Assignment Help!!!

contentserver_9.pdf

Home >Business & Finance homework help >Assignment Help!!!

The Effectiveness of Management-By-Walking-Around: A Randomized Field Study

Anita L. Tucker Harvard Business School, Soldiers Field Road, Morgan Hall 413, Boston, Massachusetts 02163, USA, [email protected]

Sara J. Singer Harvard School of Public Health, 677 Huntington Avenue, Boston, Massachusetts 02115, USA, [email protected]

M anagement-by-walking-around (MBWA) is a widely adopted technique in hospitals that involves senior managers directly observing frontline work. However, few studies have rigorously examined its impact on organizational out-

comes. This study examines an improvement program based on MBWA in which senior managers observe frontline employees, solicit ideas about improvement opportunities, and work with staff to resolve the issues. We randomly selected hospitals to implement the 18-month-long, MBWA-based improvement program; 56 work areas participated. We find that the program, on average, had a negative impact on performance. To explain this surprising finding, we use mixed methods to examine the impact of the work area’s problem-solving approach. Results suggest that prioritizing easy-to-solve problems was associated with improved performance. We believe this was because it resulted in greater action-taking. A different approach was characterized by prioritizing high-value problems, which was not successful in our study. We also find that assigning to senior managers responsibility for ensuring that identified problems get resolved resulted in better performance. Overall, our study suggests that senior managers’ physical presence in their organizations’ front lines was not helpful unless it enabled active problem solving.

Key words: health care; implementation research; patient safety; quality improvement; survey research History: Received: February 2013; Accepted: January 2014 by Edward G. Anderson, Jr. after 2 revisions

1. Introduction

Hospitals face an imperative to improve quality of care and decrease medical errors that harm patients. Healthcare thought leaders and policy makers have advocated for the adoption of “management-by-walk- ing-around” (MBWA) to achieve these goals, result- ing in widespread adoption in the United States and the United Kingdom. (Frankel 2004, National Patient Safety Agency 2011). These types of programs—in which senior managers visit the front lines to work with staff to identify and resolve obstacles—came to the attention of hospitals with the publication of one health-care system’s success at improving safety cli- mate through its MBWA-based intervention (Frankel et al. 2003). Despite the intuitive appeal of MBWA and history

of use in manufacturing organizations, empirical evi- dence on the program’s efficacy in the hospital setting is mixed. Of seven hospitals that implemented an MBWA-based program, only two were able to sustain the effort over a 3-year period (Frankel et al. 2008). Those two reported a positive impact on staffs’ per- ceptions of safety climate, but the effect on the five aborting hospitals was not reported. A study of one

Veterans Affairs hospital found that patient safety cli- mate worsened on two units that implemented the program, while it improved or stayed the same on two control units that did not implement the program (Singer et al. 2013). Another found that hospitals that implemented a general improvement program with an MBWA component did not improve on a variety of measures compared to control hospitals (Benning et al. 2011). These mixed findings provide only modest support

for widespread implementation of this program in hospitals. The lackluster performance of MBWA in health care may be that health care’s specialized and diverse disciplinary knowledge bases (e.g., cardiol- ogy, pulmonary, surgery, pharmacy, nursing, etc.) creates a complex environment where it is difficult for senior executives to effectively observe frontline work and provide improvement suggestions (Aflaki et al. 2013). In addition, the highly regulated nature of health care may minimize the marginal effectiveness of MBWA because other audit programs, such as gov- ernment-mandated inspections or incident-reporting systems, already focused senior managers’ attention on the front lines of care (Iyer et al. 2013). Further- more, the mixed results may be due to implementation

253

differences, such as the prioritization methods used to determine which problems get resolved. However, prior studies have not assessed MBWA programs at a more granular level. As a result of the contextual differences in health care and limitations of prior research, much remains to be discovered about what factors and implementation approaches are associated with the success of MBWA in hospitals. To test more systematically the impact of MBWA-

based improvement programs and to identify factors associated with its success, we implemented one such program in 19 randomly selected hospitals. We compared nurses’ perceptions of improvement in performance (PIP) in work areas that implemented the program to the same type of areas at 68 ran- domly selected control hospitals that did not imple- ment the program. A contribution of our study is thus a rigorous testing of an MBWA program. More specifically, our study design minimizes two meth- odological challenges of research on improvement programs. First, we minimize selection bias by ran- domly assigning organizations to the treatment con- dition. Our study thus provides insight into the program’s generalizability beyond those where senior managers decided on their own to implement such a program. Second, the use of control organiza- tions reduces the possibility that positive (or nega- tive) results were caused by time-dependent variables, such as changes in technology, policies, or awareness over time. Surprisingly, we find that, on average, our MBWA-based program had a negative impact on nurses’ perceptions of performance, sug- gesting that senior managers’ presence in hospital front lines to solicit improvement ideas could be det- rimental to workers’ perceptions. A second contribution of our study is developing a

categorization of problem-solving approaches that explains the conditions under which improvement solicitation programs, such as MBWA, are successful. We find that our MBWA-based program was associ- ated with improved perceptions of performance under two conditions: (1) when a higher percentage of solved problems were considered “easy” to solve, enabling more problem solving and (2) when senior managers took responsibility for ensuring that identi- fied problems were resolved. This suggests that the action-taking that results from the program, rather than the mere physical presence of the senior manag- ers, is what positively impacts the frontline staff. In section 2, we describe prior research on MBWA

programs and develop four hypotheses linking the program to performance. In section 3, we describe the intervention, the sample of hospitals that partici- pated in the research project, and our qualitative and quantitative data, measures, and analytic approach. We present the results in section 4 and discuss the

implications for research, practice, and policy in section 5.

2. MBWA-based Improvement Program’s Impact on Performance

Research has found that quality improvement pro- grams that solicit frontline workers’ ideas, such as MBWA, can have a beneficial impact on organiza- tional outcomes (Dow et al. 1999, Powell 1995). MBWA relies on managers to make frequent, learn- ing-oriented visits to their organization’s front lines to observe work and solicit employees’ opinions (Pack- ard 1995). Hewlett-Packard, the company in which MBWA originated, attributed its success using MBWA to good listening skills, willing participation, a belief that every job is important and every employee is trustworthy, and a culture where employees felt comfortable raising concerns (Packard 1995). MBWA is similar to the Toyota Production Sys- tem’s “gemba walks” (Mann 2009, Toussaint et al. 2010, Womack 2011). In a gemba walk, managers go to the location where work is performed, observe the process, and talk with the employees (Mann 2009). The purpose is to see problems in context, which aids problem solution (Mann 2009, Toussaint et al. 2010, Womack 2011). MBWA has resulted in positive organizational

change in some hospitals (Frankel et al. 2003, Prono- vost et al. 2004). One explanation is that MBWA leads to successful problem resolution because seeing a problem in context improves managers’ understand- ing of the problem, its negative impact, and its causes. This understanding increases managers’ motivation and ability to work with frontline staff and midlevel managers to resolve the issue (Mann 2009, Toussaint et al. 2010, Von Hippel 1994, Womack 2011). Theory further suggests that MBWA’s repeated cycles of identifying and resolving problems may create an organizational capability for improvement that reduces the cost of future improvement efforts, creat- ing a positive dynamic (Fine 1986, Fine and Porteus 1989, Ittner et al. 2001). This virtuous cycle is further strengthened because communication from frontline workers about problems aligns managers’ perspec- tives with customers’ experiences (Hansen et al. 2010, Hofmann and Mark 2006, Huang et al. 2010, Singer et al. 2009), enabling managers to effectively allocate scarce resources among the organization’s multiple improvement opportunities. Performance is also enhanced because managers’ presence on the front lines sends a visible signal that the organization is serious about resolving problems. This increases employees’ beliefs that leadership values improve- ment, which in turn spurs employees to engage in the discretionary behaviors necessary for process

improvement (Mcfadden et al. 2009, Zohar and Luria 2003). For these reasons, we hypothesize that MBWA will positively impact performance.

Hypothesis 1 (H1). Participation in a MBWA- type program leads to improved performance.

2.1. The Effect of Problem-Solving Approach Although we hypothesize a positive impact from MBWA, programs that solicit employee suggestions can uncover more problems than an organization can resolve, given limited problem-solving resources (Bohn 2000, Frankel et al. 2003, Repenning and Ster- man 2002). When this happens, the organization’s problem-solving support personnel must decide which of the identified issues they will work to resolve and which ones will be ignored or delayed (Keating et al. 1999, Morrison and Repenning 2011). Thus, an MBWA’s program’s success may be contin- gent upon which problems the organization decides to address. We examine two different prioritization

approaches, discuss their benefits and limitations, and develop two hypotheses. We explore two dimen- sions of problems: solution difficulty and value gained by solving the problem (Aflaki et al. 2013). To simplify the discussion, we consider only two levels of each dimension: problems are either easy to solve or difficult to solve; and they can yield either a small or large value if solved. Organizations are likely to prioritize problems that are of high value and/or problems that are easy to solve. Although we develop hypotheses based on the assumption that organiza- tions have a dominant prioritization scheme (such as addressing high-value problems), we recognize that organizations could combine the two approaches. This implies that they would emphasize high-value, easy-to-solve problems while ignoring problems that were both difficult to solve and of low value (Aflaki et al. 2013). The first prioritization approach that we consider is

one that addresses issues that are causing—or have the potential to cause—large disruptions. This high- value prioritization approach ranks problems according to a value score and solves the highest-valued prob- lems. Many structured approaches to improvement, such as six-sigma and risk management, use a high- value prioritization approach (Anderson et al. 2013a, b). In the health-care context, hospital incident-report- ing systems (Bagian et al. 2001) and MBWA-based programs (Frankel et al. 2003) advocate calculating a problem’s “value” by multiplying the problem’s score for severity with its frequency of occurrence (Bagian et al. 2001, Frankel et al. 2003). The hospital then resolves the highest-value problem first, followed by

the second highest, continuing until problem-solving resources are depleted or remaining problems fall below a threshold value (Bagian et al. 2001). Surfacing and solving the highest-valued problems should yield substantial gain in performance (Bagian et al. 2001, Girotra et al. 2010). To provide an example in the hos- pital setting, medication-related problems are often of high value because they can be fatal and can impact many patients (Bates et al. 1995). In response, many hospitals have implemented computerized physician order entry systems which reduce medication errors by preventing transcription errors and alerting physi- cians to potential drug allergies or interactions (Bates et al. 1999). This approach is beneficial because it ensures that

limited resources are preserved for problems with the highest values (Frankel et al. 2003). It also helps prevent the queue of unsolved problems from grow- ing unmanageably long by permitting the organiza- tion to discard the subset of problems that are deemed too little valued to justify solution efforts (Bohn 2000). However, there is a downside to focusing exclu-

sively on high-value problems. The ignored problems constitute the “useful many” which individually do not have a large negative impact on performance (Juran et al. 1999), but which collectively could con- tribute to serious problems such as medical errors (Reason 2000). Thus, the second approach that we consider is

prioritizing easy-to-solve problems (Johnson 2003, Repenning and Sterman 2002). An easy-to-solve priori- tization approach enables the organization to address problems that are straightforward and quick to remedy—the so-called “low-hanging fruit.” This approach may free up resources for addressing prob- lems because the more formal approach of assigning a prioritization score based on severity and occur- rence has required significant resources in the case of incident-reporting systems in both aviation and health care (Johnson 2003). An easy-to-solve prioritization approach may also

be helpful in health-care settings because the cumula- tive benefit of resolving many small problems can add up to be a significant source of improvement (Jimmerson et al. 2005). Similarly, research has found that major accidents typically result from an unpre- dictable combination of small magnitude problems rather than from a single large magnitude problem (Perrow 1984, Reason 2000). According to the “Swiss Cheese Theory,” multiple small-scale problems can align in an unfortunate way that enables an error to harm the customer (Cook and Woods 1994, Reason 2000). Consequently, resolving seemingly low-value problems can be beneficial, because they otherwise might contribute to the next major accident (Perrow

