The New Zealand health system faces the triple challenges of improving quality of care and patient experience, health outcomes and equity, and getting best value from health resources.1 Reducing hospital readmissions offers the prospect of achieving these aims and is therefore a topic of considerable interest.District Health Boards (DHBs) have been asked to make improvements in productivity and efficiency, but not at the expense of quality of care. Hospital unplanned acute readmission rates are a well-established measure of quality of care and are included in DHB performance measures.2In 2010/11, across DHBs, 7.6-11.5% of admissions are followed by an unplanned acute readmission within 28 days. However, whilst readmissions have attracted considerable interest internationally, particularly in USA recently,3 there has been comparatively little analysis in New Zealand.4 This study describes acute medical readmissions in New Zealand hospitals with a particular focus on older patients.This work was initially undertaken to support the design of a Waitemata DHB programme to prevent medical readmissions amongst older people by improving the transition from hospital to home.Readmissions definedThere is no agreed definition of readmissions in the literature. Unplanned or acute admissions following and earlier admission are usually of greatest interest.5 Authors have used time periods varying between 1 day and 12 months after discharge to define a readmission.5-7 The time period chosen may depend upon the reason for interest in readmissions.Readmissions are frequently used as a measure of hospital care quality. In this case readmissions soon after discharge are more likely to be related to deficiencies in care. 8 However, since our interest is in interventions aimed at supporting the transition of patients from hospital back into the community a longer time period also seems relevant.People at risk of readmissionReadmission rates have been shown to vary considerably between countries, between areas within countries, and between hospitals.3,6,9Internationally a number of risks have been identified for increased risk of readmission including being older, male, lower education, some ethnicities, widowed or divorced and having poor social networks or living alone3,7,10-13In New Zealand surgical readmissions in the elderly have been shown to be more common in men, older people and M ori and Pacific people.4 Chance of readmission has also been related to a number of clinical factors such as health service use, diagnosis, co-morbidities, disability, and function.3,6-9,12-15 Readmission rates for medical admissions have been found to be higher than for surgical admissions.3Readmissions are of interest because it is thought that some are potentially preventable. Higher rates of readmission are associated with lower quality of care in hospital. 16 However, assessment of the proportion of readmissions that are potentially preventable vary widely from 5-71%.7,9,14 This reflects different periods used, the difficulty in deciding whether an admission is preventable, and the widely varying methods used to make this judgement.Of greater importance is whether interventions in hospital or in the community can lead to actual reductions in readmissions. A number of systematic reviews have examined this question, and found that interventions can be effective, although not all are.17-20The aims of this study are to investigate the degree to which acute medical readmissions might be a suitable target for improving the quality and efficiency of the health system.Specifically we aim to answer a number of questions: How common are medical readmissions in older people? Which groups are at greatest risk? What is the health and health system burden of people who are readmitted? What is the potential for prevention? Methods NZ publically funded hospital admissions for the period 1 April 2009 to 31 March 2010 were examined. Data from the Ministry of Healths National Minimum Data Set (NMDS) collection was obtained on hospitalisations of people admitted between 1 Feb 2009 and 31 June 2010 to allow recognition of admissions three months before and three months after index admissions. Mortality data for the period 1 April 2009 to 31 June 2010 was also obtained and linked to the hospitalisation data using encrypted NHIs. Index admissions were defined as acute medical admissions in NZ residents where the person had stayed overnight and where they had ended in a routine or self discharge. We included self discharges as we were less interested in readmissions as a marker of quality of care than as an opportunity for intervention and people who self discharge would still be offered such an intervention. Readmissions were defined as a further acute medical admission. We particularly considered 30- and 90-day readmission rates. Ethnicity in the NMDS collection is self identified and allows multiple ethnicities to be recorded. These were prioritised according to the Ethnicity Data Protocols for the Health and Disability Sector21 and then aggregated to M ori, Pacific, Asian, and Other (Other includes Europeans). Deprivation was assigned at Census Area Unit using the NZDep2006 Index of Deprivation which is an area based index.22 Rurality was also assigned by Census Area Unit and is taken from tables provided by the Wellington School of Medicine which were in turn based up Statistics New Zealand definitions.23 For univariate analyses chi squared tests were used for dichotomous outcomes and Wilcoxon rank-sum test for continuous outcomes. Binomial regression was undertaken for multivariate analyses. All analyses were undertaken using Stata v11.2 software 00ae. Results There were 217,323 acute medical admissions amongst 164,428 patients with a subsequent routine discharge in the study period. 