Characteristics of trauma patients that leave against medical advice: An eight-year survey analysis using the National Hospital Ambulatory Medical Care Survey, 2009–2016

Abstract

Background

Leaving against medical advice (AMA) is associated with increased readmission rates, fragmented patient care, and healthcare litigation. Understanding the factors associated with trauma patients leaving AMA from acute care settings will help guide better communication with trauma patients and improve patient satisfaction. This study aims to assess the sociodemographic and in-hospital care characteristics of trauma patients that leave AMA from acute care centers across the U.S.

Methods

We pooled and analyzed eight years of data (2009–2016) from the National Hospital Ambulatory Medical Care Survey. The outcome variable was whether the patient left AMA or not. The main predictors were the triage class, weekend presentation, health insurance status, the presence of chronic diseases, and the receipt of therapeutic and diagnostic procedures. The sociodemographic characteristics -age, sex, and race/ethnicity, were measured as potential confounders in the developed model. We performed logistic regression and reported the unadjusted and adjusted odds of leaving AMA as well as the 95% confidence intervals.

Results

The weighted percent of the trauma patient population that left AMA was 1.8%. The odds of leaving AMA decreased with advancing age, and increased among non-Hispanic Blacks, compared with non-Hispanic Whites. After adjusting for age, race, and gender, the odds of leaving AMA increased among patients that lacked health insurance (AOR: 1.86; 95% CI: 1.51–2.31), and had diagnostic procedures (AOR: 2.79; 95% CI: 2.32–3.36). The odds of leaving AMA reduced among trauma patients who were classified as emergent (AOR: 0.70; 95% CI: 0.50–0.98) and had therapeutic procedures (AOR: 0.39; 95% CI: 0.32–0.47).

Conclusion

Predicting trauma patients with increased odds of leaving AMA will inform intentional communication that may reduce leaving AMA rates and improve care.

1. Introduction

Leaving against medical advice (AMA) is a substantial healthcare challenge across the world and in the United States (U.S.).¹ About 500,000 U.S. patients leave AMA annually.^2,3 While this proportion represents a small percentage of the patients’ load in the U.S.,^2,4 decisions to leave AMA affect the individual, the caregivers, and the healthcare system.⁵ Leaving AMA is associated with stigma, and incomplete medical care,^1,5 resulting in increased likelihood to be admitted in a different equidistant hospital, fragmented care,⁶ and increased cost of care.⁷

The most common reasons for leaving AMA are financial constraints, socioeconomic status, dissatisfaction with treatment, and poor communication with healthcare professionals.^1,8, 9, 10 Generally, patients who leave AMA are more likely to be male, young, and without health insurance.^3,5,10, 11, 12 The practice of leaving AMA varies across clinical departments, with high rates occurring among patients admitted to the intensive care units, and emergency department and low rates among obstetric patients.^4,10,13,14 Unlike patients of other clinical specialties, trauma patients typically present to acute care centers from accidental and unexpected acute injury.³ Additionally, trauma patients are unlikely to make choices on the hospital, type of specialist, or type of procedure they need in acute events or emergencies.³ With acute care centers characterized by the tense environment,¹⁵ and life-threatening and emergency cases appropriately prioritized above non-emergent cases,¹⁶ the type of in-hospital care a trauma patient receives might be associated with a decision to leave AMA.

Understanding the characteristics of trauma patients that leave AMA is important in reducing the rates of leaving AMA. Very few studies assessed the pattern of leaving AMA among trauma patients,^3,17 though none of the studies assessed the influence of in-hospital care such as receipt of diagnostic or medical procedures as factors associated with leaving AMA. There is yet to be any study that identifies in-hospital patient care characteristics associated with leaving AMA among U.S. trauma patients. The study aims to assess the sociodemographic and in-hospital care characteristics of trauma patients that leave AMA from acute care centers across the U.S.

2. Methods

2.1. Study population

The National Hospital Ambulatory Medical Care Survey (NHAMCS) was introduced in 1973 to gather information about ambulatory care across emergency departments (EDs), outpatient units, and ambulatory surgery centers across the United States.¹⁸ This dataset is one of the few surveys that capture nation-wide in-hospital events, providing patient and provider characteristics across the United States.¹⁸ As of 2016, the mean response rate was about 77.9%.¹⁹ Data files from 2009 to 2016 were pooled for this study.

