Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2022 May 1.
Published in final edited form as: J Nurs Scholarsh. 2021 Mar 30;53(3):333–342. doi: 10.1111/jnu.12652

Nurse’s Achilles Heel: Using Big Data to Determine Workload Factors That Impact Near Misses

Amy A Campbell 1, Todd Harlan 2, Matt Campbell 3, Madhuri S Mulekar 4, Bin Wang 5
PMCID: PMC8221452  NIHMSID: NIHMS1689335  PMID: 33786985

Abstract

Purpose:

To explore how big data can be used to identify the contribution or influence of six specific workload variables: patient count, medication count, task count call lights, patient sepsis score, and hours worked on the occurrence of a near miss (NM) by individual nurses.

Design:

A correlational and cross-section research design was used to collect over 82,000 useable data points of historical workload data from the three unique systems on a medical-surgical unit in a midsized hospital in the southeast United States over a 60-day period. Data were collected prior to the start of the Covid-19 pandemic in the United States.

Methods:

Combined data were analyzed using JMP Pro version 12. Mean responses from two groups were compared using a t-test and those from more than two groups using analysis of variance. Logistic regression was used to determine the significance of impact each workload variable had on individual nurses’ ability to administer medications successfully as measured by occurrence of NMs.

Findings:

The mean outcome of each of the six workload factors measured differed significantly (p < .0001) among nurses. The mean outcome for all workload factors except the hours worked was found to be significantly higher (p < .0001) for those who committed an NM compared to those who did not. At least one workload variable was observed to be significantly associated (p < .05) with the occurrence or nonoccurrence of NMs in 82.6% of the nurses in the study.

Conclusions:

For the majority of the nurses in our study, the occurrence of an NM was significantly impacted by at least one workload variable. Because the specific variables that impact performance are different for each individual nurse, decreasing only one variable, such as patient load, will not adequately address the risk for NMs. Other variables not studied here, such as education and experience, might be associated with the occurrence of NMs.

Clinical Relevance:

In the majority of nurses, different workload variables increase their risk for an NM, suggesting that interventions addressing medication errors should be implemented based on the individual’s risk profile.

Keywords: Big data, medication errors, near miss, safety, workload factors

The science of big data allows researchers to compare and analyze large volumes of data from multiple sources and in multiple formats to develop a better understanding of relationships and patterns among data that are useful for the prediction and estimation of future trends and needs. Data from video cameras, infrared sensors, and GPS could be used to track traffic patterns and design emission models (Zhu, Yu, Wang, Ning, & Tang, 2019), and IBM was able to forecast web traffic patterns in near real time (Baughman et al., 2016). Researchers were able to predict patient acuity scores for the next day based heavily on textual nursing notes (Konito et al., 2014). Each of these groups used big data science to better understand their environment and to make predictions on how best to address future needs.

As technology has increasingly been adopted to assist in patient care, large datasets containing a multitude of details of the patient and delivery processes are being automatically recorded by a wide array of disparate systems in the clinical environment. These datasets are useful for providing immediate patient care, but their true value is not fully realized due to their stand-alone designs. Such data could be used to identify specific situations or factors that can assist healthcare professionals to improve critical decision-making processes, resulting in providing better direct care or improving the utilization of limited resources (McGonigle & Mastrian, 2018). Brennan and Bakken (2015) have argued that the field of nursing is unique in its comprehensive approach to the patient experience, enhancing nursing researchers’ ability to create data models; however, our literature review found that current nursing research does not take into consideration the patient care factors that influence an individual nurse’s ability to provide quality care to the patient. Medication errors (MEs) result in a minimum of 7,000 deaths yearly (Institute of Medicine, 2001) and have remained a significant problem, occurring in as many as 37% of all medication administrations (Westbrook et al., 2015). This results in unacceptable levels of risk to patients. Our goal in this study was to devise a way to use big data techniques to identify factors that may lead to an increased risk for errors in the clinical setting.

When looking to use big data to address the problem of MEs, a broader picture of why the error is occurring should be examined. Previous research suggests that MEs are most often the result of individual personal issues, environmental factors, organization and management factors, team dynamics, tasks, and medications (Blackman et al., 2015; Crane et al., 2015; Li, Magrabi, & Coiera, 2012; Lipshutz, Caldwell, Robinowitz, & Gropper, 2015; Mansouri et al., 2014; Mayo & Duncan, 2004; Shahrokhi, Ebrahimpour, & Ghodousi, 2013; Speroni, Fisher, Dennis, & Daniel, 2013). When workloads (which consist of physical, environmental, and mental factors) exceed the ability of the individual to maintain focus, fatigue and exhaustion set in (Horrey et al., 2011). Excessive workloads on healthcare workers result in stressed and fatigued nursing staff, which in turn impacts the quality of patient care as reflected by a high frequency of MEs and missed care (Ball, Murrells, Rafferty, Morrow, & Griffiths, 2014; Cho, Chin, Kim, & Hong, 2016; Farquharson et al., 2013; Van Bogaert et al., 2017). Saremi and Fallah (2013) found there was a significant positive relationship between mental weariness and the severity of medical errors by providers, suggesting that healthcare professionals are likely to commit more errors with larger workloads. Despite their prevalence, MEs are notoriously underreported in practice due to inconsistent definitions (Farquharson et al., 2013), cultural expectations of perfection in health care (Wu & Marks, 2013), and cumbersome reporting processes (Vrbnjak, Denieffe, O’Gorman, & Pajnkihar, 2016).

