Causes and patterns of readmissions in patients with common comorbidities: retrospective cohort study
- “Causes and patterns of readmissions in patients with common comorbidities: retrospective cohort study” by Jacques Donzé and colleagues. (BMJ 2013;347:f7171, doi:10.1136/bmj.f7171).
Why do the study?
Work in emergency medicine or medical/surgical assessment can often yield interesting days. Patients come to hospital for various reasons; some present because they are acutely unwell, some present with exacerbations of chronic illnesses, and others present because they require social and multidisciplinary input into their care and home environments.
For a junior on the “frontline,” perhaps the most gut-wrenching moment is when you’re asked to see a patient who you’ve recently discharged from your care. You wonder whether you made a mistake in prescribing or if you discharged them too early. Then begins the attempt to diagnose the patient and, rather defensively, to find out whose fault it is that they’ve re-presented to hospital.
We know that about a fifth of patients who are discharged from hospital bounce back into hospital via GP referral or re-present through the emergency department within 30 days after discharge. Reasons for readmission include adverse drug events, avoidable infection, complications of procedures and treatments, and preventable acute exacerbations of chronic illnesses.
The authors of this study wanted to ascertain why patients get readmitted and offer solutions as to how we can reduce the numbers. They hypothesised that patients’ comorbidities could play an important role. Comorbidities can be defined as the simultaneous coexistence of more than one health condition in a single individual—for example, someone who has chronic obstructive pulmonary disease (COPD) and insulin dependent diabetes is said to have both comorbidities.
There is already evidence to suggest that a patient’s primary diagnosis on readmission (within 30 days) differs from the primary diagnosis responsible for the index—or original—admission, suggesting that there must be other factors at play. Additionally, studies have suggested that a higher incidence of comorbidity leads to an increased rate of readmission. 
What did the authors do?
The authors carried out a retrospective cohort study at a single academic 750 bed medical centre—Brigham and Women’s hospital, in Boston, MA. They included in the study all adult admissions, provided that the patient stayed in hospital for more than 24 hours.
Data were collected retrospectively; the authors collected data from the medical records of patients with an index medical admission between July 2009 and June 2010. Retrospective studies require you to look back at patient data because the exposure and outcome have already occurred—unlike in prospective studies, in which patients are followed forward. These studies tend to take less time to complete than prospective ones, making them generally less expensive to conduct.
A retrospective study design can be especially helpful for research questions with multiple outcomes of interest or in the study of diseases of low incidence. A common criticism, however, is that they can be subject to recall bias—that is, differences in the accuracy or completeness of data obtained from participants, commonly noted with interviews or questionnaires. A retrospective cohort design is entirely appropriate in studies such as this that rely on administrative data or data from patients’ records. This kind of data collection can be used in different ways—and not just for research purposes. Many audits that medical students might take part in will use patient data in this way.
Donzé and colleagues first looked at discharge summaries to determine the initial reason for a patient’s admission. Then they looked to see which patients were readmitted within 30 days of discharge. Knowing that patients do not always come back to the same hospital to which they were initially admitted, the study authors looked at the entire hospital network in Boston—the Partners network—covering 1900 beds. While not all readmissions will occur within this network, more than 80% of all readmissions after an index medical admission to Brigham and Women’s Hospital are captured.
A computer algorithm, SQLape, was applied to identify which readmissions were potentially avoidable. The SQLape software can analyse the coding for presentation and discharge for both primary index admissions and readmissions, taking into account diagnoses and procedures that took place during the visits.
Visits for follow-up treatment, elective surgery, or rehabilitation or development of disease in a new body system were seen as unavoidable. Readmissions for conditions related to those present during the index admission (including comorbidities) or complications of treatment were classified as potentially avoidable. The SQLape has previously been shown to be a validated and appropriate tool for measuring readmission rates, having a sensitivity and specificity of 96% when compared with medical chart review.