1984). To illustrate, a study of medical harm in cardiac surgery found that adverse events were more likely to be caused by multiple, simultaneous “minor” issues than by a single, “major” issue. This was because sur- geons were less able to perceive and compensate for multiple, simultaneous minor issues while they were able to recognize and remedy a single, major issue that occurred during surgery (De Leval et al. 2000). This line of research implies that it is difficult to assign a “value” to problems because their negative impact is determined in part by the specific situation in which they occur. Another situation in which the easy-to-solve priori-

tization approach may be superior is where the orga- nization has a “flat landscape” of small magnitude problems. In flat landscapes, the difference between a local high point and the global high point is too small to justify an extensive search effort (Sommer and Loch 2004). This can occur in hospitals for two reasons. First, managers typically address issues that result in patient death or other serious injury such as wrong site surgery. Thus, the only problems that remain may be small magnitude issues. Second, there are many unique opportunities for patient care to fail because work is divided among specialties, depart- ments, and shifts. Problems can occur at any of these handoffs. Thus, unlike manufacturing settings where an undetected malfunction in a machine can be the dominant source of defective product, it is less likely that there is a single, dominant source of repeated fail- ures in hospitals. When there is a flat landscape, improvement arises from solving the lower tail of problems. It may also be that organizations need to address

basic, fundamental problems before they can benefit from trying to address more complex organizational issues. For example, research suggests that problem- solving efforts are most successful when organiza- tions use relatively straightforward problems to develop sufficient problem-solving capacity before tackling larger, more complex issues (Keating et al. 1999, Morrison and Repenning 2011). Addressing easy-to-solve problems enables frequent problem- solving cycles, which develops employees’ expertise at problem solving (Adler et al. 2003). These dynam- ics suggest that organizational problem-solving capacity is more like a muscle that strengthens with exercise rather than a resource that gets depleted with use (Fine 1986, Fine and Porteus 1989, Ittner et al. 2001). We draw on the arguments outlined in the above

paragraphs to develop two hypotheses. When prob- lem-solving resources are limited and become depleted with use, the organization should focus its scarce human and financial capital on removing the problems that pose the biggest threat. Thus, a high-

value prioritization approach will be associated with improved performance.

Hypothesis 2 (H2). Work areas that resolve a higher percentage of high-value problems will have greater improvement in performance than work areas that solve a lower percentage of high-value problems.

An easy-to-solve prioritization approach should be associated with improvement because it fosters solu- tion of all problems that can be solved, regardless of their hypothetical value. In the health-care setting, this might benefit the organization because seemingly small-value problems can nonetheless negatively impact patient safety. Furthermore, the act of solving problems develops the organization’s capability to solve more problems in the future. Thus,

Hypothesis 3 (H3). Work areas that solve a higher percentage of easy-to-solve problems will have greater improvement in performance than work areas that solve a lower percentage of easy-to-solve problems.

2.2. The Role of Senior Managers in Problem Solving In addition to the prioritization approach, the success of an MBWA program depends on senior managers’ willingness to take responsibility for ensuring that problems identified through the program are resolved (Frankel et al. 2005, Pronovost et al. 2004). Senior managers can be helpful to frontline work-

ers’ resolution efforts because they control financial resources needed to address issues that involve capi- tal investment (Carroll et al. 2006). In addition, they possess the perspective necessary to resolve conflicts that arise when problems cross organizational bound- aries (MacDuffie 1997). This insight is valuable partic- ularly because high-value problems are likely to cross organizational boundaries or require financial resources to resolve. On the other hand, easy-to-solve problems impact

only one department and do not require substantial financial resources to resolve. Under these conditions, frontline employees can be empowered to identify and resolve problems (Jimmerson et al. 2005). How- ever, involving frontline workers in resolution efforts requires them to take time away from their direct pro- duction responsibilities (Repenning and Sterman 2002, Victor et al. 2000). This can be difficult for front- line employees, especially for health-care workers who provide direct patient care. Under these condi- tions, senior managers need to allocate funds for over- time or coverage so that care providers can spend time away from patient care and on resolution efforts.

As outlined in the two above paragraphs, both high-value and easy-to-solve problems require man- ager support for successful resolution. Therefore, we hypothesize that hospital work areas will achieve bet- ter results from the MBWA program when they assign to senior managers the responsibility for ensur- ing that a problem gets addressed.

Hypothesis 4 (H4). Work areas with a higher percentage of problems assigned to a senior manager to ensure resolution exhibit greater improvement than those with a lower percent- age of problems assigned to a senior manager.

These four hypotheses outline the theoretical links between our MBWA-based program and improved performance. Figure 1 depicts these relationships.

3. Methodology

We test our hypotheses in a field study of US hospi- tals randomly selected to participate in a patient safety research study, with a subset of the hospitals randomly selected (a second time) to implement our MBWA-based program. The program was launched in January 2005 and lasted for 18 months.

3.1. The MBWA-based Program We drew on prior research to design our MBWA- based program (Frankel et al. 2008, Pronovost et al. 2004, Thomas et al. 2005). It consisted of repeated cycles of senior manager–staff interaction, debriefing, problem solving, and follow-up. Senior managers such as the chief executive, operating, medical, and nursing officers (CEO, COO, CMO, and CNO, respec-

tively), interacted with frontline staff in a work area to generate, select, and solve improvement ideas. The work area manager was also involved in the selection and solution activities. Senior manager interactions took two forms: visits, called “work system visits,” to work areas to observe frontline work; and special meetings, called “safety forums,” with a larger group of frontline staff from the area to discuss safety con- cerns. The activities were coordinated with the work area manager. In work system visits, four senior managers would

spend 30 minutes to 2 hours visiting the same work area. The senior managers would each observe a dif- ferent process, such as medication administration, or a different person, such as a nurse or physician, to shed cross-disciplinary insight into the work done in the area. The purpose was to build senior managers’ understanding of the frontline work context and gather grounded information about problems (Fran- kel et al. 2008). Senior managers also facilitated a safety forum in

the work area, which was an informal meeting between senior managers and the frontline staff from the work area, held in the work area, during which the staff talked about their work area’s safety weak- nesses and strengths. We added this component to our MBWA-based intervention for two reasons. First, a San Diego children’s hospital improved its organi- zational climate by holding meetings where frontline staff spoke directly to the hospital CEO about their concerns and ideas (Sobo and Sadler 2002). Second, a prior research project on an MBWA-based pro- gram found that the program only improved the perceptions of frontline staff who participated in a work system visit (Thomas et al. 2005). Because it is not feasible for senior managers to conduct a work system visit with every single hospital employee within a short time period, Thomas’ finding suggests that work system visits on their own will be insuffi- cient to change the perceptions of most hospital employees. The MBWA-based program continued with a

“debrief meeting,” which organized information col- lected from the work system visits and safety forums. Senior managers attended, as did work area manag- ers, selected frontline workers, and the hospitals’ patient safety officers. The group compiled the improvement ideas identified, discussed and in some work areas prioritized them, and decided next steps, ranging from doing nothing to suggesting solutions and assigning responsibility. Action to address prob- lems selected for resolution followed the debriefing. Managers were encouraged to communicate with staff about implementation efforts, describing what changes, if any, were made in response to identified ideas. Patient safety officers entered the ideas

MBWA Program Performance

Problem solving activities used in MBWA

Address high- value problems

Address “easy-to- solve” problems

Managers ensure problems are

resolved

H1+

H2+

H3+

H4+

Figure 1 Model of Management-By-Walking-Around’s Impact on Per- formance

generated and actions taken into an electronic spread- sheet we provided and sent this spreadsheet to our research team for analysis. Each round of work system visits, safety forums,

debrief meeting, solution activities, and communica- tion constituted one cycle. A cycle focused on one work area and took approximately 3 months, which research has shown is the time required to solve prob- lems in an organization (Pronovost et al. 2004). See Figure 2 for a diagram of the process. After complet- ing a cycle, the management team would repeat the activities in a different work area. The program focused on the four main work areas in hospitals: operating room or postanesthesia care unit (OR/ PACU), intensive care unit (ICU), emergency depart- ment (ED), and medical or surgical ward (Med/Surg). Cycles continued over the 18-month implementation, with hospitals conducting an average of one cycle in four work areas.

3.2. Recruitment Our study employed an experimental design which included a pre-test and post-test of similar work areas in treatment and control hospitals. We randomly selected 92 US acute-care hospitals, stratified by size and geographic region, to participate in a patient safety climate survey. We provided no financial incentive, but participation in the safety climate study fulfilled a national accreditation requirement. At enrollment, all hospitals were aware that they may be invited to participate in a program to improve patient safety, but details regarding the program were with- held to prevent contamination of control hospitals. To select hospitals to participate in the MBWA-based program, we drew a second, stratified, random sam- ple of 24 hospitals from the sample of 92. The remain- ing 68 hospitals not selected were control hospitals.

