Toru Kamiya, MD

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.

One of the most common causes of hospitalization in the elderly is aspiration pneumonia related to dysphagia due to numerous underlying diseases.1 Thus, it is clinically important to identify prognostic factors associated with increased mortality in elderly patients with aspiration pneumonia. Hyponatremia is the most common electrolyte abnormality in hospitalized patients occurring in up to 11% of elderly patients in hospital.2 Previous studies have suggested that the presence and degree of hyponatremia is associated with the severity of pneumonia in adults and children, although the results have differed among studies.37

Hyponatremia is caused by various factors, including volume depletion, use of diuretics, hypothyroidism, adrenal insufficiency, heart failure, renal failure, and cirrhosis. Additionally, the syndrome of inappropriate antidiuresis (SIAD) is a frequent and heterogeneous disorder characterized by hyponatremia and impaired urinary dilution in the absence of any recognized stimulation of antidiuretic hormone secretion.8 Because not all patients with SIAD have elevated circulating levels of arginine vasopressin (AVP), the term SIAD is preferred to the term syndrome of inappropriate secretion of antidiuretic hormone (SIADH).9 One study has shown an association between the severity of pneumonia in children and the development of hyponatremia due to SIAD.10 To our knowledge, there have been no studies evaluating the impact of different causes of hyponatremia on mortality in elderly patients with aspiration pneumonia.

We therefore sought to investigate whether hyponatremia of all etiologies (all‐cause hyponatremia) was associated with mortality in elderly patients with aspiration pneumonia. Additionally, we compared the impact of hyponatremia due to SIAD, with hyponatremia of other etiologies, on mortality in this population

METHODS

RESULTS

The baseline characteristics of the study population are listed in Table 2. There were 221 elderly patients identified as having aspiration pneumonia. Of those, 65 (29%) had hyponatremia; 3 (5%) with non‐hypotonic and 62 (95%) with hypotonic hyponatremia. In the latter group, patients were characterized has having hypovolemic (39 [63%]), hypervolemic (3 [5%]), and euvolemic (20 [32%]) hyponatremia. Among the euvolemic patients, SIAD occurred in 14 (70%) of patients. Non‐SIAD euvolemic hyponatremia occurred in 6 (30%) patients and was associated with hypothyroidism (1 patient), adrenal insufficiency (1 patient), and was unclassifiable due to lack of available clinical data in 4 patients. The kappa value was 0.87 for inter‐rater agreement of the classification of hypotonic hyponatremia.

The following variables were included in multivariate logistic analyses: congestive heart failure, cirrhosis, chronic renal failure, disorientation, body temperature <35C or >40C, anemia, and serum albumin <3 g/dL (see Supporting Information, Appendix, in the online version of this article).

In the multivariate logistic analyses, all‐cause hyponatremia was not associated with increased 30‐day mortality (odds ratio [OR] 1.85, 95% confidence interval [CI] 0.635.48; P = 0.262), but was associated with a trend toward increased risk of in‐hospital mortality (OR 2.10, 95% CI 1.004.42; P = 0.050) (Table 3). Moderate and severe hyponatremia were both significantly associated with increased in‐hospital mortality (OR 6.05, 95% CI 1.4625.0; P = 0.013 and OR 5.65, 95% CI 1.1428.1; P = 0.034, respectively). The same trends were observed for 30‐day mortality, although the results were not statistically significant. No such trend was observed for mild hyponatremia.

In the multivariate logistic regression analyses, hypotonic hyponatremia due to SIAD was significantly associated with both increased risk of 30‐day mortality (OR 7.40, 95% CI 1.7331.7; P = 0.007) and increased risk of in‐hospital mortality (OR 22.3, 95% CI 4.26117; P < 0.001) (Table 4). In contrast, hypovolemic or non‐SIAD euvolemic hyponatremia was associated with neither increased risk of 30‐day mortality nor increased risk of in‐hospital mortality. There were too few hypervolemic hyponatremia patients for us to perform effective logistic analyses. The P values of the HosmerLemeshow tests were 0.45 for the multivariate logistic regression model (hypovolemic, SIAD, and non‐SIAD euvolemic vs normonatremia) with 30‐day mortality, and 0.30 for the model with in‐hospital mortality.

