Original Research

Clinical guidelines are prevalent in the field of medicine, but physicians do not consistently provide guideline‐concordant care. Nonadherence to guidelines has been documented for a variety of clinical conditions, including chronic obstructive pulmonary disease,[1, 2] pain management,[3, 4] and major depressive disorder.[5, 6]

Although several professional societies, including the Infectious Diseases Society of America (IDSA), have developed and disseminated guidelines on antibiotic use, adherence to antibiotic‐prescribing guidelines is inconsistent. Several studies have documented inappropriate antibiotic prescribing for specific infections, including acute respiratory infections,[7, 8, 9] cellulitis,[10, 11] and asymptomatic bacteriuria.[12, 13]

Improving adherence to guidelines on antibiotic use could have several benefits. For certain infections, guideline adherence has been shown to improve patient outcomes and reduce resource utilization.[10, 14, 15] In general, guidelines promote more judicious use of antibiotics by clarifying when an antibiotic is indicated, which antibiotics to prescribe, and duration of antibiotic therapy. The more judicious use of antibiotics decreases a given patient's risk of developing an antibiotic‐resistant infection and Clostridium difficileassociated diarrhea.[16] Judicious antibiotic use will also have societal benefits by slowing the spread of antibiotic‐resistant bacteria.

As part of a local effort to improve antibiotic use, we decided to present physicians with hypothetical cases of common clinical scenarios to identify barriers to following antibiotic‐prescribing guidelines. Previous investigators have used case vignettes to assess the quality of care physicians provide, including decisions about antibiotics.[17, 18, 19, 20, 21] We used case vignettes to assess physicians' familiarity with and acceptance of IDSA guidelines for 3 common infectious conditions: skin and soft tissue infections (SSTI), suspected hospital‐acquired pneumonia (HAP), and asymptomatic bacteriuria (ASB). The findings from our project were intended to inform local interventions to improve antibiotic prescribing.

METHODS

All interviews were conducted at 2 acute care hospitals in Indianapolis, Indiana: Sidney and Lois Eskenazi Hospital and the Richard Roudebush Veterans Affairs Medical Center (VAMC). Eskenazi Hospital is a 316‐bed safety‐net hospital for Marion County, Indiana. The Roudebush VAMC is a 209‐bed tertiary care facility that provides comprehensive medical care for 85,000 veterans. Both hospitals are academically affiliated with Indiana University's School of Medicine.

Both hospitals have empiric antibiotic‐prescribing guidelines printed in their annual antibiograms. These guidelines, developed by each hospital's pharmacy department and the local infectious disease (ID) physicians, are distributed annually as a pocket booklet. During this study, an antibiotic stewardship program was active at hospital A but not hospital B. As part of this program at hospital A, an ID physician reviewed inpatients on antibiotics twice a week and, with the help of inpatient team pharmacists, provided feedback to the frontline prescribers.

For this study, inpatient physicians who prescribe antibiotics at either facility were invited to participate in a 30‐minute confidential interview about their antibiotic‐prescribing habits. All invitations were sent through electronic mail. The target enrollment was 30 physicians, which is consistent with prior literature on qualitative sampling.[22] Sampling was purposeful to recruit a heterogeneous group of participants from both hospital sites. Although such a sampling strategy precluded us from making conclusions about individual subgroups, our intention was to obtain the broadest range of information and perspectives, thereby challenging our own preconceived understandings and biases.

The protocol and conduct of this study were reviewed and approved by the Indiana University Institutional Review Board. Participants read and provided signed informed consent. No compensation was provided to physician participants.

A research assistant (A.R.C.) trained in qualitative interviewing conducted all interviews.[23] These interviews covered social norms, perceptions of risk, self‐efficacy, knowledge, and acceptance of guidelines. At the end of the interview, each participant was asked to respond to 3 case vignettes (Table 1), which had been developed by an ID physician (D.L.) based on both local and IDSA guidelines.[24, 25, 26] Participants decided whether to prescribe antibiotics and, if so, which antibiotic to use. After their response, the interviewer read aloud specific recommendations from IDSA guidelines and asked, Would you feel comfortable applying this recommendation to your practice? Are there situations when you would not apply this recommendation?

All interviews were audio recorded, transcribed, and deidentified. All transcripts were reviewed by the study's research assistant (A.R.C.) for accuracy and completeness.

An ID physician (D.L.) reviewed each transcript to determine whether the participant's stated plan for each case vignette was in accordance with IDSA guidelines. Participants were evaluated on their decision to prescribe antibiotics and their choice of agents.

Transcripts were also analyzed using emergent thematic analysis.[27, 28, 29] First, 2 members of the research team (D.L., A.R.C.) reviewed all interview transcripts and discussed general impressions. Next, the analytic team reread one‐fifth of the transcripts, assigning codes to the data line by line. Codes were discussed among team members to determine the most prominent themes. During this phase, codes were added, eliminated, and combined while applying the codes to the remaining transcripts.[30] The analysts then performed focused coding: finalized codes from the first phase were applied to each transcript. The 2 analysts performed focused coding individually on each transcript in a consecutive fashion and met after every 10 transcripts to ensure consistency in their coding for the prior 10 transcripts. Analysts discussed any discrepancies to reach a consensus. Evidence was sought that may call observations and classifications into question.[31] Theoretical saturation was reached through the 30 interviews, so additional enrollment was deemed unnecessary. NVivo version 9 software (QSR International, Cambridge, MA) was used to facilitate all coding and analysis.

RESULTS

All participants were physicians who practiced inpatient medicine. Ten were women, and 20 were men. The median age of participants was 34 years (interquartile range [IQR] 3042). Twenty were attending or staff physicians and had spent a median of 10 years (IQR 315) in clinical practice. Of these attending physicians, 3 practiced pulmonary/critical care, 16 were hospitalists without subspecialty training, and 1 was a hospitalist with ID training. Seven attending physicians practiced exclusively at hospital A, 8 practiced exclusively at hospital B, and 5 practiced at both A and B. The remaining 10 participants were physicians in training or residents, who practiced at both hospitals and were either in their third or fourth year of an internal medicine or medicine/pediatrics residency program.

All participants expressed general awareness of and familiarity with clinical guidelines. Most participants also found guidelines useful in their clinical practice. According to a resident:

Despite general agreement with guidelines in principle, our interviews identified 3 major barriers to following guidelines in practice: (1) lack of awareness of specific guideline recommendations, (2) tension between adhering to guidelines and the desire to individualize patient care, and (3) skepticism of certain guideline recommendations.

DISCUSSION

In this study, we used case vignettes to identify barriers to following IDSA guidelines. Case vignettes require few resources and provide a common starting point for assessing physician decision making. Prior studies have used case vignettes to measure the quality of physicians' practice, including antibiotic prescribing.[17, 18, 19, 20, 21] Case vignettes have been used to assess antibiotic prescribing in the neonatal ICU and medical students' knowledge of upper respiratory tract infections.[21, 32] In 1 study, physicians who scored poorly on a series of case vignettes more frequently prescribed antibiotics inappropriately in actual practice.[17]

Using case vignettes, we identified 3 barriers to following IDSA guidelines on SSTI, HAP, and ASB: (1) lack of awareness of specific guideline recommendations, (2) tension between adhering to guidelines and the desire to individualize patient care, and (3) skepticism of certain guideline recommendations. These barriers were distributed unevenly across participants, highlighting the heterogeneity that exists even within a subgroup of hospital medicine physicians.

We identified lack of familiarity with guideline recommendations as a barrier in our sample of physicians. Interestingly, participants initially expressed agreement with guidelines, but when presented with case vignettes and asked for their own treatment recommendations, it became clear that their familiarity with guidelines was superficial. The disconnect between self‐reported practice and actual adherence has also been described in a separate study on healthcare‐associated pneumonia.[33] In all likelihood, participants genuinely believed that they were practicing guideline‐concordant care, but without a formal process for audit and feedback, their lack of adherence had never been raised as an issue.

A second barrier to guideline‐concordant care was the tension between individualizing patient care and adhering to standardized recommendations. On one hand, this tension is unavoidable and is inherent in the practice of medicine. However, participants' responses to our case vignettes suggested that they find their patients too different to fit into any standardized guideline. This tension was also discussed by Charani et al., who interviewed 39 healthcare professionals at 4 hospitals in the United Kingdom. These investigators found that physicians routinely consider their patients to be outside the recommendations of local evidence‐based policies.[34] Instead of referring to guidelines, physicians rely on their knowledge and clinical experience to guide their antibiotic prescribing.

The final barrier to guideline adherence that we identified was providers' skepticism of what the guidelines were recommending. Although physician discomfort with certain guideline recommendations may be alleviated by reviewing the literature informing the recommendation, education alone is often insufficient to change antibiotic prescribing practices.[35] Furthermore, part of this skepticism may reflect the lack of data from randomized controlled trials to support every guideline recommendation. For example, most guideline recommendations are based on low‐quality evidence.[36] The guideline recommendations presented in this study were based on moderate‐ to high‐quality evidence.[24, 25, 26]

To our knowledge, this study is 1 of the few to describe barriers to guideline‐concordant antibiotic use among inpatient medicine physicians in the United States. The barriers discussed above have also been described by investigators in Europe who studied antibiotic use among inpatient physicians.[34, 37, 38] These commonalities highlight the shared challenges faced by local initiatives to improve antibiotic prescribing.

Our findings suggest that the 2 hospitals we studied need more active interventions to improve antibiotic prescribing. One attractive idea is involving hospitalist physicians in future improvement efforts. Hospitalists are well positioned for this role; they care for a large proportion of hospital patients, they frequently prescribe antibiotics, andas a professionthey are committed to the efficient use of healthcare resources. Hospitalists could assist in the dissemination of local guidelines, the implementation of reliable processes to prompt antibiotic de‐escalation, and the development of local standards for documenting the indication for antibiotics and the planned duration of therapy.[39]

One limitation of this study was that we did not validate whether a physician's self‐reported response to the case vignettes correlated with his or her actual practice. Interviews were conducted by a nonphysician and kept confidential, but participants may nonetheless have been inclined to give socially desirable responses. However, this is less likely because participants readily admitted to not knowing and often not following guidelines. In addition, our case vignettes presented simplistic, hypothetical situations and were therefore less able to account for all determinants of antibiotic‐prescribing decisions. Prior research has shown that antibiotic‐prescribing decisions are influenced by a multitude of factors, including social norms and the physician's underlying beliefs and emotions.[34, 40] Antibiotic‐prescribing decisions can also be influenced by audit and feedback processes.[35] Thus, we acknowledge that our findings may have been different if this study was conducted exclusively at hospitals without an antimicrobial stewardship program.

In conclusion, case vignettes may be a useful tool to assess physician knowledge and acceptance of antibiotic‐prescribing guidelines on a local level. This study used case vignettes to identify key barriers to guideline‐concordant antibiotic use. Developing local interventions to target each of these barriers will be the next step in improving antibiotic prescribing.

Disclosure: This project was supported by a Project Development Team within the ICTSI NIH/NCRR grant number UL1TR001108. The authors report no conflicts of interest.

Clinical guidelines are prevalent in the field of medicine, but physicians do not consistently provide guideline‐concordant care. Nonadherence to guidelines has been documented for a variety of clinical conditions, including chronic obstructive pulmonary disease,[1, 2] pain management,[3, 4] and major depressive disorder.[5, 6]

Although several professional societies, including the Infectious Diseases Society of America (IDSA), have developed and disseminated guidelines on antibiotic use, adherence to antibiotic‐prescribing guidelines is inconsistent. Several studies have documented inappropriate antibiotic prescribing for specific infections, including acute respiratory infections,[7, 8, 9] cellulitis,[10, 11] and asymptomatic bacteriuria.[12, 13]

Improving adherence to guidelines on antibiotic use could have several benefits. For certain infections, guideline adherence has been shown to improve patient outcomes and reduce resource utilization.[10, 14, 15] In general, guidelines promote more judicious use of antibiotics by clarifying when an antibiotic is indicated, which antibiotics to prescribe, and duration of antibiotic therapy. The more judicious use of antibiotics decreases a given patient's risk of developing an antibiotic‐resistant infection and Clostridium difficileassociated diarrhea.[16] Judicious antibiotic use will also have societal benefits by slowing the spread of antibiotic‐resistant bacteria.

As part of a local effort to improve antibiotic use, we decided to present physicians with hypothetical cases of common clinical scenarios to identify barriers to following antibiotic‐prescribing guidelines. Previous investigators have used case vignettes to assess the quality of care physicians provide, including decisions about antibiotics.[17, 18, 19, 20, 21] We used case vignettes to assess physicians' familiarity with and acceptance of IDSA guidelines for 3 common infectious conditions: skin and soft tissue infections (SSTI), suspected hospital‐acquired pneumonia (HAP), and asymptomatic bacteriuria (ASB). The findings from our project were intended to inform local interventions to improve antibiotic prescribing.

METHODS

All interviews were conducted at 2 acute care hospitals in Indianapolis, Indiana: Sidney and Lois Eskenazi Hospital and the Richard Roudebush Veterans Affairs Medical Center (VAMC). Eskenazi Hospital is a 316‐bed safety‐net hospital for Marion County, Indiana. The Roudebush VAMC is a 209‐bed tertiary care facility that provides comprehensive medical care for 85,000 veterans. Both hospitals are academically affiliated with Indiana University's School of Medicine.