Data on staff perceptions of performance were collected at control and treatment hospitals through surveys before implementation of program activities (2004, “pre”) and again after the program was com- pleted (2006, “post”). At each hospital, we surveyed a random sample of 10% of the frontline workers, with additional oversampling in OR/PACUs, EDs, and ICUs in the post-survey period to improve sample size. The baseline “pre” response rate was 52%; and the follow-up “post” response rate was 39%. For our analyses, we used data from registered and licensed vocational nurses (n = 1117 pre and n = 903 post). Of the 24 treatment hospitals, 20 completed the pro-

gram in at least two work areas. Of the four that did not complete the treatment, one went out of business, one was purchased, and two experienced significant senior management turnover. As a result, they were unable to complete more than one cycle of activities and did not provide data. We thus excluded these four from our analysis. There was no difference in staff perceptions of performance in the pre-period between the four hospitals that dropped out of the treatment and the 20 that did not. Of the original 68 control hospitals, 48 completed the post-test survey, making an initial total sample of 68 hospitals. There was no difference in survey measures in the pre-per- iod between the 20 control hospitals that dropped out of the post-survey and the remaining hospitals. There was also no difference between treatment and control work areas on pre-period measures of staff percep- tions of performance.

3.3. Data and Measures Using the data collection spreadsheet that we pro- vided (Figure 3), treatment work areas reported 1245 patient safety problems identified during the visits and forums. Each hospital also provided a list of the

C E

C N

C M

C F

Work site visit by CEO

Time

Work site visit by CNO

Work site visit by CMO

Work site visit by CFO

Safety Form

Debrief Meeting

Solution Activities & Communication

Figure 2 Depiction of the MBWA-based Program Activities in a Work Area

senior managers, which we used to determine whether a senior manager attended the program activity and whether a senior manager was assigned responsibility for the problem. The spreadsheet also contained three columns that the work areas could use to prioritize identified problems. Twenty-four work areas in eight hospitals filled out this informa- tion.

3.3.1. Independent Variables. To test the overall impact of the MBWA-based program (H1), we created a treatment variable, “MBWA in the work area,” which indicated whether the work area received the MBWA-based treatment (=1) or was a work area from a control hospital (=0). To test the high-value prioriti- zation approach (H2), we calculated a value score for each problem by multiplying problem severity (col- umn 7 in Figure 3; 1 = low; to 10, could cause death) by estimated frequency of occurrence (column 8; 1 = very unlikely, 3 = very likely) (Bagian et al. 2001, Frankel et al. 2003). This method for calculating the potential value of solving a problem is similar to six- sigma’s risk prioritization number, which uses the product of the scores (on a scale from 1 to 10) of a

problem’s frequency of occurrence, detectability, and severity (Evans and Lindsay 2005). It is also similar to risk registers used for risk management. A risk regis- ter scores each potential risk to a project by multiply- ing the risk’s likelihood of occurrence by severity of the impact if it does occur (Anderson et al. 2013a,b). We used our value score in combination with whether or not the problem was addressed (column 10 in Fig- ure 3) to create a unit-level variable that represented the percentage of problems in the top quartile (ranked by value) that were resolved, which we call “% of top quartile that were resolved.” As an alter- nate test of H2, we also created a dummy variable, “Top ranked problem resolved?” A dichotomous variable that indicated whether or not the top-ranked problem in the work area was resolved. The alternate specification for H2 allowed us to test our prediction using innovation literature theory, which asserts that success can come from identifying and solving even just one high-value idea (Girotra et al. 2010). To test the easy-to-solve prioritization approach (H3), we calculated, from a work area’s set of problems that were resolved, the percentage that were rated “easy- to-solve,” a “1” on a 3-point scale, meaning it is was

1 2 3 4 5 6 7 8 9 10 11 12 13

Hospital #

Date of Activity

Activity Type:

Worksite Visit or Safety Town

Meeting

Participant from

Executive team

Location "Hinderers" to patient safety, or

system weaknesses observed during worksite visit, or brought up during safety town meeting (one item per row)

Safety Risk:

1: Low 3: Mild

discomfort 5: Would require

intervention 10: Could cause harm or death

Likelihood or frequency of

risk

1=Very unlikely

2=Possible 3=Very likely

Ease of implementation

1=Easy, within 30 days

2=Moderate-multiple departments (90 days) 3 = Difficult-process changes and/or major

budget (6 months)

Action items or proposed changes to hinderers

Team member(s) responsible

for follow up

C-Suite

Yes = 1 No = 0

Date change

completed

100 3/16/2 006

Worksite Visit

Betsy Green, CNO

Medical/ Surgical

Unit

New diabetics’ insurance won't pay

for glucometers. Staff concerned about patients'

inability to get the devices and their own need to learn

many different devices based upon

what the patient purchased. The delay

decreases the amount of time

nursing staff have to teach patients about

using the device.

10 2 2 Director of Laboratory

Services communicat -ed the need to a vendor of diabetic supplies.

Director of Laboratory

Services and CNO

1 Mar-06

100 Another problem of lower value would be here 2

100 3/14/ 2006

Worksite Visit

Jen Calhoun,

Safety Director

Medical/ Surgical

Unit

Overbed tables being used to hold Personal Protective Equipment

(PPE).

5 1 1 Isolation Carts have

been purchased

to hold and store PPE outside of

patient rooms.

CNO and Director of Medical/ Surgical

Unit

1 1st cart arrived

03/20/20 06

To test H2: % of the top quartile (of value) that were resolved =100%

To test H3: % of resolved problems that were “easy-to-solve” =50%

To test H4: % of problems assigned to senior manager =50%

Value = 10*2 = 20 Top quartile? = 1 (yes) Addressed? = 1 (yes) Top quartile & addressed? = 1 yes

Figure 3 Data Collection Sheet Used by Treatment Hospitals and Two Problems as Examples

“easy and could be resolved within 30 days” (col- umn 9 in Figure 3). The higher the percentage, the more the unit solved easy-to-solve problems. We called this variable “% of problems solved that were low-hanging fruit.” Finally, to test our hypothesis about senior managers (H4), at the work area level we found the percentage of problems for which a chief executive level manager was assigned responsi- bility for ensuring that the problem was resolved (column 12 in Figure 3). See Figure 3 for details on these variables.

3.3.2. Measure. In accordance with prior research (Chandrasekaran and Mishra 2012, Frankel et al. 2003, 2005, 2008), we evaluated the program’s perfor- mance using staff “PIP.” To measure PIP, we used four survey items (see Appendix A) from validated survey instruments that measured the effectiveness of quality improvement efforts (Shortell et al. 1995, Singer et al. 2009). Respondents rated each item using a 5-point scale ranging from 1 = strongly disagree to 5 = strongly agree. Agreement indicated that respon- dents thought quality and safety performance were improving. The scale exhibited high reliability (Nun- nally 1967), with a Cronbach’s alpha of 0.84 (n = 1147 nurses) in the pre-period and 0.88 (n = 1103 nurses) in the post-period. We used perception of performance for four rea-

sons. First, employee perceptions are an important outcome because they influence behaviors, which in turn impact objective measures (Zohar and Luria 2003). Second, staff perceptions of performance are a valid indicator of performance (Ketokivi and Schroe- der 2004). This is because employees are close to the work and often know if system failures are decreasing or increasing. Research has found that nurses’ percep- tions of safety are correlated with objective measures of safety outcomes, such as mortality, readmissions, and length of stay (Hansen et al. 2010, Hofmann and Mark 2006, Huang et al. 2010, Singer et al. 2009). Third, employee perceptions have been widely used as outcome measures in operations management research because they enable comparison across orga- nizations (Anderson et al. 2013a,b, Atuahene-Gima 2003, Bardhan et al. 2012, Chandrasekaran and Mish- ra 2012, Flynn et al. 1995, Kaynak 2003, Swink et al. 2006). Finally, the use of a perceptual measure was necessitated by hospitals’ unwillingness to share data on safety incidents. Our dependent variable was the change in PIP from

the pre- to the post-period. The use of change scores allowed us to examine change over time (Fitzmaurice 2001). To create a composite change score for each work area, we used the pre-data to calculate the mean of the four items for each nurse, and then aver- aged by work area. We repeated this process for the

post-data and subtracted each work area’s pre-score from its post-score. We calculated intra-class correla- tions (ICC) and a mean inter-rater agreement score (rWG) to test whether aggregation of PIP was appro- priate. Significant (ICC[1] = 0.06, F = 5.69, p < 0.000, and ICC[2] = 0.82) supported aggregation (Bliese 2000). The rWG for nurses’ rating of PIP was 0.60, which also was sufficient for aggregation (Zellmer- Bruhn 2003). Furthermore, our use of a change score as our dependent variables met the two conditions specified by Bergh and Fairbank (2002): the reliabili- ties of our survey measures for PIP in pre- and post- periods were high (0.84 and 0.86, respectively) and the correlation between the measures from the two different time periods was low (q = 0.24, p < 0.001). As is common in studies using a change score (Bergh and Fairbank 2002), the correlation between the change score and the PIP measure in the pre-period was negative (q = �0.67, p < 0.001). This indicates that there was a greater opportunity for improve- ment in PIP among work areas with a low PIP in the pre-period (Fitzmaurice 2001). Therefore, to control for impact of a work area’s starting point on the change in PIP, we included a dichotomous variable indicating whether PIP in the pre-period was in the lower quartile (“bottom quartile for 2004 PIP”). The variable was coded “1” if the work area was in the bottom quartile of work areas in PIP in the pre- period and “0” for all others. This method enabled us to test for the change in PIP while controlling for a low starting point.

3.3.3. Control Variables. For H1, which tested the overall impact of our MBWA-based program, the large sample size enabled us to include the following control variables: major teaching hospital (1 = yes, 0 = no); Dun & Bradstreet’s measure of the hospital’s financial stress, with higher numbers indicating a higher likelihood that the business will seek legal relief from creditors or cease operations without pay- ing creditors in full over the next 12 months; a set of dummy variables for the number of hospital beds (reference group = less than 100 beds; medium = 100–250 beds; large = more than 250 beds); and a set of dummy variables for type of work area (reference group = non-clinical; OR/PACU; ICU; ED; Med/ Surg unit; and other clinical unit). Data on size and teaching came from the 2004 American Hospital Association Survey of Hospitals. For the hypotheses about problem prioritization

(H2 and H3), our sample size was limited to the 24 work areas that formally prioritized their problems in the data collection spreadsheet. As a result, for these hypotheses, we did not have a large enough sample size to include non-significant control variables in our regression. However, our random selection of

hospitals helps alleviate concerns that our model may be subject to omitted variable bias (Antonakis et al. 2010). We did not include control variables for unit type (e.g., ED, ICU, and OR/PACU) as none were sig- nificant and their inclusion did not change our results. We also tested for hospital-level control variables, such as teaching status and number of beds, but none were significant and their inclusion did not change our results. We controlled for availability of “low- hanging fruit,” which was the percentage of identified problems that were rated as easy to solve. We also controlled for the average value of the top quartile of identified problems. Our regression equation for H4, the impact of a

senior manager being assigned responsibility for problem resolution, included the full set of 58 inter- vention work areas. We controlled for the percent- age of problems within a work area that were resolved (% of problems resolved) by coding a prob- lem as having had solution effort if there was evi- dence in the dataset that action had been taken to address the problem, and taking the average of this variable at the work area level. We also controlled for the fidelity of implementation with the following variables: the number of work system visits that were conducted, whether a work system visit was conducted by a senior manager (1 = yes, 0 = no), and whether a safety forum was conducted in the area (1 = yes, 0 = no).