In order to mitigate the problem of underreported MEs, many studies have also looked at near misses (NMs). While not ultimately resulting in harm to the patient, an NM represents “any event that could have had adverse consequences but did not and was indistinguishable from fully fledged adverse events in all but outcome” (Agency for Healthcare Research and Quality [AHRQ], 2019). NMs are similar in nature to MEs, allowing researchers to use NMs as a proxy for MEs in situations where MEs cannot be objectively measured (Crane et al., 2015; Ruddy et al., 2015; Vrbnjak et al., 2016). NMs also occur as the result of the same systemic, organizational, and personal problems that cause errors, but occur more frequently, with estimates of NMs occurring 300 times more frequently than full-fledged errors (Lipshutz et al., 2015; Speroni et al., 2013). This article presents a big data–based approach to study the effect of workload variables on the occurrence of NMs by nurses during medication administration.

To enable researchers to identify widely applicable causative factors for NMs, a consistent, objective form of digitally recording NMs during medication administration must be identified. Bar-code medication administration (BCMA) systems report the frequency of medication, patient, dose, and drug formulation scanning errors, medications not administered due to omission, and attempts to administer medications outside the appropriate time boundaries. These events are flagged by the BCMA as NMs. They are recorded consistently in all BCMA systems and are retrievable with a time stamp, which allows for gathering data pertaining to workload factors surrounding the NM. A survey in 2009 found that BCMAs were being used in over 40.2% of hospitals with 300 to 399 beds, making them a commonly used method for identifying NMs nationally (Helmons, Wargel, & Daniels, 2009). One of the unique benefits of big data science is that it allows for the unique examination of multiple independent variables from a variety of sources. To determine what variables to include in this study, a literature review on factors that increased the risk for medication administration errors (MAEs) and NMs was completed. Six workload factors found to be common across all in-patient clinical hospital settings were identified from literature as potentially increasing the risk for an MAE by nursing staff members: (a) number of patients assigned (Cho, Chin, Kim, & Hong, 2016; Driscoll et al., 2018; Keers, Williams, Cooke, & Ashcroft, 2013), (b) number of medications ordered (Keers et al., 2013), (c) number of tasks assigned (AHRQ, 2019; Mansouri et al., 2014), (d) frequency of interruptions (Parry, Barriball, & While, 2015), (e) acuity of patient load (Lipshutz et al., 2015), and (f) overtime hours worked (al Tehewy, Fahim, Gad, El Gafary, & Rahman, 2016; Parry, Barriball, & While, 2015).

Number of patients assigned refers to the nurse-to-patient ratio. This issue is controversial because some research has found a lower incidence of MEs in units with lower nurse-to-patient ratios (Avalere Health, 2015), but other research suggests that key quality measures such as safety are not significantly affected by increasing staffing levels alone (Hertel, 2012). Higher nurse-to-patient ratios can have additional ramifications—as the nurse’ s patient load increases, so does the number of medications and tasks associated with the patients’ care, increasing the opportunity for errors to occur (Keers et al., 2013).

Number of medications ordered is the overall volume of medications a nurse must administer within a shift. Medication administration can be a complex event, with many variables taking place at one time, adding pressure to nurses as they prepare medications for each patient. Intravenous (IV) medications include unique factors associated with MEs related to the various types of IV pumps necessary to administer these types of medications (Kuitunen, Niittynen, Airaksinen, & Homstrom, 2020). Since these secondary factors were not examined in this study, focus was kept on medications given by mouth only (i.e., per orem [PO]).

Number of tasks, much like patient count, can play an important part in increasing the risk for MEs. Tasks may range in complexity from wound care to arranging a physician referral. Similarly, time limits or ranges for completion of these tasks can be established (e.g., 3 p.m.) or vague (every shift, daily), requiring the nurse to determine the optimal time to address these needs. During a nurse’s shift, tasks fluctuate and accumulate for each patient for whom they are responsible, and as these tasks build there is a greater risk for perpetuating mistakes, which include MEs (AHRQ, 2019; Mansouri et al., 2014).

Interruptions are also believed to play a significant role in MEs. Interruptions can take the form of patients activating their call light, family members approaching staff in the hallway, and communication, both patient-related and social, from peers. These additional stimuli within a healthcare setting come in many forms, such as auditory, visual, and physical, and can have adverse effects on healthcare staff’s ability to sort through information and make a decision (Sitterding, Ebright, Broome, Patterson, & Wuchner, 2014). Call lights specifically employ the use of both lights and an auditory alarm to notify nurses of a patient need and can distract them during the medication administration process, interrupting their focus or temporarily delaying care while the call light is addressed (Hall et al., 2010).