Both hospitals have empiric antibiotic‐prescribing guidelines printed in their annual antibiograms. These guidelines, developed by each hospital's pharmacy department and the local infectious disease (ID) physicians, are distributed annually as a pocket booklet. During this study, an antibiotic stewardship program was active at hospital A but not hospital B. As part of this program at hospital A, an ID physician reviewed inpatients on antibiotics twice a week and, with the help of inpatient team pharmacists, provided feedback to the frontline prescribers.

For this study, inpatient physicians who prescribe antibiotics at either facility were invited to participate in a 30‐minute confidential interview about their antibiotic‐prescribing habits. All invitations were sent through electronic mail. The target enrollment was 30 physicians, which is consistent with prior literature on qualitative sampling.[22] Sampling was purposeful to recruit a heterogeneous group of participants from both hospital sites. Although such a sampling strategy precluded us from making conclusions about individual subgroups, our intention was to obtain the broadest range of information and perspectives, thereby challenging our own preconceived understandings and biases.

The protocol and conduct of this study were reviewed and approved by the Indiana University Institutional Review Board. Participants read and provided signed informed consent. No compensation was provided to physician participants.

A research assistant (A.R.C.) trained in qualitative interviewing conducted all interviews.[23] These interviews covered social norms, perceptions of risk, self‐efficacy, knowledge, and acceptance of guidelines. At the end of the interview, each participant was asked to respond to 3 case vignettes (Table 1), which had been developed by an ID physician (D.L.) based on both local and IDSA guidelines.[24, 25, 26] Participants decided whether to prescribe antibiotics and, if so, which antibiotic to use. After their response, the interviewer read aloud specific recommendations from IDSA guidelines and asked, Would you feel comfortable applying this recommendation to your practice? Are there situations when you would not apply this recommendation?

All interviews were audio recorded, transcribed, and deidentified. All transcripts were reviewed by the study's research assistant (A.R.C.) for accuracy and completeness.

An ID physician (D.L.) reviewed each transcript to determine whether the participant's stated plan for each case vignette was in accordance with IDSA guidelines. Participants were evaluated on their decision to prescribe antibiotics and their choice of agents.

Transcripts were also analyzed using emergent thematic analysis.[27, 28, 29] First, 2 members of the research team (D.L., A.R.C.) reviewed all interview transcripts and discussed general impressions. Next, the analytic team reread one‐fifth of the transcripts, assigning codes to the data line by line. Codes were discussed among team members to determine the most prominent themes. During this phase, codes were added, eliminated, and combined while applying the codes to the remaining transcripts.[30] The analysts then performed focused coding: finalized codes from the first phase were applied to each transcript. The 2 analysts performed focused coding individually on each transcript in a consecutive fashion and met after every 10 transcripts to ensure consistency in their coding for the prior 10 transcripts. Analysts discussed any discrepancies to reach a consensus. Evidence was sought that may call observations and classifications into question.[31] Theoretical saturation was reached through the 30 interviews, so additional enrollment was deemed unnecessary. NVivo version 9 software (QSR International, Cambridge, MA) was used to facilitate all coding and analysis.

RESULTS

All participants were physicians who practiced inpatient medicine. Ten were women, and 20 were men. The median age of participants was 34 years (interquartile range [IQR] 3042). Twenty were attending or staff physicians and had spent a median of 10 years (IQR 315) in clinical practice. Of these attending physicians, 3 practiced pulmonary/critical care, 16 were hospitalists without subspecialty training, and 1 was a hospitalist with ID training. Seven attending physicians practiced exclusively at hospital A, 8 practiced exclusively at hospital B, and 5 practiced at both A and B. The remaining 10 participants were physicians in training or residents, who practiced at both hospitals and were either in their third or fourth year of an internal medicine or medicine/pediatrics residency program.

All participants expressed general awareness of and familiarity with clinical guidelines. Most participants also found guidelines useful in their clinical practice. According to a resident:

Despite general agreement with guidelines in principle, our interviews identified 3 major barriers to following guidelines in practice: (1) lack of awareness of specific guideline recommendations, (2) tension between adhering to guidelines and the desire to individualize patient care, and (3) skepticism of certain guideline recommendations.

DISCUSSION

In this study, we used case vignettes to identify barriers to following IDSA guidelines. Case vignettes require few resources and provide a common starting point for assessing physician decision making. Prior studies have used case vignettes to measure the quality of physicians' practice, including antibiotic prescribing.[17, 18, 19, 20, 21] Case vignettes have been used to assess antibiotic prescribing in the neonatal ICU and medical students' knowledge of upper respiratory tract infections.[21, 32] In 1 study, physicians who scored poorly on a series of case vignettes more frequently prescribed antibiotics inappropriately in actual practice.[17]

Using case vignettes, we identified 3 barriers to following IDSA guidelines on SSTI, HAP, and ASB: (1) lack of awareness of specific guideline recommendations, (2) tension between adhering to guidelines and the desire to individualize patient care, and (3) skepticism of certain guideline recommendations. These barriers were distributed unevenly across participants, highlighting the heterogeneity that exists even within a subgroup of hospital medicine physicians.

We identified lack of familiarity with guideline recommendations as a barrier in our sample of physicians. Interestingly, participants initially expressed agreement with guidelines, but when presented with case vignettes and asked for their own treatment recommendations, it became clear that their familiarity with guidelines was superficial. The disconnect between self‐reported practice and actual adherence has also been described in a separate study on healthcare‐associated pneumonia.[33] In all likelihood, participants genuinely believed that they were practicing guideline‐concordant care, but without a formal process for audit and feedback, their lack of adherence had never been raised as an issue.

A second barrier to guideline‐concordant care was the tension between individualizing patient care and adhering to standardized recommendations. On one hand, this tension is unavoidable and is inherent in the practice of medicine. However, participants' responses to our case vignettes suggested that they find their patients too different to fit into any standardized guideline. This tension was also discussed by Charani et al., who interviewed 39 healthcare professionals at 4 hospitals in the United Kingdom. These investigators found that physicians routinely consider their patients to be outside the recommendations of local evidence‐based policies.[34] Instead of referring to guidelines, physicians rely on their knowledge and clinical experience to guide their antibiotic prescribing.

The final barrier to guideline adherence that we identified was providers' skepticism of what the guidelines were recommending. Although physician discomfort with certain guideline recommendations may be alleviated by reviewing the literature informing the recommendation, education alone is often insufficient to change antibiotic prescribing practices.[35] Furthermore, part of this skepticism may reflect the lack of data from randomized controlled trials to support every guideline recommendation. For example, most guideline recommendations are based on low‐quality evidence.[36] The guideline recommendations presented in this study were based on moderate‐ to high‐quality evidence.[24, 25, 26]

To our knowledge, this study is 1 of the few to describe barriers to guideline‐concordant antibiotic use among inpatient medicine physicians in the United States. The barriers discussed above have also been described by investigators in Europe who studied antibiotic use among inpatient physicians.[34, 37, 38] These commonalities highlight the shared challenges faced by local initiatives to improve antibiotic prescribing.

Our findings suggest that the 2 hospitals we studied need more active interventions to improve antibiotic prescribing. One attractive idea is involving hospitalist physicians in future improvement efforts. Hospitalists are well positioned for this role; they care for a large proportion of hospital patients, they frequently prescribe antibiotics, andas a professionthey are committed to the efficient use of healthcare resources. Hospitalists could assist in the dissemination of local guidelines, the implementation of reliable processes to prompt antibiotic de‐escalation, and the development of local standards for documenting the indication for antibiotics and the planned duration of therapy.[39]

One limitation of this study was that we did not validate whether a physician's self‐reported response to the case vignettes correlated with his or her actual practice. Interviews were conducted by a nonphysician and kept confidential, but participants may nonetheless have been inclined to give socially desirable responses. However, this is less likely because participants readily admitted to not knowing and often not following guidelines. In addition, our case vignettes presented simplistic, hypothetical situations and were therefore less able to account for all determinants of antibiotic‐prescribing decisions. Prior research has shown that antibiotic‐prescribing decisions are influenced by a multitude of factors, including social norms and the physician's underlying beliefs and emotions.[34, 40] Antibiotic‐prescribing decisions can also be influenced by audit and feedback processes.[35] Thus, we acknowledge that our findings may have been different if this study was conducted exclusively at hospitals without an antimicrobial stewardship program.

In conclusion, case vignettes may be a useful tool to assess physician knowledge and acceptance of antibiotic‐prescribing guidelines on a local level. This study used case vignettes to identify key barriers to guideline‐concordant antibiotic use. Developing local interventions to target each of these barriers will be the next step in improving antibiotic prescribing.

Disclosure: This project was supported by a Project Development Team within the ICTSI NIH/NCRR grant number UL1TR001108. The authors report no conflicts of interest.

Clinical guidelines are prevalent in the field of medicine, but physicians do not consistently provide guideline‐concordant care. Nonadherence to guidelines has been documented for a variety of clinical conditions, including chronic obstructive pulmonary disease,[1, 2] pain management,[3, 4] and major depressive disorder.[5, 6]

Although several professional societies, including the Infectious Diseases Society of America (IDSA), have developed and disseminated guidelines on antibiotic use, adherence to antibiotic‐prescribing guidelines is inconsistent. Several studies have documented inappropriate antibiotic prescribing for specific infections, including acute respiratory infections,[7, 8, 9] cellulitis,[10, 11] and asymptomatic bacteriuria.[12, 13]

Improving adherence to guidelines on antibiotic use could have several benefits. For certain infections, guideline adherence has been shown to improve patient outcomes and reduce resource utilization.[10, 14, 15] In general, guidelines promote more judicious use of antibiotics by clarifying when an antibiotic is indicated, which antibiotics to prescribe, and duration of antibiotic therapy. The more judicious use of antibiotics decreases a given patient's risk of developing an antibiotic‐resistant infection and Clostridium difficileassociated diarrhea.[16] Judicious antibiotic use will also have societal benefits by slowing the spread of antibiotic‐resistant bacteria.

As part of a local effort to improve antibiotic use, we decided to present physicians with hypothetical cases of common clinical scenarios to identify barriers to following antibiotic‐prescribing guidelines. Previous investigators have used case vignettes to assess the quality of care physicians provide, including decisions about antibiotics.[17, 18, 19, 20, 21] We used case vignettes to assess physicians' familiarity with and acceptance of IDSA guidelines for 3 common infectious conditions: skin and soft tissue infections (SSTI), suspected hospital‐acquired pneumonia (HAP), and asymptomatic bacteriuria (ASB). The findings from our project were intended to inform local interventions to improve antibiotic prescribing.

METHODS

All interviews were conducted at 2 acute care hospitals in Indianapolis, Indiana: Sidney and Lois Eskenazi Hospital and the Richard Roudebush Veterans Affairs Medical Center (VAMC). Eskenazi Hospital is a 316‐bed safety‐net hospital for Marion County, Indiana. The Roudebush VAMC is a 209‐bed tertiary care facility that provides comprehensive medical care for 85,000 veterans. Both hospitals are academically affiliated with Indiana University's School of Medicine.

Both hospitals have empiric antibiotic‐prescribing guidelines printed in their annual antibiograms. These guidelines, developed by each hospital's pharmacy department and the local infectious disease (ID) physicians, are distributed annually as a pocket booklet. During this study, an antibiotic stewardship program was active at hospital A but not hospital B. As part of this program at hospital A, an ID physician reviewed inpatients on antibiotics twice a week and, with the help of inpatient team pharmacists, provided feedback to the frontline prescribers.

For this study, inpatient physicians who prescribe antibiotics at either facility were invited to participate in a 30‐minute confidential interview about their antibiotic‐prescribing habits. All invitations were sent through electronic mail. The target enrollment was 30 physicians, which is consistent with prior literature on qualitative sampling.[22] Sampling was purposeful to recruit a heterogeneous group of participants from both hospital sites. Although such a sampling strategy precluded us from making conclusions about individual subgroups, our intention was to obtain the broadest range of information and perspectives, thereby challenging our own preconceived understandings and biases.

The protocol and conduct of this study were reviewed and approved by the Indiana University Institutional Review Board. Participants read and provided signed informed consent. No compensation was provided to physician participants.

A research assistant (A.R.C.) trained in qualitative interviewing conducted all interviews.[23] These interviews covered social norms, perceptions of risk, self‐efficacy, knowledge, and acceptance of guidelines. At the end of the interview, each participant was asked to respond to 3 case vignettes (Table 1), which had been developed by an ID physician (D.L.) based on both local and IDSA guidelines.[24, 25, 26] Participants decided whether to prescribe antibiotics and, if so, which antibiotic to use. After their response, the interviewer read aloud specific recommendations from IDSA guidelines and asked, Would you feel comfortable applying this recommendation to your practice? Are there situations when you would not apply this recommendation?

All interviews were audio recorded, transcribed, and deidentified. All transcripts were reviewed by the study's research assistant (A.R.C.) for accuracy and completeness.

An ID physician (D.L.) reviewed each transcript to determine whether the participant's stated plan for each case vignette was in accordance with IDSA guidelines. Participants were evaluated on their decision to prescribe antibiotics and their choice of agents.

Transcripts were also analyzed using emergent thematic analysis.[27, 28, 29] First, 2 members of the research team (D.L., A.R.C.) reviewed all interview transcripts and discussed general impressions. Next, the analytic team reread one‐fifth of the transcripts, assigning codes to the data line by line. Codes were discussed among team members to determine the most prominent themes. During this phase, codes were added, eliminated, and combined while applying the codes to the remaining transcripts.[30] The analysts then performed focused coding: finalized codes from the first phase were applied to each transcript. The 2 analysts performed focused coding individually on each transcript in a consecutive fashion and met after every 10 transcripts to ensure consistency in their coding for the prior 10 transcripts. Analysts discussed any discrepancies to reach a consensus. Evidence was sought that may call observations and classifications into question.[31] Theoretical saturation was reached through the 30 interviews, so additional enrollment was deemed unnecessary. NVivo version 9 software (QSR International, Cambridge, MA) was used to facilitate all coding and analysis.

