Secondary Logo

Journal Logo

Original Article

Physician Practices in Against Medical Advice Discharges

Tummalapalli, Sri Lekha; Chang, Brian A.; Goodlev, Eric R.

Author Information
doi: 10.1097/JHQ.0000000000000227

Abstract

Introduction

Discharges against medical advice (AMA) account for 1–2% of all hospital discharges, and patients discharged AMA have disproportionately high health care costs and increased morbidity, mortality, and hospital readmissions.1-5 Although patient risk factors for AMA discharge are well-described, including younger age, male sex, substance and alcohol abuse, HIV/AIDS, liver disease, psychiatric diagnoses, and previous AMA discharges, the practice patterns of physicians who discharge patients AMA are less clear.6-11

Adhering to discharge best practices has been shown to improve transitions of care in vulnerable populations; however, discharge planning in AMA discharges is challenging because patients often leave without advance warning.12,13 Although there are no dedicated guidelines for discharging patients AMA, best practices include harm reduction strategies, determination of capacity, and a discussion of the risks and benefits of hospital discharge.14-16 Documentation of these discharge best practices in AMA discharges is suboptimal.17-19 The predictors of these discharge practices, such as patient and admission characteristics, have not been fully described to determine which patients receive worse quality care.20 Furthermore, whether adherence to discharge best practices is associated with outcomes such as 30-day hospital readmission rates is not fully understood.21

Understanding physician practices surrounding AMA discharges will provide more comprehensive insights into opportunities to improve transitions of care for this population. To address this knowledge gap, we examined four physician best practices in AMA discharges: scheduling follow-up appointments, documentation of informed consent, documentation of a risk/benefit discussion, and attending physician notification. We evaluated the association of patient and admission characteristics with physician discharge practices to better understand predictors of physician best practices and identify opportunities for improved transitions of care. We also hypothesized that adherence to discharge best practices would be associated with a decreased risk of hospital readmission.

Methods

Study Design and Population

We performed a cross-sectional study of a convenience sample of adult (age ≥18 years) inpatients who were discharged AMA from the Mount Sinai Hospital between January 1, 2014, and January 1, 2015. The Mount Sinai Hospital is a 1,134-bed urban tertiary referral center serving a diverse population in New York, NY. This study was reviewed and approved for human subjects research by the Institutional Review Board at The Icahn School of Medicine at Mount Sinai.

Patient Identification and Data Extraction

Patients discharged AMA were identified through queries of Premier administrative billing databases (Premier Research Services, Charlotte, NC), searching for patients with discharge status “Against Medical Advice/Elopement.” Demographic information for each patient was also extracted from this database. Two investigators (SLT and BAC) developed a standardized protocol for conducting chart reviews of identified patients, which was prespecified before data abstraction. This protocol consisted of reviewing the discharge summary, the “after visit summary” (a patient-facing summary of the hospitalization), and consecutive medical and nursing notes before discharge. Patient elopement without completing AMA discharge paperwork was determined from review of medical and nursing notes. One investigator (BAC) completed the chart reviews, and to ensure accuracy in chart review technique, a second investigator (SLT) reviewed a subset of patient charts independently and evaluated for inconsistencies in data abstraction. Any discrepancies were resolved by consensus from the senior author (ERG).

Patient and Admission Characteristics

Our study predictors included patient characteristics: age (dichotomized to above or below the study population median age), sex, race, and type of insurance and admission characteristics: primary service team, time of discharge, documentation of an anticipated AMA discharge (an “early warning”), and length of stay. For the purposes of the analysis, primary service teams were classified as either medical or surgical. Medical service teams included general medicine, neurology, geriatrics, pediatrics (for patients ≥18 years old admitted to this primary service), family medicine, ambulatory (procedural), rehabilitation medicine, or the cardiac care ICU; surgical service teams included general surgery, vascular surgery, gynecology, orthopedic surgery, cardiothoracic surgery, otolaryngology, plastic surgery, neurosurgery, or urology. Time of discharge was classified as morning shift (7 am–7 pm) or evening shift (7 pm–7 am).