Acuity of patients has also been found to increase the risk for MAEs (Lipshutz et al., 2015); however, previous research was unable to determine if such errors occurred as a direct result of the patient’s acuity level (Millichamp & Johnston, 2020) or as an indirect result of ill patients receiving numerous medications, increasing the opportunity for error (Blignaut, Coetzee, Klopper, & Ellis, 2017). This variable is additionally complicated by the lack of an established measurement for patient acuity.

Lastly, overtime worked is the number of hours worked over 40 in the past 7 days. This factor, along with an increase in workload, has been shown to increase MAEs (Parry et al., 2015; Liu, Lee, Chia, Chi, & Yin, 2012; Muabbar & Alsharqi, 2021).

Methodology

A private, medium-sized hospital in the southeastern United States was identified to participate in this study. The chief nursing officer agreed to serve as the project’s senior sponsor, and a needs assessment on the facility was completed. Key stakeholders included the floor manager, nurses, and the hospital’s information technology department, all of whom actively participated in the project. A clear scope of the study, strategic objectives, and potential barriers were defined by researchers and discussed with stakeholders before the project was submitted to both the hospital administration and the researchers’ Institutional Review Board for approval. A medical-surgical unit was chosen for the study because these units are common to all hospitals and contain a variety of patient diagnoses, acuity, ages, turnover rates, etc.

To determine the impact of the six identified workload factors on MAEs, nursing staff were observed across different shifts and times throughout the day to create a nurse workflow model. All nurses went to a partially secluded, designated room to retrieve medications from an automated medication-dispensing system at the time of dispensation, and medications were usually delivered to the patient’s bedside immediately. However, nurses were often interrupted in the medication administration process by peers, patients, physician phone calls, and call lights as they were delivering the medication to the patient’s room.

Data were pulled from three distinct systems: electronic health records, call lights, and BCMA systems. In addition to gender and full or part-time employment status of nurses, data were collected for each of the identified six workload factors: current patient load (as determined by examining the number of patients assigned to the nurse); current medication load (as ordered for each patient); task load (as assigned to each patient); number of interruptions (as measured by the total number of pages activated by the nurse’s patient’s call light); patient acuity (as denoted by a predetermined sepsis algorithm based on the most current vital signs, patient’s level of alertness, presence of oxygen support, laboratory changes, and other additional patient-related factors); and working more than 40 hr a week (as determined by examining the previous 40 hr for sum total of hours worked each 2-hr period). Cleaned data for the six workload factors of interest were organized based on the database time stamp, compiled at the patient level, and then totaled in 2-hr time periods for each nurse based on their patient assignments. The researchers felt sufficient granularity would be achieved by organizing the data in 2-hr time periods because medications typically were not ordered more frequently than every 2 hr. Formatting for these data points were structured. Nurse– patient assignments were cross-referenced with staffing schedules (unstructured) to eliminate any errors in the workload list. The BCMA log was examined for each nurse for each 2-hr time period, and the presence or absence of NM and type of NM (wrong patient or wrong medication, dose, time, route) during that time frame were recorded (Table 1). Only data on PO NMs were used for our analysis.

Table 1.

Example of Usable Data Sorted by Nurse

Nurse Time block (date, hours) Total errors Blocks worked Call lights Medication load Patient load Task load Sepsis score Near miss occurrence
1 1/01, 06:01–08:00 0 4 19 4 6 0 1.33 0
1 1/03, 08:01–10:00 3 5 23 3 6 3 1.33 1
1 1/04, 20:01–22:00 1 6 15 5 6 5 2.7 3
1 1/04, 22:01–00:00 0 7 10 4 6 2 1.83 0
1 1/05, 00:01–02:00 0 8 8 3 6 1 1.83 0
1 1/05, 02:01–04:00 4 9 20 4 6 0 1.35 1
Open in a new tab

Historical data were cleaned by removing redundancy in patient– nurse assignments due to user error (usually caused by secretarial staff) when patient– nurse assignments were created. Per guidance from the nurse manager, any patient assigned to a nurse for less than 15 min was assumed to be an assignment error and was removed from the nurse’s workload list. If a patient was assigned to a nurse for more than 15 min within a 2-hr time frame in which data were collected, then that nurse was assumed to be at work and responsible for tasks, medications, and call lights associated with that patient for that 2-hr time period. Data gathered were consistent and complete for all variables except for sepsis. Due to complexities of the sepsis algorithm an occasional null value would be provided for a 2-hr time period instead of a sepsis score for a patient. To prevent these null values from significantly impacting the total value of the sepsis score, they were not counted towards the total sepsis value; instead, an average of the values present were taken.

Collected data were analyzed using JMP Pro 12 statistical software (SAS Inc., Cary, NC, USA). All numerical data such as patient count per 2-hr time period were summarized using mean, standard deviation, minimum, maximum, median, and interquartile range. All categorical data, such as gender, were summarized using proportion ratios. The mean response for numerical variables by those who made an error and those who did not was compared using the t test and Wilcoxon test. Mean responses between more than two groups such as among nurses were studied using analysis of variance (ANOVA). Correlation between numerical variables was studied using Pearson’s correlation coefficient, and association between categorical variables was studied using the chi-square test of independence or Fisher’s test. Nominal logistic regression with occurrence of at least one NM as a dependent variable was used to identify possible factors useful in their prediction. The area under the receiver operating curve (AUC) and correct classification rates were used to assess the value of the model. For large effects, the associated p values are often very small, reported as p < .0001, and comparison of their values can be challenging. Hence, LogWorth values were computed as –log10(p value) to provide better assessment of the level of contribution of each predictor variable. Note that highly significant p values have large LogWorths and nonsignificant p values have low LogWorths. A 5% level of significance was used to determine statistical significance of results.