RESULTS

All participants were physicians who practiced inpatient medicine. Ten were women, and 20 were men. The median age of participants was 34 years (interquartile range [IQR] 3042). Twenty were attending or staff physicians and had spent a median of 10 years (IQR 315) in clinical practice. Of these attending physicians, 3 practiced pulmonary/critical care, 16 were hospitalists without subspecialty training, and 1 was a hospitalist with ID training. Seven attending physicians practiced exclusively at hospital A, 8 practiced exclusively at hospital B, and 5 practiced at both A and B. The remaining 10 participants were physicians in training or residents, who practiced at both hospitals and were either in their third or fourth year of an internal medicine or medicine/pediatrics residency program.

All participants expressed general awareness of and familiarity with clinical guidelines. Most participants also found guidelines useful in their clinical practice. According to a resident:

Despite general agreement with guidelines in principle, our interviews identified 3 major barriers to following guidelines in practice: (1) lack of awareness of specific guideline recommendations, (2) tension between adhering to guidelines and the desire to individualize patient care, and (3) skepticism of certain guideline recommendations.

DISCUSSION

In this study, we used case vignettes to identify barriers to following IDSA guidelines. Case vignettes require few resources and provide a common starting point for assessing physician decision making. Prior studies have used case vignettes to measure the quality of physicians' practice, including antibiotic prescribing.[17, 18, 19, 20, 21] Case vignettes have been used to assess antibiotic prescribing in the neonatal ICU and medical students' knowledge of upper respiratory tract infections.[21, 32] In 1 study, physicians who scored poorly on a series of case vignettes more frequently prescribed antibiotics inappropriately in actual practice.[17]

Using case vignettes, we identified 3 barriers to following IDSA guidelines on SSTI, HAP, and ASB: (1) lack of awareness of specific guideline recommendations, (2) tension between adhering to guidelines and the desire to individualize patient care, and (3) skepticism of certain guideline recommendations. These barriers were distributed unevenly across participants, highlighting the heterogeneity that exists even within a subgroup of hospital medicine physicians.

We identified lack of familiarity with guideline recommendations as a barrier in our sample of physicians. Interestingly, participants initially expressed agreement with guidelines, but when presented with case vignettes and asked for their own treatment recommendations, it became clear that their familiarity with guidelines was superficial. The disconnect between self‐reported practice and actual adherence has also been described in a separate study on healthcare‐associated pneumonia.[33] In all likelihood, participants genuinely believed that they were practicing guideline‐concordant care, but without a formal process for audit and feedback, their lack of adherence had never been raised as an issue.

A second barrier to guideline‐concordant care was the tension between individualizing patient care and adhering to standardized recommendations. On one hand, this tension is unavoidable and is inherent in the practice of medicine. However, participants' responses to our case vignettes suggested that they find their patients too different to fit into any standardized guideline. This tension was also discussed by Charani et al., who interviewed 39 healthcare professionals at 4 hospitals in the United Kingdom. These investigators found that physicians routinely consider their patients to be outside the recommendations of local evidence‐based policies.[34] Instead of referring to guidelines, physicians rely on their knowledge and clinical experience to guide their antibiotic prescribing.

The final barrier to guideline adherence that we identified was providers' skepticism of what the guidelines were recommending. Although physician discomfort with certain guideline recommendations may be alleviated by reviewing the literature informing the recommendation, education alone is often insufficient to change antibiotic prescribing practices.[35] Furthermore, part of this skepticism may reflect the lack of data from randomized controlled trials to support every guideline recommendation. For example, most guideline recommendations are based on low‐quality evidence.[36] The guideline recommendations presented in this study were based on moderate‐ to high‐quality evidence.[24, 25, 26]

To our knowledge, this study is 1 of the few to describe barriers to guideline‐concordant antibiotic use among inpatient medicine physicians in the United States. The barriers discussed above have also been described by investigators in Europe who studied antibiotic use among inpatient physicians.[34, 37, 38] These commonalities highlight the shared challenges faced by local initiatives to improve antibiotic prescribing.

Our findings suggest that the 2 hospitals we studied need more active interventions to improve antibiotic prescribing. One attractive idea is involving hospitalist physicians in future improvement efforts. Hospitalists are well positioned for this role; they care for a large proportion of hospital patients, they frequently prescribe antibiotics, andas a professionthey are committed to the efficient use of healthcare resources. Hospitalists could assist in the dissemination of local guidelines, the implementation of reliable processes to prompt antibiotic de‐escalation, and the development of local standards for documenting the indication for antibiotics and the planned duration of therapy.[39]

One limitation of this study was that we did not validate whether a physician's self‐reported response to the case vignettes correlated with his or her actual practice. Interviews were conducted by a nonphysician and kept confidential, but participants may nonetheless have been inclined to give socially desirable responses. However, this is less likely because participants readily admitted to not knowing and often not following guidelines. In addition, our case vignettes presented simplistic, hypothetical situations and were therefore less able to account for all determinants of antibiotic‐prescribing decisions. Prior research has shown that antibiotic‐prescribing decisions are influenced by a multitude of factors, including social norms and the physician's underlying beliefs and emotions.[34, 40] Antibiotic‐prescribing decisions can also be influenced by audit and feedback processes.[35] Thus, we acknowledge that our findings may have been different if this study was conducted exclusively at hospitals without an antimicrobial stewardship program.

In conclusion, case vignettes may be a useful tool to assess physician knowledge and acceptance of antibiotic‐prescribing guidelines on a local level. This study used case vignettes to identify key barriers to guideline‐concordant antibiotic use. Developing local interventions to target each of these barriers will be the next step in improving antibiotic prescribing.

Disclosure: This project was supported by a Project Development Team within the ICTSI NIH/NCRR grant number UL1TR001108. The authors report no conflicts of interest.

Hospital readmissions have a significant impact on the healthcare system. Medicare data suggest a 19% all‐cause 30‐day readmission rate, of which 47% may be preventable.[1, 2] The Centers for Medicare & Medicaid Services continue to expand their criteria of disease states that will be penalized for readmissions, now reducing hospital reimbursement rates up to 3%. Pharmacists, by optimizing patient utilization of medications, can play a valuable role in contributing to preventing readmissions.[3]

Lack of acceptable transitional care is a serious problem that is consistently identified in the literature.[4] Transitional care involves 3 domains of transfer: information, education, and destination. A breakdown in any of these components can negatively impact patients and their caregivers.

Prior studies consistently demonstrated a high likelihood of adverse drug events (ADEs) and patients' lack of knowledge regarding medications postdischarge, both of which can lead to readmission. Forster and colleagues found that 19% to 23% of patients experienced an ADE within 5 weeks of discharge from an inpatient visit, 66% to 72% of which were drug related, and approximately one‐third were deemed preventable.[5, 6] One survey found that less than 60% of patients knew the indication for a new medication prescribed at discharge, whereas only 12% reported knowledge of an anticipated ADE.[7]

Pharmacists can play a large role in the information and education aspect of transitional care. Previous studies demonstrate that pharmacist involvement in the discharge process can reduce the incidence of ADEs and have a positive impact on patient satisfaction. There are conflicting data regarding the effect of comprehensive medication education and follow‐up calls by pharmacy team members on ADEs and medication errors (MEs).[3, 8, 9] Although overall pharmacist participation has shown positive patient‐related outcomes, the impact of pharmacists' involvement on readmissions has not been consistently demonstrated.[10, 11, 12, 13, 14]

Our study evaluated the impact of the pharmacy team in the transitions‐of‐care settings in a unique combination utilizing the pharmacist during medication reconciliation, discharge, and with 3 follow‐up phone call interactions postdischarge. Our study was designed to evaluate the impact of intensive pharmacist involvement during the acute care admission as well as for a 30‐day time period postdischarge on both ADEs and readmissions.

METHODS

All patients were admitted to hospitalist‐based internal medicine units at Northwestern Memorial Hospital, an 894‐bed academic medical center located in Chicago, Illinois. Patients were randomized by study investigators using a random number generator to either the usual care or intervention arms and then evaluated each day for eligibility to participate in the study. Patients remained blinded throughout the study. Patients met inclusion criteria if they were discharged to home and either discharged on greater than 3 scheduled prescription medications or discharged with at least 1 high‐risk medication. High‐risk medications were classified as anticoagulants, antiplatelets (eg, aspirin and clopidogrel), hypoglycemic agents (eg, insulin), immunosuppressants, or anti‐infectives. Patients also needed to participate in a minimum of 1 postdischarge phone call or experience an emergency department (ED) visit or readmission within 30 days of discharge to meet inclusion criteria. Exclusion criteria included: impaired cognition based on Mini‐Cog screening assessment scale, unable or unwilling to provide informed consent, lack of a personal phone number, nonEnglish speaking, subsequent elective readmission within 30 days of initial visit, more than 3 previous hospital admissions in the past 2 months, palliative care or home/skilled nursing hospice, anticipated length of survival less than 3 months, discharged within 24 hours of admission, discharged against medical advice, or discharged before medication education was conducted (Figure 1). Patients who met inclusion criteria provided informed consent, received a Mini‐Cog screening assessment, and were given the Rapid Estimate of Adult Literacy in Medicine revised (REALM‐R) assessment to evaluate health literacy. The REALM‐R is a word recognition test designed to identify patients at risk for poor health literacy skills. Patients with REALM‐R scores of 6 or less are considered to have low health literacy.[15] Patients were randomized to receive either the usual care or pharmacist‐directed medication evaluation and management as described in Table 1. Patients included in the study were contacted by phone postdischarge, with 3 attempts on consecutive days. Patients who were readmitted as an inpatient or had an ED visit were not contacted for the study after that point.

Figure 1

Patients enrolled in the control group received the usual standard of care by a clinical pharmacist. This included a medication reconciliation completed from the admitting physician's patient history and physical and medication counseling provided by the physician or nursing staff at discharge. Patients were not interviewed face‐to‐face on admission and did not receive discharge counseling by a pharmacy team member. Patients were assessed daily by the pharmacist for evaluation of the pharmacotherapy plans and presence of MEs or safety‐related concerns. The control group received 1 postdischarge phone call from a pharmacist at day 30 to assess for study endpoints of ADEs, MEs, ED visit, and readmission only. The endpoints of ADEs and MEs were determined by professional judgment by the clinical pharmacist based on an algorithm similar to National Coordinating Council for Medication Error Reporting and Prevention, although a specific tool was not utilized.

The study group received face‐to‐face medication reconciliation on admission by a pharmacist or a pharmacy student. Prior to discharge, a personalized medication plan was created by the pharmacist and discussed with the physician. Medication discrepancies were addressed prior to the discharge instructions being given and discussed with the patient. Medication counseling was performed at discharge by the pharmacist or pharmacy student. Patients received 3 phone calls at 3, 14, and 30 days postdischarge. The presence of ADEs and MEs were evaluated during each phone call. The patients were asked to confirm their medication regimens including drug, indication, dose, route, and frequency. They were also asked questions regarding possible side effects, new symptoms, and any changes to their current therapy. The calls focused on clarifying the pharmacy discharge plan, resolving any unanswered questions or medication‐related issues, identifying and overcoming any barriers to adherence, and assistance with providing patients access to medications by contacting pharmacies and physicians to resolve and troubleshoot further prescription claims and clarifications. Pharmacists performed all postdischarge phone calls. Pharmacy students were able to provide face‐to‐face medication reconciliation upon admission and discharge counseling under the supervision of the pharmacist for the intervention arm.

The patient Hospital Consumer Assessment of Healthcare Providers and Systems (HCAHPS) responses to the medication domain question, Did you clearly understand the purpose for taking each of your medications at the time of discharge? were collected for the 2 designated hospitalist units for both the control and study groups. HCAHPS scores were collected at the 6 months point prior to the study initiation and throughout the 6‐month study period for the control and intervention groups. A physician and 2 pharmacists, who were blinded to the study randomization and results, assessed all Northwestern Memorial Hospital readmissions to determine if the readmissions were medication‐related or not.

This study obtained institutional review board approval from Northwestern University.

RESULTS

Over the course of 7 months, 341 patients were enrolled in the study, 189 in the control arm and 152 in the study arm. Forty‐eight patients in the control group and 15 patients in the study group were lost to follow‐up. The final analysis included 278 patients, 141 in the control group and 137 in the study group. Patients were eligible for study inclusion if they received at least 1 phone call, which resulted in more patients being lost to follow‐up in the control arm due to fewer total phone call attempts. Demographic and disposition data for the control and study groups are shown in Table 2. Baseline characteristics between the 2 groups were similar with the exception of total medications at time of discharge. The control group had more total medications at discharge compared to the study group (7.2 vs 6.4, P=0.04). The number of high‐risk medications and the number of scheduled medications were similar between both groups. During medication reconciliation, 380 discrepancies (46.2%) were found in the study group compared to 205 (19.9%) in the control group (P<0.0001). The higher number of identified discrepancies in the study group was expected due to the fact that the pharmacist did not complete a face‐to‐face medication history in the control patients. The average length of stay, REALM‐R scores, and reason for admissions were similar between the 2 groups (Table 2).

