Background Discharge against medical advice (AMA) occurs infrequently but is associated with poor outcomes. There are limited descriptions of discharges AMA in national cohorts of patients with acute myocardial infarction (AMI). This study aims to evaluate discharge AMA in AMI and how it affects readmissions.
Methods We conducted a cohort study of patients with AMI in USA in the Nationwide Readmission Database who were admitted between the years 2010 and 2014. Descriptive statistics were presented for variables according to discharge home or AMA. The primary end point was all-cause 30-day unplanned readmissions and their causes.
Results 2663 019 patients were admitted with AMI of which 10.3% (n=162 070) of 1569 325 patients had an unplanned readmission within 30 days. The crude rate of discharge AMA remained stable between 2010 and 2014 at 1.5%. Discharge AMA was an independent predictor of unplanned all-cause readmissions (OR 2.27 95% CI 2.14 to 2.40); patients who discharged AMA had >twofold increased crude rate of readmission for AMI (30.4% vs 13.4%) and higher crude rate of admissions for neuropsychiatric reasons (3.2% vs 1.3%). After adjustment, discharge AMA was associated with increased odds of readmissions for AMI (OR 3.65 95% CI 3.31 to 4.03, p<0.001). We estimate that there are 1420 excess cases of AMI among patients who discharged AMA.
Conclusions Discharge AMA occurs in 1.5% of the population with AMI and these patients are at higher risk of early readmissions for re-infarction. Interventions should be developed to reduce discharge AMA in high-risk groups and initiate interventions to avoid adverse outcomes and readmission.
- acute myocardial infarction
- quality and outcomes of care
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
Discharge against medical advice (AMA), with patients leaving hospital before the treating physician’s recommendation, occurs in 1%–2% of all medical admissions1 and it poses a challenge for physicians.
Discharge AMA may be associated with lack of trust and poor patient/provider communications, which may be markers for poor outpatient treatment adherence, and decreased utilisation of necessary healthcare services.2 A variety of factors have been associated with discharge AMA including financial constraints, family pressures, dissatisfaction with the hospital routine and treatment for substance-seeking behaviour.3 Discharge AMA is important because it is associated with greater risk of 30-day mortality.4
Management of acute myocardial infarction (AMI) includes pharmacotherapy and interventional treatments shown to improve prognosis. Among patients discharged AMA there are concerns of inadequate treatment for their index AMI. Continuation of medications such as β-blockers, ACE inhibitors, statins and dual antiplatelet therapy (DAPT) postdischarge is important to lower the risk of further cardiovascular events and death.5 6 Prolonged DAPT is necessary to avoid stent thrombosis in patients who undergo percutaneous coronary intervention (PCI). To our knowledge, only one study has previously evaluated hospital discharges AMA after AMI in a single state in USA in an era when PCI was less widespread.7 This study reported a 1.1% rate of discharge AMA with no difference in the crude rate of 30-day all-cause readmission between the discharge AMA and comparison groups (9.3% vs 8.4%)7
In this study, we examined the crude rates, trends and predictors of discharge AMA, and the association between discharge AMA and unplanned 30-day readmission in a contemporary national cohort of patients following an index admission with AMI.
Study design and participants
The Nationwide Readmission Database (NRD) is a publicly available database of all-payer hospital inpatients stays, developed by the Agency for Healthcare Research and Quality (AHRQ) as a part of the US Healthcare Cost and Utilisation Project.8 The data are drawn from 21 states that account for approximately half of the total US resident population and hospitalisations.9 10
In the current study, we included men and women aged 18 years or older who were hospitalised with a primary diagnosis of AMI between 2010 and 2014 and were either discharged AMA or discharged home. A primary diagnosis of AMI was defined by the following International Classification of Disease – ninth Clinical Modification (ICD-9) codes: 4100*, 4101*, 4102*, 4103*, 4104*, 4105*, 4106*, 4107*, 4108* and 4109* which is a combination of ST-elevation myocardial infarction (STEMI) and non-ST-elevation myocardial infarction (NSTEMI). Discharge AMA was defined from the variable ‘DISPUNIFORM’, which represents the disposition of patient at discharge. We excluded patients who were discharged in December (because they may not have had 30 days of follow-up), those who died during their index admission for AMI, those who had an elective readmission and those who were not discharged home/self-care or AMA. We only considered the first AMI admission within a calendar year for a patient.
