Common Sequences of Emergency Readmissions among High-Impact Users following AAA Repair

Introduction The aim of the study was to examine common sequences of causes of readmissions among those patients with multiple hospital admissions, high-impact users, after abdominal aortic aneurysm (AAA) repair and to focus on strategies to reduce long-term readmission rate. Methods The patient cohort (2006–2009) included patients from Hospital Episodes Statistics, the national administrative data of all NHS English hospitals, and followed up for 5 years. Group-based trajectory modelling and sequence analysis were performed on the data. Results From a total of 16,973 elective AAA repair patients, 18% (n=3055) were high-impact users. The high-impact users among ruptured abdominal aortic aneurysm (rAAA) repair constituted 17.3% of the patient population (n=4144). There were 2 subtypes of high-impact users, short-term (7.2%) with initial high readmission rate following by rapid decline and chronic high-impact (10.1%) with persistently high readmission rate. Common causes of readmissions following elective AAA repair were respiratory tract infection (7.3%), aortic graft complications (6.0%), unspecified chest pain (5.8%), and gastrointestinal haemorrhage (4.8%). However, high-impact users included significantly increased number of patients with multiple readmissions and distinct sequences of readmissions mainly consisting of COPD (4.7%), respiratory tract infection (4.7%), and ischaemic heart disease (3.3%). Conclusion A significant number of patients were high-impact users after AAA repair. They had a common and distinct sequence of causes of readmissions following AAA repair, mainly consisting of cardiopulmonary conditions and aortic graft complications. The common causes of long-term mortality were not related to AAA repair. The quality of care can be improved by identifying these patients early and focusing on prevention of cardiopulmonary diseases in the community.


Introduction
Repair of abdominal aortic aneurysm (AAA) has been associated with very high readmission rates [1]. It is one of the top 7 conditions that account for 30% of all potentially preventable readmissions [1]. With the introduction of screening for AAA at a national level, the number of elective repairs has increased, and the number of repairs for ruptured abdominal aortic aneurysm (rAAA) is expected to decline. However, there are still a significant proportion of patients who suffer from rAAA and undergo repair [2]. e general patient population has a small subgroup of patients, the high-impact users, who have significantly higher rates of unplanned hospitalisations [3].
ey are shown to utilise as much as two-thirds of the health-care resources [4]. Risk profiling of the patient population to identify these patients provides health policymakers an opportunity to plan an optimal and individualised patient care by allocating appropriate resources and analysing trends in the health status of a population to prevent decline in the health status at a population level [5]. An increased readmission rate has been associated with higher mortality and discharge to care facility [6]. It may be that there are disease-specific patterns of readmissions in the high-impact users among those with AAA repair. It is important to assess causes of avoidable readmissions and whether these differ from those of the low-impact users.
Understanding the temporal order of the causes of readmissions may help us assess any repeated chain of events that occur in high-impact users. Previous studies have shown that order of events have significant impact on the outcomes of patients [7]. For example, the incidence of heart failure following atrial fibrillation was shown to be associated with a high mortality compared with those patients who were diagnosed with atrial fibrillation after heart failure [8]. What is not certain is whether distinct and common sequences of causes of readmissions are associated with the increase in readmission rate in the high-impact users. e study aimed to assess common causes of readmission and common sequences of causes of readmissions among highimpact users following AAA repair.

Hospital Administrative Data.
Hospital Episode Statistics (HES) data were used to extract information on patients diagnosed with AAA repair [9]. e data are collected by the Department of Health, Government of England, and include information on all the inpatient hospital admissions of all the patients admitted to public hospitals in NHS (National Health Service), England [9]. All patients, including private ones, who require emergency treatment are initially admitted in these hospitals. Each hospital admission is recorded as a "spell" consisting of "episodes" which denotes the care under each consultant during the patient's stay [10]. If a hospital admission requires a transfer to another hospital before the patient is discharged, then the whole hospital stay is recorded under "superspell" [10]. For the analysis, the information on each patient's spell or superspell was retrieved. All the conditions are coded using ICD-10 classification (International Classification of Diseases version 10). e procedures are coded using OPCS 4.7 (Office of Population Censuses and Surveys Classification of Interventions and Procedures version 4.7) [10].

2.2.
Patients with AAA Repair. All adult patients over the age of 18 who had primary AAA repair from the year 2006 to 2009 were included in the study. Patients who died during the follow-up period were included in the study as well. e patient cohort comprised two main types of repair, EVAR (endovascular aneurysm repair) and open repair. Initially, specific ICD-10 codes were used to identify AAA patients, as used in previous studies: elective AAA (I714, I719) and ruptured AAA (I713, I718) [11,12]. Afterwards, the type of repair of AAA was recognised using OPCS 4 codes, as used in earlier studies, and combined with ICD-10 codes to select the patient cohort [13,14]. e following OPCS codes were used: open repair (L18x, L19x, L20x, L21x, L25x) and EVAR (L26x, L27x, L28x). All the patients were followed up for a minimum of 5 years.