Physician Discharge Practices and Hospital Readmission

To identify our primary outcomes, we reviewed the Project Re-Engineered Discharge (RED) Toolkit from the Agency for Healthcare Research and Quality and Project Better Outcomes for Older adults through Safe Transitions (BOOST) intervention from the Society of Hospital Medicine to identify discharge best practices that would be relevant to our population, as there are no dedicated discharge guidelines for patients discharged AMA.12,13 Based on these interventions, we identified the following key physician discharge practices as our primary outcomes: presence of one or more scheduled follow-up appointments in discharge paperwork, documentation of an informed consent discussion before AMA discharge, risk/benefit discussion before AMA discharge, and notification of the attending physician. Risk/benefit discussion is a component of informed consent, and these were classified as two separate outcomes for the purposes of our study. Our coprimary outcome was the incidence of one or more 30-day hospital readmissions.

Data Analysis

Summary statistics were used to quantify patient characteristics, admission characteristics, and outcomes. Chi-square and Fisher exact tests were used to evaluate the association between patient characteristics (age, sex, race, and insurance status) and admission characteristics (primary service team, time of discharge, “early warning,” and length of stay) with physician discharge practices. The unadjusted prevalence of physician discharge patterns by patient and admission characteristics were reported. The analysis had 80% power to detect a prevalence ratio associated with a predictor of approximately 1.65 at a significance level of 0.05 for an outcome occurring in 30% of cases with equal numbers of exposed and unexposed cases. We then tested physician discharge practices as predictors of 30-day hospital readmission. Multivariable logistic regression was performed to examine the association between physician discharge practices and 30-day hospital readmission, controlling for other discharge practices and admission characteristics. We performed an additional sensitivity analysis excluding patients who eloped, as these patients may have left before physicians could perform discharge planning. All analyses were performed using StataIC 15 statistical software (StataCorp, College Station, TX).

Results

Characteristics of Patients Discharged Against Medical Advice

We analyzed 213 AMA discharges representing 193 unique patients; for 18 discharges (8%), patients eloped. Among patients included in our study, the mean patient age was 51 years and 67% of patients were male. The majority of patients were admitted to a medical service team (86%) and the majority of discharges took place during the morning shift (65%). Twenty-six percent of discharges had a documented “early warning” before AMA discharge. Other patient and admission characteristics are shown in Table 1.

Table 1. - Patient and Admission Characteristics of AMA Discharges (N = 213)
Patient Characteristicsa
 Age (years) 51 [17]
 Sex
  Male 142 (67)
  Female 71 (33)
 Race
  White 55 (26)
  Black 85 (40)
  Asian 2 (1)
  Other 71 (33)
 Type of insurance
  Private 62 (29)
  Medicare 72 (34)
  Medicaid 75 (35)
  Self-pay 4 (2)
Admission characteristicsa
 Primary service team
  Medicine 183 (86)
  Surgery 30 (14)
 Time of discharge
  Morning shift 138 (65)
  Evening shift 75 (35)
 Documented “Early Warning”
  Yes 56 (26)
  No 157 (74)
 Length of stay (days) 3.6 [4.4]
aContinuous variables presented as mean [SD]. Categorical variables presented as number (percentage).
AMA = against medical advice.

Physician Discharge Practices

Among AMA discharges, 33% had follow-up appointments scheduled upon discharge. There was documentation of a risk/benefit discussion before discharge (69%), informed consent (63%), and notification of the attending physician (72%) in the majority of discharges (Figure 1).

Figure 1.
Figure 1.:
Physician practice patterns in AMA discharges (N = 213). AMA = against medical advice.

Patient and Admission Characteristics Associated with Physician Discharge Practices

Patient age, sex, type of insurance, time of discharge, or length of stay were not associated with the presence of postdischarge follow-up appointments in unadjusted analyses (Table 2). Black patients were more likely to have scheduled follow-up appointments after AMA discharge compared with white patients in unadjusted analyses (44% vs. 22%, p = .005). Patients discharged from a medical service were more likely to have follow-up appointments scheduled than those discharged from a surgical service (36% vs. 17%, p = .042). In sensitivity analyses, exclusion of patients who eloped did not change these findings (see Table 1, Supplemental Digital Content, http://links.lww.com/JHQ/A102).