Variables and outcomes
We used ICD-9 and Clinical Classification Software (CCS) of diseases codes to determine comorbidities, in-hospital procedures and outcomes. Alcohol misuse was defined by the AHRQ comorbidity measure for ICD-9 codes alcohol abuse. The 30-day unplanned readmissions were defined as first rehospitalisation after discharge within 30 days from admission for AMI that was not elective. The causes of readmissions were determined from the principle diagnosis based on CCS codes (online supplementary table 1). The primary end point was all-cause 30-day readmissions and readmission cause.
Supplementary file 1
Statistical analysis was performed using Stata V.14.0 (College Station, Texas, USA). A flow diagram was used to show the proportion of patients at each stage of the analysis and those who were readmitted. Descriptive statistics were used to compare patients who were discharged AMA compared with those discharged home, with further stratification depending on whether or not they were readmitted. Statistical differences between groups for continuous variables were tested using the t-test and for categorical variables the χ2 test. For all analyses, the survey estimation commands were used (eg, svy: logistic for multiple logistic regression), following the recommendations from AHRQ for analysis of survey data to account for the complex survey design of NRD. Using the survey estimation commands, national sample sizes were determined by applying the discharge weights to the crude sample and the weight was propagated to determine the patients excluded and the estimated final sample size analysed. Multiple logistic regression was used to determine independent variables associated with discharge AMA and the influence of discharge AMA on 30-day unplanned readmission. All variables were adjusted for in the models. A multivariable model was used to determine if discharge AMA was associated with readmissions for AMI. The excess admissions associated with AMI and excess deaths from readmission among patients discharged AMA was estimated by considering the crude rate of deaths and readmissions in the non-discharge AMA group compared with the observed crude rate of deaths and readmissions in the discharge AMA group. We further performed subgroup analysis for odds of AMI readmission depending on the subgroup of patients without a coronary angiogram and among those who had PCI. We also determined the crude rate of discharge AMA by length of stay.
There were 2663 019 patients with AMI between 2010 and 2014 captured in NRD. After exclusion of patients admitted in December (n=234 702), patients who died in hospital (n=1 36 047), patients with elective readmissions (n=88 166) and patients who were discharged to short-term hospital, care home or law enforcement (n=6 34 779), there were 1569 325 patients included in the analysis of which 10.3% had a 30-day unplanned readmission. The crude rate of discharge AMA was 1.45% in 2010, which increased to 1.49% in 2014.
Patients who were discharged AMA were younger (59.9 vs 63.5 years, p<0.001), more likely to be male (75.8% vs 66.4%, p<0.001) and more likely to be uninsured (14.6% vs 7.9%) (table 1). Patients who discharged AMA were less likely to have private insurance (13.4% vs 31.5%, p<0.001) and more had Medicaid (19.8% vs 9.1%, p<0.001). Smokers (58.0% vs 45.0%, p<0.001), alcohol misusers (9.6% vs 3.5%, p<0.001), patients with chronic lung disease (25.0% vs 17.5%, p<0.001) and renal failure (18.1% vs 14.4%, p<0.001) had higher crude rates of discharge AMA. In addition, patients discharged AMA were less likely to receive coronary angiography (46.4% vs 86.6%, p<0.001), be treated with PCI (27.8% vs 63.3%, p<0.001) or receive a drug-eluting stent (15.4% vs 46.2%, p<0.001). The crude 30-day unplanned readmission rate was 24.9% among patients who discharged AMA and 10.1% among patients discharged home. The mean time to 30-day readmission was shorter in the discharge AMA group (10.6 vs 14.1 days, p<0.001) and the discharge AMA group had longer length of stay at readmission (5.1 vs 4.5 days, p<0.001) and higher crude rate of mortality at readmission (4.2% vs 3.5%, p=0.047).
Independent predictors of discharge AMA included smoking (OR 1.66 95% CI 1.57 to 1.75, p<0.001), alcohol misuse (OR 1.49 95% CI 1.35 to 1.63, p<0.001), male sex (OR 1.92 95% CI 1.81 to 2.03, p<0.001) and younger age (OR 0.97 95% CI 0.96 to 0.97, p<0.001) (table 2). Patients with STEMI were more likely to discharge AMA (OR 1.16 95% CI 1.09 to 1.23, p<0.001). Variables associated with a reduced odds of discharge AMA included private insurance (OR 0.40 95% CI 0.36 to 0.43, p<0.001), receipt of coronary artery bypass graft (CABG) (OR 0.11 95% CI 0.08 to 0.13, p<0.001) and need for ICD/pacemaker insertion (OR 0.27 95% CI 0.18 to 0.40, p<0.001). Comorbidities associated with reduced odds of discharge AMA included heart failure (OR 0.72 95% CI 0.52 to 1.00, p=0.049), atrial fibrillation (OR 0.83 95% CI 0.77 to 0.89, p<0.001), renal failure (OR 0.79 95% CI 0.73 to 0.85, p<0.001), cancer (OR 0.67 95% CI 0.57 to 0.78, p<0.001), depression (OR 0.87 95% CI 0.79 to 0.96, p=0.004) and dementia (OR 0.69 95% CI 0.58 to 0.81, p<0.001). In terms of non-patient data, medium hospital bed size and urban location were associated with discharge AMA while patients admitted to teaching hospitals were less likely to discharge AMA.