Results

The historical BCMA and NM data as well as those for the six workload factors of interest were gathered from the hospital databases for a 25-bed medical-surgical unit for a 60-day period. During this period, data on the workload of 23 nurses caring for 389 unique patients were gathered. A total of 26,150 medications were ordered, and 11,595 tasks were assigned. After the data were cleaned to remove non-PO medications, the number of medications ordered dropped to 18,104. All tasks that could be completed by the nurse assistants (i.e., 2-hr patient turns) were removed, resulting in 6,065 nurse-specific tasks. Sepsis scores were calculated for each patient every 2 hr, resulting in 15,576 usable sepsis scores. Patient load varied from one to seven patients assigned in every 2-hr time period as a result of discharges, admissions, and other roles (e.g., manager, charge nurse, infusion specialist) assigned to the nurse. 33,484 unique call lights were directed immediately to the assigned nurse through the hospital’s pager system.

Of 23 nurses in the study, 20 (86.96%) were female and accounted for 2,850 (87.77%) of the 2-hr blocks worked and 619 (86.21%) of the 2-hr blocks with NMs. Three male nurses accounted for 397 (12.23%) of the 2-hr blocks worked and 99 (13.79%) of the 2-hr blocks with NMs. Presence of such gender disparity is very common in the nursing profession; however, no significant association was observed between the gender of the nurse and NMs (p = .1556). NMs were reported in 99 two-hour blocks worked by male nurses (24.94%) and 619 two-hour blocks worked by female nurses (21.72%). All 3 male nurses and 15 of the 20 female nurses were employed full-time. Five female nurses were employed part-time. A Fisher test found no significant association between occurrence of NMs and full or part-time employment (p = .5675).

The BCMA database flagged 1,174 NMs associated with PO medications (6.48%), of which 1,065 (90.7%) were of scanning wrong dose-time-medication and 109 (9.28%) were flagged as wrong patient. Over a 60-day time period, data for a total of 3,247 two-hour time periods on 23 nurses were collected, of which 2,529 two-hour time periods (77.82%) were devoid of any NM and 720 two-hour time periods (22.17%) contained one or more NM. The number of 2-hr time periods worked in the past 7 days ranged from 1 to 42. Of those 2-hr time periods that contained at least one NM, 434 (60.4%) contained only one NM, 193 (26.9%) contained two NMs, 43 (6.0%) contained three NMs, and 48 (6.7%) contained four or more NMs.

Summary statistics for the six workload factors (Table 2) show a considerable amount of variation among different 2-hour time periods. Medications to be administered numbered as few as 0 and as many as 41 in a 2-hr time period. To determine if there were any correlations between each variable (previous research suggested an increase in patients may result in an increase in medications to be administered), a correlation analysis and scatterplots were completed. Neither the correlation analysis nor the scatterplots showed a significant correlation among the six independent workload variables. The mean of workload factors per 2-hr time period for all 23 nurses were compared using ANOVA (Table 3), and all showed significant differences (p < .0001) for each of the six factors among nurses indicating considerable variation in factors recorded from nurse to nurse. When the mean responses for the six workload factors over 2-hr time periods that had NMs were compared to those that did not using a t-test (Table 4), five of the six workload factors showed significantly higher average workload during the 2-hr time periods in which NMs occurred than when none occurred (p < .0001 for each). Only the mean number of 2-hr time periods worked was not found to be significantly different between those with NMs and those without.

Table 2.

Summary Statistics for Six Workload Factors Over All 2-Hour Time Periods

n Mean SD Min Max Median Interquartile range
Patient count 3,247 4.80 1.128 1 7 5 2
Medication count 3,247 5.58 6.61 0 41 3 7
Task count 3,247 1.87 2.13 0 13 1 3
Call light count 3,247 10.31 8.80 0 53 10 16
Average sepsis score 3,247 1.91 1.18 0 8 2 1.55
2-hr time periods worked 3,247 22.67 7.26 1 42 23 10
Open in a new tab

Table 3.

Summary Statistics for the Average Response of Each Nurse for Six Workload Factors

n Mean SD Min Max p value (ANOVA)
Patient count 23 4.67 0.77 2.00 6.00 <.0001
Medication count 23 5.34 1.77 2.14 8.05 <.0001
Task count 23 1.78 0.57 0.57 2.86 <.0001
Call light count 23 9.63 5.00 0.00 15.35 <.0001
Average sepsis score 23 2.06 0.55 1.63 4.43 <.0001
2-hr time periods worked 23 18.83 7.75 3.00 26.42 <.0001
Open in a new tab

Note. ANOVA = analysis of variance.

Table 4.

