This study presents a new model for identifying health conditions among the RACF population using existing electronic health record data. We found that RACF residents had many comorbidities, and certain conditions were more likely to be under-reported in ACFI data than others. Agreement between the ACFI and EHR data was none to fair for most conditions we examined. However, agreement between the ACFI and EHR was substantial for several conditions that typically require a high level of assistance due to physical disability and/or have complex medication management. Based on these results, we caution policymakers and researchers from drawing strong conclusions about condition prevalence based on ACFI data for most conditions with the exceptions of Huntington’s disease, multiple sclerosis, diabetes, Parkinson’s disease, dementia and schizophrenia, paranoid or psychotic states. We also identified seven distinct clusters of multiple chronic conditions and found that the sickest most complex cluster are the residents who have the longest stays (C2). We identified a young cluster (C7) of residents whose conditions are associated with high healthcare and social service needs likely extending prior to RACF admission based on their high prevalence of mental and behavioural conditions and conditions associated with substance use disorders (e.g. liver disease, lung cancer), and also their potential lack of social support (largest proportion of unmarried people). The clusters also varied in terms of functional and cognitive status, reflecting different care needs.
Circulatory diseases affect a large majority of residents – which was expected based on previous studies – however, unexpectedly constipation had the highest prevalence of all the conditions we analysed. Although constipation can be an acute condition, it can also occur chronically, and we found that it was indeed a recurring problem for residents. Many residents identified as having constipation were flagged due to laxative use. We examined laxative use trends in a post-hoc analysis and found that among residents who used laxatives, the median number of laxative doses administered per resident during their stay was 684. RACF residents are at increased risk of constipation as a side effect of medications and other conditions, but constipation may be modifiable with ensuring optimal adequate hydration, physical activity and fibre intake. Unfortunately, there has been a trend of increasing use of supplements paired with decreasing spending on fresh produce in this setting(14), which may be exacerbating the problem of chronic constipation. These data highlight the need for attention to this overlooked issue that can have a substantial impact on the quality of life of residents; the focus needs to shift from treatment of constipation to prevention. Many of the other conditions that were revealed as highly prevalent, such as depression, arthritis, dementia and chronic lower respiratory disease, could be improved through a holistic approach to management that includes improvements in lifestyle behaviours, diet, physical activity and social participation in meaningful activities(15) which are currently not funded in Australia.
Compared to previous studies that have used ACFI data alone, our study found generally higher rates of health conditions among aged care residents(16). This was expected since the ACFI limits the number of conditions that can be reported. Furthermore, ACFI is not regularly updated, thus new conditions starting after RACF admission are less likely to be recorded in ACFI compared to the EHR. These differences were more pronounced for some conditions. For example, we found higher prevalence rates based on residents’ EHR data for the three most common conditions compared to those identified with ACFI data alone: dementia (58% vs. 48%), depression (54% vs. 23%) and arthritis (61% vs. 14%)(17). Our results suggest that ACFI data alone may generally underestimate condition prevalence, but the magnitude of underestimation varies by condition. Consistent with prior research, dementia can be reasonably well identified from ACFI data(18), but we suspect that dementia may still be underdiagnosed and underreported in this setting as previous research has shown(19). This trend in underreporting of conditions is supported by condition prevalence reported in Aged Care Assessment Program (ACAP) data(16) (an assessment conducted before RACF admission – i.e. before the ACFI). ACAP allows up to ten conditions to be reported, and studies that used ACAP(16) data have found generally higher prevalence rates for conditions than those based on ACFI data, but both sources report generally lower rates than our estimates from EHR data(17). As for survey data, our prevalence estimates for most conditions also tended to be higher; compared to estimates from the Australian Bureau of Statistics Survey of Disability, Ageing and Carers in 2015 for respondents in residential aged care, our prevalence estimates are similar for arthritis, but higher for osteoporosis, anxiety, depression, hypertension and diabetes(20).
Our study has several strengths. The use of EHR data with notes on conditions, medication administration data and ACFI assessments exploited all sources of existing relevant data in the aged care setting to provide the most comprehensive identification of conditions possible without primary data collection or time intensive chart reviews. This approach to identifying conditions is practical since it requires no new data collection in the facility. Another strength is that our large sample is demographically similar to the Australian aged care population and we expect that our estimates are generalisable nationally. Also, our regression approach corrected for potential correlation in facility-level degree of reporting. Finally, our comorbidity cluster analysis provided a much more detailed illustration of co-occurring conditions beyond what simple counts of conditions can provide, and it highlighted the complexity and diversity of needs of the RACF population.
Our findings have some limitations rooted in the fact that the ACFI and the EHR are designed respectively for funding purposes and for clinical care – not for epidemiological surveillance. There is likely some degree of underreporting of health conditions in the EHR, but the rate of underreporting and how it varies by condition is unknown. We suspect that some conditions may simply be more likely to be under-diagnosed and/or under-reported, such as osteoporosis. Our estimates for osteoporosis prevalence were well below the 86% estimate in this setting in the US(16), and osteoporosis was reported at lower rates than fractures for some clusters, although it is likely that most of these fractures are low-impact osteoporotic fractures. Lower prevalence of osteoporosis also occurred in clusters that were relatively more complex, which supports the hypothesis that more complex residents are less likely to have their osteoporosis documented and treated(22). The gold standard for identifying conditions would entail examining each resident and reviewing historical external GP records (since the GP is likely to change once a person enters residential care) and hospital records, however, this was not practical or feasible for this large of a sample. We have taken the first steps to fully utilise data collected within aged care systems; linking hospital and historical GP data should be pursued in future studies.