Comparison with Other Studies
A moderate 30-day unplanned readmission rate of 16.7% was identified, compared with an overall 30-day readmission rate of 5-29% in the United States [
40], an overall 28-day readmission rate of 15.3% in the United Kingdom [
41], and a 42-day readmission rate of 39-59% for patients discharged from a department of internal medicine in Switzerland [
42]. The differences probably arise from variations in methods and definitions, study criteria, and population groups with unplanned readmission. The differences may also be related to the quality and organization of health care, and the availability of social support services. In line with other studies, risk factors associated with 30-day unplanned readmission were male sex, older age, and lower socioeconomic status. Patients who were discharged to elderly care homes had a higher readmission rate than those discharged to their own homes. This could be related to this population being older and having a poorer health status, with a need for more psychosocial support, but could also be a reflection of the quality of care in elderly homes. This problem of unplanned readmission from elderly homes has been referred to as the "revolving door syndrome" in one review [
43]. An in-depth study on the quality of care in elderly homes is needed to address this issue. Our study also showed variations in the unplanned readmission rates by geographical hospital cluster, with a higher rate of readmission in Clusters 1, 3 and 7, in which the patients were older and poorer, and in need of residential care, thereby confirming the value of demographic factors in predicting the risk of readmission.
Our study showed that there was a significant difference in LOS and mortality rates between 30-day unplanned readmissions and all other admissions. The longer LOS and higher mortality rate suggests that patients readmitted within 30-days of discharge had a poorer functional status. After malignant neoplasms, the top three high-mortality medical conditions were aortic aneurysms, pneumonia, and chronic liver disease and cirrhosis. The results are in line with other studies suggesting that these conditions were associated with higher mortality as a result of unmet clinical needs and clinical complications (e.g., the need for surgery, infection, postoperative complications) [
44‐
47]. However, Silber et al found that the mortality rate may also be correlated with the failure rate (the tendency for patients with complications to die) as well as unmet clinical needs and complication rate [
48,
49]. Thus, it seems that the ability to discharge surviving patients who have experienced complications represent two different complementary components of quality that should both be identified when assessing quality of care. Comparing the 10 medical conditions, patients with cerebrovascular disease had the longest LOS, especially in rehabilitation institutions, probably because they required more residential medical rehabilitation and psychosocial support before discharge. While this medical condition had the longest hospital stay, the OR of hospital readmission was, however, less than 1. To maximize the quality of care, future studies need to determine why patients with cerebrovascular disease required a longer hospital stay but had lower unplanned readmission rates, and whether a longer hospital stay may facilitate better discharge planning, community support for psychosocial and rehabilitation care, and subsequently lower readmission rates. Echoing this thought, while patients with heart diseases had the shortest hospital stay in both acute and rehabilitation care, the OR of hospital readmission for heart diseases was greater than 1. Thus, future studies may need to review whether early discharge contributed to hospital readmission in some conditions.
After controlling for significant patient risk factors and geographical cluster, chronic liver disease and cirrhosis, malignant neoplasms, septicaemia, and heart diseases had high attributable risk for 30-day unplanned readmission. Stroke, diabetes mellitus, injury and poisoning, and aortic aneurysm had a lower attributable risk for 30-day unplanned readmission. In the past 10 years, there have been many hospital-based and community support programs introduced to help patients with stroke and diabetes mellitus, including specific "stroke wards" to enhance the rehabilitation of patients, and diabetes mellitus clinics to provide more frequent post-discharge support for patients. The outcomes indicate that a range of disease-specific programs can affect the rate at which patients are admitted for a second time. It is surprising that heart disease still ranks high in the 30-day unplanned readmission rates, even though there are many in-hospital and community programs established for such patients. The outcomes may be related to the life-threatening nature of heart disease. Further study is needed to explore the relationship between readmission rates and heart disease because the disease code includes a number of distinct forms of heart disease. For efficient use of resources in rehabilitation, it would be desirable to identify the patients at risk of readmission, and focused care, including comprehensive discharge planning, community support, and medical follow-up, should be reviewed for the five medical conditions with high risk-adjusted ORs for 30-day unplanned readmission.
Finally, we found that the attributed cost of 30-day unplanned readmission was high. The HA spent approximately US$8 million on unplanned readmissions within 30 days of discharge for five medical conditions in Hong Kong 2007. With a conservative approach, an unchanged readmission rate and inflation of 4.5%, it is projected that the HA will spend US$13 million on 30-day readmissions in 2017, with the cost being much higher when considering readmissions that occurred within 60 and 90 days. Over 10 years, the projected cost increase is 63%, mainly related to the growth of the Hong Kong population, which is projected to increase from 6.9 million in 2007 to 7.5 million in 2017 [
50], and HA expenditure on readmissions in future could well exceed our current estimates. Across the 10 common medical conditions, heart disease, pneumonia and malignant neoplasms contributed the greatest financial pressure to the healthcare system. The patient population and LOS were highest for heart disease (
n = 8,450; LOS 6.2 days), pneumonia (
n = 6,670; LOS 7.6 days), and malignant neoplasms (
n = 2,225; LOS 9.8 days). The expenditure could be markedly reduced by even small changes in the readmission rate. Scrutiny of the factors associated with these readmissions may lead to identification of unmet clinical, educational, and psychosocial needs. An audit of 811 readmissions concluded that hospital system factors accounted for 37%, followed by clinical factors at 38%, and patient factors at 21% [
32]. Once the causes of readmission are defined, research can focus on possible ways to improve the quality of care to reduce readmission rates, and consequently reduce the costs of readmission allowing relocation of resources.
Strengths and Weakness of the Study
The main strength of this research lies in its large population from the CMS database of all public hospitals in Hong Kong provided by the HA which is responsible for 90% of hospital services in Hong Kong. The data are cross checked for completeness and accuracy.
Several limitations of this review should be considered. First, disease codes based on the primary discharge diagnosis were used, and the severity of disease and other co-morbidities were not available. Also, there may be a potential systematic difference or error in data entry of codes, as error rates are higher with more specific diagnostic codes [
51]. However, it was believed that errors would be minimal in this study because all of the diagnostic codes were entered by the discharge physician. Second, while we explored the health outcome for various medical conditions in relation to readmission, data on the patients' quality of life, health status, functional status and satisfaction were unavailable. Finally, the cost data were incomplete because we were unable to access costs for follow-up, outreach community services, and human and other resources across various medical conditions; the overall costs may be higher than calculated. Nevertheless, the hospital cost, which contributed to the major portion of the total cost, was included.