Speech Recognition Technology for Increasing Nursing Documentation Efficiency

Citation
Everett, M., Redner, J., Kalenscher, A., Durso, D. & Nguyen, S. (Fall 2022). Speech recognition technology for increasing nursing documentation efficiency. Online Journal of Nursing Informatics (OJNI), 26 (2), https://www.himss.org/resources/online-journal-nursing-informatics

Abstract

Speech recognition technology (SRT) has increased in prevalence as a solution for providers to improve documentation but has not been widely implemented for nursing. This retrospective study suggests that SRT improves nursing documentation efficiency by decreasing time spent in flowsheets.

Background: Speech Recognition Technology (SRT) has increased in prevalence as a solution for providers to improve their documentation efficiency. Because nursing documentation consists primarily of discrete data entry into flowsheets, SRT has not been widely implemented for nurses (Blackley et al., 2019). Nursing documentation, especially in flowsheets, is time consuming. Acute care nurses spend 19% to 35% of a 12-hour shift documenting in flowsheets (Collins et al., 2018).

Aims: The purpose of this retrospective study was to analyze the use of SRT for discrete documentation in medical-surgical (med-surg) nursing flowsheets and to assess the impact on documentation efficiency.

Method: The SRT program was implemented for nurses, including training and access to a library of voice commands for flowsheet documentation. The time spent in the shift assessment for SRT users was compared to non-SRT users at baseline and post implementation.

Results: SRT users spent 9.0% to 9.7% less time documenting within the nursing shift assessment flowsheet.

Limitations and Implications: The study was limited to retrospective quantitative data with implications for further research relating to nurse satisfaction, patient perceptions, and accuracy of documentation.

Conclusion: The results of this study indicate that SRT can provide efficiencies for nursing documentation beyond voice-to-text dictation.

Clinical documentation presents a significant time burden for nurses, detracts from patient care, and contributes to clinician burnout (Ommaya et al., 2018). Acute care nurses spend 19% to 35% of a 12-hour shift documenting in flowsheets (Collins et al., 2018). Adding to the documentation time are optimization requests to create more flowsheet rows to capture standards of care. Collins et al. (2018) urged nursing informaticists to address the increasing documentation burden for nurses through technology innovations.

Speech Recognition Technology (SRT) refers to software that converts speech into text. In healthcare, SRT is traditionally used by providers to dictate free-text clinical notes into the electronic health record (EHR). Providers report that SRT saves them time, increases their efficiency, and allows them to quickly document more relevant details (Blackley et al., 2020). SRT has limited implementation and research beyond providers (Blackley et al., 2019) and use cases for SRT have focused primarily on narrative notes.

One of the features of SRT programs is the ability to create voice-activated macroinstructions (macros) that execute a series of keystrokes with a single voice command, such as “Create Progress Note”; however, there have been limited use cases for nursing. SRT has not been widely implemented for nurses because nursing documentation predominantly consists of discrete data entry into flowsheets. The investigators discovered a method to utilize voice commands to enter discrete nursing documentation within EHR fields for example, “Chart Normal IV Assessment”.  

Purpose/Research Question/Hypothesis

Use cases for SRT beyond voice-to-text were developed to include discrete data entry into nursing flowsheets. The objective was to take advantage of the organization’s enterprise SRT license, expanding the scope beyond providers to nursing. The purpose of the study was to evaluate the effect of SRT on documentation efficiency. The investigators predicted that nurses using SRT voice commands would demonstrate improved documentation efficiency.

Review of the Literature

Prior to implementing the speech recognition program for nurses, a literature search was performed using CINAHL, Google Scholar, Ovid, and PubMed databases using the Boolean search terms "speech recognition technology" and "nursing". The search was repeated at the conclusion of the implementation in 2021 to review more recent studies. The literature search yielded 230 results, of which 22 were deemed appropriate. Studies were excluded for the following reasons: duplications, the setting was not healthcare, published more than five years prior to the search, or the subjects using SRT were not clinicians.