95,318 of these admissions in 66,983 patients were in people 65 years and older. These are the focus of this study. Readmission by age groupReadmission was very common. Up to one-third of acute medical admissions were followed by another acute medical admission within the next 3 months in some population groups. Rates of readmission increased with age until the seventies and then tended to plateau. M ori and Pacific people had higher readmission rates for most age groups than Asians and Others. Readmission rates in M ori, Pacific, and Asians after the age of 90 years are not shown because of the small numbers of people in these older age groups. 30-day readmission rates showed a similar pattern (not shown but available on request). Figure 1. Adult medical readmission within 90 days of discharge by age and ethnic group, New Zealand 2009/10 The higher rates of readmission for older people led us to focus on older people when planning potential interventions. The remainder of the analysis in this paper focuses on people 65 years and over. Figure 2 Cumulative readmission rates for people 65 years and over by time after discharge, New Zealand 2009/10 The cumulative chance of readmission for an older person who is discharged after an acute medical admission increases over time reaching 10.8% (95%CIs 10.6-10.9%)by 30 days after discharge and 18.3% (95%CI 18.1-18.3%)after 90 days. As shown in Figure 2 second readmissions within 90 days were also not rare. Whereas 10.8% of individuals who were admitted went on to be readmitted within 30 days, 16.1% (95%CI 15.8-16.3%) of admissions were followed by a readmission within 30 days. This difference is due to people with high readmission rates being counted only once in the first analysis but multiple times in the second. Within 90 days, 27.8% (95%CI 27.5-28.1%) of admissions were followed by a readmission. When considering the impact of readmissions on the health system the proportion of admissions that are readmissions is also important. 13.8% (95%CI 13.6-14.0%) of all acute medical admissions in over 65 year olds in the study period were 30-day readmissions and 25.5% (95%CI 25.5%-26.0%) were 90-day readmissions. Groups at risk of readmissionM ori and Pacific were more likely to be readmitted than those in other ethnic groups. Table 1 shows the result of a multivariate model which included age, ethnicity, gender, deprivation and rurality as predictors of readmission within 30 days. Increasing age, increased deprivation and male gender were also associated with increased risk of readmission. Interestingly, people living in rural areas were slightly less likely to be readmitted. Table 1. Risk factors for readmission within 30 days of discharge for people 65 years and over, New Zealand 2009/10 (*Note: years over the age of 65 was use in the analysis rather than age) Morbidity and mortality of people who were readmittedWhilst it is not surprising, it is still worthy of comment that people who are being readmitted with acute medical problems are more unwell and have worse outcomes than people being admitted for their first admission. Table 2. Comparison of readmissions (within 30 days of discharge) and first admissions for mean admission complexity and outcomes for people 65 years and older Readmissions were more likely to be complex (coded as having a patient clinical complexity level other than 0), and had a longer mean length of stay, and a higher mean cost weight than first admissions. This indicates that people being readmitted were more unwell and/or more complex than first admissions. Readmission outcomes were also worse than first admissions. They were almost twice as likely to die within hospital or soon after discharge. Readmissions were also almost twice as likely to result in a further readmission. PreventionAs we were interested in designing interventions to prevent readmissions we were interested in what proportion of readmissions were potentially preventable. Unfortunately this is very difficult to determine from hospitalisation data. However, we have looked at two indicators that provide some information. A readmission might be more likely to have been preventable by better hospital or transition care if it was for the same disease as the preceding admission. Our analysis shows that 30.9% (95%CI 30.1-31.6%) of 30-day readmissions had the same diagnosis as the preceding admission (defined by having the same DRG code e.g. F62 heart failure) and 48.0% (95%CI 47.1%-48.8%) were in the same diagnostic group (defined by having the same first letter in the DRG code e.g. F circulatory). The second indicator we considered was whether the readmission was for an ambulatory sensitive hospitalisation (ASH). These admissions might potentially be preventable by linking people back into community care more effectively. Overall, 34.7% (95%CI 33.8-35.5%) of 30-day readmissions and 35.6% (95%CI 34.9-37.2%) of 90-day readmissions were for ASH conditions. When considering possible interventions to reduce readmissions the question arises whether we should target a few high risk diagnoses such as congestive heart failure (CHF) and chronic obstructive pulmonary disease (COPD) or instead design an intervention that reaches all people at high risk of readmission. We therefore looked at to what extent particular diagnostic groups or diagnoses made up the majority of readmissions. Table 3. Main diagnostic groups and diagnoses of readmissions within 30 days of an acute medical admissions in people 65 years and older, New Zealand 