Six patients with SIAD were classified as having an A‐DROP severity class of mild, 4 as moderate, and 4 as severe (P = 0.908, Wilcoxon‐type test for trend). There was no association between the occurrence of SIAD and the severity of pneumonia.

DISCUSSION

We demonstrated that mortality in elderly patients with aspiration pneumonia was significantly associated with SIAD, but not with all‐cause hyponatremia. Unlike SIAD, other etiologies of hyponatremia were not associated with mortality in elderly patients with aspiration pneumonia. A recent study by Waikar and colleagues concluded that hyponatremia subgrouped by severity was not significantly associated with in‐hospital mortality in pneumonia patients, although a trend between severe hyponatremia and mortality was observed.16 Likewise, a study by Zilberberg and colleagues reported no significant increased risk of death with hyponatremia compared with normonatremia.4 These results are similar to our results for all‐cause hyponatremia and for hyponatremia subgrouped by severity. In contrast, a study by Nair and colleagues reported some increased risk of death with hyponatremia.5 Our results suggest that the heterogeneity of these previous results was probably due to the fact that SIAD was not identified in these other studies.

While the rationale for increased mortality in patients with pneumonia associated with SIAD is not known, it may be that there is a direct deleterious effect of elevated AVP. AVP has 3 distinct receptor subtypes, V1A, V1B, and V2. Stimulation of the V1A receptor in vascular smooth muscle promotes an increase in systemic vascular resistance, and stimulation of the same receptor in cardiac myocytes promotes myocyte hypertrophy. Stimulation of the V1B receptor in the anterior pituitary promotes adrenocorticotropic hormone release, and stimulation of the V2 receptor in the renal collecting ducts promotes an increase in water retention, which plays the main role in SIAD.1719 Our hypothesis in elderly SIAD patients with aspiration pneumonia is that increased AVP levels may lead not only to water retention and hyponatremia, but also to other effects such as vasoconstriction and myocyte hypertrophy, which may adversely influence the cardiovascular systems of elderly patients (Figure 1).

Figure 1

In our study, SIAD in elderly patients with aspiration pneumonia was more strongly associated with in‐hospital mortality than with 30‐day mortality. The average length of stay (LOS) of all patients dying in hospital (42 days) was significantly longer than the average LOS of those dying within 30 days of admission (15 days; P < 0.001, MannWhitney Test; Table 2). These findings suggest that SIAD was associated more strongly with longer‐term mortality than with acute‐stage mortality. The reason for the association between SIAD and longer‐term mortality remains unclear, although there may be some association between longer‐term mortality and the pathophysiologic mechanisms of AVP.

Our study has some limitations. First, because of the retrospective observational design, there is a potential for bias. We used multivariate analyses adjusted for confounding factors, however, other residual confounding factors may have remained. In addition, since the diagnosis of pneumonia was based on chart review, there may have been imprecision in the accuracy of diagnosing aspiration pneumonia. Aspiration pneumonia sometimes occurs without apparent episodes of aspiration, and this would have led to underdiagnosis. In contrast, aspiration pneumonitis can be mistaken for aspiration pneumonia; this would have led to overdiagnosis.

Second, volume status is difficult to evaluate prospectively, and thus by nature of our design, appropriate assignment of volume status was difficult. Several studies have used test infusions of isotonic saline to discriminate between these alternatives, but because our study was retrospective, we were unable to use this test.11, 20 Some studies have reported that, in patients in a state of volume depletion, volume repletion removes the stimulus for antidiuretic hormone release, allowing excess water to be excreted in a dilute urine and the serum sodium concentration to return toward normal.21, 22 According to this theory, instead of using an isotonic test infusion, we added in our study a criterion of volume depletion in which patients with a sustained increase in serum sodium concentration of 5 mEq/L and a sustained decrease in blood urea nitrogen, even with administration of hypotonic solution, were classified as volume depleted.

Third, all patients were analyzed according to status on admission, although some patients with hypovolemic hyponatremia at admission were found to have hyponatremia due to SIAD after admission.

