had access to their own data as well as unit-specific HH data through the Sanibit app. The unit manager and hospital administrators had access to the blinded trial data and advanced analysis in the Sanibit software to make management decisions on infection control. To protect the privacy and welfare of participants, all data presented to the unit manager and hospital administrators was rendered anonymous through back end processing.

The electronic HH system

System overview

Sanibit is an electronic HH tracking platform with real-time private reminders, HH motion tracking, individual-level compliance analytics, and a gamified app. When a health care worker enters or exits the patient room without washing hands, the Sanibit wristband will vibrate to remind the provider. If the provider takes sanitizer and rubs hands for the desired time, the wristband displays a green notification light. All data concerning HH opportunities are reported online 24/7 and analyzed at the desired level of detail (room, provider, group, etc.). Personalized web and mobile dashboards support game-based team competition and self-motivation for providers (Fig 1A).

Fig 1.

The Sanibit system overview. (A) Sanibit is an electronic HH tracking platform with real-time private reminders, HH motion tracking, individual-level compliance analytics, and a gamification app. When a health care worker enters or exits the patient room without washing hands, the Sanibit wristband will vibrate to remind the provider. If the provider takes sanitizer and rubs hands for the desired time, the wristband displays a green notification light. All data concerning HH opportunities are reported online 24/7 and analyzed at the desired level of detail (room, provider, group, etc.). Personalized web and mobile dashboards support game-based team competition and self-motivation for providers. (B) Three Sanibit controllers were installed inside each patient room and one installed outside the room. Additional controllers were installed at the nursing station, soiled utility room and other common areas. Each controller has an ultrasonic sensor and can communicate with the wristbands via Bluetooth and upload data to the cloud server through WiFi.

Hardware

Sanibit consists of user-specific wristband sensors and patient-room-specific controllers. Each health care worker wears his/her Sanibit sensor wristband, with main components including a motion sensor, a Bluetooth chip, 2 LEDs and a vibrating motor. Three Sanibit controllers were installed inside each patient room and one installed outside the room. Additional controllers were installed at the nursing station, soiled utility room and other common areas. Each controller has an ultrasonic sensor and can communicate with the wristbands via Bluetooth and upload data to the cloud server through WiFi. Controller installation in a typical patient room is illustrated in Figure 1B.

HH opportunity detection

The Sanibit wristband determines its location (outside or inside a room) based on the strength of the Bluetooth signals received from nearby controllers. When the wristband detects a location change (eg, from outside a patient room to inside the room), it records a HH opportunity. A HH opportunity is in general defined as when a health care worker enters or exits a patient room, except for the following scenarios:

• To/from the soiled utility room. There is one soiled utility room in the ICU, which is used to dispose dirty equipment or contaminated items. To reflect the realistic clinical situation when providers cannot wash hands while holding dirty items in and out of a patient room, providers have 30 seconds to leave the patient room to go to the soiled utility room, without required HH.

• Quick entry/exit of patient room. To reflect the realistic situation when providers enter the room without touching the patient, health care workers have 30 seconds after a full HH compliance to go in or out of patient rooms without requiring additional HH (see “Hand Hygiene Compliance Definitions” below).

We ensured the wristbands be worn by comparing daily staff work schedule with HH event time stamp in the database.

HH event detection

When a health care worker takes alcohol-based hand rub (ABHR) or soap from a dispenser, the controller installed next to the dispenser detects the health care worker’s hand with its built-in ultrasonic sensor, and sends commands to notify the health care worker’s wristband of the detection of a HH event. The wristband then activates its motion sensor to determine the duration of hand rubbing.

HH compliance definitions

For each HH opportunity, HH compliance is calculated based upon an associated HH event. A provider can perform HH either before or after a HH opportunity (entering/exiting a room). In this trial, a time window is set from 30 seconds before to 10 seconds after a HH opportunity to detect HH events. HH compliances are evaluated for each HH opportunity based on the following definitions:

1. A noncompliance HH is defined as no HH events detected, or a HH event with zero hand rubbing duration;

2. A partial compliance is defined as a HH event with a hand rubbing duration less than 5 seconds for ABHR and 15 seconds for soap/water;

3. A full compliance is defined as a HH event with a hand rubbing duration of at least 5 seconds for alcohol-based solutions or 15 seconds for soap/water.

Of note, the 5 seconds for ABHR and 15 seconds for soap/water were determined by the hospital infection control team based on their current goals for HH at University of Louisville Hospital.