Comparison of Mean ± Standard Deviation of Six Workload Factors Between Blocks With and Without Near Misses (NMs)

NM in 2-hr time period p value for t-test
Workload factor No (n = 2,527) Yes (n = 720) Mean for yes > mean for no
Patient ount 4.74 ± 1.56 4.99 ± 1.00 <.0001
Medication ount 4.48 ± 5.68 9.42 ± 8.08 <.0001
Task ount 1.67 ± 2.02 2.57 ± 2.41 <.0001
Call light count 9.76 ± 8.65 12.23 ± 9.04 <.0001
Average sepsis score 253 ± 1.20 2.05 ± 1.07 <.0001
2-hr time periods worked 22.71 ± 7.22 22.55 ± 7.42 .7001
Open in a new tab

A nominal logistic model (Table 5) identified medication count (p < .0001, LogWorth = 36.31) as the most influential variable for prediction of occurrence of NMs, followed by call light count (p < .0001, LogWorth = 7.33), nurse (p < .0001, LogWorth = 4.88), and task count (p = .0371, LogWorth = 1.43). In the presence of these predictors, the average sepsis score, 2-hr time periods worked last week, and patient count did not provide significant amount of additional information towards the prediction of occurrence of NMs. The AUC for this model was 0.73 with a correct classification rate of 78.72%. For some nurses, odds of having at least one NM in a given 2-hr time period were as much as 2.6 times higher compared to the other nurses. Figure 1 shows the increasing occurrence of at least one NM per 2-hr time period with increasing call light count and medication count.

Table 5.

Outcome of Logistic Regression to Predict Occurrence of at Least One Near Miss in a Given 2-Hour Time Period Using Nurse and Workload Factors as Predictors

Predictors Chi-square statistic p value (likelihood ratio test) Log worth
Medication count 161.64 <.0001 36.305
Call light count 29.86 <.0001 7.334
Nurse 61.55 <.0001 4.881
Task count 4.34 .0372 1.429
Average sepsis score 1.54 .2153 0.667
2-hr time periods worked 0.52 .4707 0.327
Patient count 0.20 .6570 0.182
Open in a new tab

Figure 1.

Figure 1.

Open in a new tab

Trends in percentages for 2-hr blocks with at least one near miss as a function of the number of call light counts and the number of medication counts in a given 2-hr time period.

Noting a large variation in predictor factors among nurses, an individual logistic model was developed for each nurse to predict the occurrence of NMs in a given 2-hr time period using six workload variables as possible predictors. Outcomes reported in Table 6 show that the significance of each workload factor’s impact on predicting the occurrence of at least one NM varied from nurse to nurse, with as few as zero and as many as three workload factors having a significant impact on the occurrence of NMs. Medication count impacted the occurrence of NMs significantly in 12 (52.17%) of the 23 the nurses in the study, followed by both patient and task count in 4 (21.73%), call light count in 3 (13.04%), average sepsis score in 2 (8.69%), and 2-hr time periods worked in 1 (4.35%). Five of the 23 nurses did not have enough data on at least one of the workload factors to determine significance of its effect. The AUC and correct classification rates are indicators of usefulness of such models for predicting the occurrence of at least one NM in a given 2-hr time period using workload factors.

Table 6.

Outcome of Individual Logistic Regression Fits to Predict Occurrence of Near Misses (NMs) in a Given 2-Hour Time Period Using Workload Variables as Possible Predictors

No. of 2-hr time periods p value
Nurse Worked last week With at least 1 NM Whole mode Patient count Medication count Task count Call light count Average sepsis score 2-hr time periods worked AUC Correct Classification Rate
1 147 40 .0003 .8769 .0001 .9415 — .8090 .6629 .7902 .7483
2 202 52 <.0001 .3804 <.0001 .7499 .3342 .0639 .4627 .7509 .7475
3 119 24 .0002 .6376 .0582 .3922 .0109 .0003 .4124 .7952 .8319
4 7 1 2193 — .9997 .9993 — .9994 .9991 1.0000 1.0000
5 198 35 .0619 .2199 .4151 .5327 .1989 .1610 .5024 .6778 .8232
6 188 40 .0001 .3319 .0657 .0145 .1836 .7424 .9093 .7666 .8085
7 167 28 .1413 .0494 .2256 .2282 .0893 .4242 .9235 .6650 .8383
8 189 36 .0001 .0498 <.0001 .7589 — .7766 .8797 .7629 .8042
9 215 40 .0036 .8779 .2852 .3841 .0005 .3231 .6781 .7287 .8047
10 184 34 .0004 .0458 .0118 .0187 .9792 .1281 .8117 .7698 .8152
11 168 55 <.0001 .2061 <.0001 .2798 .3559 .5127 .4751 .7525 .7381
12 53 11 .0594 .4459 .0023 .0626 .2573 .1172 .3482 .8160 .8302
13 180 44 <.0001 .7495 .0016 .0255 — .9378 .3605 .7669 .8000
14 197 57 .0017 .1703 .0573 .9179 .0043 .2646 .8781 .7188 .6954
15 14 4 .0102 .9961 <.0001 .9960 .8260 .9967 .3232 1.0000 1.0000
16 20 3 .0096 <.0001 .9951 <.0001 1.0000 .9904 <.0001 1.0000 1.0000
17 217 62 <.0001 .3392 .0001 .0861 .7248 .7916 .6748 .7722 .7373
18 158 30 <.0001 .8318 .0001 .5483 .0958 .6720 .0575 .7878 .8228
19 200 43 <.0001 .1825 <.0001 .9607 .1196 .5673 .8457 .7910 .8150
20 36 10 .6277 .8952 .8697 .5024 .2958 .4316 .2965 .7308 .6667
21 168 30 .0372 .7609 .0799 .3777 .3772 .0328 .7560 .7002 .8274
22 5 2 .1509 — .9998 .9995 .9989 — 1.0000 1.0000 1.0000
23 215 37 .0181 .6413 .0068 .0648 .2928 .7288 .3644 .7038 .8279
Open in a new tab