3.4. Sample Size and Analysis We used linear regression with robust standard errors and clustered by hospital (Rabe-Hesketh and Everitt 2004) in Stata 11.1TM to test our hypotheses. The Shap- iro–Wilk test for all regressions showed that the resid- uals were normally distributed (V close to 1 and p > 0.10) (Royston 1992). Multicollinearity was also not an issue as all variance inflation factors for all of our equations were less than 2.5, well below the upper threshold of 10 (Chatterjee and Hadi 1986). To test the overall impact of our MBWA-based

program (H1), we use data from the four main clini- cal work areas (OR/PACU, ICU, ED, and Med/ Surg). We had data for both pre- and post-PIP mea- sures from 58 intervention work areas in 20 treat- ment hospitals and 138 work areas in 48 control hospitals. However, missing data for a control vari- able (financial stress) in two intervention work areas resulted in a final sample size of 56 intervention work areas. To test the impact of problem selection (H2 and H3), we used data from the 24 work areas from eight treatment hospitals that formally priori- tized their problems. Finally, to test the impact of senior manager assignment to problem resolution (H4), we used the full set of intervention work areas (n = 58).

3.5. Qualitative Data Collection and Analysis During the intervention, we visited each treatment hospital to tour the clinical areas and to observe MBWA activities, including work system visits, safety forums, and debrief meetings. In addition, we dis- cussed and observed examples of changes imple- mented in response to problems identified through the program to verify accuracy of the data submitted. There were no discrepancies. We also conducted semi-structured interviews with a frontline staff member, a department manager, and the CEO from each hospital (see Appendix B). Interviews addressed the nature of performance improvement in the hospi- tal in general and as it related to implementing the MBWA-based program. Interviews and notes from the meetings were recorded and transcribed. Investi- gators also wrote a journal of the day’s activities from notes taken during the day. The journal and tran- scripts from each hospital were combined into a sin- gle document, which served as our source of qualitative data. After the intervention was complete, we used

these qualitative data in combination with the prob- lem data submitted by the work areas to illuminate differences among work areas in the types of issues identified, actions taken to resolve them, and manag- ers’ attitudes. We analyzed transcripts using the pro- cedure described in Miles and Huberman (1994, pp. 58–62). We initially used a list of codes based on our interview questions. We read the transcripts multiple times, revising the codes as we deepened our under- standing of similarities and contrasts among the implementation of the program. How the managers prioritized problems for solution efforts emerged as a main theme. One author went through the qualita- tive data to select all relevant quotes for this theme. Both authors independently reviewed the quotes while blinded from the performance results. We compared our perceptions to come to a consensus. We use the quotations to illustrate differences in implementation approach that impacted the effec- tiveness of the intervention. Table 6 in the results section displays representative quotations from the five work areas that improved the most over the course of the intervention and the five that decreased the most.

4. Results

4.1. Summary Statistics Average PIP in the 56 treatment work areas was 3.78 in the pre-period and 3.69 in the post-period. The dif- ference of �0.09 was not statistically significant at the 10% significance level. The same four types of work areas (n = 138) in control hospitals had a mean PIP of 3.8 in both time periods. Table 1 shows descriptive

statistics. Using the subset of work areas that priori- tized their problems (n = 24), the mean value score for all identified problems was seven on the scale of 1 (lowest) to 30 (highest). Descriptive statistics and cor- relations are shown in Table 2. On average, the mean value score was 17 for the top quartile of identified problems. The highest score, on average, was about 19.

4.2. Regression Results Contrary to our prediction, the MBWA-based treat- ment was associated with a statistically significant decrease in PIP (�0.17, p < 0.05) compared to the same types of work areas in control hospitals (H1, Table 3, Model 1). A possible explanation is that some treat- ment work areas failed to conduct the recommended activities (Nembhard et al., 2009). However, the fol- lowing statistics provide evidence that treatment areas did indeed implement the MBWA-based pro- gram: 91% had a work system visit; each treatment work area received a mean of 3.41 visits (SD = 3.16, maximum of 12); 50% had a safety forum; on average, they identified 19 problems and took action on 11 (Table 1). The effectiveness of the program did vary, how-

ever, among work areas. As shown in Model 1, our control variable for whether or not the work area was in the bottom quartile for pre-period PIP was signifi- cant (b = 0.75, p < 0.001), suggesting that work areas with the lowest PIP scores in the pre-intervention per- iod exhibited a positive change in PIP over the course of the intervention. Additional analysis revealed that the work areas that were in the bottom quartile for our dependent. variable, change in PIP, had a decline in PIP rang-

ing from 0.375 to 2.25. Of these 15 work areas that experienced the greatest decline in PIP, four were already below median in the pre-period, suggesting that their decline was not merely a regression to the mean effect. The work areas in the top quartile of change in PIP experienced an increase in PIP ranging from 0.38 to 1.33 points. This large variation in results prompted us to examine factors associated with success.

Model 1 in Table 4 shows results from testing H2 and H3. A higher percentage of problems solved that were rated as “easy-to-solve” was associated with higher% change in PIP (coefficient = 1.00, p < 0.05), providing support for H3. A one standard deviation (27%) increase in the percent of solved problems that were easy-to-solve was associated with a 1.0 point increase in change in PIP, which was a 26% improve- ment. However, the percentage of problems rated in the top quartile for value that were solved was not significant. Thus, H2 is not supported. Testing H2 using highest-value score instead of the

mean priority of the top quartile and a dummy for whether the top-ranked problem for value was resolved instead of the percentage of problems rated in the top quartile for value that was solved was also not significant (Table 4, Model 2). This result fails to support theory from the innovation literature suggest- ing that solving the highest-value idea drives perfor- mance in our context. However, the percentage of problems resolved that were rated “easy-to-solve” remained significant in this model (coefficient = 0.82, p < 0.01), providing additional support for H3. Priori- tizing easy-to-solve problems appeared to increase PIP. An alternate explanation for our finding could be

that work areas were more successful because they spent more money on problem solving rather than because they prioritized easy problems. To control for this “spend more” explanation, the authors individu- ally rated the rough cost of each solved problem on a scale of 1–3 with 1 = low (cost ≤ $500), 2 = medium (cost > $500 < $150,000), and 3 = high (cost ≥ $150,000) based on the description of how work areas solved the problem and independent research to check the cost of products or services mentioned in the description. We used these ranges because they represented different categories of solutions. The cheapest category was solutions that involved a one- time purchase of a relatively low-cost supply (<$500). An example is applying a coating to one window to improve patient privacy. The second category was intended to cover mid-range solutions such as the purchase of equipment or consumable supplies. An

Table 1 Mean, Standard Deviation (SD), and Correlations for Treatment Work Areas (N = 56 work areas)

Variable Mean SD Min Max 1 2 3 4 5 6

1 Postperiod PIP 3.69 0.61 1.92 5.00 2 Change in PIP �0.09 0.67 �2.25 1.33 0.639*** 3 Had work system visit 91% 29% 0 1 �0.195 �0.197 4 Number of work system visits in area 3.41 3.16 0 12 0.055 �0.1 0.342* 5 Had safety forum 50% 50% 0 1 0.056 0.028 �0.313* 0.097 6 Percent of problems addressed 62% 31% 0 1 �0.088 0.079 0.083 0.043 �0.074 7 Percent of problems assigned to

senior manager 10.4% 23.7% 0 93% 0.186 0.175 0.114 0.359** 0.176 0.065

***p < 0.001, **p < 0.01, *p < 0.05.

example is the purchase and installation of new light- ing in a catheterization laboratory to illuminate proce- dures. The most expensive category was for solutions that involved construction or hiring of multiple peo- ple. An example is a solution that involved hiring multiple people to transport patients within the hos- pital. We compared scores and discussed our ratio- nale until we reached consensus for all solved problems. We then summed the total estimated solu- tion costs, estimating 1 = $250; 2 = $5000; and 3 = $150,000, for all of the solved problems in each work area. Another possible explanation is that variation in

quality of solution efforts impacted the results (e.g., some work areas might have engaged in only superfi-

cial steps while others might have systematically resolved underlying causes). We also controlled for this “higher quality” explanation by hiring 10 nurses not affiliated with our study hospitals to rate the solu- tion effectiveness of the proposed solution for each

Table 2 Mean, Standard Deviation (SD), and Correlations for Treatment Work Areas and Identified Problems (n = 24)

Variable Mean SD Min Max 1 2 3 4 5 6 7

1 Change in PIP 0.02 0.53 �1.17 1.1 2 Avg value of top quartile of

identified problems 17.23 6.67 6 30 0.298 1

3 Highest-valued score 18.75 7.43 6 30 0.325 0.952*** 1 4 Availability of

low-hanging fruit 36% 26% 0% 100% �0.016 �0.305 �0.289 1

5 Percentof top quartile problems solved

88% 29% 0% 100% 0.186 0.091 0.109 �0.045 1

6 Highest-valued problem was solved

88% 34% 0 1 0.209 0.110 0.039 �0.086 0.799***

7 Percent of solved problems that were low-hanging fruit

33% 27% 0% 83% 0.327 0.097 0.099 0.551** 0.432* 0.350†

8 Percent of problems assigned to senior manager

22.5% 32.4% 0% 93% 0.308 0.582** 0.576** 0.136 0.054 0.242 0.457*

***p < 0.001, **p < 0.01, *p < 0.05, †p < 0.10.

Table 3 Linear Regression testing Hypothesis 1 (the Change in PIP in Treatment Work Areas vs. the Same Types of Work Areas from Control Hospitals) Clustered by Hospital with Robust Standard Errors in Parentheses

Model 1

H1. Treatment work area (1 = yes) �0.17*(0.08) Bottom quartile PIP (pre-period) 0.75*** (0.10) Major teaching hospital (1 = yes) 0.21† (0.13) Financial stress 0.00 (0.00) Medium-size hospital (100–250 beds) �0.43*(0.10) Large-size hospital (>250 beds) (1 = yes) �0.26* (0.12) OR/PACU (1 = yes) �0.08 (0.11) ICU (1 = yes) 0.00 (0.13) ED (1 = yes) �0.15 (0.13) Was a work system visit conducted? Not in model Was a safety forum conducted? Not in model Constant �0.02 (0.20) Observations 194 Treatment and control work areas 56 & 138 Degrees of freedom F (9, 55) F-statistic 9.06*** Adjusted R2 0.20