2009/10 Cardiovascular disease and respiratory disease together made up over 40% of readmissions. However, the four commonest diagnoses (angina/chest pain, CHF, COPD, and pneumonia) made up only 27% of all readmissions. Discussion Medical readmissions were common amongst older people and are a potentially worthwhile focus for intervention. A quarter of all acute medical admissions were preceded by another acute medical admission within 3 months. These admissions were more costly and complex than other admissions. If these readmissions could be reduced substantially there would be significant savings to the health system. In addition people being readmitted are at risk of poor outcomes as measured by mortality and further readmissions. Interventions to provide this group better care might reduce these poor outcomes. It is of concern that there were groups of our population who were at increased risk of these poor outcomes. These included M ori, Pacific people, people living in deprived areas and men. Possible reasons for this disparity are that these groups were less healthy, that the care we provided, either in hospital or in the community, was less effective for their needs, or that they had lower access to other community resources that enabled them to recover their health. We believe that these are serious concerns and warrant further study. It also behoves any one developing an intervention to prevent readmissions to make particular efforts to ensure that the interventions meet these groups needs. Readmissions are of interest because some are thought to be preventable by improvements in the health care system. However, as previously discussed, there is no agreement on what proportion of readmissions are preventable or indeed which groups of readmissions are preventable. A readmission may be preventable by better care whilst the patient is in hospital, by better organised and supported care of the transition back into the community, or by better ongoing care once the patient is established back in the community. Data presented here are limited in examining the potential for prevention of readmissions. Patients who are quickly readmitted to hospital with the same illness might be more likely to benefit from better hospital or transition care. This group made up nearly a third of readmissions over 30 days in our study. However, not all of these readmissions will be preventable, and many readmissions with unrelated illnesses might well benefit from the same improvements in care. Ambulatory sensitive hospitalisations (ASH) made up just over a third of readmissions. ASH may be related to quality of primary care and higher quality of care is thought to reduce admission from this group of diagnoses. This category of admission has not previously been examined in relationship to readmission. This high proportion of readmissions in this category suggests that primary care intervention may be an area worth considering. This study examines a population with a potential to improve care and reduce health system costs. However, to achieve these aims we must first be able to identify reliably those people at risk of readmission and then find interventions that can be shown to reduce readmissions and other adverse outcomes. CHF and COPD are well recognised chronic conditions that contribute significantly to readmissions in older people. There is ample evidence from systematic reviews that a range of interventions can reduce hospitalisations in these groups.24-28 Another approach may be to focus on those patients at high risk of readmission and support their transition back into the community. Panattoni et al, in examining predictive risk models to identify patients at high risk of emergency hospitalisation in New Zealand suggest development of cost-effective strategies.29 A predictive risk model was developed for Waitemata DHB in 2009 and a further model focused on older people is currently under development. Similar approaches have being reported internationally.11,30 A number of systematic reviews have examined the effectiveness of 2018transition interventions.17-20 One concluded that programmes that begin in hospital, are continued in the community and are multidimensional are more likely to be effective. 31 Unfortunately, whilst there is reasonable evidence for reducing hospitalisations there is less for reducing other poor outcomes such as mortality. A limitation and strength of this study is that it relies on national routinely collected data. Whilst this means we are able to present a national picture of medical readmissions in older people it means we are limited to data that is part of this collection. We have further limited our study to only considering demographic risk factors for readmission. As mentioned previously a range of other social and clinical risk factors are known to be highly correlated with readmission and these are being examined in the development of a predictive risk model for Waitemata DHB. Whilst considerable attention is given to ensuring the accuracy of the NMDS collection it is unlikely to be as accurate as data collected for a research study. For example it is known that hospital records continue to mis-record peoples ethnicities which may lead to inaccurate estimates of risk of readmission related to ethnicity.32 In conclusion, medical readmissions in older people in New Zealand are common and, if predicted and effectively prevented, represent an opportunity to improve peoples outcomes, reduce disparities and reduce health service costs. Other studies will be needed to show which interventions in these, or subgroups of these patients, are effective and cost-effective.