A total of 55 patients (39%) in the control arm were readmitted to an inpatient hospital or had an ED visit within 30‐days postdischarge compared to 34 patients (24.8%) in the study group (P=0.001) (Table 3). Of the patients readmitted to the ED, 21 were enrolled in the control arm (14.8%) compared to only 6 patients in the study arm (4.4%) (P=0.005). Reviewers concluded that 24% of the control group readmissions were medication‐related versus 23% of the study group (P=1.0). In total, 78 out of 89 readmissions were to Northwestern Memorial Hospital. Medication‐related causes to outside institutions were not evaluated. The causes for all readmissions were not evaluated.

A sensitivity analysis was undertaken to understand the impact of the lost to follow‐up rate in both the control and study groups. Undertaking an assumption that all 15 patients lost to follow‐up in the study group were readmitted and that 15 of 48 patients lost to follow‐up in the control group were readmitted, the intervention continued to show a significant benefit in reduction of composite ED and inpatient readmissions (35.7% study group vs 49.6% control group, P=0.022)

Multivariate logistic regression analysis that controlled for Charlson Comorbidity Index score, length of stay, total number of medications on discharge, and payer type showed an adjusted odds ratio of 0.55 (95% confidence interval [CI]: 0.32‐0.94) in the intervention cohort compared to controls for the combined endpoint of readmission and ED visit within 30‐days postdischarge. The adjusted odds ratio for 30‐day readmission alone was 0.88 (95% CI: 0.49‐1.61).

Eighteen of the 86 control patients who received a 30‐day postdischarge phone call experienced an ADE or ME compared to 11 of the 83 study patients (P=0.22). Patient satisfaction scores of both designated units as represented by the HCAHPS score in the medication knowledge domain increased from the prestudy period. Patients selected agree or strongly agree only 47% of the time at the 6‐month prestudy point compared to 56% of the time during the 6‐month study period.

DISCUSSION

Although previous studies show conflicting results regarding the impact of pharmacist interventions on readmissions, our study demonstrated a decrease in the composite measure of inpatient readmissions and ED visits. Its success stresses the need for a comprehensive approach that contains continuity of care by healthcare providers to reconcile and manage medications throughout the hospital stay, extending up to a full month postdischarge with multiple phone calls. This included (1) face‐to‐face medication reconciliation on admission, (2) development of a personalized medication plan discussed with the patient's physician, (3) addressing any medication discrepancies to the discharge instructions being given to the patient, (4) medication counseling performed at discharge, and (5) 3 postdischarge phone calls at 3, 14, and 30 days.

A study conducted in 2001 analyzed the Medicare Current Beneficiary Survey (MCBS) and found that living alone, having limited education, and lack of self‐management skills have significant associations with early readmission.[16] Approximately 80 million Americans have limited health literacy and are associated with poor health outcomes and healthcare utilization as seen in a review completed by Berkman and colleagues.[17] Because no difference was found between both groups, it would suggest health literacy did not influence or bias the study group. Additionally, no statistically different medication issues, such as total number of medications or rates of ADEs and MEs, were identified in the patients of this study. This may be explained by the small, final population size at the 30‐day period or that the impact of the pharmacist intervention did not reach the threshold that this study was powered to detect. Also, a lack of statistical significance may be due to the subjective nature of ADEs/MEs and the prevention of ADEs/MEs throughout all patients' hospitalizations from the clinical pharmacist's involvement in care, which was not collected. Although a combined endpoint collecting readmission to either the ED or rehospitalization was lower in the intervention cohort, the isolated rehospitalization endpoint was not significantly different between the 2 groups. ED utilization was markedly decreased, but we may have lacked the power to show a statistically significant decrease in rehospitalization. These results mirror those of the Project RED (Re‐Engineered Discharge) intervention.[17]

HCAHPS surveys are sent to only a small percent of randomly selected patients who are discharged from the hospital. Thus, respondents may or may not have been included in the study, indicating a possible greater impact of the intervention on individual patients than collected. Importantly, the described interventions appeared to improve patients' perception of understanding the purpose of their medications. We found that HCAHPS scores across the 2 units improved, though the intervention only impacted 16.8% of all patients discharged from these units due to the nature of the survey distribution.

The pharmacists' abilities to educate all eligible patients prior to discharge from 7:30 _am to 4:00 _pm each day of the week was a limitation of this study, as some patients were discharged outside of the duty hours. This may have allowed for a differential exclusion and could have led to selection bias. Another limitation is that a large number of patients were lost to follow‐up in the control group, likely because the first postdischarge contact with patients was not until the day 30 phone calls. The extensive exclusion criteria caused many patients not to be enrolled. Though the intervention arm received postdischarge phone calls at days 3 and 14, only postdischarge call‐backs at day 30 of the intervention arm were compared to the control arm, which could have led to bias in the 30‐day analysis of the intervention arm, as patients may have not reported previous issues that were resolved in earlier phone calls. Medication‐related readmissions were not statistically different between the groups, which could suggest that the difference in readmissions were not solely due to the intervention, and a decrease in healthcare utilization may be due to chance. The subjective nature of how ADEs and MEs were collected also serves as a limitation, as they were only screened for presence or absence and not classified by severity or category. This study was at a single‐center academic institution, which may limit the ability to apply the results to other institutions. Last, outcome assessments relied on participant report, including ADE and ME occurrence and presentation at outside hospitals. Future study evaluation conducted as a multicenter design while continuing to strengthen the continuity of the healthcare provider and patient relationship at each intervention would be ideal. Also, having an objective measure of ADEs and MEs with severity categorization would be beneficial.

Compared to previous literature, our study design was unique in the number of phone calls made to patients postdischarge and its prospective, randomized design. In the previously mentioned study by Walker et al., phone calls were made only at days 3 and 30.[13] Although the majority of readmissions occurred within the first 14 days of discharge, additional visits to the ED and readmissions may have been avoided by contacting patients twice within the critical 14‐day period. Another distinction of this study design was the expansion of a rather limited and peripheral pharmacist role in transitions of care to a much more integrated participation. We believe the relationship developed between patients and their pharmacy care team provided coordination and the continuity of communication regarding their care. Additionally, our study was unique through the use of pharmacy extenders via fourth‐year pharmacy students who were completing their advanced pharmacy practice rotations. Pharmacy extenders can also be certified and trained pharmacy technicians, which many hospitals utilize to perform medication reconciliations at a lower cost than pharmacists. As hospitals face increased demands to shrink budgets due to decreasing reimbursements, healthcare systems will be forced to find creative new ways to use existing resources.

In conclusion, transition of care is a high‐risk situation for many patients. A comprehensive approach by healthcare providers, including pharmacists and pharmacy extenders, may have a positive impact in reducing or preventing ADEs/MEs, inpatient admissions, and ED visits. Although our study focused directly on the impact of a pharmacy care team on transitions‐of‐care, we cannot conclude this applies strictly to pharmacists. Across the nation, the role of various disciplines of healthcare providers in admission, hospitalization, discharge, and postdischarge is not standardized and varies significantly by institution. Importantly, no mechanism currently exists to directly reimburse for such efforts, but demonstration of cost effectiveness through reduced posthospital utilization may justify this investment for accountable care organizations.[18]

Hospital readmissions have a significant impact on the healthcare system. Medicare data suggest a 19% all‐cause 30‐day readmission rate, of which 47% may be preventable.[1, 2] The Centers for Medicare & Medicaid Services continue to expand their criteria of disease states that will be penalized for readmissions, now reducing hospital reimbursement rates up to 3%. Pharmacists, by optimizing patient utilization of medications, can play a valuable role in contributing to preventing readmissions.[3]

Lack of acceptable transitional care is a serious problem that is consistently identified in the literature.[4] Transitional care involves 3 domains of transfer: information, education, and destination. A breakdown in any of these components can negatively impact patients and their caregivers.

Prior studies consistently demonstrated a high likelihood of adverse drug events (ADEs) and patients' lack of knowledge regarding medications postdischarge, both of which can lead to readmission. Forster and colleagues found that 19% to 23% of patients experienced an ADE within 5 weeks of discharge from an inpatient visit, 66% to 72% of which were drug related, and approximately one‐third were deemed preventable.[5, 6] One survey found that less than 60% of patients knew the indication for a new medication prescribed at discharge, whereas only 12% reported knowledge of an anticipated ADE.[7]

Pharmacists can play a large role in the information and education aspect of transitional care. Previous studies demonstrate that pharmacist involvement in the discharge process can reduce the incidence of ADEs and have a positive impact on patient satisfaction. There are conflicting data regarding the effect of comprehensive medication education and follow‐up calls by pharmacy team members on ADEs and medication errors (MEs).[3, 8, 9] Although overall pharmacist participation has shown positive patient‐related outcomes, the impact of pharmacists' involvement on readmissions has not been consistently demonstrated.[10, 11, 12, 13, 14]

Our study evaluated the impact of the pharmacy team in the transitions‐of‐care settings in a unique combination utilizing the pharmacist during medication reconciliation, discharge, and with 3 follow‐up phone call interactions postdischarge. Our study was designed to evaluate the impact of intensive pharmacist involvement during the acute care admission as well as for a 30‐day time period postdischarge on both ADEs and readmissions.

METHODS

All patients were admitted to hospitalist‐based internal medicine units at Northwestern Memorial Hospital, an 894‐bed academic medical center located in Chicago, Illinois. Patients were randomized by study investigators using a random number generator to either the usual care or intervention arms and then evaluated each day for eligibility to participate in the study. Patients remained blinded throughout the study. Patients met inclusion criteria if they were discharged to home and either discharged on greater than 3 scheduled prescription medications or discharged with at least 1 high‐risk medication. High‐risk medications were classified as anticoagulants, antiplatelets (eg, aspirin and clopidogrel), hypoglycemic agents (eg, insulin), immunosuppressants, or anti‐infectives. Patients also needed to participate in a minimum of 1 postdischarge phone call or experience an emergency department (ED) visit or readmission within 30 days of discharge to meet inclusion criteria. Exclusion criteria included: impaired cognition based on Mini‐Cog screening assessment scale, unable or unwilling to provide informed consent, lack of a personal phone number, nonEnglish speaking, subsequent elective readmission within 30 days of initial visit, more than 3 previous hospital admissions in the past 2 months, palliative care or home/skilled nursing hospice, anticipated length of survival less than 3 months, discharged within 24 hours of admission, discharged against medical advice, or discharged before medication education was conducted (Figure 1). Patients who met inclusion criteria provided informed consent, received a Mini‐Cog screening assessment, and were given the Rapid Estimate of Adult Literacy in Medicine revised (REALM‐R) assessment to evaluate health literacy. The REALM‐R is a word recognition test designed to identify patients at risk for poor health literacy skills. Patients with REALM‐R scores of 6 or less are considered to have low health literacy.[15] Patients were randomized to receive either the usual care or pharmacist‐directed medication evaluation and management as described in Table 1. Patients included in the study were contacted by phone postdischarge, with 3 attempts on consecutive days. Patients who were readmitted as an inpatient or had an ED visit were not contacted for the study after that point.

Figure 1

Patients enrolled in the control group received the usual standard of care by a clinical pharmacist. This included a medication reconciliation completed from the admitting physician's patient history and physical and medication counseling provided by the physician or nursing staff at discharge. Patients were not interviewed face‐to‐face on admission and did not receive discharge counseling by a pharmacy team member. Patients were assessed daily by the pharmacist for evaluation of the pharmacotherapy plans and presence of MEs or safety‐related concerns. The control group received 1 postdischarge phone call from a pharmacist at day 30 to assess for study endpoints of ADEs, MEs, ED visit, and readmission only. The endpoints of ADEs and MEs were determined by professional judgment by the clinical pharmacist based on an algorithm similar to National Coordinating Council for Medication Error Reporting and Prevention, although a specific tool was not utilized.

The study group received face‐to‐face medication reconciliation on admission by a pharmacist or a pharmacy student. Prior to discharge, a personalized medication plan was created by the pharmacist and discussed with the physician. Medication discrepancies were addressed prior to the discharge instructions being given and discussed with the patient. Medication counseling was performed at discharge by the pharmacist or pharmacy student. Patients received 3 phone calls at 3, 14, and 30 days postdischarge. The presence of ADEs and MEs were evaluated during each phone call. The patients were asked to confirm their medication regimens including drug, indication, dose, route, and frequency. They were also asked questions regarding possible side effects, new symptoms, and any changes to their current therapy. The calls focused on clarifying the pharmacy discharge plan, resolving any unanswered questions or medication‐related issues, identifying and overcoming any barriers to adherence, and assistance with providing patients access to medications by contacting pharmacies and physicians to resolve and troubleshoot further prescription claims and clarifications. Pharmacists performed all postdischarge phone calls. Pharmacy students were able to provide face‐to‐face medication reconciliation upon admission and discharge counseling under the supervision of the pharmacist for the intervention arm.

The patient Hospital Consumer Assessment of Healthcare Providers and Systems (HCAHPS) responses to the medication domain question, Did you clearly understand the purpose for taking each of your medications at the time of discharge? were collected for the 2 designated hospitalist units for both the control and study groups. HCAHPS scores were collected at the 6 months point prior to the study initiation and throughout the 6‐month study period for the control and intervention groups. A physician and 2 pharmacists, who were blinded to the study randomization and results, assessed all Northwestern Memorial Hospital readmissions to determine if the readmissions were medication‐related or not.

This study obtained institutional review board approval from Northwestern University.