Finally, as the authors were interested in whether patients with specific comorbidities were at higher risk of potentially avoidable readmission, they calculated crude and adjusted relative risks for patients with each comorbidity (compared with those without the specific comorbidity). The comorbidities were neoplasm, diabetes mellitus, chronic heart failure, ischaemic heart disease, atrial fibrillation, COPD, and chronic kidney disease.
What did the study find?
The study confirmed what is already known about hospital readmission rates: about 20% of patients are readmitted within 30 days. In this hospital, 12 383 patients were discharged over the year, and, after exclusion of death before discharge, transfer to another hospital, and discharge against medical advice, 10 731 were included in the analysis. Of these, 2398 (22.3%) patients were readmitted within 30 days.
Nearly 36% of readmissions captured in the analysis (representing 8% of the total cohort) were classified as potentially avoidable. In the potentially avoidable readmissions, the most common primary readmission diagnoses were infection (11.6%), neoplasm (8.4%), and heart failure (7.1%) (table).
|Comorbidity||Five most likely primary readmission diagnoses|
|Neoplasm (n=441)||Neoplasm (16.3%; n=72)||Infection (12.9%; n=57)||Metabolic disorder (4.3%; n=19)||Gastrointestinal disorder (3.9%; n=17)||Renal failure (2.7%; n=12)|
|Diabetes mellitus (n=236)||Heart failure (12.7%; n=30)||Infection (8.9%; n=21)||Neoplasm (6.4%; n=15)||Ischaemic heart disease (4.7%; n=11)||Liver disorder (4.2%; n=10)|
|Chronic heart failure (n=207)||Heart failure (26.1%; n=54)||Infection (8.2%; n=17)||Ischaemic heart disease (7.7%; n=16)||Arrhythmia (2.9%; n=6)||Renal failure (2.9%; n=6)|
|Ischaemic heart disease (n=208)||Heart failure (13.5%; n=28)||Infection (11.5%; n=24)||Ischaemic heart disease (9.6%; n=20)||Arrhythmia (4.8%; n=10)||Renal failure (3.4%; n=7)|
|Atrial fibrillation (n=132)||Heart failure (18.2%; n=24)||Infection (11.4%; n=15)||Ischaemic heart disease (8.3%; n=11)||Arrhythmia (8.3%; n=11)||Gastrointestinal disorder (3.8%; n=5)|
|COPD (n=86)||Infection (16.3%; n=14)||Heart failure (16.3%; n=14)||Neoplasm (9.3%; n=8)||COPD (5.8%; n=5)||Venous thromboembolism (3.5%; n=3)|
|Chronic kidney disease (n=184)||Heart failure (20.7%; n=38)||Infection (8.2%; n=15)||Renal failure (7.1%; n=13)||Ischaemic heart disease (6.0%; n=11)||Liver disorder (2.7%; n=5)|
|Entire cohort (n=854)||Infection (11.6%; n=99)||Neoplasm (8.4%; n=72)||Heart failure (7.1%; n=61)||Gastrointestinal disorder (4.7%; n=40)||Liver disorder (3.9%; n=33)|
Heart failure was the most common reason for potentially avoidable readmission among patients with comorbidities such as chronic heart failure, ischaemic heart disease, atrial fibrillation, diabetes mellitus, and chronic kidney disease. Patients with cancer as a comorbidity were most likely to be readmitted because of neoplasm compared with other primary readmission diagnoses; and patients with COPD as a comorbidity were most likely to return with infection compared with other primary readmission diagnoses (table).
And how do we know it is possible to prevent these readmissions? Because the top five primary readmission diagnoses of avoidable readmissions were known complications of their respective comorbidity. For example, of 86 patients with a comorbidity of COPD, the most likely cause of readmission was an infective exacerbation, seen in 14 patients (16.3%), which is a known complication of COPD (table).