The New Zealand health system faces the triple challenges of improving quality of care and patient experience, health outcomes and equity, and getting best value from health resources.1 Reducing hospital readmissions offers the prospect of achieving these aims and is therefore a topic of considerable interest.District Health Boards (DHBs) have been asked to make improvements in productivity and efficiency, but not at the expense of quality of care. Hospital unplanned acute readmission rates are a well-established measure of quality of care and are included in DHB performance measures.2In 2010/11, across DHBs, 7.6-11.5% of admissions are followed by an unplanned acute readmission within 28 days. However, whilst readmissions have attracted considerable interest internationally, particularly in USA recently,3 there has been comparatively little analysis in New Zealand.4 This study describes acute medical readmissions in New Zealand hospitals with a particular focus on older patients.This work was initially undertaken to support the design of a Waitemata DHB programme to prevent medical readmissions amongst older people by improving the transition from hospital to home.Readmissions definedThere is no agreed definition of readmissions in the literature. Unplanned or acute admissions following and earlier admission are usually of greatest interest.5 Authors have used time periods varying between 1 day and 12 months after discharge to define a readmission.5-7 The time period chosen may depend upon the reason for interest in readmissions.Readmissions are frequently used as a measure of hospital care quality. In this case readmissions soon after discharge are more likely to be related to deficiencies in care. 8 However, since our interest is in interventions aimed at supporting the transition of patients from hospital back into the community a longer time period also seems relevant.People at risk of readmissionReadmission rates have been shown to vary considerably between countries, between areas within countries, and between hospitals.3,6,9Internationally a number of risks have been identified for increased risk of readmission including being older, male, lower education, some ethnicities, widowed or divorced and having poor social networks or living alone3,7,10-13In New Zealand surgical readmissions in the elderly have been shown to be more common in men, older people and M ori and Pacific people.4 Chance of readmission has also been related to a number of clinical factors such as health service use, diagnosis, co-morbidities, disability, and function.3,6-9,12-15 Readmission rates for medical admissions have been found to be higher than for surgical admissions.3Readmissions are of interest because it is thought that some are potentially preventable. Higher rates of readmission are associated with lower quality of care in hospital. 16 However, assessment of the proportion of readmissions that are potentially preventable vary widely from 5-71%.7,9,14 This reflects different periods used, the difficulty in deciding whether an admission is preventable, and the widely varying methods used to make this judgement.Of greater importance is whether interventions in hospital or in the community can lead to actual reductions in readmissions. A number of systematic reviews have examined this question, and found that interventions can be effective, although not all are.17-20The aims of this study are to investigate the degree to which acute medical readmissions might be a suitable target for improving the quality and efficiency of the health system.Specifically we aim to answer a number of questions: How common are medical readmissions in older people? Which groups are at greatest risk? What is the health and health system burden of people who are readmitted? What is the potential for prevention? Methods NZ publically funded hospital admissions for the period 1 April 2009 to 31 March 2010 were examined. Data from the Ministry of Healths National Minimum Data Set (NMDS) collection was obtained on hospitalisations of people admitted between 1 Feb 2009 and 31 June 2010 to allow recognition of admissions three months before and three months after index admissions. Mortality data for the period 1 April 2009 to 31 June 2010 was also obtained and linked to the hospitalisation data using encrypted NHIs. Index admissions were defined as acute medical admissions in NZ residents where the person had stayed overnight and where they had ended in a routine or self discharge. We included self discharges as we were less interested in readmissions as a marker of quality of care than as an opportunity for intervention and people who self discharge would still be offered such an intervention. Readmissions were defined as a further acute medical admission. We particularly considered 30- and 90-day readmission rates. Ethnicity in the NMDS collection is self identified and allows multiple ethnicities to be recorded. These were prioritised according to the Ethnicity Data Protocols for the Health and Disability Sector21 and then aggregated to M ori, Pacific, Asian, and Other (Other includes Europeans). Deprivation was assigned at Census Area Unit using the NZDep2006 Index of Deprivation which is an area based index.22 Rurality was also assigned by Census Area Unit and is taken from tables provided by the Wellington School of Medicine which were in turn based up Statistics New Zealand definitions.23 For univariate analyses chi squared tests were used for dichotomous outcomes and Wilcoxon rank-sum test for continuous outcomes. Binomial regression was undertaken for multivariate analyses. All analyses were undertaken using Stata v11.2 software 00ae. Results There were 217,323 acute medical admissions amongst 164,428 patients with a subsequent routine discharge in the study period. 