RESULTS

Over the course of 7 months, 341 patients were enrolled in the study, 189 in the control arm and 152 in the study arm. Forty‐eight patients in the control group and 15 patients in the study group were lost to follow‐up. The final analysis included 278 patients, 141 in the control group and 137 in the study group. Patients were eligible for study inclusion if they received at least 1 phone call, which resulted in more patients being lost to follow‐up in the control arm due to fewer total phone call attempts. Demographic and disposition data for the control and study groups are shown in Table 2. Baseline characteristics between the 2 groups were similar with the exception of total medications at time of discharge. The control group had more total medications at discharge compared to the study group (7.2 vs 6.4, P=0.04). The number of high‐risk medications and the number of scheduled medications were similar between both groups. During medication reconciliation, 380 discrepancies (46.2%) were found in the study group compared to 205 (19.9%) in the control group (P<0.0001). The higher number of identified discrepancies in the study group was expected due to the fact that the pharmacist did not complete a face‐to‐face medication history in the control patients. The average length of stay, REALM‐R scores, and reason for admissions were similar between the 2 groups (Table 2).

A total of 55 patients (39%) in the control arm were readmitted to an inpatient hospital or had an ED visit within 30‐days postdischarge compared to 34 patients (24.8%) in the study group (P=0.001) (Table 3). Of the patients readmitted to the ED, 21 were enrolled in the control arm (14.8%) compared to only 6 patients in the study arm (4.4%) (P=0.005). Reviewers concluded that 24% of the control group readmissions were medication‐related versus 23% of the study group (P=1.0). In total, 78 out of 89 readmissions were to Northwestern Memorial Hospital. Medication‐related causes to outside institutions were not evaluated. The causes for all readmissions were not evaluated.

A sensitivity analysis was undertaken to understand the impact of the lost to follow‐up rate in both the control and study groups. Undertaking an assumption that all 15 patients lost to follow‐up in the study group were readmitted and that 15 of 48 patients lost to follow‐up in the control group were readmitted, the intervention continued to show a significant benefit in reduction of composite ED and inpatient readmissions (35.7% study group vs 49.6% control group, P=0.022)

Multivariate logistic regression analysis that controlled for Charlson Comorbidity Index score, length of stay, total number of medications on discharge, and payer type showed an adjusted odds ratio of 0.55 (95% confidence interval [CI]: 0.32‐0.94) in the intervention cohort compared to controls for the combined endpoint of readmission and ED visit within 30‐days postdischarge. The adjusted odds ratio for 30‐day readmission alone was 0.88 (95% CI: 0.49‐1.61).

Eighteen of the 86 control patients who received a 30‐day postdischarge phone call experienced an ADE or ME compared to 11 of the 83 study patients (P=0.22). Patient satisfaction scores of both designated units as represented by the HCAHPS score in the medication knowledge domain increased from the prestudy period. Patients selected agree or strongly agree only 47% of the time at the 6‐month prestudy point compared to 56% of the time during the 6‐month study period.

DISCUSSION

Although previous studies show conflicting results regarding the impact of pharmacist interventions on readmissions, our study demonstrated a decrease in the composite measure of inpatient readmissions and ED visits. Its success stresses the need for a comprehensive approach that contains continuity of care by healthcare providers to reconcile and manage medications throughout the hospital stay, extending up to a full month postdischarge with multiple phone calls. This included (1) face‐to‐face medication reconciliation on admission, (2) development of a personalized medication plan discussed with the patient's physician, (3) addressing any medication discrepancies to the discharge instructions being given to the patient, (4) medication counseling performed at discharge, and (5) 3 postdischarge phone calls at 3, 14, and 30 days.

A study conducted in 2001 analyzed the Medicare Current Beneficiary Survey (MCBS) and found that living alone, having limited education, and lack of self‐management skills have significant associations with early readmission.[16] Approximately 80 million Americans have limited health literacy and are associated with poor health outcomes and healthcare utilization as seen in a review completed by Berkman and colleagues.[17] Because no difference was found between both groups, it would suggest health literacy did not influence or bias the study group. Additionally, no statistically different medication issues, such as total number of medications or rates of ADEs and MEs, were identified in the patients of this study. This may be explained by the small, final population size at the 30‐day period or that the impact of the pharmacist intervention did not reach the threshold that this study was powered to detect. Also, a lack of statistical significance may be due to the subjective nature of ADEs/MEs and the prevention of ADEs/MEs throughout all patients' hospitalizations from the clinical pharmacist's involvement in care, which was not collected. Although a combined endpoint collecting readmission to either the ED or rehospitalization was lower in the intervention cohort, the isolated rehospitalization endpoint was not significantly different between the 2 groups. ED utilization was markedly decreased, but we may have lacked the power to show a statistically significant decrease in rehospitalization. These results mirror those of the Project RED (Re‐Engineered Discharge) intervention.[17]

HCAHPS surveys are sent to only a small percent of randomly selected patients who are discharged from the hospital. Thus, respondents may or may not have been included in the study, indicating a possible greater impact of the intervention on individual patients than collected. Importantly, the described interventions appeared to improve patients' perception of understanding the purpose of their medications. We found that HCAHPS scores across the 2 units improved, though the intervention only impacted 16.8% of all patients discharged from these units due to the nature of the survey distribution.

The pharmacists' abilities to educate all eligible patients prior to discharge from 7:30 _am to 4:00 _pm each day of the week was a limitation of this study, as some patients were discharged outside of the duty hours. This may have allowed for a differential exclusion and could have led to selection bias. Another limitation is that a large number of patients were lost to follow‐up in the control group, likely because the first postdischarge contact with patients was not until the day 30 phone calls. The extensive exclusion criteria caused many patients not to be enrolled. Though the intervention arm received postdischarge phone calls at days 3 and 14, only postdischarge call‐backs at day 30 of the intervention arm were compared to the control arm, which could have led to bias in the 30‐day analysis of the intervention arm, as patients may have not reported previous issues that were resolved in earlier phone calls. Medication‐related readmissions were not statistically different between the groups, which could suggest that the difference in readmissions were not solely due to the intervention, and a decrease in healthcare utilization may be due to chance. The subjective nature of how ADEs and MEs were collected also serves as a limitation, as they were only screened for presence or absence and not classified by severity or category. This study was at a single‐center academic institution, which may limit the ability to apply the results to other institutions. Last, outcome assessments relied on participant report, including ADE and ME occurrence and presentation at outside hospitals. Future study evaluation conducted as a multicenter design while continuing to strengthen the continuity of the healthcare provider and patient relationship at each intervention would be ideal. Also, having an objective measure of ADEs and MEs with severity categorization would be beneficial.

Compared to previous literature, our study design was unique in the number of phone calls made to patients postdischarge and its prospective, randomized design. In the previously mentioned study by Walker et al., phone calls were made only at days 3 and 30.[13] Although the majority of readmissions occurred within the first 14 days of discharge, additional visits to the ED and readmissions may have been avoided by contacting patients twice within the critical 14‐day period. Another distinction of this study design was the expansion of a rather limited and peripheral pharmacist role in transitions of care to a much more integrated participation. We believe the relationship developed between patients and their pharmacy care team provided coordination and the continuity of communication regarding their care. Additionally, our study was unique through the use of pharmacy extenders via fourth‐year pharmacy students who were completing their advanced pharmacy practice rotations. Pharmacy extenders can also be certified and trained pharmacy technicians, which many hospitals utilize to perform medication reconciliations at a lower cost than pharmacists. As hospitals face increased demands to shrink budgets due to decreasing reimbursements, healthcare systems will be forced to find creative new ways to use existing resources.

In conclusion, transition of care is a high‐risk situation for many patients. A comprehensive approach by healthcare providers, including pharmacists and pharmacy extenders, may have a positive impact in reducing or preventing ADEs/MEs, inpatient admissions, and ED visits. Although our study focused directly on the impact of a pharmacy care team on transitions‐of‐care, we cannot conclude this applies strictly to pharmacists. Across the nation, the role of various disciplines of healthcare providers in admission, hospitalization, discharge, and postdischarge is not standardized and varies significantly by institution. Importantly, no mechanism currently exists to directly reimburse for such efforts, but demonstration of cost effectiveness through reduced posthospital utilization may justify this investment for accountable care organizations.[18]

Hospital readmissions have a significant impact on the healthcare system. Medicare data suggest a 19% all‐cause 30‐day readmission rate, of which 47% may be preventable.[1, 2] The Centers for Medicare & Medicaid Services continue to expand their criteria of disease states that will be penalized for readmissions, now reducing hospital reimbursement rates up to 3%. Pharmacists, by optimizing patient utilization of medications, can play a valuable role in contributing to preventing readmissions.[3]

Lack of acceptable transitional care is a serious problem that is consistently identified in the literature.[4] Transitional care involves 3 domains of transfer: information, education, and destination. A breakdown in any of these components can negatively impact patients and their caregivers.

Prior studies consistently demonstrated a high likelihood of adverse drug events (ADEs) and patients' lack of knowledge regarding medications postdischarge, both of which can lead to readmission. Forster and colleagues found that 19% to 23% of patients experienced an ADE within 5 weeks of discharge from an inpatient visit, 66% to 72% of which were drug related, and approximately one‐third were deemed preventable.[5, 6] One survey found that less than 60% of patients knew the indication for a new medication prescribed at discharge, whereas only 12% reported knowledge of an anticipated ADE.[7]

Pharmacists can play a large role in the information and education aspect of transitional care. Previous studies demonstrate that pharmacist involvement in the discharge process can reduce the incidence of ADEs and have a positive impact on patient satisfaction. There are conflicting data regarding the effect of comprehensive medication education and follow‐up calls by pharmacy team members on ADEs and medication errors (MEs).[3, 8, 9] Although overall pharmacist participation has shown positive patient‐related outcomes, the impact of pharmacists' involvement on readmissions has not been consistently demonstrated.[10, 11, 12, 13, 14]

Our study evaluated the impact of the pharmacy team in the transitions‐of‐care settings in a unique combination utilizing the pharmacist during medication reconciliation, discharge, and with 3 follow‐up phone call interactions postdischarge. Our study was designed to evaluate the impact of intensive pharmacist involvement during the acute care admission as well as for a 30‐day time period postdischarge on both ADEs and readmissions.

METHODS

All patients were admitted to hospitalist‐based internal medicine units at Northwestern Memorial Hospital, an 894‐bed academic medical center located in Chicago, Illinois. Patients were randomized by study investigators using a random number generator to either the usual care or intervention arms and then evaluated each day for eligibility to participate in the study. Patients remained blinded throughout the study. Patients met inclusion criteria if they were discharged to home and either discharged on greater than 3 scheduled prescription medications or discharged with at least 1 high‐risk medication. High‐risk medications were classified as anticoagulants, antiplatelets (eg, aspirin and clopidogrel), hypoglycemic agents (eg, insulin), immunosuppressants, or anti‐infectives. Patients also needed to participate in a minimum of 1 postdischarge phone call or experience an emergency department (ED) visit or readmission within 30 days of discharge to meet inclusion criteria. Exclusion criteria included: impaired cognition based on Mini‐Cog screening assessment scale, unable or unwilling to provide informed consent, lack of a personal phone number, nonEnglish speaking, subsequent elective readmission within 30 days of initial visit, more than 3 previous hospital admissions in the past 2 months, palliative care or home/skilled nursing hospice, anticipated length of survival less than 3 months, discharged within 24 hours of admission, discharged against medical advice, or discharged before medication education was conducted (Figure 1). Patients who met inclusion criteria provided informed consent, received a Mini‐Cog screening assessment, and were given the Rapid Estimate of Adult Literacy in Medicine revised (REALM‐R) assessment to evaluate health literacy. The REALM‐R is a word recognition test designed to identify patients at risk for poor health literacy skills. Patients with REALM‐R scores of 6 or less are considered to have low health literacy.[15] Patients were randomized to receive either the usual care or pharmacist‐directed medication evaluation and management as described in Table 1. Patients included in the study were contacted by phone postdischarge, with 3 attempts on consecutive days. Patients who were readmitted as an inpatient or had an ED visit were not contacted for the study after that point.

Figure 1

Patients enrolled in the control group received the usual standard of care by a clinical pharmacist. This included a medication reconciliation completed from the admitting physician's patient history and physical and medication counseling provided by the physician or nursing staff at discharge. Patients were not interviewed face‐to‐face on admission and did not receive discharge counseling by a pharmacy team member. Patients were assessed daily by the pharmacist for evaluation of the pharmacotherapy plans and presence of MEs or safety‐related concerns. The control group received 1 postdischarge phone call from a pharmacist at day 30 to assess for study endpoints of ADEs, MEs, ED visit, and readmission only. The endpoints of ADEs and MEs were determined by professional judgment by the clinical pharmacist based on an algorithm similar to National Coordinating Council for Medication Error Reporting and Prevention, although a specific tool was not utilized.

The study group received face‐to‐face medication reconciliation on admission by a pharmacist or a pharmacy student. Prior to discharge, a personalized medication plan was created by the pharmacist and discussed with the physician. Medication discrepancies were addressed prior to the discharge instructions being given and discussed with the patient. Medication counseling was performed at discharge by the pharmacist or pharmacy student. Patients received 3 phone calls at 3, 14, and 30 days postdischarge. The presence of ADEs and MEs were evaluated during each phone call. The patients were asked to confirm their medication regimens including drug, indication, dose, route, and frequency. They were also asked questions regarding possible side effects, new symptoms, and any changes to their current therapy. The calls focused on clarifying the pharmacy discharge plan, resolving any unanswered questions or medication‐related issues, identifying and overcoming any barriers to adherence, and assistance with providing patients access to medications by contacting pharmacies and physicians to resolve and troubleshoot further prescription claims and clarifications. Pharmacists performed all postdischarge phone calls. Pharmacy students were able to provide face‐to‐face medication reconciliation upon admission and discharge counseling under the supervision of the pharmacist for the intervention arm.