Table 2. - Prevalence (Unadjusted) of Physician Practice Patterns by Patient and Admission Characteristics (N = 213)
Follow-up appointment, N (%) Informed consent, N (%) Risk/benefit discussion, N (%) Attending notified, N (%)
Patient characteristics
 Age (years)
  >52 (N = 103) 39 (38) 66 (65) 73 (71) 70 (68)
  ≤52 (N = 110) 31 (28) 67 (61) 73 (66) 83 (75)
 Sex
  Male (N = 142) 43 (30) 85 (60) 94 (66) 103 (73)
  Female (N = 71) 27 (38) 48 (69) 52 (73) 50 (70)
 Race
  White (N = 55) 12 (22)a 33 (60) 37 (67) 30 (55)b
  Black (N = 85) 37 (44)a 53 (63) 59 (69) 66 (78)b
  Other (N = 71) 19 (27)a 45 (63) 48 (68) 55 (77)b
 Type of insurance
  Private (N = 62) 15 (24) 36 (58) 42 (68) 48 (77)
  Medicare (N = 72) 32 (44) 47 (66) 52 (72) 45 (63)
  Medicaid (N = 75) 22 (29) 48 (64) 50 (67) 56 (75)
Admission characteristics
 Primary service team
  Medicine (N = 183) 65 (36)c 117 (64) 127 (69) 134 (73)
  Surgery (N = 30) 5 (17)c 16 (55) 19 (63) 19 (63)
 Time of discharge
  Morning shift (N = 138) 48 (35) 87 (64) 100 (72) 102 (74)
  Evening shift (N = 75) 22 (29) 46 (61) 46 (61) 51 (68)
 Documented “Early Warning”
  Yes (N = 56) 22 (39) 35 (63) 39 (70) 47 (84)d
  No (N = 157) 48 (31) 98 (63) 107 (68) 106 (68)d
 Length of stay (days)
  >2 (N = 89) 34 (38) 53 (60) 56 (63) 63 (71)
  ≤2 (N = 124) 36 (29) 80 (65) 90 (73) 90 (73)
ap = .005.
bp = .01.
cp = .042.
dp = .019.

Documented attending notification after AMA discharge was more common among black patients compared with white patients (78% vs. 55%, p = .01). Attending notification was more common among patients with a documented “early warning” (84% vs. 68%, p = .019). None of the predictors were associated with documentation of risk/benefit discussion or informed consent before AMA discharge (Table 2). When patients who eloped were excluded, those discharged in the morning shift were more likely to have a risk/benefit discussion documented (80% vs. 64%, p = .016, see Table 1, Supplemental Digital Content, http://links.lww.com/JHQ/A102).

Hospital Readmission Rate

Of patients discharged AMA, 23% were readmitted to the hospital one or more times within 30 days. Patients with scheduled follow-up appointments were more likely to be readmitted to the hospital within 30 days in unadjusted analyses (31% vs. 19%, p = .041). When adjusted for other discharge practices and admission characteristics, there was no difference in hospital readmission between patients with and without each physician discharge practice (Table 3).

Table 3. - Unadjusted 30-Day Hospital Readmission Rate and Adjusted Odds Ratio of 30-Day Hospital Readmission by Physician Practice Patterns and Admission Characteristics (N = 213)
Readmission ratea, N (%) aORb (CI)
Physician discharge practices
 Follow-up appointment
  Yes (N = 70) 22 (31)c 1.77 (0.90–3-46)
  No (N = 143) 27 (19)c 1 (ref)
 Informed consent
  Yes (N = 133) 33 (25) 2.09 (0.62–7.04)
  No (N = 79) 16 (20) 1 (ref)
 Risk/benefit discussion
  Yes (N = 146) 34 (23) 0.52 (0.15–1.84)
  No (N = 67) 15 (22) 1 (ref)
 Attending notified
  Yes (N = 153) 37 (24) 1.10 (0.51–2.39)
  No (N = 60) 12 (20) 1 (ref)
Admission characteristics
 Primary service team
  Medicine (N = 183) 45 (25) 1.59 (0.51–4.98)
  Surgery (N = 30) 4 (14) 1 (ref)
 Time of discharge
  Morning shift (N = 138) 35 (25) 1.53 (0.75–3.14)
  Evening shift (N = 75) 14 (19) 1 (ref)
 “Early Warning”
  Yes (N = 56) 18 (32) 1.77 (0.87–3.60)
  No (N = 157) 31 (20) 1 (ref)
aReadmission rates are presented as unadjusted number (percentage).
bOdds ratio adjusted for other discharge practices and admission characteristics (primary service team, time of discharge, and “early warning”).
cp = .041.