Real-time and private reminders

The following reminders on the wristbands were used to provide real-time feedback and interventions to the providers:

1. When a provider takes sanitizer and rubs hands for 5 seconds, the wristband shows green notification light;

2. At 5 seconds after a HH opportunity, if the wristband does not detect an associated HH event, it will vibrate to remind the health care worker. After the vibration reminder, the provider has 5 seconds to start a HH event, otherwise this HH opportunity will be marked as noncompliance.

Cloud-based software for compliance analysis and motivation

All HH opportunity data are collected and analyzed at the desired level of detail (room, provider, group, etc.). Each individual provider receives a daily update and detailed analysis on their performance from a mobile-friendly web app. Built into these analytic visualizations are motivational factors taken from well-established gamification practices, including competition with peers, satisfying animations, and an achievements and rewards system.

All participants are encouraged to access the Sanibit website (ulh.sanibit.com) and login with their own credentials to view their personal HH opportunities and compliance rate as well as the ICU’s total compliance rate. The unit manager and infection control team receive weekly analyses consisting of anonymous individual and unit reports on HH compliance. These data are used for coaching, performance tracking, and reporting purposes.

Incentive programs

From October 12, 2020 to November 14, 2020, all participants were entered into a daily $20 gift card raffle. From November 15, 2020 to December 14, 2020, a scoring system was instituted: each full compliance HH event earned 2 points; each partial compliance event earned 1 point; each noncompliance earned −0.5 points. When one accumulated 100 points, he/she could redeem points for a $10 gift card. Starting December 15, 2020, a new point system was adopted: each full compliance event earned 1 point; each partial-compliance event earned 0.5 points; each noncompliance earned 0 points. Gift cards ($10) were once again redeemable once reaching 100 points.

Statistical analysis

The HH data captured by the Sanibit system during the first 16 weeks of trial were summarized on a weekly basis by total HH opportunities, full compliance rates, partial compliance rates, and combined compliance rates (full compliance rate plus partial compliance rate). The event date and time, room number, and in/out event were recorded for each manual observation via SpeedyAudit application. For each manual observation, the event date and time, room number, and in/out event were used to see whether there was a matched HH opportunity in the Sanibit system, where the matched time window was within 1 minute, and all other factors including date, room number and in/out event were exactly matched. χ² tests were applied to assess for a significant time effect. In addition to total unit review, weekly analysis was performed on individual provider’s HH opportunities per shift, full compliance rate, partial compliance rate, and combined compliance rate in order to examine the individual-level variations. Average duration of hand rubbing for the entire ICU was summarized on a weekly basis, where 5 seconds for rubbing time was used to define whether a full compliance was achieved. One-way analysis of variance was used to test whether there was a time effect. A statistical test was claimed significant if the associated P value < .05. All statistical analyses were carried out in the statistical software R version 3.6.2 (https://www.r-project.org/).

RESULTS

We initially compared HH compliance data detected by the Sanibit system with manual observations by trained infection control specialists and researchers. Based on the location and time, we identified 213 matched opportunities which were detected by both the Sanibit system and manual observation. Sixty-six percent of these HH opportunities were consistent between the 2 methods. The discrepancy was most likely from 3 sources. First, in manual observation, the action to get ABHR/soap is not standardized and manual observers will record the HH event whenever there is any motion to get the ABHR/soap. However, the Sanibit system will only record the HH event when the hand motion actually activates the sensor in the controllers and will not record the HH event when it does not activate the sensor. Second, the Sanibit system only records HH events with enough wrist motions during hand rubbing while manual observation will record any hand rubbing motions. Third, the Sanibit system requires providers to start rubbing hands with 3 seconds after getting ABHR/soap to be counted as a HH event while manual observation does not have this requirement.

The Sanibit system detected 1362 HH opportunities in week 1 (October 12–18, 2020) and 747 opportunities in week 16 (January 25–31, 2021) with week 8 being the highest period (1833 opportunities) (Fig 2). Among all HH opportunities, 92.5% of them were hand rubbing with ABHR and 7.5% were hand washing with soap and water. The average number of HH opportunities per provider per shift ranged from 36.1 to 95.6. The full HH compliance rate increased significantly from 8.4% in week 1 to 20.5% in week 16 with week 10 being the highest (27.4%). The partial compliance rate maintained relative consistency between 13.2% and 20.0%. The combined compliance rate (full compliance rate + partial compliance rate) increased from 23.5% in week 1 to 34.6% in week 16 with week 10 being the highest (41.4%). Interestingly, the holiday weeks (weeks 7, 11, 12) showed significant declines in HH compliance. The incentive program with penalty for noncompliance increased compliance significantly and it appeared that the incentive program without penalty for noncompliance resulted in a decrease in overall HH compliance when compared to the incentives with penalty (Fig 2).