Blackley et al., (2019), performed a systematic review of speech recognition for clinical documentation, which included 122 articles published from the years 1990 to 2018. Only eight (6.6%) of these articles focused on SRT use by nursing and were primarily qualitative, related to nursing preferences for using SRT. Across all studies included in this systematic review, Blackley et al. found that the most common outcome variable reported was documentation time as it related to cost and productivity analysis. Some studies reported decreases in documentation time while others found no effect or a slight increase in documentation time while clinicians used SRT.

Joseph et al., (2020), published a systematic review that examined the impact of SRT on accuracy and efficiency specific to nursing documentation. The authors identified ten articles for inclusion that were of quantitative design and involved nurses using SRT in clinical, community or simulated healthcare settings. The included studies reported primary outcome measures of voice-to-text accuracy and efficiency of SRT. Secondary outcome measures identified were documentation quality, user satisfaction, and cost-effectiveness. Accuracy of SRT varied across the studies; this variance could be attributed to the use of different speech recognition software and microphones.

Only two studies that involved nurses using SRT voice commands, or voice macros, were found in the literature (Alapetite, 2008; Northern Sydney Local Health District [NSLHD], 2012). Although outside of the date range of this literature search, these studies were included in Joseph et al.’s (2020) systematic review. In the experimental study by Alapetite (2008), anesthesia teams consisting of one doctor and one nurse participated in simulated scenarios and used a speech input interface, which allowed free-text narration and voice commands for documentation entry. The group using SRT spent an average of 18 seconds on time dedicated to fill the record, compared to an average of 3 min 45 sec (p < 0.003) for the group using the traditional interface (Alapetite, 2008). The Northern Sydney Local Health District (NSLHD, 2012) reported that emergency department nurses used SRT voice commands and macros for documentation entry and chart navigation. NSLHD administered qualitative surveys to evaluate patient satisfaction with their care team’s use of SRT. Patients reported feeling “very comfortable” with clinicians using SRT (77%) and 64% of patients felt that hearing the dictation increased their confidence that the clinician correctly transcribed their information (NSLHD, 2012).

Although the literature predominantly focused on physician use of SRT, there is some evidence that SRT provides benefits to nursing in terms of efficiency and satisfaction with EHR use. No studies were found in the literature that reported medical-surgical nurses using SRT to file discrete data into fixed-entry fields with voice-activated macroinstructions; the literature mainly focused on note dictation. Only two older studies (Alapetite, 2008; NSLHD, 2012) involved nurses utilizing SRT voice commands; however, these studies occurred in perioperative and emergency department settings.

Ethics

This study did not contain any protected health information and was approved by Virtua Health’s Investigational Review Board (IRB).

Design and Method

This study included retrospective analysis of baseline and post-implementation data and cross-sectional secondary analysis of data from the EHR to compare documentation efficiency among users and non-users of SRT over three time periods among nurses. The three time periods include baseline, five months after training/implementation of SRT, and 12 months after training/implementation of SRT (see Figure 1).

Figure Notes: Timeline reflecting the implementation and data collection periods from 2018 through 2020.

Data were analyzed from the Epic Nursing Efficiency Assessment Tool (NEAT) report at baseline to determine time spent in flowsheets. Based on the data, the shift assessment was selected as the target assessment. The time spent in the shift assessment was analyzed at baseline and compared at each collection interval using the NEAT report to assess the effectiveness of SRT on time spent in the shift assessment flowsheet.

In addition to the data from the EHR, the investigators also worked with the SRT vendor (Nuance), which provided utilization reports. Data were extracted to identify nurses who used at least one SRT command. Nurses who had been using voice commands were separated from those who were not to create the sample groups (see Figure 2).

Figure Notes. The graph illustrates that SRT users had a 9.7% documentation time savings in August 2019 and 9% in February 2020. SRT users' documentation time increased by 0.6 minutes compared to non-SRT users, whose documentation time increased by 2.4 minutes.