Limitations

There are several limitations to our results. This is a single-center study, and we were limited by small sample size given the low rate of AMA discharges. Given that a risk/benefit discussion is a component of informed consent, there was overlap between these two outcomes. We were unable to control for physician characteristics in our analysis, so it is possible that variability in discharge practices could have been attributable to physician instead of patient or admission characteristics. We did not account for patient comorbidities in our analysis, such as adjusting for Charleston Comorbidity Index, which may have resulted in residual confounding. Given multiple comparisons tested, it is possible that differences at a significance level of 0.05 may have been due to statistical chance alone. Finally, we were unable to assess adherence to follow-up appointments and were not able to differentiate between appointments made before hospitalization and those made as part of a hospital discharge plan. Nevertheless, our data do support the conclusion that discharge best practices were not uniformly adopted, and patients discharged AMA remain at high risk for hospital readmission. These data underscore the need for increased attention and research to best support this vulnerable population.

Discussion

We examined physician discharge best practices in AMA discharges and found that a minority of patients discharged AMA had follow-up appointments scheduled at discharge. Most discharges had documentation of informed consent, a risk/benefit discussion, and attending notification. Our analysis found that black patients and patients admitted to the medical service were more likely to have scheduled follow-up appointments, and black patients leaving AMA were more likely to have the attending notified. In unadjusted analysis, we found patients with follow-up appointments were more likely to be readmitted to the hospital but otherwise did not find an association between physician discharge practices and 30-day readmission.

Our finding that few patients discharged AMA had scheduled follow-up appointments is consistent with previous estimates.17 Interestingly, in our population, we found that documentation of informed consent, risk/benefit discussion, and attending notification was higher than previously described, whereas “early warning” was lower, which may reflect institutional variability in documentation or local practice patterns.17 An analysis of patients leaving AMA from the emergency department found most patients had follow-up care, whereas documentation of capacity and a risk/benefit discussion was suboptimal, which may indicate variability in documentation based on care setting.18

We found an increased likelihood of scheduled postdischarge follow-up among black patients that left AMA, which was also seen in a study of heart failure patients.22 However, this study and our current study did not assess adherence to scheduled follow-up appointments, which may be lower in non-white patients.23 However, another study of patients with heart failure found black patients were less likely to receive discharge instructions.24 The variability in discharge practices between racial groups reported in this study, and these previous studies suggests that there may be additional unmeasured factors that influence the relationship between patient race and discharge best practices.

We hypothesized that certain patient and admission characteristics would be associated with discharge best practices but did not find consistent evidence for this in our study. For example, we hypothesized morning shift discharges would be associated with improved adherence to discharge best practices, as hospitals and care teams are generally more well-staffed during the day. Although unadjusted analyses did not demonstrate such a pattern, analyses excluding patients who eloped did support increased documentation of risk-benefit discussions during the morning shift, consistent with our hypothesis, however, other best practices were not impacted in this analysis. Similarly, we hypothesized advance notice with an “early warning” of an AMA discharge documented in the medical record would improve discharge practices, given more time to adequately plan for discharge. Although “early warning” was associated with an increase in attending notification, it was not associated with an increase in follow-up appointments scheduled. These findings suggest that the time-limited nature of AMA discharges is not the only factor driving lack of adherence to discharge best practices.