Fourth, because the sample size of this study was small with our results revealing wide confidence intervals, an effect between other causes of hyponatremia and mortality might not have been identified. However, for 80% power, the calculated sample size was 100 non‐SIAD patients with aspiration pneumonia versus 10 SIAD patients, given that the mortality rate of elderly patients with aspiration pneumonia was, at a moderate estimate, 15% according to the studies of both Stukenborg and colleagues and Oliver and colleagues, and the mortality rate of SIAD patients was increased by 400% compared with that of non‐SIAD patients according to the study of Song and colleagues, with an alpha error of 0.05.7, 23, 24 Our sample size was therefore greater than the required size.

Fifth, because the APD dataset was compiled in 2007 for another study, it was not concurrent, and this may have led to other limitations in interpreting the data.

Finally, in Japan, the average length of hospital stay was 36.3 days in 2004 and 34.1 days in 2007much longer than other developed countries.25 Because of this situation, in‐hospital mortality, and not 30‐day mortality, represented long‐term mortality. Therefore, our results may not be easily applicable to the situation in other developed countries.

In conclusion, our results suggest that the presence of SIAD on admission in elderly patients with aspiration pneumonia is associated with increased mortality. This novel finding should be re‐evaluated, but it does raise the question of a direct, negative impact of AVP on patients' clinical outcomes. In the future, a larger prospective cohort study should be conducted to confirm the findings of this study, given the small sample size and the retrospective nature of the study. Additionally, a different population of pneumonia patients, such as those with community‐acquired pneumonia, should be examined to further evaluate the etiologies of hyponatremia in pneumonia and the association between hyponatremia of these different etiologies and mortality.

One of the most common causes of hospitalization in the elderly is aspiration pneumonia related to dysphagia due to numerous underlying diseases.1 Thus, it is clinically important to identify prognostic factors associated with increased mortality in elderly patients with aspiration pneumonia. Hyponatremia is the most common electrolyte abnormality in hospitalized patients occurring in up to 11% of elderly patients in hospital.2 Previous studies have suggested that the presence and degree of hyponatremia is associated with the severity of pneumonia in adults and children, although the results have differed among studies.37

Hyponatremia is caused by various factors, including volume depletion, use of diuretics, hypothyroidism, adrenal insufficiency, heart failure, renal failure, and cirrhosis. Additionally, the syndrome of inappropriate antidiuresis (SIAD) is a frequent and heterogeneous disorder characterized by hyponatremia and impaired urinary dilution in the absence of any recognized stimulation of antidiuretic hormone secretion.8 Because not all patients with SIAD have elevated circulating levels of arginine vasopressin (AVP), the term SIAD is preferred to the term syndrome of inappropriate secretion of antidiuretic hormone (SIADH).9 One study has shown an association between the severity of pneumonia in children and the development of hyponatremia due to SIAD.10 To our knowledge, there have been no studies evaluating the impact of different causes of hyponatremia on mortality in elderly patients with aspiration pneumonia.

We therefore sought to investigate whether hyponatremia of all etiologies (all‐cause hyponatremia) was associated with mortality in elderly patients with aspiration pneumonia. Additionally, we compared the impact of hyponatremia due to SIAD, with hyponatremia of other etiologies, on mortality in this population

METHODS

RESULTS

The baseline characteristics of the study population are listed in Table 2. There were 221 elderly patients identified as having aspiration pneumonia. Of those, 65 (29%) had hyponatremia; 3 (5%) with non‐hypotonic and 62 (95%) with hypotonic hyponatremia. In the latter group, patients were characterized has having hypovolemic (39 [63%]), hypervolemic (3 [5%]), and euvolemic (20 [32%]) hyponatremia. Among the euvolemic patients, SIAD occurred in 14 (70%) of patients. Non‐SIAD euvolemic hyponatremia occurred in 6 (30%) patients and was associated with hypothyroidism (1 patient), adrenal insufficiency (1 patient), and was unclassifiable due to lack of available clinical data in 4 patients. The kappa value was 0.87 for inter‐rater agreement of the classification of hypotonic hyponatremia.

The following variables were included in multivariate logistic analyses: congestive heart failure, cirrhosis, chronic renal failure, disorientation, body temperature <35C or >40C, anemia, and serum albumin <3 g/dL (see Supporting Information, Appendix, in the online version of this article).