***p < 0.001, **p < 0.01, *p < 0.05, †p < 0.10.

Table 4 Regression Comparing Change in PIP in Treatment Work Areas that Rated the Severity, Frequency, and Ease of Solution of the Problems, Clustered by Hospital with Robust Standard errors in parentheses (H2 and H3)

Model 1 Model 2 Model 3

Mean value of top quartile of identified problems

0.02 (0.02) — —

Highest-value score of identified problems

— 0.02 (0.02) —

Availability of low-hanging fruit

�0.60 (0.49) �0.45 (0.45) �0.90† (0.42)

H2. Percent oftop quartile value resolved

�0.22 (0.23) — —

H2. Was top-ranked value problem resolved (1 = yes)

— �0.01 (0.26) —

H3. Percent of solved problems that were low-hanging fruit

1.00* (0.30) 0.82** (0.21) 1.22* (0.46)

Bottom quartile 2004 PIP pre (1 = yes)

0.39* (0.16) 0.36^ (0.19) 0.38* (0.13)

Cum. cost of solving problems

— — 0.00 (0.00)

Avg effectiveness of solution effort

— — �0.11 (0.10)

Constant �0.25 (0.48) �0.47 (0.46) 0.61 (0.62) Observations 24 24 24 Degrees of freedom F (5, 7) F (5, 7) F (5, 7) F-statistic 10.99** 5.28* 7.08* Adjusted R2 0.06 0.07 0.08

***p < 0.001, **p < 0.01, *p < 0.05, †p < 0.10.

problem using a scale from 1 to 10. The low end of the scale was used for problems that were not resolved (1 = “no information given”; 2 = management dis- missed the issue or it was not a safety issue; and 3 = issue not considered due to lack of funds or issue passed off to someone else without any follow-up). The higher the number, the more substantial and sys- tematic the solution (e.g., 9 = major investment or change; 10 = systemic fix that would prevent recur- rence). The scale is available from authors. Agreement among nurses on their ratings was acceptable (j = 0.23) (Landis and Koch 1977). The mean rating for solution effectiveness was higher at 5.9 for solved problems (“solution action in progress” on our scale) than 2.7 (“no solution implemented”) for unsolved problems, which validates their coding. Given our small sample size, in this secondary

analysis we omitted the high-value prioritization vari- ables, as they were not significant in our primary analyses. As Model 3 shows, the variable for the cumulative “cost of solving problems” was not signif- icant. This may be because work areas could improve PIP without having to spend a lot of money on solu- tions. Solution effectiveness was also not significant. The percentage of solved problems that were low- hanging fruit remained significant (coefficient = 1.22, p < 0.05), indicating that the results are similar after accounting for spending and solution effectiveness. The evidence in the three models supports H3, which predicted that prioritizing easy-to-solve problems would be associated with higher PIP. Table 5 shows the results from testing H4, which

proposed that senior managers taking responsibility for ensuring that identified problems get resolved would be associated with higher% change in PIP. H4 was supported (coefficient = 0.79, p < 0.05). Increas- ing the percent of problems assigned to senior manag- ers by one standard deviation (23%) was associated

with a 0.79 increase in PIP. This equates to a 21% increase in PIP.

4.3. Robustness Check Other scholars have used a different approach for testing improvement over time by using the post- measure as the outcome variable and the pre-measure as a control variable (Fitzmaurice 2001). We tested our hypotheses using this method and the results were the same (results not shown).

4.4. Qualitative Results To provide insight into the nature of implementation of MBWA-based programs, Table 6 presents qualita- tive data from the five work areas that improved the most and the five that decreased the most. Between pre- and post-periods, on average PIP improved by 0.85 for the top five work areas and decreased by 1.4 for the bottom five. Our examination of issues identi- fied and actions taken suggests that the top work areas identified meaningful problems and managers took these problems seriously. For example, hospital 88s Med/Surg unit was one of the most improved work areas. One of the identified issues was that the small size of the medication room prevented two nurses from preparing medications simultaneously, which was an inconvenience and delayed patient care. Senior managers discussed the issue with staff and they collectively made a plan to move the medi- cation room to a larger space. The COO commented, “It’s a little thing, but when you actually see them doing the process, you say, ‘Wait a minute, that is dif- ficult for them.’” An interview with a nurse high- lighted management’s willingness to address issues. She commented, “These people address safety issues. It may not always get addressed the way you want, but it still gets addressed.” Conversely, in the bottom work areas, an emphasis

on prioritizing the highest-valued problems limited solution efforts. For example, hospital 129s ED identi- fied valid issues, such as long lead times to receive lab results. However, in the safety forum, we observed the manager spend the entire time getting staff input on prioritizing the items, leaving no time to discuss how the issues might be resolved. This work area did not solve any of the problems they had identified, despite investing substantial time in identifying and prioritizing them. As Table 6 shows, this pattern was common. Two of the six bottom work areas did not resolve any problems, another’s “solutions” were lar- gely to re-educate staff, and a fourth area provided us with no information about solved problems. These implementation details suggest an inability to make meaningful progress on solving the problems. The lack of solution efforts illustrates how relying too heavily on a high-value prioritization approach can

Table 5 Impact of the Percentage of Problems Assigned to Senior Managers on Change in PIP in Treatment Work Areas (H4)

Model 1

H4. Percentage of problems assigned to senior managers for resolving

0.79* (0.32)

Bottom quartile PIP pre (1 = yes) 0.56** (0.15) Percentage of problems solved 0.12 (0.33) Number of work system visits in the area �0.04† (0.02) Senior manager participated in work system visit (1 = yes)

�0.12 (0.23)

Safety forum in the area (1 = yes) �0.12 (0.14) Constant �0.08 (0.31) Observations 58 Degrees of freedom F (6, 19) F-statistic 2.96* Adjusted R2 0.10

***p < 0.001, **p < 0.01, *p < 0.05, †p < 0.10.

T a b le

6 Il lu st ra ti ve

P ro b le m s,

S o lu ti o n s,

a n d Q u o te s fr o m

T o p a n d B o tt o m

Q u a rt il e W o rk

A re a s

H o sp it al

ID W o rk

ar ea

2 0 0 4

S co re

2 0 0 6

S co re

% C h an g e

E xa m p le s

S o lu ti o n ef fo rt s

Il lu st ra ti ve

q u o te s an d ex am

p le s ab o u t p ri o ri ti za ti o n

1 1 6

O R /P A C U

3 .3

4 .6

4 1 %

N ee d m o re

cl in ic sp ac e

M ad e n ew

cl in ic ro o m s

T h e as so ci at es

w ill p ri o ri ti ze

w it h th e m an ag er s,

w h o

h av e a g o o d id ea

o f w h at

th e st af f w an t to

d o

1 0 0

M ed /S u rg

2 .6

3 .6

3 8 %

N ew

ly d ia g n o se d d ia b et ic p at ie n ts

ca n n o t g et

g lu co m et er s fr o m

in su ra n ce ; b u y d if fe re n t ki n d s,

h ar d fo r n u rs es

to te ac h

V en d o r d o n at ed

g lu co m et er s,

in -s er vi ce d n u rs es ,

m ad e ki ts

fo r n ew

ly d ia g n o se d d ia b et ic p at ie n ts

M an ag er

o rd er ed

n ew

is o la ti o n ca rt s to

ke ep

su p p lie s

fo r ea ch

p at ie n t o u ts id e th e d o o r to

p re ve n t sp re ad

o f M R S A

8 8

M ed /S u rg

3 .6

4 .7

3 1 %

M ed ic at io n ro o m

is ve ry

sm al l fo r tw o p eo p le

A ft er

d is cu ss in g w it h st af f, ch an g ed

m ed ic at io n

p re p ar at io n to

a la rg er

ro o m .

T h es e p eo p le ad d re ss

sa fe ty

is su es . It m ay

n o t al w ay s g et

ad d re ss ed

th e w ay

yo u w an t it to , b u t it st ill g et s ad d re ss ed .

4 7

E D

3 .0

3 .8

2 8 %

N ee d p ro m p t re sp o n se

fr o m

p h ar m ac y fo r se le ct ed

m ed s;

n ee d lif t eq u ip m en t fo r o b es e

p at ie n ts ; P yx is * IT

d is p la y

d is p o se d to

m ed ic at io n er ro rs

In st al le d p h o n e sy st em

w it h p ri o ri ty

ac ce ss

to p h ar m ac y;

id en ti fi ed

o r ad d ed

lif t eq u ip m en t;

re p ro g ra m m ed

P yx is IT

d is p la y

W e u n d er st an d w h at

n ee d s to

b e d o n e - tr yi n g to

g et

ri d o f

ve rb al

o rd er s,

tr yi n g to

se t u p o u r P yx is

m ac h in e d if fe re n tl y

3 9

E D

4 .0

5 .0

2 5 %

Fe el lik e “d u m p in g g ro u n d ”

w h en

th e cl in ic cl o se s;

R o o f le ak s,

n ee d m o re

b lo o d p re ss u re

m ac h in es

R el o ca te d cl in ic in

to ex p an d E D p at ie n ts ; h ir ed

ad d it io n al E D st af f; fi xe d ro o f; p ro vi d ed

b lo o d

p re ss u re

eq u ip m en t

N u rs e al m o st

g av e w ro n g m ed ic at io n b ec au se

tw o si m ila r

d ru g s n ex t to

ea ch

o th er

in P yx is . T o ld

C N O . P h ar m ac is t

ca m e u p ri g h t aw

ay an d ch an g ed

d ra w er

3 4

O R /P A C U

5 .0

3 .8

�2 5 %

O R ta b le n o t sa fe

fo r b ar ia tr ic

p at ie n ts ; in su ffi ci en t ch ec ki n g

o f p at ie n t la b s p ri o r to

su rg er y

N o so lu ti o n s lis te d

A n yo n e ca n su b m it sa fe ty

id ea

to th ei r vi ce

p re si d en t. It g et s

se n t o u t fo r re vi ew

to ap p lic ab le d ep ar tm

en ts

1 1 9

O R /P A C U

3 .8

2 .6

�3 1 %

N ee d ex h au st

ai r, so m e eq u ip m en t

(c h ai rs ), b ac ku p o f p at ie n ts

in E D ,

b ed s n o t re ad y

C h an g es

to im p ro ve

ai r, eq u ip m en t o rd er ed

o r re p ai re d , w o rk in g o n fl o w in

E D

It is h ar d to

fi n d th e ti m e an d en er g y [t o su st ai n th is p ro g ra m ]

b ec au se

th er e ar e o th er

d em

an d s th at

p o u r in .