Adherence to discharge best practices was not associated with a decreased risk of 30-day readmission, which may reflect a high baseline level of health care utilization in this population that is difficult to modify. In fact, patients with scheduled follow-up appointments were more likely to be readmitted to the hospital, which likely reflects a high burden of illness and utilization of health care in the patients who had high rates of readmission. The association between having a follow-up appointment and being readmitted may have been confounded by the possibility that patients with scheduled follow-up appointments may have had higher comorbidities and thus were more likely to be readmitted. Alternatively, patients who presented to a follow-up appointment could have been found there to be sick and were sent to the hospital for readmission. Implementation of a bundle of discharge best practice interventions has been effective in reducing readmissions for hospital inpatients and warrants further study in the AMA population.12,25

Our results highlight the difficulty in facilitating safe transitions of care for patients discharged AMA. Discharge AMA occurs in clinically high-risk groups, including patients with acute myocardial infarction, heart failure, and traumatic brain injury, which underscores the importance of thorough discharge planning.26-29 Although patient-focused interventions for patients leaving AMA have been described, such as integrated programs with harm reduction and support services, our results highlight the possible need for physician-focused interventions in this high-risk transition of care period.30 Involvement of all members of the interdisciplinary care team is crucial to the discharge planning process.31 Physician empathy for patients during AMA discharges may be further explored as a motivation to improve care for this population.32

Conclusions

In summary, we found adherence to discharge best practices in AMA discharges was inconsistent and suboptimal, particularly for scheduling follow-up appointments. We did not find an association between physician discharge practices and 30-day hospital readmission.

Implications

Physician-focused interventions to improve discharge best practices in this population deserve further study.

Author's Biographies

Sri Lekha Tummalapalli, MD, MBA, is a Nephrology Fellow in the Division of Nephrology, Department of Medicine, University of California, San Francisco, California.

Brian A. Chang, MD, is a Resident in the Department of Anesthesiology, Columbia University, New York, New York.

Eric R. Goodlev, MD, is Clinical Assistant Professor in the Division of Geriatrics and Palliative Care, Hospice and Palliative Medicine, Wynnewood, Pennsylvania.

Acknowledgments

The authors thank the Mount Sinai Advancing Clinical Excellence in Medicine grant program, which supported this research.