Statistical Methods.
e trajectory model was applied to the modified dataset which categorised individuals into different subgroups. e outcome was the annual number of emergency readmissions for each patient for each successive year during the 5-year follow-up. In order to determine the optimum number of subgroups within a population, the choice of model was based on the following criteria: smallest value of Bayesian Information Criteria (BIC), largest value for average posterior probability for each group, odds of correct classification (OCC) >5, and each trajectory with significant parameter estimates (p < 0.05). ese criteria are usually chosen to test for the model with best estimate of number of groups and predictors associated with them [15][16][17]. BIC is based, in part, on the likelihood function to measure the efficiency of the model to predict different groups in the data. Each is given a probability score for one's membership in the group. For each group, the mean of the probability scores of the individuals in the group is calculated and used as an indicator for adequate internal reliability if the value is more than 0.7 [18]. Odds of correct classification measure how improved the membership probability of individuals belonging to the in-group is as compared with other groups.
Sequence analysis in this research was conducted by software "TraMineR" in R statistical language [19]. Multiple clinical events of particular interest are fed into the program which allows it to search and identify order of them for each patient.
e administrative data were manipulated and shuffled so that the time and diagnosis of every emergency hospital admission during the follow-up period were aligned in successive columns in the data [20]. e list of common causes of emergency readmissions is mentioned in Appendix. Each row in the dataset demonstrated the causes of hospital admissions that occurred with each patient, in a chronological order. Sequence analysis was performed on the transformed dataset as it can search, identify, and visualise sequences of events with each patient [19]. Each diagnosis was coded with a unique alphabet. For each patient, the string of sequence of alphabets was created based on their chronological order. Common strings of events were identified within each group.

Elective AAA Repair.
e best-fit model (BIC-61509, AIC-61474) classified the patient population (n � 16, 973) into 2 groups based on their nonelective readmissions: Group 1 (82.0%) and Group 2 (18%) (Figure 1). Group 1 had persistently low rate of readmission and, therefore, was classified as low-impact, while, group 2 had constant high rate of readmission and was labelled as high-impact.
Within the patient population, the common causes of nonelective readmissions over 5-year period were respiratory tract infection (n � 748, 7.7%), chest pain (n � 543, 5.6%), aortic graft complications (n � 465, 4.7%), gastrointestinal haemorrhage (n � 462, 4.7%), and external injuries (n � 461, 4.7%). Of the total population, 57.6% had emergency readmission (n � 9791). Within low-impact users, 49.7% of them had emergency readmission (n � 6918), none of these patients had multiple readmissions but had similar common causes of emergency readmissions. Of the high-impact users, 82.8% of them had emergency readmissions (n � 2531). e common causes and sequences of readmissions are mentioned in Table 1. e time interval between each emergency readmission is displayed in Figure 2.

Ruptured AAA Repair.
e best-fit model (BIC-9936, AIC-9895) classified the patient population (n � 4144) into 3 subgroups based on their nonelective annual readmission rates: Group 1 (82.7%), Group 2 (10.1%), and Group 3 (7.2%) ( Figure 3). Group 1 had persistently low rate of readmission and, therefore, was classified as low-impact. ose with high readmission rates (high-impact users) were part of Group 2 and Group 3. Group 2 included chronic high-impact users because they had persistently high readmission rate. Group 3 were short-term high-impact users who initially had high readmission rate but then had rapid decline in readmission rate.