Fig 2.

Entire ICU hand hygiene opportunities, full hand hygiene compliance rate, partial hand hygiene compliance rate, and combined hand hygiene compliance rate from week 1 to week 16.

We then analyzed the individual HH compliance over time. We found significant variations among providers in terms of HH opportunities per shift, full compliance, partial compliance, and combined compliance rates. The most improved provider increased his/her HH combined compliance rate from 14.29% in week 1 to 50.00% in week 16 and the least improved provider actually showed deterioration over time (40.91% in week 4 to 28.77% in week 16) (Fig 3).

Fig 3.

Individual level hand hygiene opportunities, full hand hygiene compliance rate, partial hand hygiene compliance rate, and combined hand hygiene compliance rate from week 1 to week 16.

Lastly, we studied the average duration of hand rubbing over time in partial compliance occurrences, and we found that the average hand rubbing duration in week 1 was 2.99 seconds and increased to 3.05 seconds in week 16. The range over the entire study was between 2.51 and 3.09 seconds; however, there were no significant differences between weeks. We further divided hand rubbing durations into 5 groups: 0–1 second, 1–2 seconds, 2–3 seconds, 3–4 seconds, and 4–5 seconds over the 16 weeks and calculated their percentages (Fig 4). We found no significant difference for hand rubbing durations over time. For this trial, a full compliance was defined as a HH event with a hand rubbing duration of at least 5 seconds. The green light on the wristband appeared when 5 seconds of hand rubbing was achieved. No one rubbed hands after the green light was on and we considered the average duration of hand rubbing in the full compliance group as 5 seconds.

Fig 4.

Partial compliance hand rubbing duration (5 groups: 0–1 second, 1–2 seconds, 2–3 seconds, 3–4 seconds, and 4–5 seconds) changes over the 16 weeks, and no significant difference was found.

DISCUSSION

This is the first report on the use of the Sanibit system, an automated HH compliance/quality monitoring system, in the ICU. This system provides real-time feedback with detailed individual level analysis and a reward system. The main findings include: (1) the implementation of this system improved HH compliance (both full compliance and combined compliance rates) significantly; (2) there were significant variations among different providers, demanding precision interventions to improve HH compliance; (3) different incentive programs could affect the compliance rates directly, providing insights on how to promote desired behaviors among health care workers; and (4) the consistently short hand rubbing time in partial compliance reflects the baseline (as no reminder or intervention is applied for partial compliance) and indicates the strong needs for intervention in partial compliance.

Manual observation of HH has been criticized regarding data validity for multiple reasons. First, the logistical challenges of manual observation allow for the capture of only a very small percentage of total HH opportunities within a unit. Second, manual observation requires a significant amount of work and time from the infection control team, which can be extremely costly. Third, the reactivity to direct observation, the so-called observation or Hawthorne effect, presents a significant risk of bias.^14,15 In addition, human observers may apply shortcuts in perceiving and processing information and are another important source of potential bias or interpretational error.¹⁶ The Sanibit system is able to systematically evaluate HH performance 24 hours a day, 7 days a week avoiding many of these obstacles. All the data are uploaded automatically to the cloud and analyzed immediately for real-time feedback. Of note, wearing the Sanibit wristband might have the potential to create a Hawthorne effect by participants realizing that they are monitored.

Most of the previous research on HH focused on group level compliance rate. Only a few articles using identification (ID) badge-based technology were able to provide individual level data. We found significant variations in the HH compliance rates among different providers. The changes among different providers after the implementation of this electronic HH system varied significantly over time as well. The crucial benefit of individual-level HH data is that the impact of psychological predictors and mechanisms of HH performance can be analyzed at individual levels. Some providers showed very low compliance and some demonstrated significant improvement over time. Behavioral interventions for low compliance and low improvement performers may require alternative techniques compared to groups with better compliance rates. More research is needed to explore the potentially beneficial outcomes of individualized interventions or incentives. From the data gathered by the automated HH monitoring system, it could give hospitals an essential tool in understanding where poor compliance behaviors occur by location, provider, and time. Specifically, compliance rates varied dramatically by patient room, specific dispensers, duration of room entry to exit, time of the day, and specific week. The targeted interventions should be more precise to design programs and incentives for individuals to achieve the best results. Other well-known factors associated with the levels of HH compliance are staff profession, workload, and location of sanitizers,^17–19 which could be analyzed and used for designing precision interventions.