The patient Hospital Consumer Assessment of Healthcare Providers and Systems (HCAHPS) responses to the medication domain question, Did you clearly understand the purpose for taking each of your medications at the time of discharge? were collected for the 2 designated hospitalist units for both the control and study groups. HCAHPS scores were collected at the 6 months point prior to the study initiation and throughout the 6‐month study period for the control and intervention groups. A physician and 2 pharmacists, who were blinded to the study randomization and results, assessed all Northwestern Memorial Hospital readmissions to determine if the readmissions were medication‐related or not.

This study obtained institutional review board approval from Northwestern University.

RESULTS

Over the course of 7 months, 341 patients were enrolled in the study, 189 in the control arm and 152 in the study arm. Forty‐eight patients in the control group and 15 patients in the study group were lost to follow‐up. The final analysis included 278 patients, 141 in the control group and 137 in the study group. Patients were eligible for study inclusion if they received at least 1 phone call, which resulted in more patients being lost to follow‐up in the control arm due to fewer total phone call attempts. Demographic and disposition data for the control and study groups are shown in Table 2. Baseline characteristics between the 2 groups were similar with the exception of total medications at time of discharge. The control group had more total medications at discharge compared to the study group (7.2 vs 6.4, P=0.04). The number of high‐risk medications and the number of scheduled medications were similar between both groups. During medication reconciliation, 380 discrepancies (46.2%) were found in the study group compared to 205 (19.9%) in the control group (P<0.0001). The higher number of identified discrepancies in the study group was expected due to the fact that the pharmacist did not complete a face‐to‐face medication history in the control patients. The average length of stay, REALM‐R scores, and reason for admissions were similar between the 2 groups (Table 2).

A total of 55 patients (39%) in the control arm were readmitted to an inpatient hospital or had an ED visit within 30‐days postdischarge compared to 34 patients (24.8%) in the study group (P=0.001) (Table 3). Of the patients readmitted to the ED, 21 were enrolled in the control arm (14.8%) compared to only 6 patients in the study arm (4.4%) (P=0.005). Reviewers concluded that 24% of the control group readmissions were medication‐related versus 23% of the study group (P=1.0). In total, 78 out of 89 readmissions were to Northwestern Memorial Hospital. Medication‐related causes to outside institutions were not evaluated. The causes for all readmissions were not evaluated.

A sensitivity analysis was undertaken to understand the impact of the lost to follow‐up rate in both the control and study groups. Undertaking an assumption that all 15 patients lost to follow‐up in the study group were readmitted and that 15 of 48 patients lost to follow‐up in the control group were readmitted, the intervention continued to show a significant benefit in reduction of composite ED and inpatient readmissions (35.7% study group vs 49.6% control group, P=0.022)

Multivariate logistic regression analysis that controlled for Charlson Comorbidity Index score, length of stay, total number of medications on discharge, and payer type showed an adjusted odds ratio of 0.55 (95% confidence interval [CI]: 0.32‐0.94) in the intervention cohort compared to controls for the combined endpoint of readmission and ED visit within 30‐days postdischarge. The adjusted odds ratio for 30‐day readmission alone was 0.88 (95% CI: 0.49‐1.61).

Eighteen of the 86 control patients who received a 30‐day postdischarge phone call experienced an ADE or ME compared to 11 of the 83 study patients (P=0.22). Patient satisfaction scores of both designated units as represented by the HCAHPS score in the medication knowledge domain increased from the prestudy period. Patients selected agree or strongly agree only 47% of the time at the 6‐month prestudy point compared to 56% of the time during the 6‐month study period.

DISCUSSION

Although previous studies show conflicting results regarding the impact of pharmacist interventions on readmissions, our study demonstrated a decrease in the composite measure of inpatient readmissions and ED visits. Its success stresses the need for a comprehensive approach that contains continuity of care by healthcare providers to reconcile and manage medications throughout the hospital stay, extending up to a full month postdischarge with multiple phone calls. This included (1) face‐to‐face medication reconciliation on admission, (2) development of a personalized medication plan discussed with the patient's physician, (3) addressing any medication discrepancies to the discharge instructions being given to the patient, (4) medication counseling performed at discharge, and (5) 3 postdischarge phone calls at 3, 14, and 30 days.

A study conducted in 2001 analyzed the Medicare Current Beneficiary Survey (MCBS) and found that living alone, having limited education, and lack of self‐management skills have significant associations with early readmission.[16] Approximately 80 million Americans have limited health literacy and are associated with poor health outcomes and healthcare utilization as seen in a review completed by Berkman and colleagues.[17] Because no difference was found between both groups, it would suggest health literacy did not influence or bias the study group. Additionally, no statistically different medication issues, such as total number of medications or rates of ADEs and MEs, were identified in the patients of this study. This may be explained by the small, final population size at the 30‐day period or that the impact of the pharmacist intervention did not reach the threshold that this study was powered to detect. Also, a lack of statistical significance may be due to the subjective nature of ADEs/MEs and the prevention of ADEs/MEs throughout all patients' hospitalizations from the clinical pharmacist's involvement in care, which was not collected. Although a combined endpoint collecting readmission to either the ED or rehospitalization was lower in the intervention cohort, the isolated rehospitalization endpoint was not significantly different between the 2 groups. ED utilization was markedly decreased, but we may have lacked the power to show a statistically significant decrease in rehospitalization. These results mirror those of the Project RED (Re‐Engineered Discharge) intervention.[17]

HCAHPS surveys are sent to only a small percent of randomly selected patients who are discharged from the hospital. Thus, respondents may or may not have been included in the study, indicating a possible greater impact of the intervention on individual patients than collected. Importantly, the described interventions appeared to improve patients' perception of understanding the purpose of their medications. We found that HCAHPS scores across the 2 units improved, though the intervention only impacted 16.8% of all patients discharged from these units due to the nature of the survey distribution.

The pharmacists' abilities to educate all eligible patients prior to discharge from 7:30 _am to 4:00 _pm each day of the week was a limitation of this study, as some patients were discharged outside of the duty hours. This may have allowed for a differential exclusion and could have led to selection bias. Another limitation is that a large number of patients were lost to follow‐up in the control group, likely because the first postdischarge contact with patients was not until the day 30 phone calls. The extensive exclusion criteria caused many patients not to be enrolled. Though the intervention arm received postdischarge phone calls at days 3 and 14, only postdischarge call‐backs at day 30 of the intervention arm were compared to the control arm, which could have led to bias in the 30‐day analysis of the intervention arm, as patients may have not reported previous issues that were resolved in earlier phone calls. Medication‐related readmissions were not statistically different between the groups, which could suggest that the difference in readmissions were not solely due to the intervention, and a decrease in healthcare utilization may be due to chance. The subjective nature of how ADEs and MEs were collected also serves as a limitation, as they were only screened for presence or absence and not classified by severity or category. This study was at a single‐center academic institution, which may limit the ability to apply the results to other institutions. Last, outcome assessments relied on participant report, including ADE and ME occurrence and presentation at outside hospitals. Future study evaluation conducted as a multicenter design while continuing to strengthen the continuity of the healthcare provider and patient relationship at each intervention would be ideal. Also, having an objective measure of ADEs and MEs with severity categorization would be beneficial.

Compared to previous literature, our study design was unique in the number of phone calls made to patients postdischarge and its prospective, randomized design. In the previously mentioned study by Walker et al., phone calls were made only at days 3 and 30.[13] Although the majority of readmissions occurred within the first 14 days of discharge, additional visits to the ED and readmissions may have been avoided by contacting patients twice within the critical 14‐day period. Another distinction of this study design was the expansion of a rather limited and peripheral pharmacist role in transitions of care to a much more integrated participation. We believe the relationship developed between patients and their pharmacy care team provided coordination and the continuity of communication regarding their care. Additionally, our study was unique through the use of pharmacy extenders via fourth‐year pharmacy students who were completing their advanced pharmacy practice rotations. Pharmacy extenders can also be certified and trained pharmacy technicians, which many hospitals utilize to perform medication reconciliations at a lower cost than pharmacists. As hospitals face increased demands to shrink budgets due to decreasing reimbursements, healthcare systems will be forced to find creative new ways to use existing resources.

In conclusion, transition of care is a high‐risk situation for many patients. A comprehensive approach by healthcare providers, including pharmacists and pharmacy extenders, may have a positive impact in reducing or preventing ADEs/MEs, inpatient admissions, and ED visits. Although our study focused directly on the impact of a pharmacy care team on transitions‐of‐care, we cannot conclude this applies strictly to pharmacists. Across the nation, the role of various disciplines of healthcare providers in admission, hospitalization, discharge, and postdischarge is not standardized and varies significantly by institution. Importantly, no mechanism currently exists to directly reimburse for such efforts, but demonstration of cost effectiveness through reduced posthospital utilization may justify this investment for accountable care organizations.[18]

Healthcare‐associated infections are a major cause of illness and death in hospitalized patients, and preventing healthcare‐associated infection is a global challenge.[1] Worldwide, the prevalence of healthcare‐associated infections in developed and undeveloped countries ranges from 5.1% to 11.6% and 5.7% to 19.1%, respectively.[2] In the United States, roughly 2 million such infections occur annually, resulting in approximately 99,000 deaths[3] and estimated annual direct medical costs between $28.4 and $33.8 billion.[4] In Japan, nearly 9% of patients admitted to the intensive care unit (ICU) develop an infection during hospitalization,[5] and 5% of all patients hospitalized become infected with methicillin‐resistant Staphylococcus aureus.[6] The management of healthcare‐associated infections in Japan accounts for up to 5% of total annual healthcare costs, with an estimated $6.8 billion estimated to be potentially preventable.[7] In addition, healthcare‐associated infections are associated with increased length of stay in the hospital. Studies estimate surgical site infections extend length of stay by 9.7 days,[8] and bloodstream infections increase length of stay by 10 days.[9]

Improving hand hygiene practice for healthcare workers is considered a core strategy to decrease the incidence of healthcare‐associated infection.[6, 10] Specifically, the use of alcohol‐based hand rub is strongly recommended in acute care hospitals by both the World Health Organization (WHO) and the US Centers for Disease Control and Prevention.[11, 12] Improving hand hygiene adherence may reduce healthcare‐associated infection by 9% to 50%,[13, 14] and multiple studies have reported that greater use of alcohol‐based hand rubs results in significant reductions in healthcare‐associated infections.[14, 15]

Due to the difficulty in improving hand hygiene in various settings across the world, the WHO strategy for improving hand hygiene has been adopted and implemented by several studies in varying locations, such as Costa Rica, Italy, Mali, Pakistan, and Saudi Arabia.[16] Implementations of these multimodal strategies, following WHObased guidelines, have been shown to increase the level of hand hygiene adherence among healthcare workers and reduce infections at these locations.[14, 17, 18] This study expands upon that work by extending the same implementation strategy to assess the effectiveness of the introduction of alcohol‐based hand rub on hand hygiene practice at multiple hospitals in Japan.

In a previous article[19] we reported results from an observational study assessing healthcare worker hand hygiene adherence before touching the patient in 4 geographically diverse hospitals in Japan. The study reported that hand hygiene adherence in Japanese hospitals was lower than reported mean values from other international studies, and that greater adherence to hand hygiene should be encouraged. In this article, we present the results of a multimodal intervention intended to improve levels of healthcare worker hand hygiene in 3 of these hospitals.

METHODS

RESULTS

Data were collected from May 2013 to July 2013 in hospital A, in October 2012 in hospital B, and June 2013 in hospital C to ensure data were collected after the 6‐month intervention at each site. A total of 2982 observations of hand hygiene were performed in 10 distinct units across the 3 participating hospitals during the postintervention periods. Hand hygiene observations were performed during the day Monday through Friday between 8:30 _am and 7:30 _pm, with the majority occurring prior to 1:00 _pm.

The overall postintervention hand hygiene adherence rate (in all 3 hospitals) was significantly higher at 32.7% (974/2982) adherence compared to 18.0% (482/2679) adherence in the preintervention period (P<0.001). An increased hand hygiene adherence rate in each participating hospital in the postintervention period was observed (Figure 1). Similar trends of higher overall hand hygiene adherence rates for both nurses and physicians in the postintervention period were seen. Use of alcohol‐based hand rub among those with appropriate hand hygiene was significantly higher, with 90.0% (880/974) using hand rub in the postintervention period versus 67.0% (322/482) in the preintervention period (P<0.001). Comparison of overall hand hygiene adherence rates by unit type and healthcare worker subgroup between the pre‐ and postintervention periods are shown in Table 4. Detailed comparisons of hand hygiene adherence rates for each hospital are available in the supplementary appendix. Although a significant improvement of hand hygiene practice was observed in the majority of participating units (6/10), there was a significant decline in hand hygiene practice in 2 units for nurses and 1 unit for physicians. Hand hygiene adherence rates by healthcare worker subgroups (both physicians and nurses) were significantly higher in the postintervention period than those in the preintervention period. Trends toward higher hand hygiene adherence rate of nurses in the postintervention period were observed (34.8% adherence for nurses compared to 30.4% adherence for physicians); the difference between nurses and physicians were not statistically significant (P=0.07).