SRT was available to all nurses who elected to attend a voluntary one-hour, instructor-led training course. The class focused on fundamentals for dictation and building commands for nursing documentation. Training was conducted in a technology training classroom, using the EHR training environment. Each learner had access to a microphone at their learning station and participated in classroom exercises for dictation, navigating the chart using voice, and executing voice commands.

The class also reviewed the organization’s guidelines for SRT use, which are applicable to all clinicians using SRT. The guidelines dictate that clinicians should not build commands that automatically save data or sign orders; instead, clinicians must manually review SRT documentation before filing. The guidelines stress the importance of clinicians taking responsibility for entering accurate documentation by SRT.

Setting & Sample

The sample for this study included medical-surgical nurses from 20 units within three of Virtua Health’s acute care hospitals who worked a 12-hour or greater shift and who had attended SRT training. Shifts that did not contain medication administrations were excluded to ensure that only nurses taking a patient assignment were included in the study. Participants were labeled either typical documenters (non-SRT users) or SRT users. Nursing stations were equipped with microphones for documentation.

While 125 nurses attended SRT training, an SRT user was defined as someone who used at least one SRT command during the data collection period. At data collection period #1, 55 nurses were identified as SRT users; 49 of these nurses appeared in the 2018 baseline data. The six nurses who were not in the baseline data set had not worked a medical-surgical shift for that period. At data collection period #2, 10 nurses no longer appeared on the utilization report. This can be attributed to transfers to non-medical-surgical units or to other organizations.

Variables & Measurement

The main outcome measure of the study was documentation efficiency, which was measured by the average minutes spent in the shift assessment flowsheet at each time point for active SRT and non-SRT nurses.

The investigators collaborated with the EHR vendor (Epic) to customize the NEAT report and have the data sent monthly. Once Epic sets up the NEAT report, it will automatically collect data in the background of the EHR to record the time taken to perform specific nursing activities within the EHR. The report contains information for all inpatient nurses; however, this study specifically looked at the data from medical-surgical nurses in three acute care hospitals and only at time spent in the shift assessment flowsheet. Utilization data from the SRT vendor was also used to identify the nurses who used voice commands during the period of NEAT data collection. Clicks saved for each SRT command were physically calculated. Each click of a mouse or keyboard was counted as one click saved.

Findings

The baseline data from 2018 indicated that non-SRT users (n=748) spent an average of 16.4 minutes in the shift assessment. SRT users (n=49) spent an average of 16.5 minutes. Six months into the implementation, the non-SRT users (n=696) spent an average of 18.6 minutes in the shift assessment while the SRT users (n=55) spent an average of 16.8 minutes. The SRT users spent 9.7% less time. At one-year post-implementation, the data indicates that non-SRT users (n=668) spent 18.8 minutes documenting in the shift assessment while the SRT users (n=45) spent 17.1 minutes, a 9%-time savings (see Figure 2).  While these results indicated a time savings, there was no statistically significant difference in mean minutes per shift between the SRT and non-SRT users. However, these commands can save medical-surgical nurses up to 40 clicks per voice command (see Table 1).

Table Notes. The top medical-surgical commands range from 10 to 40 clicks saved with each use, for an an average of 23 clicks saved per use.

Discussion

During the course of the study, documentation efficiency increased slightly among SRT users. In addition to the technology providing a means of saving an average of 23 clicks per command (Table 1), SRT users also saved time while documenting. At baseline, the non-SRT and SRT users spent nearly the same amount of time documenting within the shift assessment.

After the baseline data were collected, the organization increased the number of mandatory fields in the shift assessment with an EHR upgrade in February 2019. The additional required documentation in the shift assessment during the data collection periods was a potential confounding variable. It was expected that the overall documentation time would increase for both groups. However, the SRT group experienced minimal impact to their documentation time while the non-SRT group experienced an increase of greater than two minutes.