This study measured not only HH quantity but also HH quality. ID badge-based technology suffers from significant drawbacks including the inability to detect whether there is an actual alcohol/soap dispensing event because detection is not based on action, only distance. ID badge technology also suffers from an inability to detect duration of hand washing action, which the WHO recommends for a minimum of 20–30 seconds. Our study used a sequence of actions to be more accurate in defining the actual HH event: (1) confirmed HH opportunity; (2) confirmed alcohol/soap dispensing; (3) confirmed initiation of hand rubbing; and (4) defined hand rubbing duration. We found that despite incentive programs and full compliance reminders, the duration of hand rubbing for providers in partial HH compliance remained relatively consistent over 16 weeks, which presents a huge opportunity for improvement in the design and interventions.

This is the first study to track HH behaviors for both alcohol-based solution and soap/water solution as we installed sensors at both dispensers across the entire ICU. Iversen et al evaluated a similar system using ID badge and sensor-equipped alcohol solution dispensers and found that providers’ HH compliance was consistent over time, emphasizing an opportunity for hospitals to reach hygiene goals with individualized interventions based on data generated by their system.¹³ Dufour et al used radio-frequency ID badges and sensor-equipped alcohol solution dispensers and found that HH compliance with alcohol was dependent upon the providers’ occupation and personal behavior, number of providers, time spent in the room, and dispenser location.²⁰ No study have addressed the HH behaviors with soap/water solution, which is recommended by WHO when hands are visibly soiled.²¹

Our study applied several interventions to improve HH compliance. (1) The light on the controller indicated successful and adequate alcohol/soap dispensing. We noticed that many providers paid attention to this signal to ensure adequate dispensing. Before the trial, providers used nonstandard techniques to obtain solutions. When not enough solution was dispensed, it often leads to hand drying after only a few seconds and prevents further hand rubbing. (2) After consulting with providers, managers, and infection control experts, we decided to use hand rubbing of 5 seconds for alcohol solutions and 15 seconds for soap/water solutions as our full compliance HH measurements considering the current ICU practice and feedback. The initial goal was to first improve HH duration to a minimum of 5 seconds and then improve compliance toward WHO recommended time goals. (3) We applied 2 incentive programs, the first one included penalty for noncompliance and a higher reward for full compliance; the second one included no penalty for noncompliance and a lower reward for full compliance. These 2 incentive programs had significantly different impacts on HH compliance over time. (4) We noticed a significant drop in HH opportunities and compliance during holiday weeks, which could be related to staffing, behavior relaxation during nonstandard work periods, or change in patient acuity. (5) Login to the Sanibit system portal/app promoted personal reflection on the providers’ HH, individual point totals, and overall compliance of the entire ICU. However, some health care workers simply ignored the system. To further improve the HH compliance, hospital administration engagement, infection control initiatives, mandatory wearing of the wristbands for all providers, targeted HH education, and precision HH interventions will be keys to achieve the full benefits of this system.

There are several limitations in this prospective, observational study. First is the lack of control groups without using an electronic HH monitoring system. We studied the impact of an electronic HH system in the ICU without comparing to historical manual observation data due to significant differences in methodology. Second, not all health care workers of this ICU participated in the study. The behaviors of nonparticipating providers were not known. Third, we only enrolled dayshift providers and the compliance rate and behavioral HH data of night shifts were not recorded and this is our next aim for this trial. Fourth, adaptive gamification has been shown to improve learning by tailoring the reward to individual learners’ preferences.²² A study to compare 2 electronic HH systems found that function of real-time reminder and feedback has a more noticeable effect on promoting HH.²³ Our future research will focus on personalizing the feedbacks based on individual performance to maintain engagement.

CONCLUSIONS

A sensor-based platform with automated HH compliance and quality monitoring, real-time feedback and comprehensive individual level analysis, improved providers’ HH compliance in an ICU. There were significant variations among individual providers. Baseline hand rubbing time seemed to be difficult to influence over time without a time indicator.