In the multivariate logistic analyses, all‐cause hyponatremia was not associated with increased 30‐day mortality (odds ratio [OR] 1.85, 95% confidence interval [CI] 0.635.48; P = 0.262), but was associated with a trend toward increased risk of in‐hospital mortality (OR 2.10, 95% CI 1.004.42; P = 0.050) (Table 3). Moderate and severe hyponatremia were both significantly associated with increased in‐hospital mortality (OR 6.05, 95% CI 1.4625.0; P = 0.013 and OR 5.65, 95% CI 1.1428.1; P = 0.034, respectively). The same trends were observed for 30‐day mortality, although the results were not statistically significant. No such trend was observed for mild hyponatremia.

In the multivariate logistic regression analyses, hypotonic hyponatremia due to SIAD was significantly associated with both increased risk of 30‐day mortality (OR 7.40, 95% CI 1.7331.7; P = 0.007) and increased risk of in‐hospital mortality (OR 22.3, 95% CI 4.26117; P < 0.001) (Table 4). In contrast, hypovolemic or non‐SIAD euvolemic hyponatremia was associated with neither increased risk of 30‐day mortality nor increased risk of in‐hospital mortality. There were too few hypervolemic hyponatremia patients for us to perform effective logistic analyses. The P values of the HosmerLemeshow tests were 0.45 for the multivariate logistic regression model (hypovolemic, SIAD, and non‐SIAD euvolemic vs normonatremia) with 30‐day mortality, and 0.30 for the model with in‐hospital mortality.

Six patients with SIAD were classified as having an A‐DROP severity class of mild, 4 as moderate, and 4 as severe (P = 0.908, Wilcoxon‐type test for trend). There was no association between the occurrence of SIAD and the severity of pneumonia.

DISCUSSION

We demonstrated that mortality in elderly patients with aspiration pneumonia was significantly associated with SIAD, but not with all‐cause hyponatremia. Unlike SIAD, other etiologies of hyponatremia were not associated with mortality in elderly patients with aspiration pneumonia. A recent study by Waikar and colleagues concluded that hyponatremia subgrouped by severity was not significantly associated with in‐hospital mortality in pneumonia patients, although a trend between severe hyponatremia and mortality was observed.16 Likewise, a study by Zilberberg and colleagues reported no significant increased risk of death with hyponatremia compared with normonatremia.4 These results are similar to our results for all‐cause hyponatremia and for hyponatremia subgrouped by severity. In contrast, a study by Nair and colleagues reported some increased risk of death with hyponatremia.5 Our results suggest that the heterogeneity of these previous results was probably due to the fact that SIAD was not identified in these other studies.

While the rationale for increased mortality in patients with pneumonia associated with SIAD is not known, it may be that there is a direct deleterious effect of elevated AVP. AVP has 3 distinct receptor subtypes, V1A, V1B, and V2. Stimulation of the V1A receptor in vascular smooth muscle promotes an increase in systemic vascular resistance, and stimulation of the same receptor in cardiac myocytes promotes myocyte hypertrophy. Stimulation of the V1B receptor in the anterior pituitary promotes adrenocorticotropic hormone release, and stimulation of the V2 receptor in the renal collecting ducts promotes an increase in water retention, which plays the main role in SIAD.1719 Our hypothesis in elderly SIAD patients with aspiration pneumonia is that increased AVP levels may lead not only to water retention and hyponatremia, but also to other effects such as vasoconstriction and myocyte hypertrophy, which may adversely influence the cardiovascular systems of elderly patients (Figure 1).

Figure 1

In our study, SIAD in elderly patients with aspiration pneumonia was more strongly associated with in‐hospital mortality than with 30‐day mortality. The average length of stay (LOS) of all patients dying in hospital (42 days) was significantly longer than the average LOS of those dying within 30 days of admission (15 days; P < 0.001, MannWhitney Test; Table 2). These findings suggest that SIAD was associated more strongly with longer‐term mortality than with acute‐stage mortality. The reason for the association between SIAD and longer‐term mortality remains unclear, although there may be some association between longer‐term mortality and the pathophysiologic mechanisms of AVP.