Figure 1

Hospital A achieved the highest postintervention adherence rates (39.9% adherence postintervention), as well as the greatest absolute improvement in hand hygiene (increase of 29.0%). There were significant improvements in 3 of the 4 participating units in hospital A, with the emergency department showing improvements only in the nurse subgroup. In hospital B, total hand hygiene adherence increased from 24.7% to 30.0% (P=0.01); however, this increase was mainly due to increase in hand hygiene adherence rates of nurses. There were significant increases in hand hygiene adherence rates for nurses in the medicine (+11%, P=0.04) and surgery wards (+14%, P=0.01), with nonsignificant increases for physicians (+10% medicine, P=0.07;+2% surgery, P=0.78). However, in the emergency department, nurses showed no significant improvement, and physicians had a significant decrease in adherence (15.7% preintervention vs 7.4% postintervention; P=0.02). In hospital C, total hand hygiene practice rates were significantly improved (from 18.9% to 26.5%; P<0.001); however, this was driven by improvements only in the surgical ward (14.6% preintervention to 42.3% postintervention; P<0.001). The rates for nurses declined significantly in both the medicine and ICU wards, leading to no observed improvements on those wards.

DISCUSSION

Our multicenter intervention study in Japan included observations from almost 3000 encounters between clinicians and patients. Before the intervention, the overall rate of hand hygiene adherence was 18%. After the multimodal intervention, the absolute increase in healthcare worker hand hygiene adherence was 15%. Although there was overall improvement, the adherence rates varied by hospital, with hospital A increasing by 29% and hospital B and C only attaining increases of 5% and 7%, respectively.

Despite the importance of hand hygiene of healthcare workers, it is challenging to increase hand hygiene adherence because it requires behavioral modification. Moreover, it remains uncertain what factors will affect healthcare worker behavior. We implemented pragmatic strategies to evaluate the efficacy of hand hygiene multimodal interventions based on internationally recognized WHO hand hygiene adherence strategies[11] and an institutional‐level contest with financial incentives. The findings in the current study help us understand not only how a multimodal intervention importantly improves hand hygiene adherence, but also what factors potentially make healthcare workers modify their behaviors.

In this study, we evaluated whether an institutional‐level contest with financial incentives contributed to improved hand hygiene adherence of healthcare workers. This study demonstrated improvement of hand hygiene practice after implementation of a multimodal hand hygiene intervention combined with an institutional‐level contest with financial incentives. The contest might have had a modest effect to help motivate the participating hospitals to improve their hand hygiene adherence rate. This is consistent with a previous study that demonstrated financial incentives were associated with modifying healthcare workers' hand hygiene practice.[21] However, we did not strictly standardize how the contest information was distributed in each participating institution and the objective assessment for changes in motivation by the contest was lacking in this study. Thus, changes in motivation by the contest with financial incentives likely varied by each participating institution. Further studies are needed to assess if this type of approach is worth pursuing.

We observed several noteworthy associations between the intervention components that were implemented at each facility and their improvement in hand hygiene adherence. Among the participating hospitals, hospital A was most successful with improving hand hygiene adherence, although all participating hospitals achieved a similar number of the 15 recommended intervention components during the intervention (8 to 10 per hospital). Interestingly, hospital A initiated the most new components during the intervention period (8 new components for a total of 10 out of 15), whereas hospital B and hospital C initiated only 1 or 2 new components during the intervention period. Hospital A also successfully involved hospital executives, and elicited the commitment of a nurse manager and physician leader. Consistent with a previous study,[22] we believe that involvement of hospital executives appears to be important to increase overall hand hygiene rate among healthcare workers.

In contrast, hospitals B and C did not involve senior executives or identify nurse or physician champions for all participating units. Based on the results in this study, we believe that the involvement of hospital executives is likely a key for the penetration of hospital‐wide hand hygiene culture among healthcare workers.

Although this study was unable to determine which components are precisely associated with improving hand hygiene adherence, the findings suggest initiating multiple intervention components at the same time may provide more motivation for change than initiating only 1 or 2 components at a time. It is also possible that certain intervention components were more beneficial than others. For example, hospital A, which achieved the most success, was the only hospital to obtain leadership support. Other studies have demonstrated that the presence of leadership appeared to play a key role in improving hand hygiene adherence.[23, 24] Moreover, a recent Japanese nationwide survey demonstrated higher safety centeredness was associated with regular use of standard infection prevention practice.[25] Consistent with a previous study, improving hand hygiene adherence cannot be simply achieved by improving infrastructure (eg, introduction of portable alcohol‐based hand rub) alone, but it depends on altering healthcare worker behavior.[26]

This study has several limitations. Because participating hospitals could tailor the specific interventions chosen for their facility, the improvement in hand hygiene adherence was likely multifactorial. We are unable in the existing study to determine a direct causal relationship between any of the individual intervention components and hand hygiene adherence. We are also unable to determine whether the improvements seen in hospital A were due to participation in the contest or due to the specific intervention components that were implemented. However, WHO hand hygiene guidelines point out that recognition of the importance of hand hygiene varies in different regions and countries, and the goal for hand hygiene interventions is to establish a culture of hand hygiene practice through pragmatic intervention strategies, frequent evaluation, and feedback to healthcare workers.[27] Thus, we prioritized pragmatic strategies to include in our intervention to promote hand hygiene adherence. Another limitation was the date of implementation of the multimodal intervention was slightly different at each facility. It was challenging to implement the intervention simultaneously across institutions due to competing priorities at each facility. Although the primary goal of hand hygiene is to reduce the burden of healthcare‐associated infection, we were unable to measure infection rates at the participating facilities. It is possible the presence of an external observer had an impact on the healthcare workers' behavior.[28] However, the healthcare workers were not informed as to what the observer was monitoring to minimize this potential effect. Lastly, the findings in this study provide immediate intervention effects but further study will be required to determine if these effects are sustainable.

Altering healthcare worker behavior is likely the key element to improve hand hygiene adherence, and behavioral modification may be achieved with the support of leadership at the unit and facility level. However, even though we found significant improvements in healthcare worker hand hygiene adherence after the intervention, the adherence rates are still relatively low compared to reported adherence rates from other countries,[29] suggesting further intervention is needed in this setting to optimize and hygiene practice. Because hand hygiene practice is a crucial strategy to prevent healthcare‐associated infections, every effort should be made to enhance the hand hygiene practice of healthcare workers.

Healthcare‐associated infections are a major cause of illness and death in hospitalized patients, and preventing healthcare‐associated infection is a global challenge.[1] Worldwide, the prevalence of healthcare‐associated infections in developed and undeveloped countries ranges from 5.1% to 11.6% and 5.7% to 19.1%, respectively.[2] In the United States, roughly 2 million such infections occur annually, resulting in approximately 99,000 deaths[3] and estimated annual direct medical costs between $28.4 and $33.8 billion.[4] In Japan, nearly 9% of patients admitted to the intensive care unit (ICU) develop an infection during hospitalization,[5] and 5% of all patients hospitalized become infected with methicillin‐resistant Staphylococcus aureus.[6] The management of healthcare‐associated infections in Japan accounts for up to 5% of total annual healthcare costs, with an estimated $6.8 billion estimated to be potentially preventable.[7] In addition, healthcare‐associated infections are associated with increased length of stay in the hospital. Studies estimate surgical site infections extend length of stay by 9.7 days,[8] and bloodstream infections increase length of stay by 10 days.[9]

Improving hand hygiene practice for healthcare workers is considered a core strategy to decrease the incidence of healthcare‐associated infection.[6, 10] Specifically, the use of alcohol‐based hand rub is strongly recommended in acute care hospitals by both the World Health Organization (WHO) and the US Centers for Disease Control and Prevention.[11, 12] Improving hand hygiene adherence may reduce healthcare‐associated infection by 9% to 50%,[13, 14] and multiple studies have reported that greater use of alcohol‐based hand rubs results in significant reductions in healthcare‐associated infections.[14, 15]

Due to the difficulty in improving hand hygiene in various settings across the world, the WHO strategy for improving hand hygiene has been adopted and implemented by several studies in varying locations, such as Costa Rica, Italy, Mali, Pakistan, and Saudi Arabia.[16] Implementations of these multimodal strategies, following WHObased guidelines, have been shown to increase the level of hand hygiene adherence among healthcare workers and reduce infections at these locations.[14, 17, 18] This study expands upon that work by extending the same implementation strategy to assess the effectiveness of the introduction of alcohol‐based hand rub on hand hygiene practice at multiple hospitals in Japan.

In a previous article[19] we reported results from an observational study assessing healthcare worker hand hygiene adherence before touching the patient in 4 geographically diverse hospitals in Japan. The study reported that hand hygiene adherence in Japanese hospitals was lower than reported mean values from other international studies, and that greater adherence to hand hygiene should be encouraged. In this article, we present the results of a multimodal intervention intended to improve levels of healthcare worker hand hygiene in 3 of these hospitals.

METHODS

RESULTS

Data were collected from May 2013 to July 2013 in hospital A, in October 2012 in hospital B, and June 2013 in hospital C to ensure data were collected after the 6‐month intervention at each site. A total of 2982 observations of hand hygiene were performed in 10 distinct units across the 3 participating hospitals during the postintervention periods. Hand hygiene observations were performed during the day Monday through Friday between 8:30 _am and 7:30 _pm, with the majority occurring prior to 1:00 _pm.

The overall postintervention hand hygiene adherence rate (in all 3 hospitals) was significantly higher at 32.7% (974/2982) adherence compared to 18.0% (482/2679) adherence in the preintervention period (P<0.001). An increased hand hygiene adherence rate in each participating hospital in the postintervention period was observed (Figure 1). Similar trends of higher overall hand hygiene adherence rates for both nurses and physicians in the postintervention period were seen. Use of alcohol‐based hand rub among those with appropriate hand hygiene was significantly higher, with 90.0% (880/974) using hand rub in the postintervention period versus 67.0% (322/482) in the preintervention period (P<0.001). Comparison of overall hand hygiene adherence rates by unit type and healthcare worker subgroup between the pre‐ and postintervention periods are shown in Table 4. Detailed comparisons of hand hygiene adherence rates for each hospital are available in the supplementary appendix. Although a significant improvement of hand hygiene practice was observed in the majority of participating units (6/10), there was a significant decline in hand hygiene practice in 2 units for nurses and 1 unit for physicians. Hand hygiene adherence rates by healthcare worker subgroups (both physicians and nurses) were significantly higher in the postintervention period than those in the preintervention period. Trends toward higher hand hygiene adherence rate of nurses in the postintervention period were observed (34.8% adherence for nurses compared to 30.4% adherence for physicians); the difference between nurses and physicians were not statistically significant (P=0.07).

Figure 1

Hospital A achieved the highest postintervention adherence rates (39.9% adherence postintervention), as well as the greatest absolute improvement in hand hygiene (increase of 29.0%). There were significant improvements in 3 of the 4 participating units in hospital A, with the emergency department showing improvements only in the nurse subgroup. In hospital B, total hand hygiene adherence increased from 24.7% to 30.0% (P=0.01); however, this increase was mainly due to increase in hand hygiene adherence rates of nurses. There were significant increases in hand hygiene adherence rates for nurses in the medicine (+11%, P=0.04) and surgery wards (+14%, P=0.01), with nonsignificant increases for physicians (+10% medicine, P=0.07;+2% surgery, P=0.78). However, in the emergency department, nurses showed no significant improvement, and physicians had a significant decrease in adherence (15.7% preintervention vs 7.4% postintervention; P=0.02). In hospital C, total hand hygiene practice rates were significantly improved (from 18.9% to 26.5%; P<0.001); however, this was driven by improvements only in the surgical ward (14.6% preintervention to 42.3% postintervention; P<0.001). The rates for nurses declined significantly in both the medicine and ICU wards, leading to no observed improvements on those wards.

DISCUSSION

Our multicenter intervention study in Japan included observations from almost 3000 encounters between clinicians and patients. Before the intervention, the overall rate of hand hygiene adherence was 18%. After the multimodal intervention, the absolute increase in healthcare worker hand hygiene adherence was 15%. Although there was overall improvement, the adherence rates varied by hospital, with hospital A increasing by 29% and hospital B and C only attaining increases of 5% and 7%, respectively.

Despite the importance of hand hygiene of healthcare workers, it is challenging to increase hand hygiene adherence because it requires behavioral modification. Moreover, it remains uncertain what factors will affect healthcare worker behavior. We implemented pragmatic strategies to evaluate the efficacy of hand hygiene multimodal interventions based on internationally recognized WHO hand hygiene adherence strategies[11] and an institutional‐level contest with financial incentives. The findings in the current study help us understand not only how a multimodal intervention importantly improves hand hygiene adherence, but also what factors potentially make healthcare workers modify their behaviors.

In this study, we evaluated whether an institutional‐level contest with financial incentives contributed to improved hand hygiene adherence of healthcare workers. This study demonstrated improvement of hand hygiene practice after implementation of a multimodal hand hygiene intervention combined with an institutional‐level contest with financial incentives. The contest might have had a modest effect to help motivate the participating hospitals to improve their hand hygiene adherence rate. This is consistent with a previous study that demonstrated financial incentives were associated with modifying healthcare workers' hand hygiene practice.[21] However, we did not strictly standardize how the contest information was distributed in each participating institution and the objective assessment for changes in motivation by the contest was lacking in this study. Thus, changes in motivation by the contest with financial incentives likely varied by each participating institution. Further studies are needed to assess if this type of approach is worth pursuing.

We observed several noteworthy associations between the intervention components that were implemented at each facility and their improvement in hand hygiene adherence. Among the participating hospitals, hospital A was most successful with improving hand hygiene adherence, although all participating hospitals achieved a similar number of the 15 recommended intervention components during the intervention (8 to 10 per hospital). Interestingly, hospital A initiated the most new components during the intervention period (8 new components for a total of 10 out of 15), whereas hospital B and hospital C initiated only 1 or 2 new components during the intervention period. Hospital A also successfully involved hospital executives, and elicited the commitment of a nurse manager and physician leader. Consistent with a previous study,[22] we believe that involvement of hospital executives appears to be important to increase overall hand hygiene rate among healthcare workers.