Limitations

This study had a sample size limited to medical-surgical nurses engaged in using SRT to document in the shift assessment flowsheet using voice commands. Demographic variables such as age, race and gender were not collected. Due to the small number of SRT users, there was a variable sample size over the course of the study. Being early adopters, SRT nurses possibly had more computer experience or willingness to try innovative technologies. SRT adoption may have been limited by the requirement to attend an in-person training class prior to gaining access. In addition, this study is limited to the flowsheet with the highest documentation time, the shift assessment, and does not include SRT use for EHR navigation, EHR search, note writing, order placement, or other flowsheets.

Documentation quality and accuracy were not reviewed to determine if discrete documentation included all required data and clinical exceptions pertinent to specific patients. SRT vendor data analytic tools were limited in their ability to quantify the frequency at which system and user-created commands were executed. Finally, this study does not include qualitative research on the nurse's perceived efficiency and satisfaction with using SRT.

Recommendations for Future Research

This study identified several ways in which SRT improves efficiency and workflows within the EHR. Although a full literature review was done prior to performing this study, the available research during the time of the study was limited to voice-to-text dictation. Few studies existed on the use of voice-activated macros through SRT for nursing. The investigators identified several areas for future research. The first area of interest for future study is in using real-time data collection within the EHR rather than collecting the data retrospectively. Next, the current study focused solely on medical-surgical nurses, but other nursing specialties and allied health roles within the organization are also utilizing the technology.

Another area for consideration involves looking at the efficiency of using SRT as it relates to the accuracy of documentation. The financial implications are difficult to quantify; however, the investigators are working with the organization’s EHR vendor to determine how SRT use may impact end-of-shift overtime. This specific study focused only on quantitative data surrounding the use of SRT within Virtua Health’s EHR, so a qualitative study of end user satisfaction, feedback, and patient perceptions is an area for future research. Finally, a future study should be conducted using an experimental design with a larger sample size.

Nursing Implications

The most significant implication of these findings for medical-surgical nurses is the time saving aspect of SRT use. Documentation burden remains a key driver in burnout for clinicians, including nurses, and this poses a threat to the American healthcare system (Goss et al., 2019). This documentation burden has been identified as a common concern for nurses. Using SRT allows nurses to redistribute their time saved towards the patient, whether it be in direct patient care or other areas like patient education. Time saved and clicks saved are objective data points that attest to the benefits of SRT for nurses.

Satisfaction with SRT use is correlated with efficiency, decreased prevalence of errors, and reduced editing times (Goss et al., 2019). Overall, increasing nurse satisfaction can yield increased employee engagement, a healthier work environment, increased productivity, and higher staff retention (Joseph et al., 2020). Improvement in documentation speed can promote documentation in real time, benefitting both patients and the interdisciplinary team. Patients have reported increased confidence in clinicians who documented with SRT at the bedside (NSLHD, 2012). However, nursing may require a cultural shift to be comfortable documenting verbally in front of patients. SRT is evolving to include the ability to listen to natural speech for documentation within EHRs (Locke et al., 2021), so the possibilities for future use could transform nursing documentation at the bedside.

 Conclusion

Healthcare organizations are continually challenged to decrease nursing documentation burden and allow more time for patient care. Nursing informaticists can assist with nurse documentation efficiency by developing innovative solutions using existing technology. The results demonstrate that utilizing SRT to create and apply voice activated macros for medical-surgical nurses practicing in acute care hospitals saves 9% to 9.7% of documentation time in the shift assessment. The top medical-surgical commands range from 10 to 40 clicks saved with each use. This represents an average of 23 clicks saved per use, per user. Therefore, it is important that clinical informatics – in partnership with clinical system users, operational leaders, EHR and technology vendors – continue to advocate for the use of macros within the EHR, as well as for SRT technology beyond provider workflows.