Discussion

Evidence suggests that addressing workload issues may reduce the risk for NMs, but previous research that examined mitigation strategies implemented at the unit level resulted in inconsistent results (Lapkin, Levett-Jones, Chenoweth, & Johnson, 2016). To better understand the role of workload factors on patient care, it is necessary to conduct an examination of factors at the individual nurse level. Our research suggests that an individual nurse’s risk for an NM is impacted by a unique combination of variables. These results suggest that systemic and organizational changes that address only one or two of these workload factors will have only a moderate impact on NMs for only a subset of the nursing staff. To maximize effectiveness, clinical interventions to reduce NMs must be customized to address each nurse’s specific Achilles’ heel or specific risk profile. Once identified, this Achilles’ heel could additionally be mitigated through strategic scheduling, patient assignment, and load-balancing interventions.

A limitation of our study is that we were not able to measure additional work associated with admissions and discharges and consults, as well as additional interruptions that were not captured digitally, such as phone calls and peer, physician, or family interruptions. Considerable variation among nurses (see Table 3) indicating individual nurses’ experience, education, work habits, etc. may play a considerable role in occurrence or nonoccurrence of NMs, but this was not addressed in this study. Additionally, examining personal factors that are likely to impact the nurse’s awareness, such as emotional condition (Patel et al., 2011) and home-life environment at the beginning of each shift, could also be important in forming a complete and more precise picture of the nurse’s state of mind at the beginning of each shift.

As discussed earlier, NMs are a serious and common problem that impact patient care. Our results suggest that addressing a single factor in isolation (the one-size-fits-all approach) as is generally done by researchers and hospital management may not be sufficient to address this problem. The findings of this study suggest that data that are already being gathered and stored by different hospital systems concerning nurse workload can be combined and used to determine the risk for NMs. It is time that researchers and managers harness this vast amount of data and use it to improve the quality of care nurses are providing. This can be done successfully if the data that are currently fragmented and locked away inside different clinical systems can be brought together and used to improve patient care.

Clinical Resources.

Acknowledgements

A patent for this process and subsequent technology has been applied for by Drs. Amy Campbell, Matt Campbell, and Todd Harlan. Statistical analysis by Dr. Mulekar and Dr. Wang reported in this publication was supported in part by the National Center for Advancing Translational Sciences of the National Institutes of Health under award number UL1TR001417.