Discharge AMA was an independent predictor of unplanned readmissions (OR 2.27 95% CI 2.14 to 2.40). Causes of 30-day unplanned readmissions according to discharge AMA status are depicted in figure 1 and online supplementary table 2. Patients with discharge AMA had more than a twofold increased crude rate of readmission for AMI (30.4% (95% CI 30.4% to 30.5%) vs 13.4% (95% CI 13.4% to 13.5%)) and a higher crude rate of admissions for neuropsychiatric reasons (3.2% (95% CI 3.2% to 3.2%) vs 1.3% (95% CI 1.3% to 1.3%)). Multivariable analysis reveals that discharge AMA was associated with increased odds of readmissions for AMI (OR 3.65 95% CI 3.31 to 4.03, p<0.001). We estimate that there are 1420 excess cases of AMI among patients who discharged AMA and 58 excess deaths during readmissions among patients discharged AMA.
Among patients who discharged AMA and did not receive an angiogram, the OR of readmission for AMI was 3.59 95% CI 3.18 to 4.04, p<0.001. For patients who received PCI, the OR of readmission for AMI was 3.65 95% CI 2.94 to 4.53, p<0.001.
The crude rate of discharge AMA peaked at 1 day length of stay while the corresponding peak for patients who were not discharged AMA was 2 days (online supplementary table 3).
Our analysis reveals that discharge AMA in patients admitted with an AMI occurs in less than 2% of patients. Nevertheless, these patients represent a high-risk cohort with a twofold increase in odds of 30-day unplanned readmission and a one in three chance of readmission with re-infarction within 30 days. Once differences in baseline characteristics are adjusted for, patients who discharged AMA have a fourfold increased odds of admission with re-infarction within 30 days. Our study adds to the current understanding of discharge AMA in AMI based on the study of Fiscella et al conducted nearly 20 years ago7 and show that the crude rates of discharge AMA have remained similar (1.5% vs 1.1% in a previous study).
We identify several variables associated with discharge AMA including smoking, alcohol misuse and younger men, although prevalent comorbidities such as obesity, heart failure, atrial fibrillation, renal failure, cancer and depression reduced the odds of discharge AMA. Patients who discharge AMA differ from those who are discharged home in that they are more likely to receive Medicaid or be uninsured, live in lower income areas, smoke, misuse alcohol and tend to be younger men. The study in California by Fiscella et al reported similar findings that discharge AMA was more common in patients who were younger, men, of low income, black, insured through Medicaid or uninsured, and had less physical comorbidity and greater mental health comorbidity.7 In addition, we observe important health service system-level elements associated with discharge AMA such as bed size, urban location and teaching hospital.11–13 Our results suggest that medium hospital bed size and those from an urban location are more likely to discharge AMA. Interestingly, we observed that patients from teaching hospitals were less likely to discharge AMA. The effect of teaching hospitals was considered in a previous state-wide study in California by Fiscella et al which found no significant differences in discharge AMA among teaching hospitals compared with non-teaching hospitals (11.1% vs 9.8%).7 Another qualitative study of 9 patients, 10 physicians and 23 nurses/social workers, despite being limited because the study was not generalisable to community or smaller hospitals, suggests that discharge AMA was more common in teaching hospital settings because patients felt confused and frustrated.14 However, the current study of patients all across USA showed that patients from teaching hospitals were less likely to discharge AMA (53.4% vs 57.8%), although the absolute difference was relatively small, hence the clinical relevance of this observation is unclear. Future work would need to both confirm this finding, and understand the mechanisms that underlie potential differences in discharge AMA among different institutional structures.