Our study has some limitations. First, because of the retrospective observational design, there is a potential for bias. We used multivariate analyses adjusted for confounding factors, however, other residual confounding factors may have remained. In addition, since the diagnosis of pneumonia was based on chart review, there may have been imprecision in the accuracy of diagnosing aspiration pneumonia. Aspiration pneumonia sometimes occurs without apparent episodes of aspiration, and this would have led to underdiagnosis. In contrast, aspiration pneumonitis can be mistaken for aspiration pneumonia; this would have led to overdiagnosis.

Second, volume status is difficult to evaluate prospectively, and thus by nature of our design, appropriate assignment of volume status was difficult. Several studies have used test infusions of isotonic saline to discriminate between these alternatives, but because our study was retrospective, we were unable to use this test.11, 20 Some studies have reported that, in patients in a state of volume depletion, volume repletion removes the stimulus for antidiuretic hormone release, allowing excess water to be excreted in a dilute urine and the serum sodium concentration to return toward normal.21, 22 According to this theory, instead of using an isotonic test infusion, we added in our study a criterion of volume depletion in which patients with a sustained increase in serum sodium concentration of 5 mEq/L and a sustained decrease in blood urea nitrogen, even with administration of hypotonic solution, were classified as volume depleted.

Third, all patients were analyzed according to status on admission, although some patients with hypovolemic hyponatremia at admission were found to have hyponatremia due to SIAD after admission.

Fourth, because the sample size of this study was small with our results revealing wide confidence intervals, an effect between other causes of hyponatremia and mortality might not have been identified. However, for 80% power, the calculated sample size was 100 non‐SIAD patients with aspiration pneumonia versus 10 SIAD patients, given that the mortality rate of elderly patients with aspiration pneumonia was, at a moderate estimate, 15% according to the studies of both Stukenborg and colleagues and Oliver and colleagues, and the mortality rate of SIAD patients was increased by 400% compared with that of non‐SIAD patients according to the study of Song and colleagues, with an alpha error of 0.05.7, 23, 24 Our sample size was therefore greater than the required size.

Fifth, because the APD dataset was compiled in 2007 for another study, it was not concurrent, and this may have led to other limitations in interpreting the data.

Finally, in Japan, the average length of hospital stay was 36.3 days in 2004 and 34.1 days in 2007much longer than other developed countries.25 Because of this situation, in‐hospital mortality, and not 30‐day mortality, represented long‐term mortality. Therefore, our results may not be easily applicable to the situation in other developed countries.

In conclusion, our results suggest that the presence of SIAD on admission in elderly patients with aspiration pneumonia is associated with increased mortality. This novel finding should be re‐evaluated, but it does raise the question of a direct, negative impact of AVP on patients' clinical outcomes. In the future, a larger prospective cohort study should be conducted to confirm the findings of this study, given the small sample size and the retrospective nature of the study. Additionally, a different population of pneumonia patients, such as those with community‐acquired pneumonia, should be examined to further evaluate the etiologies of hyponatremia in pneumonia and the association between hyponatremia of these different etiologies and mortality.

One of the most common causes of hospitalization in the elderly is aspiration pneumonia related to dysphagia due to numerous underlying diseases.1 Thus, it is clinically important to identify prognostic factors associated with increased mortality in elderly patients with aspiration pneumonia. Hyponatremia is the most common electrolyte abnormality in hospitalized patients occurring in up to 11% of elderly patients in hospital.2 Previous studies have suggested that the presence and degree of hyponatremia is associated with the severity of pneumonia in adults and children, although the results have differed among studies.37

Hyponatremia is caused by various factors, including volume depletion, use of diuretics, hypothyroidism, adrenal insufficiency, heart failure, renal failure, and cirrhosis. Additionally, the syndrome of inappropriate antidiuresis (SIAD) is a frequent and heterogeneous disorder characterized by hyponatremia and impaired urinary dilution in the absence of any recognized stimulation of antidiuretic hormone secretion.8 Because not all patients with SIAD have elevated circulating levels of arginine vasopressin (AVP), the term SIAD is preferred to the term syndrome of inappropriate secretion of antidiuretic hormone (SIADH).9 One study has shown an association between the severity of pneumonia in children and the development of hyponatremia due to SIAD.10 To our knowledge, there have been no studies evaluating the impact of different causes of hyponatremia on mortality in elderly patients with aspiration pneumonia.