2.2. Inclusion and exclusion criteria

The NHAMCS has undergone changes over time.¹⁸ Prior to 2012, a second trauma variable that separated injury cases from poison and adverse reactions was not available. We eliminated cases of poison and adverse reactions by manually assessing the diagnosis between 2009 and 2011 and eliminating injury and poison cases. Additionally, chronic disease reporting increased from 5 reported diseases (2009–2011) to 22 reportable diseases (2012–2016). We, therefore, calculated the index of chronic diseases as a four-point categorical variable (no chronic disease, one, two, and three or more chronic diseases) as earlier used in the injury research literature.²⁰

The eight-year pooled data generated a sample population of 219,564 (Fig. 1). The primary inclusion criteria for our study were patients with physical injury. We excluded non-injury cases (n = 140,405), cases of poisoning and adverse reactions (n = 8284), and cases whose injury status was missing (n = 4915) or coded as unsure (n = 393). We further excluded cases that were dead on arrival (n = 18), died in the emergency room (n = 55), and left without being seen (n = 101). We limited the data to the participants aged 18 years and older (n = 49,561). The resulting dataset had 899 cases that left against medical advice and 48,672 that had normal discharges.

Fig. 1.

Data Selection Steps in generating a sample population of trauma cases from the National Hospital Ambulatory Medical Care Survey (NHAMCS): 2009–2016.

2.3. Outcome variable

The outcome variable was whether the patient left AMA or not. Trauma patients that presented at the acute care centers and emergency departments but were not duly discharged from hospital care were classified as having left AMA. Specifically, we classified cases that left AMA as trauma patients that left AMA in one of these two scenarios: left after they were triaged but had not received definitive care, or left after they commenced definitive care but before completing care. Leaving AMA was measured as a binary variable; 1 represented cases that left AMA in any of the two scenarios mentioned above, and 0, otherwise.

2.4. Exposure variable

The predictor variables were the social and health characteristics of the trauma patients. The events of interest were weekend admission, health insurance status, the triage class of the patient, previous discharge status, and the diagnostic and therapeutic procedures provided to the patient. Health insurance status was measured as a dichotomous variable. Weekend admission, measured as a dichotomous variable, was defined as trauma patients that presented at the emergency department on Saturdays and Sundays. The triage class was measured in five categories: non-urgent, semi-urgent, urgent, emergent, and immediate. Previous discharge status, defined as trauma cases that were discharged 72 hours prior to index presentation, was measured as a dichotomous variable. Diagnostic procedure was defined as a patient having at least one of any blood tests (such as complete blood count, arterial blood gases etc.), non-blood-based tests (such as urinalysis, electrocardiogram etc.) and imaging tests (such as x-rays, ultrasound etc.). Therapeutic procedure was defined as a patient having medical procedures such as bladder catheterization, cast application, limb splint, central line insertion, suturing/staple application, and other procedures. A list of the assessed diagnostic and therapeutic procedures (reported as diagnostic and medical procedures) is available in the NHAMCS documentation.^21,22 Diagnostic and therapeutic procedures were measured as dichotomous variables. We selected age, gender, and race as a priori confounders.

2.5. Analytical plan

We reported the unweighted frequency and weighted percentage of the variables of interest. A chi-square analysis was used to assess the association between the outcome and independent variables, and statistical significance was set at a p-value of less than 0.05. Logistic regression, using survey weights, was used to compute the unadjusted and adjusted odds of leaving AMA, with statistical significance set at a 95% confidence interval.

Data analysis was performed using SAS version 9.4/SAS Studio 3.7 (SAS Institute Inc, 2013). Since we pooled data from eight years, the final weight was computed by dividing the yearly sample weight by eight as guided by the formula described in the NHAMCS documentation.²³

3. Results

In this study, 1.8% of the trauma patients, representing 539,648 US trauma patient population, left against medical advice. The remaining population, an estimated 29,663,064 trauma patient population, had normal discharge (Table 1). Most of the trauma patients were non-Hispanic whites (66%), aged 18–35 years (39%) with equal gender distribution. About 29% of the patients visited the emergency department on weekends, while about 16% of the patients had no form of health insurance. Most of the trauma patients (37%) were triaged as emergent cases, while 6% were triaged as immediate. Approximately 27% of the patients had diagnostic procedures, while 51% had therapeutic procedures. The age and race of trauma patients were significantly associated with leaving AMA. Additionally, health insurance status, the presence of chronic disease, and receipt of diagnostic and therapeutic procedures were associated with leaving AMA (Table 1).

Table 1.

Frequency distribution of the patients’ characteristics from the National Hospital Ambulatory Medical Care Survey (NHAMCS): 2009–2016.