The authors found that the relative risk of potentially avoidable readmission was highest among patients with three specific comorbidities: cancer (relative risk 1.83), chronic heart failure (1.16), and chronic kidney disease (1.23). That is, patients with cancer had an 83% chance, chronic heart failure patients had a 16% chance, and patients with chronic kidney disease had a 23% chance of having a potentially avoidable readmission compared with patients who did not have the comorbidity.
Strengths and weaknesses
This study examines an important research question using a large sample of patients discharged from various hospital services. The retrospective nature of the analysis allowed researchers to extract data on several study outcomes (13 primary diagnoses associated with readmission). The authors used Medicare Severity-Diagnosis Related Groups (MS-DRG) codes to define the primary diagnosis for the index admission and any readmissions. These codes are said to correlate more closely with the reason for hospital admissions than the commonly used international classification of diseases (ICD) codes. According to the authors, the inclusion of data on comorbidities, in addition to primary diagnosis, makes this study unique.
The classification of readmissions as either unavoidable or potentially avoidable is a key feature of this study. The reliance on a single statistical algorithm, however, is likely to result in some degree of misclassification. Ideally, at least three practicing clinicians (two assessors, one to adjudicate) would go through each patient’s notes to determine whether the readmission was avoidable or not, paying attention to history and clinical presentation as well as to differential diagnoses at the time of admission and the discharge diagnosis. This would be extremely time consuming, cost intensive, and unrealistic to complete for 2398 patients. Therefore, it was appropriate to use a computer algorithm.
The authors indicate that as many as 20% of readmissions might not be captured in the Partners network, which could be a source of potential bias. Finally, these data come from a tertiary care centre in Boston, MA. Readmission patterns might differ in other jurisdictions, particularly in non-academic centres.
What does the study mean?
This study illustrates that over a third of hospital readmissions are potentially avoidable and reminds us that reasons why patients are readmitted are often related to their comorbidities. In particular, patients with comorbidities such as cancer, heart failure, and kidney disease are at the greatest risk of readmission, and greater care should be taken not only in managing their presenting complaint but also in managing their comorbidities at the same time.
There is a strong demand for interventions to reduce readmissions clinically, administratively, and financially. Yet hospital doctors have been trained to focus on the presenting complaint of a patient and might, as a result, pay less attention to their comorbidities. This study shows us that we may need to rethink the way we manage patients while they are in hospital, as well as how well we manage their care on discharge.Neil Chanchlani, foundation year 1 doctor1, Erin Russell, assistant editor2
1Whipps Cross University Hospital, London, UK, 2Canadian Medical Association Journal, Ottawa, ON, Canada
Correspondence to: firstname.lastname@example.org
Competing interests: None declared.
Provenance and peer review: Commissioned; not externally peer reviewed.
- Valderas JM, Starfield B, Sibbald B, Salisbury C, Roland M. Defining comorbidity: implications for understanding health and health services. Ann Fam Med 2009;7:357-63.
- Dharmarajan K, Hsieh AF, Lin Z, Bueno H, Ross JS, Horwitz LI, et al. Diagnoses and timing of 30-day readmissions after hospitalization for heart failure, acute myocardial infarction, or pneumonia. JAMA 2013;309:355-63.
- Van Walraven WC, Dhalla IA, Bell C, Etchells E, Stiell IG, Zarnke K, et al. Derivation and validation of an index to predict early death or unplanned readmission after discharge from hospital to the community. CMAJ 2010;182:551-7.
- Kripalani S, Roumie CL, Dalal AK, Cawthon C, Businger A, Eden SK, et al. Effect of a pharmacist intervention on clinically important medication errors after hospital discharge: a randomized trial. Ann Intern Med 2012;157:1-10.
- Halfon P, Eggli Y, Pretre-Rohrbach I, Meylan D, Marazzi A, Burnand B. Validation of the potentially avoidable hospital readmission rate as a routine indicator of the quality of hospital care. Med Care 2006;44:972-81.
Cite this as: Student BMJ 2014;22:g1667