We therefore sought to investigate whether hyponatremia of all etiologies (all‐cause hyponatremia) was associated with mortality in elderly patients with aspiration pneumonia. Additionally, we compared the impact of hyponatremia due to SIAD, with hyponatremia of other etiologies, on mortality in this population

METHODS

RESULTS

The baseline characteristics of the study population are listed in Table 2. There were 221 elderly patients identified as having aspiration pneumonia. Of those, 65 (29%) had hyponatremia; 3 (5%) with non‐hypotonic and 62 (95%) with hypotonic hyponatremia. In the latter group, patients were characterized has having hypovolemic (39 [63%]), hypervolemic (3 [5%]), and euvolemic (20 [32%]) hyponatremia. Among the euvolemic patients, SIAD occurred in 14 (70%) of patients. Non‐SIAD euvolemic hyponatremia occurred in 6 (30%) patients and was associated with hypothyroidism (1 patient), adrenal insufficiency (1 patient), and was unclassifiable due to lack of available clinical data in 4 patients. The kappa value was 0.87 for inter‐rater agreement of the classification of hypotonic hyponatremia.

The following variables were included in multivariate logistic analyses: congestive heart failure, cirrhosis, chronic renal failure, disorientation, body temperature <35C or >40C, anemia, and serum albumin <3 g/dL (see Supporting Information, Appendix, in the online version of this article).

In the multivariate logistic analyses, all‐cause hyponatremia was not associated with increased 30‐day mortality (odds ratio [OR] 1.85, 95% confidence interval [CI] 0.635.48; P = 0.262), but was associated with a trend toward increased risk of in‐hospital mortality (OR 2.10, 95% CI 1.004.42; P = 0.050) (Table 3). Moderate and severe hyponatremia were both significantly associated with increased in‐hospital mortality (OR 6.05, 95% CI 1.4625.0; P = 0.013 and OR 5.65, 95% CI 1.1428.1; P = 0.034, respectively). The same trends were observed for 30‐day mortality, although the results were not statistically significant. No such trend was observed for mild hyponatremia.

In the multivariate logistic regression analyses, hypotonic hyponatremia due to SIAD was significantly associated with both increased risk of 30‐day mortality (OR 7.40, 95% CI 1.7331.7; P = 0.007) and increased risk of in‐hospital mortality (OR 22.3, 95% CI 4.26117; P < 0.001) (Table 4). In contrast, hypovolemic or non‐SIAD euvolemic hyponatremia was associated with neither increased risk of 30‐day mortality nor increased risk of in‐hospital mortality. There were too few hypervolemic hyponatremia patients for us to perform effective logistic analyses. The P values of the HosmerLemeshow tests were 0.45 for the multivariate logistic regression model (hypovolemic, SIAD, and non‐SIAD euvolemic vs normonatremia) with 30‐day mortality, and 0.30 for the model with in‐hospital mortality.

Six patients with SIAD were classified as having an A‐DROP severity class of mild, 4 as moderate, and 4 as severe (P = 0.908, Wilcoxon‐type test for trend). There was no association between the occurrence of SIAD and the severity of pneumonia.

DISCUSSION

We demonstrated that mortality in elderly patients with aspiration pneumonia was significantly associated with SIAD, but not with all‐cause hyponatremia. Unlike SIAD, other etiologies of hyponatremia were not associated with mortality in elderly patients with aspiration pneumonia. A recent study by Waikar and colleagues concluded that hyponatremia subgrouped by severity was not significantly associated with in‐hospital mortality in pneumonia patients, although a trend between severe hyponatremia and mortality was observed.16 Likewise, a study by Zilberberg and colleagues reported no significant increased risk of death with hyponatremia compared with normonatremia.4 These results are similar to our results for all‐cause hyponatremia and for hyponatremia subgrouped by severity. In contrast, a study by Nair and colleagues reported some increased risk of death with hyponatremia.5 Our results suggest that the heterogeneity of these previous results was probably due to the fact that SIAD was not identified in these other studies.