1 2 9

E D

4 .4

3 .0

�3 1 %

L o n g le ad

ti m es

fo r ra d io lo g y an d

la b , ti es

u p ro o m s,

lo n g w ai ts

in E D , u n it s n o t ta ki n g p at ie n ts

N o so lu ti o n s lis te d

S p en t 3 0 m in u te s d ec id in g o n p ri o ri ty

sc o re s w it h n o d is cu ss io n

o f ac ti o n s to

re so lv e th em

9 E D

2 .9

1 .9

�3 3 %

1 3 /2 2 p ro b le m s w er e au d it it em

s b y m an ag er s su ch

as : N o t

w as h in g h an d s,

le av in g

ca b in et

u n lo ck ed

N in e so lu ti o n s w er e to

“e d u ca te

st af f”

N o d at a ab o u t th ei r so lu ti o n ef fo rt s.

6 5

E D

4 .3

2 .0

�5 3 %

P o lic e b ri n g in g in

d an g er o u s

p at ie n ts

w it h o n ly tw o p eo p le

o n at

n ig h t

T al k to

p o lic e d ep ar tm

en t ab o u t p at ie n ts ,

h av e se cu ri ty

ca m er as , an d p an ic b u tt o n s

Y o u ca n n o t fi x th em

al l, b u t yo u h av e to

p ri o ri ti ze . O u r p at ie n t

sa fe ty

co m m it te e w ill en d u p d o in g th at

P yx is

T M is an

au to m at ed

m ed ic at io n -d is p en si n g d ev ic e u se d b y n u rs es

to ad m in is te r m ed ic at io n s to

th ei r p at ie n ts .

preclude taking action. Furthermore, in some of the work areas in the bottom quartile for change in PIP scores, such as hospital 34s OR/PACU and hospital 65s ED, identified issues had to be validated by an external group, such as the patient safety committee, before resolution efforts would be authorized. This additional step substantially slowed the pace of change. Hospital 65s CEO explained his prioritization philosophy, “You can’t fix them all, but you have to prioritize. Our patient safety committee will end up doing that.” However, the safety officer from that hos- pital explained the negative effect this had on staffs’ perceptions, “What happens is you heighten the awareness among people and then, if they don’t see resolutions, then it becomes a bone of contention.”

5. Discussion, Implications, and Limitations

In this study, we investigated the effectiveness of an MBWA-based program in randomly selected hospi- tals. We found evidence that participating in this par- ticular program decreased performance on average. Given that many quality-improvement initiatives fail to achieve expected gains (Beer 2003, Nair 2006, Repenning and Sterman 2002), it is perhaps not sur- prising that our program failed to yield positive results for all work areas. Nonetheless, this is an important result because many hospitals throughout the United States and United Kingdom have imple- mented—and continue to implement—similar pro- grams. Our study provides a cautionary tale that visits by senior managers to the front lines of the organiza- tion to solicit improvement ideas will not necessarily increase staffs’ perceptions of performance improve- ment. There may be negative repercussions if senior managers attempt, but fail, to engage meaningfully with frontline staff. We suspect that the negative conse- quences arose from soliciting, but not sufficiently addressing, frontline staffs’ concerns (Keating et al. 1999, Morrison and Repenning 2011). Failure to meet expectations, once raised, can frustrate employees, negatively impact organizational climate, and dampen employees’ willingness to provide future input (Tucker 2007). Thus, our study suggests that there is a hidden, psychological cost of asking employees for ideas that are subsequently disregarded. To understand why some units had better results

than others, we examined two approaches to problem solving. Solving a higher percentage of the highest- valued problems was not associated with increased PIP. This result is similar to an earlier finding in the TQM literature that formalization could overwhelm actual improvement efforts, leading to employee dis- satisfaction with the program (Mathews and Katel 1992). Conversely, solving a higher percentage of

easy-to-solve problems was successful, lending sup- port for approaches that create a bias toward action. This signals the value in addressing “low-hanging fruit,” at least in the short term (Keating et al. 1999, Morrison and Repenning 2011). Our research does not find that a focus on surfacing and resolving only high-value problems yields improved staff percep- tions. Senior managers can facilitate a bias for action. We

found that having senior managers assume responsi- bility for ensuring that problems get resolved was associated with increased PIP. One explanation for this finding is that organizational change often requires senior managers to provide financial resources to pay for required equipment, materials, or labor; and organizational support to get an upstream department in the organization to change how they do their work if benefits accrue downstream. In other words, senior managers can help ensure that action happens. Given the improvement literature’s empha- sis on empowering frontline employees to solve prob- lems (Powell 1995), our finding may be interpreted as highlighting the importance of empowering frontline employees to identify and solve problems while sup- porting those efforts by ensuring that organizational obstacles to improvement are removed.

5.1. Implications for Theory Manager commitment is associated with successful implementation of performance improvement pro- grams that rely on frontline employee participation (Ahire and O’Shaughnessy 1998, Coronado and Antony 2002, Kaynak 2003, Nair 2006, Worley and Doolen 2006). We found that a program that stimu- lated managerial involvement was productive for some, but not all, work areas. An explanation of the negative result of our MBWA-based program was that asking employees for their suggestions and then not implementing them sent the message that employees’ ideas were not valued and that the pro- gram was symbolic. Research by Miles supports this explanation (1965). He postulated that managers hold one of two beliefs about the value of employee partici- pation programs. One belief was that frontline staff participation was valuable because it increased mor- ale, though the actual ideas they contributed were unhelpful. These managers believed in the symbolic value of employee participation programs, such as MBWA. Miles (1965) found that improvement pro- grams failed when managers held this belief. The sec- ond belief—which was associated with success in Miles’ study—was that interactions with frontline staff were valuable because their ideas were actually useful. The belief in the substantive value of employ- ees’ ideas underlies a core TPS principle: respect for people (Liker 2004). Miles’ study suggests that senior

managers’ respect for frontline employees’ concerns may have been an important but unmeasured moder- ator variable for our MBWA program. An implication is that rather than just seeking to increase manager involvement, it may be critical first to ensure that managers value the ideas raised by frontline staff. An explanation for the lack of positive impact from

the high-value prioritization approach may be that problem values in the hospital work areas in our study had a relatively flat landscape. As a result, pur- suing a high-value prioritization approach did not yield a substantial improvement over focusing on easy-to-solve problems. The flat landscape may be because the work areas had already addressed their large-value problems or because the fragmented ser- vice environment of health care creates a wide range of small-scale problems. The easy-to-solve prioritiza- tion approach may have been successful in our study, because the work areas needed to first tackle funda- mental, lower-value problems before advancing to more complex problems (Keating et al. 1999, Morri- son and Repenning 2011). Taking care of the basic infrastructure and requirements is a necessary precur- sor to more comprehensive organizational change required by higher priority score problems (Keating et al. 1999, Morrison and Repenning 2011). There are likely circumstances under which priori-

tizing high-value problems is helpful, such as when only one idea can be fully developed, like implemen- tation of an enterprise-wide information system. We also believe that organizations benefit from resolving high-value problems, which tend to be top-down, strategic improvements, as well as easy-to-solve prob- lems, which tend to be bottom-up, tactical initiatives. Organizations should try to nurture both kinds of problem-solving capabilities. For example, organiza- tions may have experts working on identifying and solving high-value problems through six-sigma pro- jects, while frontline employees simultaneously work on resolving smaller scale issues in their local work area through lean initiatives. Furthermore, it may be that organizations begin their improvement journey by successfully resolving relatively easy problems, but then need to develop new capabilities to resolve more complex problems (Keating et al. 1999, Morri- son and Repenning 2011). For example, reducing the time required to find vital sign monitor equipment on a nursing unit likely requires different problem-solv- ing skills than reducing patients’ lengths of stay in the hospital.

5.2. Implications for Policy Our study suggests that policy makers can play an important role in improving safety in hospitals by encouraging organizations to build problem-solving capacity. Rather than requiring hospitals to partici-

pate in a specific change program, such as MBWA, that may not be fully validated, policy makers could instead provide incentives for hospitals to build the generic capacity to solve frontline problems. Given the trend toward requiring hospital to implement multiple quality-improvement initiatives concur- rently, we suspect that it is likely that many programs are being implemented superficially and in ways that lead to harmful results similar to those we observed in this study. This could be contributing to the oft- reported failure to achieve gains through improve- ment initiatives that frustrate the health-care industry (Landrigan et al. 2010). Our study provides a warning about mandating implementation of improvement programs before fully understanding the conditions required for the programs to yield successful out- comes. The financial incentives used to encourage adop-

tion of electronic health records in the United States may be instructive. Policy makers rewarded “mean- ingful use,” as demonstrated by the functionality that was achieved, rather than rewarding implementation of a particular software (Blumenthal 2010). Similarly, policy makers could provide incentives for building problem-solving capabilities that improve patient- centered performance rather than advocate for a spe- cific improvement program.

5.3. Implications for Practice Many initiatives to improve safety begin by trying to increase employees’ reports of near misses, errors, and incidents (Bagian et al. 2001, Evans et al. 2007). Implied assumptions are that increasing the number of reports enables organizations to conduct trend analysis that illuminates high-value problems which can then be solved; and that many issues will be of sufficiently low value that they can be ignored at low or no cost to the organization. In contrast, our study suggests that there may be little benefit, and some potential harm, to this approach. Rather than increas- ing reporting, organizations might be better served by addressing known problems, which builds problem- solving capabilities, which in turn enables action- taking on more problems. Our finding corroborates prior research that highlighted the importance of problem-solving capacity for successful improvement programs (Adler et al. 2003, Keating et al. 1999, Mor- rison and Repenning 2011). This advice is consistent with the vision for a continuously learning health-care system articulated by the US Institute of Medicine, requirements for which include systematic problem solving. Our study also resembles Kaizen, a struc- tured problem-solving approach involving managers and frontline workers. However, important differ- ences that may make Kaizen more successful than our program are that Kaizen occurs after managers and

frontline staff have been trained on a standardized problem-solving technique and that it emphasizes taking action to solve as many problems as possible within the given time period (Imai 1986). Thus, it pre- vents resource depletion by limiting the time spent identifying and solving problems rather than by selecting among them.