References

1. Baptist AP, Warrier I, Arora R, Ager J, Massanari RM. Hospitalized patients with asthma who leave against medical advice: Characteristics, reasons, and outcomes. J Allergy Clin Immunol. 2007;119:924-929.
2. Hwang SW, Li J, Gupta R, Chien V, Martin RE. What happens to patients who leave hospital against medical advice? CMAJ. 2003;168:417-420.
3. Aliyu ZY. Discharge against medical advice: Sociodemographic, clinical and financial perspectives. Int J Clin Pract. 2002;56:325-327.
4. Alfandre D, Yang J, Harwood K, et al. “Against medical advice” discharges among HIV-infected patients: Health and health services outcomes. J Assoc Nurses AIDS Care. 2017;28:95-104.
5. Garland A, Ramsey CD, Fransoo R, et al. Rates of readmission and death associated with leaving hospital against medical advice: A population-based study. CMAJ. 2013;185:1207-1214.
6. Alfandre DJ. “I'm going home”: Discharges against medical advice. Mayo Clinic Proc. 2009;84:255-260.
7. Saitz R. Discharges against medical advice: Time to address the causes. CMAJ. 2002;167:647-648.
8. Franks P, Meldrum S, Fiscella K. Discharges against medical advice: Are race/ethnicity predictors? J Gen Intern Med. 2006;21:955-960.
9. Spooner KK, Salemi JL, Salihu HM, Zoorob RJ. Discharge against medical advice in the United States, 2002-2011. Mayo Clinic Proc. 2017;92:525-535.
10. Ling S, Cleverley K, Brennenstuhl S, Bindseil K. Predictors of leaving an inpatient medical withdrawal service against medical advice: A retrospective analysis. J Addict Med. 2018;12:453-458.
11. Ti L, Ti L. Leaving the hospital against medical advice among people who use illicit drugs: A systematic review. Am J Public Health. 2015;105:e53-9.
12. Hansen LO, Greenwald JL, Budnitz T, et al. Project BOOST: Effectiveness of a multihospital effort to reduce rehospitalization. J Hosp Med. 2013;8:421-427.
13. Berkowitz RE, Fang Z, Helfand BK, Jones RN, Schreiber R, Paasche-Orlow MK. Project ReEngineered Discharge (RED) lowers hospital readmissions of patients discharged from a skilled nursing facility. J Am Med Directors Assoc. 2013;14:736-740.
14. Alfandre D, Brenner J, Onukwugha E. Against medical advice discharges. J Hosp Med. 2017;12:843-845.
15. Clark MA, Abbott JT, Adyanthaya T. Ethics seminars: A best-practice approach to navigating the against-medical-advice discharge. Acad Emerg Med. 2014;21:1050-1057.
16. Alfandre D, Schumann JH. What is wrong with discharges against medical advice (and how to fix them). JAMA. 2013;310:2393-2394.
17. Edwards J, Markert R, Bricker D. Discharge against medical advice: How often do we intervene? J Hosp Med. 2013;8:574-577.
18. Schaefer MR, Monico EP. Documentation proficiency of patients who leave the emergency department against medical advice. Conn Med. 2013;77:461-466.
19. Stearns CR, Bakamjian A, Sattar S, Weintraub MR. Discharges against medical advice at a county hospital: Provider perceptions and practice. J Hosp Med. 2017;12:11-17.
20. Alfandre D. Advancing the science of discharges against medical advice: Taking a deeper dive. Heart. 2018;268–269.
21. Kumar N. Burden of 30-day readmissions associated with discharge against medical advice among inpatients in the United States. Am J Med. 2019;132:708-717.e4.
22. DeVore AD, Cox M, Eapen ZJ, et al. Temporal trends and variation in early scheduled follow-up after a hospitalization for heart failure: Findings from get with the guidelines-heart failure. Circ Heart Fail. 2016;9:pii:e002344.
23. Leukhardt WH, Golob JF, McCoy AM, Fadlalla AM, Malangoni MA, Claridge JA. Follow-up disparities after trauma: A real problem for outcomes research. Am J Surg. 2010;199:348-352; discussion 53.
24. Yancy CW, Abraham WT, Albert NM, et al. Quality of care of and outcomes for African Americans hospitalized with heart failure: Findings from the OPTIMIZE-HF (organized program to initiate lifesaving treatment in hospitalized patients with heart failure) registry. J Am Coll Cardiol. 2008;51:1675-1684.
25. Adams CJ, Stephens K, Whiteman K, Kersteen H, Katruska J. Implementation of the Re-Engineered Discharge (RED) toolkit to decrease all-cause readmission rates at a rural community hospital. Qual Manag Health Care. 2014;23:169-177.
26. Kwok CS, Bell M, Anderson HV, et al. Discharge against medical advice after percutaneous coronary intervention in the United States. JACC Cardiovasc Interv. 2018;11:1354-1364.
27. Patel B, Prousi G, Shah M, et al. Thirty-day readmission rate in acute heart failure patients discharged against medical advice in a matched cohort study. Mayo Clin Proc. 2018;93:1397-1403.
28. Kwok CS, Walsh MN, Volgman A, et al. Discharge against medical advice after hospitalisation for acute myocardial infarction. Heart. 2019;105:315–321.
29. Marcoux J, Alkutbi M, Lamoureux J, Feyz M, Saluja RS, de Guise E. Discharge against medical advice in traumatic brain injury: Follow-up and readmission rate. Can J Neurol Sci. 2017;44:311-317.
30. Ti L, Milloy MJ, Turje RB, Montaner J, Wood E, Kerr T. The impact of an HIV/AIDS adult integrated health program on leaving hospital against medical advice among HIV-positive people who use illicit drugs. J Public Health (Oxf). 2017;39:e33–e9.
31. Baumrucker SJ, Carter GT, Stolick M, et al. Critically ill patient leaving against medical advice. Am J Hosp Palliat Care. 2018;35:1461-1466.
32. Machin LL, Goodwin D, Warriner D. An alternative view of self-discharge against medical advice: An opportunity to demonstrate empathy, empowerment, and care. Qual Health Res. 2018;28:702-710.
Keywords:

discharge planning; against medical advice; transitions of care; patient safety

Supplemental Digital Content

© 2019 National Association for Healthcare Quality