References

  1. Agency for Healthcare Research and Quality, Patient Safety Network. (2019). Adverse events, near misses, and errors. Retrieved from https://psnet.ahrq.gov/primers/primer/34/Adverse-Events-Near-Misses-and-Errors
  2. al Tehewy M, Fahim H, Gad NI, El Gafary M, & Rahman SA (2016). Medication administration errors in a university hospital. Journal of Patient Safety, 12(1), 34. 10.1097/PTS.0000000000000196 [DOI] [PubMed] [Google Scholar]
  3. Avalere Health. (2015). Optimal nurse staffing to improve quality of care and patient outcomes: Executive summary (white paper). American Nurses Association. Retrieved from https://www.nursingworld.org/~4ae116/globalassets/practiceandpolicy/advocacy/ana_optimal-nurse-staffing_white-paper-es_2015sep.pdf [Google Scholar]
  4. Ball JE, Murrells T, Rafferty AM, Morrow E, & Griffiths P. (2014). ‘Care left undone’ during nursing shifts: Associations with workload and perceived quality of care. BMJ Quality & Safety, 23(2), 116–125. 10.1136/bmjqs-2012-001767 [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Baughman A, Bogdany R, Harrison B, O’Connell B, Pearthree H, Frankel B, … Upton C. (2016). IBM predicts cloud computing demand for sports tournaments. Informs Journal on Applied Analytics, 46(1). 10.1287/inte.2015.0820 [DOI] [Google Scholar]
  6. Blackman I, Henderson J, Willis E, Hamilton P, Toffoli L, Verrall C, … Harvey C. (2015). Factors influencing why nursing care is missed. Journal of Clinical Nursing, 24(1–2), 47–56. 10.1111/jocn.12688 [DOI] [PubMed] [Google Scholar]
  7. Blignaut A, Coetzee S, Klopper H, & Ellis S. (2017). Medication administration errors and related deviations from safe practice: An observational study. Journal of Clinical Nursing, 26(21–22), 3610–3623. https://doi-org.libproxy.usouthal.edu/10.1111/jocn.13732 [DOI] [PubMed] [Google Scholar]
  8. Brennan P, & Bakken S. (2015). Nursing needs big data and big data needs nursing. Journal of Nursing Scholarship, 47(5), 477–484. 10.1111/jnu.12159 [DOI] [PubMed] [Google Scholar]
  9. Cho E, Chin DL, Kim S, & Hong O. (2016). The relationships of nurse staffing level and work environment with patient adverse events. Journal of Nursing Scholarship, 48(1), 74–82. 10.1111/jnu.12183 [DOI] [PubMed] [Google Scholar]
  10. Crane S, Sloane PD, Elder N, Cohen L, Laughtenschlaeger N, Walsh K, & Zimmerman S. (2015). Reporting and using near-miss events to improve patient safety in diverse primary care practices: A collaborative approach to learning from our mistakes. Journal of the American Board of Family Medicine, 28(4), 452–460. 10.3122/jabfm.2015.04.140050 [DOI] [PubMed] [Google Scholar]
  11. Driscoll A, Grant M, Carroll D, Dalton S, Deaton C, Jones I, … Astin F. (2018). The effect of nurse-to-patient ratios on nurse-sensitive patient outcomes in acute specialist units: A systematic review and meta-analysis. European Journal of Cardiovascular Nursing, 17(1), 6–22. https://doi-org.libproxy.usouthal.edu/10.1177/1474515117721561 [DOI] [PubMed] [Google Scholar]
  12. Farquharson B, Bell C, Johnston D, Jones M, Schofield P, Allan J, … Johnston M. (2013). Nursing stress and patient care: real-time investigation of the effect of nursing tasks and demands on psychological stress, physiological stress, and job performance: Study protocol. Journal of Advanced Nursing, 69(10), 2327–2335. 10.1111/jan.12090 [DOI] [PubMed] [Google Scholar]
  13. Hall M, Ferguson-Pare M, Peter E, White D, Besner J, Chisolm A, … Hemingway A. (2010). Going blank: Factors contributing to interruptions to nurses’ work and related outcomes. Journal of Nursing Management, 18, 1040–1047. 10.1111/j.1365-2834.2010.01166.x [DOI] [PubMed] [Google Scholar]
  14. Helmons PJ, Wargel LN, & Daniels CE (2009). Effect of bar-code-assisted medication administration on medication administration errors and accuracy in multiple patient care areas. American Journal of Health-System Pharmacy, 66(13), 1202–1210. 10.2146/ajhp080357 [DOI] [PubMed] [Google Scholar]
  15. Hertel R. (2012). Regulating patient staffing: A complex issue. Matters, 21(1), 3–7. Retrieved from https://www.amsn.org/sites/default/files/documents/practice-resources/healthy-work-environment/resources/MSM-Hertel-Jan12.pdf [Google Scholar]
  16. Horrey WJ, Noy YI, Folkard S, Popkin SM, Howarth HD, & Courtney TK (2011). Research needs and opportunities for reducing the adverse safety consequences of fatigue. Accident Analysis & Prevention, 43(2), 591–594. 10.1016/j.aap.2010.01.014 [DOI] [PubMed] [Google Scholar]
  17. Institute of Medicine, Committee on Quality of Health Care in America. (2001). Crossing the quality chasm: A new health system for the 21st century. Washington, DC: National Academies Press. Retrieved from http://www.ncbi.nlm.nih.gov/books/NBK222274/ [Google Scholar]
  18. Keers RN, Williams SD, Cooke J, & Ashcroft DM (2013). Prevalence and nature of medication administration errors in health care settings: A systematic review of direct observational evidence. Annals of Pharmacotherapy, 47(2), 237–256. 10.1345/aph.1R147 [DOI] [PubMed] [Google Scholar]
  19. Konito E, Airola A, Pahikkala T, Lundgren-Laine H, Junttila K, Korvenranta H, … Salantera S. (2014). Predicting patient acuity from electronic patient records. Journal of Biomedical Informatics, 51, 35–40. 10.1016/j.jbi.2014.04.001 [DOI] [PubMed] [Google Scholar]