5.4. Limitations Our findings must be considered in light of study limitations. First, our small sample size limited our analysis. Our sample was small for several reasons. The cost- and time-intensive nature of conducting an experiment with hospitals over 18 months made it challenging to conduct our field-based, intervention- al program with 24 organizations, and we would have struggled if there were more. In addition, despite our providing a method of prioritizing prob- lems, many organizations chose not to assign priori- tization values and therefore work-area coded data on problem value were not available for all treatment work areas. Future research with larger sample sizes could test more nuanced theory. For example, an easy-to-solve prioritization approach may be most successful for work areas that start from a weak position and can benefit most from action, whereas a high-value prioritization approach may be most helpful for experienced work areas that can be more selective. A second limitation is the perceptual measure of

improvement. Hospitals were unwilling to share actual safety incident measures with us. In addition, publicly available clinical measures, such as mortal- ity, readmissions, and process of care measures, started being reported publicly only after the initia- tion of this study. Although we conducted analyses using these “post study” clinical outcome data, the regressions were not significant in explaining varia- tion. However, for reasons detailed above, a percep- tual measure is an important indicator of the impact of the intervention we tested. Furthermore, prior research on an MBWA-based intervention that did have access to clinical outcome data did not find links between multiple clinical outcomes and the interven- tion (Benning et al. 2011), corroborating our study results. Third, hospitals did not track resources spent on

solution efforts. Therefore, estimation was the only way of testing the alternate explanation that spending more money on process improvement yielded better outcomes. Future research could contribute to improvement theory by examining the cost of improvement efforts compared to benefits. A fourth limitation is that we did not randomize an easy-to- solve prioritization approach vs. a high-value prioriti- zation approach among work areas. Instead, those

differences emerged naturally. A randomized assign- ment of these two prioritization approaches would provide a stronger test of the hypotheses.

5.5. Conclusions Understanding the impact of MBWA-based programs is helpful for organizations that may be considering implementing them. In our study, organizations whose managers ensured that problems were addressed achieved better results. This suggests that improvement programs are more likely to change employees’ perceptions when they result in action being taken to resolve problems than when they are a symbolic show of manager interest. On the basis of study findings, we recommend that organizations focus on increasing their capacity to act on improve- ment suggestions rather than expending further effort on generating more suggestions and prioritizing them.

Acknowledgments

Funding was provided by Agency for Healthcare Research and Quality RO1 HSO13920. Additional funding was obtained from Fishman Davidson Center at Wharton. Jenni- fer E. Hayes provided valuable data coding assistance.

Appendix A: Survey Questions for Perceived Improvement in Performance The quality of services I help provide is currently the best it has ever been. We are getting fewer complaints about our work. Overall, the level of patient safety at this facility is improving.

The overall quality of service at this facility is improving.

Appendix B: Interview Questions

B.1. FrontLine Personnel Interview Protocol I wanted to ask you some questions about the patient safety culture at this hospital. We recognize that most hospital personnel experience problems in the course of their work and that these are not a reflection of their skill level or of the quality of care provided at their facility. My goal is to understand differences in safety culture among organizations.

1. Do personnel on this unit talk openly about safety issues and errors?

2. Who (or what) provides impetus for patient safety at this hospital? How do they do it? Reporting of near miss or safety-related inci- dents has received some attention lately. We would like to better understand how health- care professionals that actually provide direct

care to patients think about reporting incidents related to patient safety.

3. Have you ever reported something? What made you decide to report that incident? What happened as a result of reporting? Did you ever learn the outcome? Can you recall a specific adverse event that was caused by an error or series of errors? What happened? Can you describe the investi- gation process (i.e., what happened to people involved, what changes, if any, resulted from the investigation)? My last question relates to a major change in a care process at your hospital.

4. Thinking about a recent major change related to patient care processes, can you describe how this change was introduced in your unit? Probe: What training did you receive? Has implementation required any workarounds of the built-in features of the system?

B.2. Manager Interview Protocol I wanted to ask you some questions about the patient safety culture at this hospital. We recognize that most hospital personnel experience problems in the course of their work and that these are not a reflection of their skill level or of the quality of care provided at their facility. My goal is to under- stand differences in safety culture among organiza- tions.

1. Do you feel comfortable talking about safety issues and errors in your manager meetings with senior leadership?

2. Do you encourage your staff to speak up? How?

3. Who (or what) provides impetus for patient safety at this hospital? How do they do it? Reporting of near miss or safety-related inci- dents has received some attention lately. We would like to better understand how health- care professionals that actually provide direct care to patients think about reporting incidents related to patient safety.

4. Can you step through a recent “near-miss” safety report that you addressed? Briefly (do not need details) what was the situation and what was the response, if any?

5. Can you recall a specific adverse event that was caused by an error or series of errors? Briefly, what happened? Can you describe the investigation process (i.e., what happened to people involved, what changes, if any, resulted from the investigation)? My last question relates to a major change in a care process at your hospital.

6. Thinking about a recent major change related to patient care processes, can you describe how this change was introduced to a unit? Probe: What training was provided? Has implementa- tion required any workarounds of the built-in features of the system?

B.3. Hospital Administrator Interview Protocol I wanted to ask you some questions about your daily activities as a hospital executive and your views on the patient safety culture at your hospital. We recog- nize that leadership styles and organizational cultures are unique at every institution and none is necessarily better than any other. My goal is to understand the full variation among organizations.

1. What are your primary priorities for the hospi- tal? [Prompt if it is not mentioned] Where does patient safety fall in your list of priorities?

2. How do you see your role in patient safety? In what ways do you provide leadership in this area?

3. How would you describe the general attitude of health-care professionals and employees within the hospital toward patient safety?

4. It is well known that middle managers are a key to implementation, and these people are often extremely pressed due to budget con- straints. What is the situation with middle managers in your hospital?

5. How do you obtain information about the haz- ards present at the front lines of your organiza- tion?

6. Thinking about the most recent major organi- zational change related to patient safety, can you describe the change, your decision-making process, and its implementation? Probe: Did some event or new piece of information prompt your decision to implement the change? Did you evaluate the business case before making the change?

References Adler, P. S., P. Riley, S.-W. Kwon, J. Signer, B. Lee, R. Satrasala.

2003. Performance improvement capability: Keys to accelerat- ing performance improvement in hospitals. Calif. Manage. Rev. 45(2): 12–33.

Aflaki, S., P. R. Kleindorfer, V. S. De Miera Polvorinos. 2013. Find- ing and implementing energy efficiency projects in industrial facilities. Prod. Oper. Manag. 22(3): 503–517.

Ahire, S. L., K. C. O’Shaughnessy. 1998. The role of top manage- ment commitment in quality management: An empirical analysis of the auto parts industry. Int. J. Qual. Sci. 3(1): 5–37.

Anderson, E. G., A. Chandrasekaran, A. Davis-Blake, G. G. Par- ker. 2013a. Managing distributed product development pro- jects: Integration strategies for language and geographic

barriers. Working Paper. University of Texas at Austin, Aus- tin, Texas.

Anderson, M. A., E. G. Anderson, G. G. Parker. 2013b. Operations Management for Dummies. John Wiley and Sons Inc, Hoboken, NJ.

Antonakis, J., S. Bendahan, P. Jacquart, R. Lalive. 2010. On mak- ing causal claims: A review and recommendations. Leadersh. Quart. 21(6): 1086–1120.

Atuahene-Gima, K. 2003. The effects of centrifugal and centripe- tal forces on product development speed and quality: How does problem solving matter? Acad. Manage. J. 46(3): 359– 373.

Bagian, J. P., C. Lee, J. Gosbee, J. Derosier, E. Stalhandske, N. Eld- ridge, R. Williams, M. Burkhardt. 2001. Developing and deploying a patient safety program in a large health care delivery system: You can’t fix what you don’t know about. Jt. Comm. J. Qual. Patient Saf. 27(10): 522–532.

Bardhan, I., V. V. Krishnan, S. Lin. 2012. Team dispersion, infor- mation technology, and project performance. Prod. Oper. Manag. 22(6): 1478–1493.

Bates, D. W., D. L. Boyle, M. B. Vander Vleit, J. Schneider, L. L. Leape. 1995. Relationship between medication errors and adverse drug events. J. Gen. Intern. Med. 10: 199–205.

Bates, D. W., J. M. Teich, J. Lee, D. Seger, G. J. Kuperman, N. Ma’Luf, D. Boyle, L. Leape. 1999. The impact of computerized physician order entry on medication error prevention. J. Am. Med. Inform. Assoc. 6(4): 313–321.

Beer, M. 2003. Why total quality management programs do not persist: The role of management quality and implications for leading a TQM transformation. Dec. Sci. 34(4): 623–642.

Benning, A., M. Dixon-Woods, U. Nwulu, M. Ghaleb, J. Dawson, N. Barber, B. D. Franklin, A. Girling, K. Hemming, M. Car- malt, G. Rudge, T. Naicker, A. Kotecha, M. C. Derrington, R. Lilford. 2011. Multiple component patient safety intervention in English hospitals: Controlled evaluation of second phase. BMJ 342: d199.

Bergh, D. D., J. F. Fairbank. 2002. Measuring and testing change in strategic management research. Strateg. Manag. J. 23(4): 359–366.

Bliese, P. D. 2000. Within-group agreement, non-independence, and reliability: Implications for data aggregation and analysis. K. J. Klein, S. W. J. Kozlowski, eds. Multilevel Theory, Research, and Methods in Organizations. Jossey-Bass, San Francisco, CA, 349–381.

Blumenthal, D. D. 2010. Launching HITECH. N. Engl. J. Med. 362(5): 382–385.

Bohn, R. 2000. Stop fighting fires. Harv. Bus. Rev. 78(4): 82–91.

Carroll, J. S., S. Hatakenaka, J. W. Rudolph. 2006. Naturalistic decision making and organizational learning in nuclear power plants: Negotiating meaning between managers and problem investigation teams. Organ. Stud. 27(7): 1037–1057.

Chandrasekaran, A., A. Mishra. 2012. Task design, team context, and psychological safety: An empirical analysis of R&D pro- jects in high technology organizations. Prod. Oper. Manag. 21(6): 977–996.

Chatterjee, S., A. S. Hadi. 1986. Influential observations, high leverage points, and outliers in linear regression. Stat. Sci. 1(3): 379–416.

Cook, R., D. Woods. 1994. Operating at the sharp end: The com- plexity of human error. M. S. Bogner, ed. Human Error in Medicine. Lawrence Erlbaum Associates, Inc., Hillsdale, NJ, 255–310.

Coronado, R. B., J. Antony. 2002. Critical success factors for the successful implementation of six sigma projects in organiza- tions. TQM Mag. 14(2): 92–99.

De Leval, M. R., J. Carthey, D. J. Wright, V. T. Farewell, J. T. Rea- son. 2000. Human factors and cardiac surgery: A multicenter study. J. Thorac. Cardiovasc. Surg. 119(4): 661–672.

Dow, D., D. Samson, S. Ford. 1999. Exploding the myth: Do all quality management practices contribute to superior quality performance? Prod. Oper. Manag. 8(1): 1–27.