95,318 of these admissions in 66,983 patients were in people 65 years and older. These are the focus of this study. Readmission by age groupReadmission was very common. Up to one-third of acute medical admissions were followed by another acute medical admission within the next 3 months in some population groups. Rates of readmission increased with age until the seventies and then tended to plateau. M ori and Pacific people had higher readmission rates for most age groups than Asians and Others. Readmission rates in M ori, Pacific, and Asians after the age of 90 years are not shown because of the small numbers of people in these older age groups. 30-day readmission rates showed a similar pattern (not shown but available on request). Figure 1. Adult medical readmission within 90 days of discharge by age and ethnic group, New Zealand 2009/10 The higher rates of readmission for older people led us to focus on older people when planning potential interventions. The remainder of the analysis in this paper focuses on people 65 years and over. Figure 2 Cumulative readmission rates for people 65 years and over by time after discharge, New Zealand 2009/10 The cumulative chance of readmission for an older person who is discharged after an acute medical admission increases over time reaching 10.8% (95%CIs 10.6-10.9%)by 30 days after discharge and 18.3% (95%CI 18.1-18.3%)after 90 days. As shown in Figure 2 second readmissions within 90 days were also not rare. Whereas 10.8% of individuals who were admitted went on to be readmitted within 30 days, 16.1% (95%CI 15.8-16.3%) of admissions were followed by a readmission within 30 days. This difference is due to people with high readmission rates being counted only once in the first analysis but multiple times in the second. Within 90 days, 27.8% (95%CI 27.5-28.1%) of admissions were followed by a readmission. When considering the impact of readmissions on the health system the proportion of admissions that are readmissions is also important. 13.8% (95%CI 13.6-14.0%) of all acute medical admissions in over 65 year olds in the study period were 30-day readmissions and 25.5% (95%CI 25.5%-26.0%) were 90-day readmissions. Groups at risk of readmissionM ori and Pacific were more likely to be readmitted than those in other ethnic groups. Table 1 shows the result of a multivariate model which included age, ethnicity, gender, deprivation and rurality as predictors of readmission within 30 days. Increasing age, increased deprivation and male gender were also associated with increased risk of readmission. Interestingly, people living in rural areas were slightly less likely to be readmitted. Table 1. Risk factors for readmission within 30 days of discharge for people 65 years and over, New Zealand 2009/10 (*Note: years over the age of 65 was use in the analysis rather than age) Morbidity and mortality of people who were readmittedWhilst it is not surprising, it is still worthy of comment that people who are being readmitted with acute medical problems are more unwell and have worse outcomes than people being admitted for their first admission. Table 2. Comparison of readmissions (within 30 days of discharge) and first admissions for mean admission complexity and outcomes for people 65 years and older Readmissions were more likely to be complex (coded as having a patient clinical complexity level other than 0), and had a longer mean length of stay, and a higher mean cost weight than first admissions. This indicates that people being readmitted were more unwell and/or more complex than first admissions. Readmission outcomes were also worse than first admissions. They were almost twice as likely to die within hospital or soon after discharge. Readmissions were also almost twice as likely to result in a further readmission. PreventionAs we were interested in designing interventions to prevent readmissions we were interested in what proportion of readmissions were potentially preventable. Unfortunately this is very difficult to determine from hospitalisation data. However, we have looked at two indicators that provide some information. A readmission might be more likely to have been preventable by better hospital or transition care if it was for the same disease as the preceding admission. Our analysis shows that 30.9% (95%CI 30.1-31.6%) of 30-day readmissions had the same diagnosis as the preceding admission (defined by having the same DRG code e.g. F62 heart failure) and 48.0% (95%CI 47.1%-48.8%) were in the same diagnostic group (defined by having the same first letter in the DRG code e.g. F circulatory). The second indicator we considered was whether the readmission was for an ambulatory sensitive hospitalisation (ASH). These admissions might potentially be preventable by linking people back into community care more effectively. Overall, 34.7% (95%CI 33.8-35.5%) of 30-day readmissions and 35.6% (95%CI 34.9-37.2%) of 90-day readmissions were for ASH conditions. When considering possible interventions to reduce readmissions the question arises whether we should target a few high risk diagnoses such as congestive heart failure (CHF) and chronic obstructive pulmonary disease (COPD) or instead design an intervention that reaches all people at high risk of readmission. We therefore looked at to what extent particular diagnostic groups or diagnoses made up the majority of readmissions. Table 3. Main diagnostic groups and diagnoses of readmissions within 30 days of an acute medical admissions in people 65 years and older, New Zealand 