  20. Kuitunen S, Niittynen I, Airaksinen M, & Homstrom A-R (2020). Systemic causes of in-hospital intravenous medication errors: A systematic review. Journal of Patient Safety. Advanced online publication. 10.1097/PTS.0000000000000632 [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Lapkin S, Levett-Jones T, Chenoweth L, & Johnson M. (2016). The effectiveness of interventions designed to reduce medication administration errors: A synthesis of findings from systematic reviews. Journal of Nursing Management, 24(7), 845–858. https://doi-org.libproxy.usouthal.edu/10.1111/jonm.12390 [DOI] [PubMed] [Google Scholar]
  22. Li SYW, Magrabi F, & Coiera E. (2012). A systematic review of the psychological literature on interruption and its patient safety implications. Journal of the American Medical Informatics Association, 19(1), 6–12. 10.1136/amiajnl-2010-000024 [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Lipshutz AKM, Caldwell JE, Robinowitz DL, & Gropper MA (2015). An analysis of near misses identified by anesthesia providers in the intensive care unit. BMC Anesthesiology, 15(1), 93. 10.1186/s12871-015-0075-z [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Liu L, Lee S, Chia P, Chi S, & Yin Y. (2012). Exploring the association between nurse workload and nurse-sensitive patient safety outcome indicators. Journal of Nursing Research, 20(4), 300–309. 10.1097/jnr.0b013e3182736363 [DOI] [PubMed] [Google Scholar]
  25. Mansouri A, Ahmadvand A, Hadjibabaie M, Javadi M, Khoee SH, Dastan F, & Gholami K. (2014). A review of medication errors in Iran: Sources, underreporting reasons and preventive measures. Iranian Journal of Pharmaceutical Research, 13(1), 3–17.Retrieved from https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3985240/ [PMC free article] [PubMed] [Google Scholar]
  26. Mayo AM, & Duncan D. (2004). Nurse perceptions of medication errors: What we need to know for patient safety. Journal of Nursing Care Quality, 19(3), 209–217. [DOI] [PubMed] [Google Scholar]
  27. Millichamp T, & Johnston A. (2020). Interventions to support safe medication administration by emergency department nurses: An integrative review. International Emergency Nursing, 49, 100811. 10.1016/j.ienj.2019.100811 [DOI] [PubMed] [Google Scholar]
  28. Muabbar H, & Alsharqi O. (2021). The impact of short-term solutions of nursing shortage on nursing outcome, nurse perceived quality of care, and patient safety. American Journal of Nursing Research, 9(2), 35–44. 10.12691/ajnr-9-2-1 [DOI] [Google Scholar]
  29. Parry AM, Barriball KL, & While AE (2015). Factors contributing to registered nurse medication administration error: A narrative review. International Journal of Nursing Studies, 52(1), 403–420. 10.1016/j.ijnurstu.2014.07.003 [DOI] [PubMed] [Google Scholar]
  30. Patel V, Cohen T, Murarka T, LJosen J, Kagita S, Myneni S, Buchamna T. & Ghaemmaghami V. (2011). Recovery at the edge of error: Debunking the myth of the infallible expert. Journal Biomedical Informatics, 44(3), 413–424. 10.1016/j.jbi.2010.09.005 [DOI] [PubMed] [Google Scholar]
  31. Ruddy RM, Chamberlain JM, Mahajan PV, Funai T, O’Connell KJ, Blumberg S, … Pediatric Emergency Care Applied Research Network. (2015). Near misses and unsafe conditions reported in a pediatric emergency research network. BMJ Open, 5(9), e007541. 10.1136/bmjopen-2014-007541 [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Saremi M, & Fallah MR (2013). Subjective fatigue and medical errors among nurses in an educational hospital. Iran Occupational Health, 10(4), 1–8.Retrieved from http://ioh.iums.ac.ir/artic0le-1-805-en.html [Google Scholar]
  33. Shahrokhi A, Ebrahimpour F, & Ghodousi A. (2013). Factors effective on medication errors: A nursing view. Journal of Research in Pharmacy Practice, 2(1), 18–23. 10.4103/2279-042X.114084 [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Sitterding M, Ebright P, Broome M, Patterson E, & Wuchner S. (2014). Situation awareness and interruption handling during medication administration. Western Journal of Nursing Research, 36(7), 891–916. https://doi-org.libproxy.usouthal.edu/10.1177/0193945914533426 [DOI] [PubMed] [Google Scholar]
  35. Speroni KG, Fisher J, Dennis M, & Daniel M. (2013). What causes near-misses and how are they mitigated? Nursing, 43(4), 19–24. 10.1097/01.NURSE.0000427995.92553.ef [DOI] [PubMed] [Google Scholar]
  36. Van Bogaert P, Peremans L, Van Heusden D, Verspuy M, Kureckova V, Van de Cruys Z, & Franck E. (2017). Predictors of burnout, work engagement and nurse reported job outcomes and quality of care: A mixed method study. BMC Nursing, 16(5). [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Vrbnjak D, Denieffe S, O’Gorman C, & Pajnkihar M. (2016). Barriers to reporting medication errors and near misses among nurses: A systematic review. International Journal of Nursing Studies, 63, 162–178. 10.1016/j.ijnurstu.2016.08.019 [DOI] [PubMed] [Google Scholar]
  38. Westbrook J, Li L, Lehnbom E, Baysari M, Braithwaite J, Burke R, & Day R. (2015). What are incident reports telling us? A comparative study at two Australian hospitals of medication errors identified at audit, detected by staff and reported to an incident system. International Journal for Quality in Health Care, 27(1), 1–9. https://doi-org.libproxy.usouthal.edu/10.1093/intqhc/mzu098 [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Wu AW, & Marks CM (2013). Close calls in patient safety: Should we be paying closer attention? Canadian Medical Association Journal, 185(13), 1119–120. 10.1503/cmaj.130014 [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Zhu L, Yu F, Wang Y, Ning B, & Tang T. (2019). Big data analytics in intelligent transportation systems: A Survey. IEEE, 20(1), 383–398. https://ieeexplore-ieee-org.libproxy.usouthal.edu/abstract/document/8344848 [Google Scholar]

RESOURCES