While the rationale for increased mortality in patients with pneumonia associated with SIAD is not known, it may be that there is a direct deleterious effect of elevated AVP. AVP has 3 distinct receptor subtypes, V1A, V1B, and V2. Stimulation of the V1A receptor in vascular smooth muscle promotes an increase in systemic vascular resistance, and stimulation of the same receptor in cardiac myocytes promotes myocyte hypertrophy. Stimulation of the V1B receptor in the anterior pituitary promotes adrenocorticotropic hormone release, and stimulation of the V2 receptor in the renal collecting ducts promotes an increase in water retention, which plays the main role in SIAD.1719 Our hypothesis in elderly SIAD patients with aspiration pneumonia is that increased AVP levels may lead not only to water retention and hyponatremia, but also to other effects such as vasoconstriction and myocyte hypertrophy, which may adversely influence the cardiovascular systems of elderly patients (Figure 1).

Figure 1

In our study, SIAD in elderly patients with aspiration pneumonia was more strongly associated with in‐hospital mortality than with 30‐day mortality. The average length of stay (LOS) of all patients dying in hospital (42 days) was significantly longer than the average LOS of those dying within 30 days of admission (15 days; P < 0.001, MannWhitney Test; Table 2). These findings suggest that SIAD was associated more strongly with longer‐term mortality than with acute‐stage mortality. The reason for the association between SIAD and longer‐term mortality remains unclear, although there may be some association between longer‐term mortality and the pathophysiologic mechanisms of AVP.

Our study has some limitations. First, because of the retrospective observational design, there is a potential for bias. We used multivariate analyses adjusted for confounding factors, however, other residual confounding factors may have remained. In addition, since the diagnosis of pneumonia was based on chart review, there may have been imprecision in the accuracy of diagnosing aspiration pneumonia. Aspiration pneumonia sometimes occurs without apparent episodes of aspiration, and this would have led to underdiagnosis. In contrast, aspiration pneumonitis can be mistaken for aspiration pneumonia; this would have led to overdiagnosis.

Second, volume status is difficult to evaluate prospectively, and thus by nature of our design, appropriate assignment of volume status was difficult. Several studies have used test infusions of isotonic saline to discriminate between these alternatives, but because our study was retrospective, we were unable to use this test.11, 20 Some studies have reported that, in patients in a state of volume depletion, volume repletion removes the stimulus for antidiuretic hormone release, allowing excess water to be excreted in a dilute urine and the serum sodium concentration to return toward normal.21, 22 According to this theory, instead of using an isotonic test infusion, we added in our study a criterion of volume depletion in which patients with a sustained increase in serum sodium concentration of 5 mEq/L and a sustained decrease in blood urea nitrogen, even with administration of hypotonic solution, were classified as volume depleted.

Third, all patients were analyzed according to status on admission, although some patients with hypovolemic hyponatremia at admission were found to have hyponatremia due to SIAD after admission.

Fourth, because the sample size of this study was small with our results revealing wide confidence intervals, an effect between other causes of hyponatremia and mortality might not have been identified. However, for 80% power, the calculated sample size was 100 non‐SIAD patients with aspiration pneumonia versus 10 SIAD patients, given that the mortality rate of elderly patients with aspiration pneumonia was, at a moderate estimate, 15% according to the studies of both Stukenborg and colleagues and Oliver and colleagues, and the mortality rate of SIAD patients was increased by 400% compared with that of non‐SIAD patients according to the study of Song and colleagues, with an alpha error of 0.05.7, 23, 24 Our sample size was therefore greater than the required size.

Fifth, because the APD dataset was compiled in 2007 for another study, it was not concurrent, and this may have led to other limitations in interpreting the data.

Finally, in Japan, the average length of hospital stay was 36.3 days in 2004 and 34.1 days in 2007much longer than other developed countries.25 Because of this situation, in‐hospital mortality, and not 30‐day mortality, represented long‐term mortality. Therefore, our results may not be easily applicable to the situation in other developed countries.

In conclusion, our results suggest that the presence of SIAD on admission in elderly patients with aspiration pneumonia is associated with increased mortality. This novel finding should be re‐evaluated, but it does raise the question of a direct, negative impact of AVP on patients' clinical outcomes. In the future, a larger prospective cohort study should be conducted to confirm the findings of this study, given the small sample size and the retrospective nature of the study. Additionally, a different population of pneumonia patients, such as those with community‐acquired pneumonia, should be examined to further evaluate the etiologies of hyponatremia in pneumonia and the association between hyponatremia of these different etiologies and mortality.