We found that a few comorbidities were associated with discharge AMA but there may be differences in why they show the association. Smoking and alcohol misuse may make it challenging for patients to seek the substance they desire so they discharge AMA. We also observed increased discharge AMA among patients with hypertension, diabetes and chronic lung disease. One possible reason may be that these chronic conditions may have been managed by community physicians or care teams and patients may feel more comfortable and supported by these teams so are more willing to leave hospital and return to their usual care providers once the acute problem is treated. In addition, previous myocardial infarction and PCI were associated with increased discharge AMA. We speculate that this may be related to a patient’s previous experience and felt that a period of observation once the problem was treated was not necessary so they choose to discharge AMA once the acute problem is treated. The literature suggests that discharge AMA stigmatises patients, reduces their access to care and can reduce the quality of informed consent discussions in discharge planning.15
The complexity of discharge AMA in AMI may be related to the extent of care and resulting outcomes will obviously depend on the point at which the patient is discharged. A patient who undergoes PCI or CABG and chooses to leave hospital will at least receive partial treatment compared with a patient who leaves prior to coronary revascularisation. It may be important for clinicians to be aware that despite differences in the extent to which patients are treated prior to discharge AMA, a physician still has a responsibility to advocate for a patient’s well-being so discharges AMA should be accompanied by reasonable efforts to coordinate a patient’s ongoing care.16
An interesting observation in the current study is that some comorbidities including heart failure, atrial fibrillation, cancer, dementia and renal failure reduce the odds of discharge AMA. Possible explanations include that patients recognise that their condition will require long-term care and discharge AMA may damage the relationship with care professionals. Dementia may remove the patient’s autonomy to make decisions to leave against advice. Reasons for discharge AMA in cardiovascular diseases have been previously explored in a qualitative study.14 Communications were identified as an area that required improvement and healthcare providers should be trained in cultural diversity, interpersonal skills, customer service and also be accurate and open about wait times.
In the setting of an AMI patients require treatment with potent antithrombotic agents and may undergo percutaneous revascularisation procedures that require prolonged DAPT including lifelong single antiplatelet therapy. Our study demonstrates that the performance of a coronary angiogram or PCI occurs less frequently in cases that discharge AMA. This may relate to the fact that patients leave prior to the possibility of performing cardiac catheterisation or that the clinician don’t offer these procedures due to a concern of non-compliance with post-PCI DAPT. Patients with NSTEMI may be more likely to leave prior to PCI but this is less likely in STEMI because PCI is an emergency procedure. The risks associated with discharge AMA among patients with NSTEMI includes the risks related to not getting PCI as well as PCI-related complications like stent thrombosis while those with STEMI are more likely to only have readmissions due to complications from PCI. We observed lower crude rates of readmissions for bleeding and renal failure among patients that discharge AMA. This may be related to patients not taking their medications if they discharged AMA, and due to the fact that these patients were younger and their baseline risk of bleeding is likely to be less. Among cases that did undergo PCI, patients discharged AMA were more likely to receive bare metal stents that require shorter DAPT regimes despite the evidence of poorer outcomes compared with drug-eluting stents.17 18
A key finding of the current study is the fourfold increase in odds of readmissions for AMI among patients who discharge AMA. There are a few possible reasons for this finding. Patients can discharge AMA at any point in their care so there will be heterogeneity in the treatments received by patients. We found that evidence-based diagnostic procedures and treatments including coronary angiograms, percutaneous coronary intervention (PCI) and coronary artery bypass graft (CABG) were lower in patients who discharged AMA. The potential unstable coronary lesion poses a significant risk when left untreated, compounded by the lack of secondary prevention with statins and antithrombotic therapies. Second, if a patient receives treatment but discharges AMA they may not receive secondary prevention medications such as antithrombotic medications and statins that would place them at increased risk of re-infarction. In addition, patients who undergo PCI may be discharged without appropriate antiplatelet therapy increasing the risk of stent thrombosis. Finally, the discharge AMA group may be discharged without the necessary tools for smoking cessation.
In the current study, we report an unplanned readmissions crude rate of 10.3% which is lower that the 19.3% described by Fiscella et al.7 This is likely because of better quality of care, more widespread adoption of PCI, antithrombotics and better provision of evidence-based therapies in contemporary practice. In the study by Fiscella et al, the authors did not study specific causes of readmission (apart from (acute coronary syndrome) ACS or non-ACS), or differentiate between the non-ACS causes of readmission.7 In contrast, our analysis provides more granular insight into both cardiovascular and non-cardiovascular causes for readmissions and shows that there are a broad range of causes of unplanned readmissions with important differences between the discharge AMA and non-discharge AMA groups.