2009/10 Cardiovascular disease and respiratory disease together made up over 40% of readmissions. However, the four commonest diagnoses (angina/chest pain, CHF, COPD, and pneumonia) made up only 27% of all readmissions. Discussion Medical readmissions were common amongst older people and are a potentially worthwhile focus for intervention. A quarter of all acute medical admissions were preceded by another acute medical admission within 3 months. These admissions were more costly and complex than other admissions. If these readmissions could be reduced substantially there would be significant savings to the health system. In addition people being readmitted are at risk of poor outcomes as measured by mortality and further readmissions. Interventions to provide this group better care might reduce these poor outcomes. It is of concern that there were groups of our population who were at increased risk of these poor outcomes. These included M ori, Pacific people, people living in deprived areas and men. Possible reasons for this disparity are that these groups were less healthy, that the care we provided, either in hospital or in the community, was less effective for their needs, or that they had lower access to other community resources that enabled them to recover their health. We believe that these are serious concerns and warrant further study. It also behoves any one developing an intervention to prevent readmissions to make particular efforts to ensure that the interventions meet these groups needs. Readmissions are of interest because some are thought to be preventable by improvements in the health care system. However, as previously discussed, there is no agreement on what proportion of readmissions are preventable or indeed which groups of readmissions are preventable. A readmission may be preventable by better care whilst the patient is in hospital, by better organised and supported care of the transition back into the community, or by better ongoing care once the patient is established back in the community. Data presented here are limited in examining the potential for prevention of readmissions. Patients who are quickly readmitted to hospital with the same illness might be more likely to benefit from better hospital or transition care. This group made up nearly a third of readmissions over 30 days in our study. However, not all of these readmissions will be preventable, and many readmissions with unrelated illnesses might well benefit from the same improvements in care. Ambulatory sensitive hospitalisations (ASH) made up just over a third of readmissions. ASH may be related to quality of primary care and higher quality of care is thought to reduce admission from this group of diagnoses. This category of admission has not previously been examined in relationship to readmission. This high proportion of readmissions in this category suggests that primary care intervention may be an area worth considering. This study examines a population with a potential to improve care and reduce health system costs. However, to achieve these aims we must first be able to identify reliably those people at risk of readmission and then find interventions that can be shown to reduce readmissions and other adverse outcomes. CHF and COPD are well recognised chronic conditions that contribute significantly to readmissions in older people. There is ample evidence from systematic reviews that a range of interventions can reduce hospitalisations in these groups.24-28 Another approach may be to focus on those patients at high risk of readmission and support their transition back into the community. Panattoni et al, in examining predictive risk models to identify patients at high risk of emergency hospitalisation in New Zealand suggest development of cost-effective strategies.29 A predictive risk model was developed for Waitemata DHB in 2009 and a further model focused on older people is currently under development. Similar approaches have being reported internationally.11,30 A number of systematic reviews have examined the effectiveness of 2018transition interventions.17-20 One concluded that programmes that begin in hospital, are continued in the community and are multidimensional are more likely to be effective. 31 Unfortunately, whilst there is reasonable evidence for reducing hospitalisations there is less for reducing other poor outcomes such as mortality. A limitation and strength of this study is that it relies on national routinely collected data. Whilst this means we are able to present a national picture of medical readmissions in older people it means we are limited to data that is part of this collection. We have further limited our study to only considering demographic risk factors for readmission. As mentioned previously a range of other social and clinical risk factors are known to be highly correlated with readmission and these are being examined in the development of a predictive risk model for Waitemata DHB. Whilst considerable attention is given to ensuring the accuracy of the NMDS collection it is unlikely to be as accurate as data collected for a research study. For example it is known that hospital records continue to mis-record peoples ethnicities which may lead to inaccurate estimates of risk of readmission related to ethnicity.32 In conclusion, medical readmissions in older people in New Zealand are common and, if predicted and effectively prevented, represent an opportunity to improve peoples outcomes, reduce disparities and reduce health service costs. Other studies will be needed to show which interventions in these, or subgroups of these patients, are effective and cost-effective.