In contrast, hospitals B and C did not involve senior executives or identify nurse or physician champions for all participating units. Based on the results in this study, we believe that the involvement of hospital executives is likely a key for the penetration of hospital‐wide hand hygiene culture among healthcare workers.

Although this study was unable to determine which components are precisely associated with improving hand hygiene adherence, the findings suggest initiating multiple intervention components at the same time may provide more motivation for change than initiating only 1 or 2 components at a time. It is also possible that certain intervention components were more beneficial than others. For example, hospital A, which achieved the most success, was the only hospital to obtain leadership support. Other studies have demonstrated that the presence of leadership appeared to play a key role in improving hand hygiene adherence.[23, 24] Moreover, a recent Japanese nationwide survey demonstrated higher safety centeredness was associated with regular use of standard infection prevention practice.[25] Consistent with a previous study, improving hand hygiene adherence cannot be simply achieved by improving infrastructure (eg, introduction of portable alcohol‐based hand rub) alone, but it depends on altering healthcare worker behavior.[26]

This study has several limitations. Because participating hospitals could tailor the specific interventions chosen for their facility, the improvement in hand hygiene adherence was likely multifactorial. We are unable in the existing study to determine a direct causal relationship between any of the individual intervention components and hand hygiene adherence. We are also unable to determine whether the improvements seen in hospital A were due to participation in the contest or due to the specific intervention components that were implemented. However, WHO hand hygiene guidelines point out that recognition of the importance of hand hygiene varies in different regions and countries, and the goal for hand hygiene interventions is to establish a culture of hand hygiene practice through pragmatic intervention strategies, frequent evaluation, and feedback to healthcare workers.[27] Thus, we prioritized pragmatic strategies to include in our intervention to promote hand hygiene adherence. Another limitation was the date of implementation of the multimodal intervention was slightly different at each facility. It was challenging to implement the intervention simultaneously across institutions due to competing priorities at each facility. Although the primary goal of hand hygiene is to reduce the burden of healthcare‐associated infection, we were unable to measure infection rates at the participating facilities. It is possible the presence of an external observer had an impact on the healthcare workers' behavior.[28] However, the healthcare workers were not informed as to what the observer was monitoring to minimize this potential effect. Lastly, the findings in this study provide immediate intervention effects but further study will be required to determine if these effects are sustainable.

Altering healthcare worker behavior is likely the key element to improve hand hygiene adherence, and behavioral modification may be achieved with the support of leadership at the unit and facility level. However, even though we found significant improvements in healthcare worker hand hygiene adherence after the intervention, the adherence rates are still relatively low compared to reported adherence rates from other countries,[29] suggesting further intervention is needed in this setting to optimize and hygiene practice. Because hand hygiene practice is a crucial strategy to prevent healthcare‐associated infections, every effort should be made to enhance the hand hygiene practice of healthcare workers.

Healthcare‐associated infections are a major cause of illness and death in hospitalized patients, and preventing healthcare‐associated infection is a global challenge.[1] Worldwide, the prevalence of healthcare‐associated infections in developed and undeveloped countries ranges from 5.1% to 11.6% and 5.7% to 19.1%, respectively.[2] In the United States, roughly 2 million such infections occur annually, resulting in approximately 99,000 deaths[3] and estimated annual direct medical costs between $28.4 and $33.8 billion.[4] In Japan, nearly 9% of patients admitted to the intensive care unit (ICU) develop an infection during hospitalization,[5] and 5% of all patients hospitalized become infected with methicillin‐resistant Staphylococcus aureus.[6] The management of healthcare‐associated infections in Japan accounts for up to 5% of total annual healthcare costs, with an estimated $6.8 billion estimated to be potentially preventable.[7] In addition, healthcare‐associated infections are associated with increased length of stay in the hospital. Studies estimate surgical site infections extend length of stay by 9.7 days,[8] and bloodstream infections increase length of stay by 10 days.[9]

Improving hand hygiene practice for healthcare workers is considered a core strategy to decrease the incidence of healthcare‐associated infection.[6, 10] Specifically, the use of alcohol‐based hand rub is strongly recommended in acute care hospitals by both the World Health Organization (WHO) and the US Centers for Disease Control and Prevention.[11, 12] Improving hand hygiene adherence may reduce healthcare‐associated infection by 9% to 50%,[13, 14] and multiple studies have reported that greater use of alcohol‐based hand rubs results in significant reductions in healthcare‐associated infections.[14, 15]

Due to the difficulty in improving hand hygiene in various settings across the world, the WHO strategy for improving hand hygiene has been adopted and implemented by several studies in varying locations, such as Costa Rica, Italy, Mali, Pakistan, and Saudi Arabia.[16] Implementations of these multimodal strategies, following WHObased guidelines, have been shown to increase the level of hand hygiene adherence among healthcare workers and reduce infections at these locations.[14, 17, 18] This study expands upon that work by extending the same implementation strategy to assess the effectiveness of the introduction of alcohol‐based hand rub on hand hygiene practice at multiple hospitals in Japan.

In a previous article[19] we reported results from an observational study assessing healthcare worker hand hygiene adherence before touching the patient in 4 geographically diverse hospitals in Japan. The study reported that hand hygiene adherence in Japanese hospitals was lower than reported mean values from other international studies, and that greater adherence to hand hygiene should be encouraged. In this article, we present the results of a multimodal intervention intended to improve levels of healthcare worker hand hygiene in 3 of these hospitals.

METHODS

RESULTS

Data were collected from May 2013 to July 2013 in hospital A, in October 2012 in hospital B, and June 2013 in hospital C to ensure data were collected after the 6‐month intervention at each site. A total of 2982 observations of hand hygiene were performed in 10 distinct units across the 3 participating hospitals during the postintervention periods. Hand hygiene observations were performed during the day Monday through Friday between 8:30 _am and 7:30 _pm, with the majority occurring prior to 1:00 _pm.

The overall postintervention hand hygiene adherence rate (in all 3 hospitals) was significantly higher at 32.7% (974/2982) adherence compared to 18.0% (482/2679) adherence in the preintervention period (P<0.001). An increased hand hygiene adherence rate in each participating hospital in the postintervention period was observed (Figure 1). Similar trends of higher overall hand hygiene adherence rates for both nurses and physicians in the postintervention period were seen. Use of alcohol‐based hand rub among those with appropriate hand hygiene was significantly higher, with 90.0% (880/974) using hand rub in the postintervention period versus 67.0% (322/482) in the preintervention period (P<0.001). Comparison of overall hand hygiene adherence rates by unit type and healthcare worker subgroup between the pre‐ and postintervention periods are shown in Table 4. Detailed comparisons of hand hygiene adherence rates for each hospital are available in the supplementary appendix. Although a significant improvement of hand hygiene practice was observed in the majority of participating units (6/10), there was a significant decline in hand hygiene practice in 2 units for nurses and 1 unit for physicians. Hand hygiene adherence rates by healthcare worker subgroups (both physicians and nurses) were significantly higher in the postintervention period than those in the preintervention period. Trends toward higher hand hygiene adherence rate of nurses in the postintervention period were observed (34.8% adherence for nurses compared to 30.4% adherence for physicians); the difference between nurses and physicians were not statistically significant (P=0.07).

Figure 1

Hospital A achieved the highest postintervention adherence rates (39.9% adherence postintervention), as well as the greatest absolute improvement in hand hygiene (increase of 29.0%). There were significant improvements in 3 of the 4 participating units in hospital A, with the emergency department showing improvements only in the nurse subgroup. In hospital B, total hand hygiene adherence increased from 24.7% to 30.0% (P=0.01); however, this increase was mainly due to increase in hand hygiene adherence rates of nurses. There were significant increases in hand hygiene adherence rates for nurses in the medicine (+11%, P=0.04) and surgery wards (+14%, P=0.01), with nonsignificant increases for physicians (+10% medicine, P=0.07;+2% surgery, P=0.78). However, in the emergency department, nurses showed no significant improvement, and physicians had a significant decrease in adherence (15.7% preintervention vs 7.4% postintervention; P=0.02). In hospital C, total hand hygiene practice rates were significantly improved (from 18.9% to 26.5%; P<0.001); however, this was driven by improvements only in the surgical ward (14.6% preintervention to 42.3% postintervention; P<0.001). The rates for nurses declined significantly in both the medicine and ICU wards, leading to no observed improvements on those wards.

DISCUSSION

Our multicenter intervention study in Japan included observations from almost 3000 encounters between clinicians and patients. Before the intervention, the overall rate of hand hygiene adherence was 18%. After the multimodal intervention, the absolute increase in healthcare worker hand hygiene adherence was 15%. Although there was overall improvement, the adherence rates varied by hospital, with hospital A increasing by 29% and hospital B and C only attaining increases of 5% and 7%, respectively.

Despite the importance of hand hygiene of healthcare workers, it is challenging to increase hand hygiene adherence because it requires behavioral modification. Moreover, it remains uncertain what factors will affect healthcare worker behavior. We implemented pragmatic strategies to evaluate the efficacy of hand hygiene multimodal interventions based on internationally recognized WHO hand hygiene adherence strategies[11] and an institutional‐level contest with financial incentives. The findings in the current study help us understand not only how a multimodal intervention importantly improves hand hygiene adherence, but also what factors potentially make healthcare workers modify their behaviors.

In this study, we evaluated whether an institutional‐level contest with financial incentives contributed to improved hand hygiene adherence of healthcare workers. This study demonstrated improvement of hand hygiene practice after implementation of a multimodal hand hygiene intervention combined with an institutional‐level contest with financial incentives. The contest might have had a modest effect to help motivate the participating hospitals to improve their hand hygiene adherence rate. This is consistent with a previous study that demonstrated financial incentives were associated with modifying healthcare workers' hand hygiene practice.[21] However, we did not strictly standardize how the contest information was distributed in each participating institution and the objective assessment for changes in motivation by the contest was lacking in this study. Thus, changes in motivation by the contest with financial incentives likely varied by each participating institution. Further studies are needed to assess if this type of approach is worth pursuing.

We observed several noteworthy associations between the intervention components that were implemented at each facility and their improvement in hand hygiene adherence. Among the participating hospitals, hospital A was most successful with improving hand hygiene adherence, although all participating hospitals achieved a similar number of the 15 recommended intervention components during the intervention (8 to 10 per hospital). Interestingly, hospital A initiated the most new components during the intervention period (8 new components for a total of 10 out of 15), whereas hospital B and hospital C initiated only 1 or 2 new components during the intervention period. Hospital A also successfully involved hospital executives, and elicited the commitment of a nurse manager and physician leader. Consistent with a previous study,[22] we believe that involvement of hospital executives appears to be important to increase overall hand hygiene rate among healthcare workers.

In contrast, hospitals B and C did not involve senior executives or identify nurse or physician champions for all participating units. Based on the results in this study, we believe that the involvement of hospital executives is likely a key for the penetration of hospital‐wide hand hygiene culture among healthcare workers.

Although this study was unable to determine which components are precisely associated with improving hand hygiene adherence, the findings suggest initiating multiple intervention components at the same time may provide more motivation for change than initiating only 1 or 2 components at a time. It is also possible that certain intervention components were more beneficial than others. For example, hospital A, which achieved the most success, was the only hospital to obtain leadership support. Other studies have demonstrated that the presence of leadership appeared to play a key role in improving hand hygiene adherence.[23, 24] Moreover, a recent Japanese nationwide survey demonstrated higher safety centeredness was associated with regular use of standard infection prevention practice.[25] Consistent with a previous study, improving hand hygiene adherence cannot be simply achieved by improving infrastructure (eg, introduction of portable alcohol‐based hand rub) alone, but it depends on altering healthcare worker behavior.[26]

This study has several limitations. Because participating hospitals could tailor the specific interventions chosen for their facility, the improvement in hand hygiene adherence was likely multifactorial. We are unable in the existing study to determine a direct causal relationship between any of the individual intervention components and hand hygiene adherence. We are also unable to determine whether the improvements seen in hospital A were due to participation in the contest or due to the specific intervention components that were implemented. However, WHO hand hygiene guidelines point out that recognition of the importance of hand hygiene varies in different regions and countries, and the goal for hand hygiene interventions is to establish a culture of hand hygiene practice through pragmatic intervention strategies, frequent evaluation, and feedback to healthcare workers.[27] Thus, we prioritized pragmatic strategies to include in our intervention to promote hand hygiene adherence. Another limitation was the date of implementation of the multimodal intervention was slightly different at each facility. It was challenging to implement the intervention simultaneously across institutions due to competing priorities at each facility. Although the primary goal of hand hygiene is to reduce the burden of healthcare‐associated infection, we were unable to measure infection rates at the participating facilities. It is possible the presence of an external observer had an impact on the healthcare workers' behavior.[28] However, the healthcare workers were not informed as to what the observer was monitoring to minimize this potential effect. Lastly, the findings in this study provide immediate intervention effects but further study will be required to determine if these effects are sustainable.

Altering healthcare worker behavior is likely the key element to improve hand hygiene adherence, and behavioral modification may be achieved with the support of leadership at the unit and facility level. However, even though we found significant improvements in healthcare worker hand hygiene adherence after the intervention, the adherence rates are still relatively low compared to reported adherence rates from other countries,[29] suggesting further intervention is needed in this setting to optimize and hygiene practice. Because hand hygiene practice is a crucial strategy to prevent healthcare‐associated infections, every effort should be made to enhance the hand hygiene practice of healthcare workers.