Evans, J. R., W. M. Lindsay. 2005. The Management and Control of Quality. Thompson, South-Western, Mason, OH.

Evans, S. M., B. J. Smith, A. Esterman, W. B. Runciman, G. Madd- ern, K. Stead, P. Selim, J. O’Shaughnessy, S. Muecke, S. Jones. 2007. Evaluation of an intervention aimed at improving vol- untary incident reporting in hospitals. Qual. Saf. Health Care 16(3): 169–175.

Fine, C. H. 1986. Quality improvement and learning in productive systems. Manage. Sci. 32(10): 1301–1315.

Fine, C. H., E. L. Porteus. 1989. Dynamic process improvement. Oper. Res. 37(4): 580–591.

Fitzmaurice, G. 2001. A conundrum in the analysis of change. Nutrition 17(4): 360–361.

Flynn, B. B., R. G. Schroeder, S. Sakakibara. 1995. The impact of quality management practices on performance and competi- tive advantage. Dec. Sci. 26(5): 659–691.

Frankel, A. 2004. Patient safety leadership walkrounds. White Paper. Institute for Healthcare Improvement, Boston, MA.

Frankel, A., E. Graydon-Baker, C. Neppl, T. Simmonds, M. Gus- tafson, T. Gandhi. 2003. Patient safety leadership walkrounds. Jt. Comm. J. Qual. Saf. 29(1): 16–26.

Frankel, A., S. P. Grillo, E. G. Baker, C. N. Huber, S. Abookire, M. Grenham, P. Console, M. O’Quinn, G. Thibault, T. K. Gandhi. 2005. Patient safety leadership walkrounds at Partners Health Care: Learning from implementation. Joint Comm. J. Qual. Patient Saf. 31(8): 423–437.

Frankel, A., S. Pratt Grillo, M. Pittman, E. J. Thomas, L. Horowitz, M. Page, B. Sexton. 2008. Revealing and resolving patient safety defects: The impact of leadership walkrounds on front- line caregiver assessments of patient safety. Health Serv. Res. 43(6): 2050–2066.

Girotra, K., C. Terwiesch, K. T. Ulrich. 2010. Idea generation and the quality of the best idea. Manage. Sci. 56(4): 591–605.

Hansen, L., M. Williams, S. J. Singer. 2010. Perceptions of hospital safety climate and incidence of readmission. Health Serv. Res. 42(2): 596–616.

Hofmann, D. A., B. Mark. 2006. An investigation of the relationship between safety climate and medication errors as well as other nurse and patient outcomes. Pers. Psychol. 59(4): 847–869.

Huang, D. T., G. Clermont, L. Kong, L. A. Weissfeld, J. B. Sexton, K. M. Rowan, D. C. Angus. 2010. Intensive care unit safety culture and outcomes: a US multicenter study. Int. J. Qual. Health Care 22(3): 151–161.

Imai, M. 1986. Kaizen: The Key to Japan’s Competitive Success. Ran- dom House Business Division, New York, NY.

Ittner, C. D., V. Nagar, M. V. Rajan. 2001. An empirical examina- tion of dynamic quality-based learning models. Manage. Sci. 47(4): 563.

Iyer, A., H. Saranga, S. Seshadri. 2013. Effect of quality manage- ment systems and total quality management on productivity before and after: Empirical evidence from the Indian auto component industry. Prod. Oper. Manag. 22(2): 283–301.

Jimmerson, C., D. Weber, D. K. Sobek. 2005. Reducing waste and errors: Piloting lean principles at Intermountain Healthcare. Jt. Comm. J. Qual. Patient Saf. 31(5): 249–257.

Johnson, C. W. 2003. How will we get the data and what will we do with it then? Issues in the reporting of adverse healthcare events. Qual. Saf. Health Care 12(suppl. II): ii64–ii67.

Juran, J. M., A. B. Godfrey, R. E. Hoogstoel, E. G. Schilling. 1999. Juran’s Quality Handbook. McGraw Hill, New York, NY.

Kaynak, H. 2003. The relationship between total quality manage- ment practices and their effects on firm performance. J. Oper. Manage. 21(4): 405–435.

Keating, E. K., R. Oliva, N. P. Repenning, S. Rockart, J. D. Ster- man. 1999. Overcoming the improvement paradox. Eur. Man- age. J. 17(2): 120–134.

Ketokivi, M. A., R. G. Schroeder. 2004. Perceptual measures of performance: Fact or fiction? J. Oper. Manage. 22(3): 247–264.

Landis, J. R., G. G. Koch. 1977. The measurement of observer agreement for categorical data. Biometrics 33(1): 159–174.

Landrigan, C. P., G. J. Parry, C. B. Bones, A. D. Hackbarth, D. A. Goldmann, P. J. Sharek. 2010. Temporal trends in rates of patient harm resulting from medical care. N. Engl. J. Med. 363 (22): 2124–2134.

Liker, J. 2004. The Toyota Way. McGraw-Hill, New York, NY.

Macduffie, J. P. 1997. The road to root cause: Shop-floor prob- lem-solving at three auto assembly plants. Manage. Sci. 43(4): 479–502.

Mann, D. 2009. The missing link: Lean leadership. Front. Health Serv. Manag. 26(1): 15–26.

Mathews, J., P. Katel. 1992. The cost of quality. Newsweek, Septem- ber 7, 1992, 48-49.

Mcfadden, K. L., S. C. Henagan, C. R. Gowen III. 2009. The patient safety chain: Transformational leadership’s effect on patient safety culture, initiatives, and outcomes. J. Oper. Manag. 27(5): 390–404.

Miles, R. E.. 1965. Human relations or human resources. Harvard Bus. Rev. 43(4): 148–157.

Miles, M. B., A. M. Huberman. 1994. Qualitative Data Analysis: An expanded Sourcebook. Sage, Thousand Oaks, CA.

Morrison, J. B., N. P. Repenning. 2011. Sustaining employee par- ticipation: The challenge of tipping point dynamics. Working Paper. Brandeis University, Waltham, MA.

Nair, A. 2006. Meta-analysis of the relationship between quality management practices and firm performance–implications for quality management theory development. J. Oper. Manage. 24(6): 948–975.

National Patient Safety Agency. 2011. Patient Safety First: 2008– 2010, The Campaign Review. National Patient Safety Agency, London, UK.

Nembhard, I. M., J. A. Alexander, T. Hoff, R. Ramanujam. 2009. Understanding implementation failure in health care delivery: A role for organizational research and theory. Acad. Manag. Perspect. 23(1): 1–27.

Nunnally, J. C. 1967. Psychometric Theory. McGraw-Hill, New York.

Packard, D. 1995. The HP Way: How Bill Hewlett and I Built our Company. Harper Business, New York, NY.

Perrow, C. 1984. Normal Accidents: Living with high-risk technologies. Princeton University Press, Princeton, NJ.

Powell, T. C. 1995. Total quality management as competitive advantage: A review and empirical study. Strateg. Manag. J. 16(1): 15–37.

Pronovost, P. J., B. Weast, K. Bishop, L. Paine, R. Griffith, B. L. Rosenstein, R. P. Kidwell, K. B. Haller, R. Davis. 2004. Senior

executive: Adopt-a-work unit: A model for safety improve- ment. Jt Comm. J. Qual. Improv. 30(2): 59–68.

Rabe-Hesketh, S., B. Everitt. 2004. A Handbook of Statistical Analysis using Stata. Chapman & Hall/CRC Press, Boca Raton, FL.

Reason, J. 2000. Human error: Models and management. Br. Med. J. 320(7237): 768–770.

Repenning, N. P., J. D. Sterman. 2002. Capability traps and self-confirming attribution errors in the dynamics of process improvement. Admin. Sci. Q. 47(2): 265–295.

Royston, J. P. 1992. Approximating the Shapiro-Wilk W-test for non-normality. Stat. Comput. 2(3): 117–119.

Shortell, S. M., J. L. O’brien, J. M. Carman, R. W. Foster. 1995. Assessing the impact of continuous quality improvement/ total quality management: Concept versus implementation. Health Serv. Res. 30(2): 377.

Singer, S. J., S. Lin, A. Falwell, D. Gaba, L. Baker. 2009. Relation- ship of safety climate and safety performance in hospitals. Health Serv. Res. 44(2 (Part 1)), 399–421.

Singer, S. J., P. E. Rivard, J. E. Hayes, P. Shokeen, D. Gaba, A. Rosen. 2013. Improving patient care through leadership engagement with frontline staff: A department of Veterans affairs case study. Jt. Comm. J. Qual. Patient Saf. 39(8): 349–360.

Sobo, E. J., B. L. Sadler. 2002. Improving organizational communi- cation and cohesion in a health care setting through employee-leadership exchange. Hum. Organ. 61(3): 277–287.

Sommer, S. C., C. H. Loch. 2004. Selectionism and learning in pro- jects with complexity and unforeseeable uncertainty. Manage. Sci. 50(10): 1334–1347.

Swink, M., S. Talluri, T. Pandejpong. 2006. Faster, better, cheaper: A study of NPD project efficiency and performance tradeoffs. J. Oper. Manage. 24(5): 542–562.

Thomas, E. J., J. B. Sexton, T. B. Neilands, A. Frankel, R. L. Helmreich. 2005. The effect of executive walk rounds on nurse safety climate attitudes: A randomized trial of clinical units. BMC Health Serv. Res. 5(1): 28–36.

Toussaint, J., R. A. Gerard, E. Adams. 2010. On the Mend: Revolu- tionizing Healthcare to Save Lives and Transform the Industry. Lean Enterprise Institute, Cambridge, MA.

Tucker, A. L. 2007. An empirical study of system improvement by frontline employees in hospital units. Manuf. Serv. Oper. Manag. 9(4): 492–505.

Victor, B., A. Boynton, T. Stephens-Jahng. 2000. The effective design of work under total quality management. Organ. Sci. 11(1): 102–117.

Von Hippel, E. 1994. “Sticky information” and the locus of prob- lem solving: Implications for innovation. Manage. Sci. 40(4): 429–440.

Womack, J. P. 2011. Gemba Walks. Lean Enterprise Institute, Cam- bridge, MA.

Worley, J. M., T. L. Doolen. 2006. The role of communication and management support in a lean manufacturing implementa- tion. Manage. Dec. 44(2): 228–245.

Zellmer-Bruhn, M. E. 2003. Interruptive events and team knowl- edge acquisition. Manage. Sci. 49(4): 514–528.

Zohar, D., G. Luria. 2003. The use of supervisory practices as leverage to improve safety behaviour: A cross-level interven- tion model. J. Saf. Res. 34(5): 567–577.

Copyright of Production & Operations Management is the property of Wiley-Blackwell and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.