Discussion
Following AAA repair, high-impact users follow a distinct pathway of hospital care use. ey have persistently high readmission rate as compared with low-impact users. ey have significant number of patients with multiple emergency hospital admissions. Within rAAA repair patients, there was a third group, short-term high-impact, that initially had very high readmission rate followed by rapid Low-impact High-impact decline. ese patients were those with poor prognosis and did not survive after initial high readmission rate. Highimpact users following AAA repair had repeated hospital admissions for cardiopulmonary conditions. e common causes of long-term readmissions in patient populations were respiratory tract infection, exacerbation of COPD, external injuries, and aortic graft complications. ere has been a debate about the proportion of emergency readmission that can be prevented in the community. e implementation of penalty for hospitals with higher than expected 30-day all-cause readmission rate among medical patients sparked research into preventative measures for causes of readmission. However, it was found that most of the readmissions were not preventable [21,22]. It was not investigated what proportion of readmissions among vascular patients could be classified as preventable.
is study has shown that patients with multiple hospital admissions mainly suffer from cardiopulmonary and aortic graft complications which can be potentially avoidable. Furthermore, 30-day readmission is routinely assessed in the clinical practice following this policy, and most research is conducted around it. It is based on earlier studies which showed that most of the readmissions occur within 30 days of discharge from the hospital. However, long-term followup of the patients refuted previous evidence and indicated that patients continue to have high readmission beyond 30      days and readmissions can occur even after one year. What happens to the readmission rate of high-impact users among vascular surgery patients? Is it different from the rest of the patients? Do they continue to have high readmission rate in the long-term? ese questions are important, and the study attempted to answer them for better personalised care of the patients and possible role of management program among these patients. All causes of emergency admissions during the 5-year period following rAAA repair were examined. Common causes of readmissions were cardiopulmonary conditions, aortic graft complications, and external injuries: these were common in all subgroups. Similar causes of readmission were found in earlier studies evaluating causes of short-term readmissions following AAA repair, but most of them are based on patients with elective repair [23]. e common causes of readmission after elective repair were wound complication, chest infection, sepsis, and myocardial infarction [24]. ere was a higher rate of aortic graft complications and reintervention with EVAR use, but bowel obstruction, hernia repair, and gastrointestinal conditions were more common with open repair [24].
Since cardiopulmonary conditions are prevalent among sequences of multiple readmissions in high-impact users, a need for improved care to prevent exacerbation and progression of chronic cardiopulmonary conditions and infections in the community may be required to prevent multiple readmissions in high-impact users following AAA repair.
e sequence analysis identified that multiple readmissions mainly consisted of a vicious cycle of COPD, respiratory tract infection, and ischaemic heart disease. It may suggest that the primary care team should be vigilant to assess patients once they are discharged back to the community after AAA repair. Meticulous preventative measures such as regular flu vaccination to prevent chest infection, secondary preventative medical therapy for ischaemic heart disease, and regular follow-up for COPD should be followed in these patients to prevent them from becoming highimpact. Moreover, the high-impact users had higher readmissions for aortic graft complications than the other group. In contrast, majority of low-impact users did not have any multiple readmissions. All subgroups had similar common causes of readmissions, but high-impact users had significantly higher proportion of patients with multiple readmissions compared to other groups. e common sequences of causes of multiple readmissions in these patients consisted of exacerbation of COPD and chest infections. is was particularly important since it indicated that mere observation of common causes of readmission were similar in all subgroups. However, sequence analysis identified distinct sequences of readmissions that can be targeted by policymakers to prevent patients from having multiple readmissions.
Sequence analysis is a novel approach to study chronological order of events which can impact clinical outcome.
is technique had been used in social science and psychology to understand pattern of events during the lifecourse of participants in the study. Previous studies have not evaluated temporal sequence of readmissions but only provided cross-sectional crude analysis of the common causes of readmissions.
is technique can be applied to other adverse events or health-care services to assess deterioration in patient's health status. Hospital data did not contain information on the community events which led to hospital admissions. Further studies using primary care data will be helpful to understand all factors that lead to hospital readmissions among high-risk users.
is study had certain limitations despite the efforts to understand trends in readmission rates among subgroups of the patient population. e study was limited by the use of ICD and OPCS for identification of patient cohort, which is prone to coding errors in the administrative data collection [25]. With any retrospective cohort study, selection bias could not be ignored. e number of patients undergoing EVAR was small as compared with open repair. e patient cohort was selected to achieve follow-up of 5 years as it is counted as a minimum standard for the longterm follow-up by the Society for Vascular Surgeons [23]. During this period, the use of EVAR in rAAA was not widespread. Its use has increased in the last few years. e EVAR technique has significantly evolved with new catheters, graft stents, wires, and balloons for implantation. e clinicians are also more experienced in patient selection and procedural techniques [26]. Hence, analysis in the future is required to assess long-term morbidity associated with the procedure with advanced instrumentation. Furthermore, that the decision to perform EVAR is based on a complex interplay of availability of service, anatomical  complexity, and patient comorbidity, and so there are biases in the selection of patients [27]. erefore the differences between EVAR and open rAAA repair may be accounted for by this bias.
In conclusion, high-impact users form a significant number of patients who follow a distinct pathway of hospital care use following AAA repair. ey mainly suffer from certain cardiopulmonary conditions in the community that lead to their recurrent hospital admissions. Prevention of these conditions can improve their health status. e potential role of cardiac rehabilitation after aneurysm repair and separate management pathway from the rest of the patient population should be further explored.