Despite this lower overall lower crude rate of readmission our analysis reveals a nearly fourfold increase in odds of unplanned readmission for AMI compared with the twofold increase reported by Fiscella et al.7 This suggests that re-infarction for patients who discharge AMA has worsened over time, perhaps relating to the more widespread use of PCI as a treatment strategy, and hence a greater potential for stent thrombosis with the premature discontinuation of DAPT.
Among patients that choose to discharge AMA, measures should be developed to obviate potential risks such as prescription of DAPT and other secondary prevention medications prior to discharge or a means to deliver these in the community. Interventions should be developed across healthcare providers spanning secondary and primary care interface including pharmacy outreach programmes to enable prescription and/continuation of therapies in the community. We observed higher crude rates of neuropsychiatric reasons for readmission in patients who discharge AMA and care may be improved by early involvement of psychiatric services particularly in patients with a history of mental health conditions or substance abuse.
We speculate that one of the factors that may influence patients' decisions to discharge AMA is how ill or symptomatic they are which is influenced by the severity of the AMI. We found that patients who discharge AMA had less circulatory support (1.8% vs 2.9%), vasopressor use (0.3% vs 0.5%) and intra-aortic balloon pump use (1.7% vs 2.8%). In addition, these patients had fewer complications such as complete heart block (0.6% vs 0.9%), ventricular fibrillation (1.2% vs 2.3%), ventricular tachycardia (4.3% vs 5.4%), cardiogenic shock (2.1% vs 2.5%) and cardiac arrest (1.2% vs 1.5%). These findings may suggest that patients who have less severe AMI are more likely to discharge AMA.
There are several limitations to this study. The overall data are derived from five unique data sets corresponding to each year between 2010 and 2014 so there is no possible linkage between years and the same patient can appear more than once in different years. The data set does not capture pharmacotherapy data and the compliance/prescription of medications is unknown. The population at risk of readmission may be overestimated because of survivorship. We do not have data on out-of-hospital mortality which would reduce the population at risk of unplanned readmissions. Causes of readmissions were identified using the primary discharge diagnosis codes which may be subject to reporting biases. In the interest of reducing potential confounding, we determined adjusted ORs to estimate the association between collected variables and discharge AMA. However, ORs, which can approximate for rate ratio, have limitations because ORs may overestimate associations especially the case for events which are not rare like 30-day readmissions. Another limitation of the study is that we are unable to comment definitively about reasons for our findings due to the observational nature of this study. Furthermore, even though we are able to adjust for a variety of variables such as comorbidities, hospital and system related factors, and socioeconomic factors, these adjustments may not fully account for the extent of their effect on the models of the current study because of unmeasured confounders.
In conclusion, discharge AMA occurs in 1.5% of the population treated for an AMI and is associated with a greater risk of 30-day unplanned readmissions. These patients are at particularly high risk for readmission due to AMI, with a nearly fourfold independent increase in odds for these readmissions. Our multivariable analysis suggests that patients who are more likely to discharge AMA appear to be younger, male, uninsured, from low-income areas who were also smokers and misused alcohol. We estimate that there are 1420 excess cases of AMI among patients that discharge AMA. Interventions should be developed to reduce discharge AMA in high-risk groups to avoid adverse outcomes and readmission.
What is already known on this subject?
Discharge against medical advice (AMA) with patients leaving hospital before the treating physician’s recommendation, occurs in 1%–2% of all medical admissions and it poses a challenge for physicians.
There are limited descriptions of discharges AMA in national cohorts of patients with acute myocardial infarction (AMI).
What might this study add?
Discharge AMA in patients admitted with an AMI occurs in less than 2% of patients.
These patients represent a high-risk cohort with a twofold increase in odds of 30-day unplanned readmission and a fourfold increase in re-infarction within 30 days.
How might this impact on clinical practice?
Interventions should be developed to reduce discharge AMA in high-risk groups and avoid adverse outcomes and readmission when it occurs.
We are grateful to the Healthcare Cost and Utilization Project (HCUP) and the HCUP Data Partners for providing the data used in the analysis.
Contributors CSK and MAM were responsible for the study design and concept. CSK performed the data cleaning and analysis. CSK wrote the first draft of the manuscript and all authors contributed to the writing of the paper.
Funding The study was supported by a grant from the Research and Development Department at the Royal Stoke Hospital. This work is conducted as a part of PhD for CSK which is supported by Biosensors International.
Competing interests None declared.
Ethics approval This is anonymised data and there is no patient identifiable information. This study was determined to be exempt from review by the Medical Research Council’s “Does our study need NHS REC approval form” and was conducted in accordance with the HCUP Data Use Agreement.
Provenance and peer review Not commissioned; externally peer reviewed.