Variable	Totala		Left Against Medical Adviceb		Discharged Normallyc		p-value
	Unweighted Frequency (N = 49,561)	Weighted %	Unweighted Frequency (n = 889)	Weighted %	Unweighted Frequency (n = 48,672)	Weighted %
Patient’s Class
Non-Urgent	538	0.9	10	0.9	528	0.9	0.277
Semi-Urgent	3594	6.9	59	5.5	3535	7.0
Urgent	15,452	29.9	297	30.3	15,155	29.9
Emergent	18,539	37.4	301	35.6	18,238	37.4
Immediate	3169	6.0	69	8.3	3100	6.0
Unknown	8269	18.9	153	19.4	8116	18.8
Weekend Presentation
Yes	14,351	28.9	267	27.1	14,084	28.9	0.363
No	35,210	71.1	622	72.9	34,588	71.1
Health Insurance
No insurance	7892	16.2	225	25.5	7667	16.0	<0.001∗
Has insurance	35,796	71.5	505	54.0	35,291	71.9
Unknown	5873	12.3	159	20.5	5714	12.1
Previously discharged
Yes	2036	4.0	42	5.1	1994	3.9	0.154
No	42,008	85.0	735	81.7	41,273	85.1
Unknown	5517	11.0	112	13.2	5405	10.9
Presence of chronic diseases
Yes	36,572	70.5	648	68.4	35,924	70.6	<0.001∗
No	12,358	27.9	201	25.8	12,157	27.9
Unknown	631	1.6	40	5.8	591	1.5
Diagnostic procedures
Yes	13,830	27.4	441	50.1	13,389	27.0	<0.001∗
No	35,095	71.1	409	45.2	34,686	71.6
Unknown	636	1.5	39	4.7	597	1.4
Therapeutic procedures
Yes	25,644	51.0	242	27.5	25,402	51.4	<0.001∗
No	22,711	46.4	604	66.1	22,107	46.1
Unknown	1206	2.6	43	6.4	1163	2.5
Age categories
18–35 years	19,190	39.4	401	46.1	18,789	39.3	<0.001∗
36–55 years	16.397	32.6	341	35.4	16,056	32.5
56–75 years	9037	18.2	125	14.5	8912	18.2
>75 years	4.937	9.9	22	4.0	4915	10.0
Gender
Male	24,644	50.0	444	51.1	24,200	50.0	0.603
Female	24,917	50.0	445	48.9	24,472	50.0
Race/Ethnicity
Non-Hispanic White	32,670	66.2	501	56.5	32,169	66.4	<0.001∗
Non-Hispanic Black	9047	18.6	235	27.2	8812	18.4
Hispanics	6090	12.2	119	13.2	5971	12.1
Other Races	1754	3.0	34	3.2	1720	3.1

^aRepresents estimated 30,202,712 hospital population.

^bRepresents an estimated 539,648 hospital population.

^cRepresents an estimated 29,663,064 hospital population.

For most of the years between 2009 and 2016, trauma patients aged 18–35 had the highest proportion of leaving AMA while patients aged 55 years and older had the lowest proportion of leaving AMA (Fig. 2). Additionally, for most years between 2009 and 2016, non-Hispanic Blacks had the highest leaving AMA rates while non-Hispanic Whites had the lowest leaving AMA rates for most years (Fig. 3). Although all the racial/ethnic trends show staggered patterns of leaving AMA, non-Hispanic Blacks demonstrated an increasing pattern of leaving AMA trend.

Fig. 2.

Trend of the weighted proportion of trauma patients that leave against medical advice (AMA) between 2009 and 2016 by age groups.

Fig. 3.

Trend of the weighted proportion of trauma patients that leave against medical advice (AMA) between 2009 and 2016 by race/ethnicity.

In the unadjusted model, trauma patients who were triaged as semi-urgent and emergent were 43% (Odds Ratio (OR): 0.57; 95% CI: 0.36–0.91) and 31% (OR: 0.69; 95% CI: 0.49–0.97) less likely to leave AMA compared to those triaged as immediate cases, assuming all other factors remained constant (Table 2). Additionally, trauma patients without health insurance had two-fold increased odds of leaving AMA compared to those with health insurance (OR: 2.12; 95% CI:1.73–2.61), assuming all other factors remained constant. Trauma patient that had diagnostic procedures had about 3-fold increased odds of leaving AMA (OR: 2.95; 95% CI: 2.46–3.53) while those that had therapeutic procedures had 63% reduced odds of leaving AMA (OR: 0.37; 95% CI: 0.31–0.46), assuming all other factors remained constant. The odds of leaving AMA decreased as age increased, with patients aged 18–35 years, 36–55 years, and 56–75 years having a 2.9 fold (95% CI: 1.63–5.21), 2.7 fold (95% CI: 1.50–4.84), and 2.0 fold (1.07–3.62) increased odds of leaving AMA, respectively as compared to patients aged 75 years and older. Additionally, non-Hispanic Blacks had 74% (OR: 1.74; 95% CI: 1.41–2.15) increased odds of leaving AMA as compared to non-Hispanic Whites, assuming all other factors remained constant.

Table 2.

Unadjusted odds ratio of the patients’ characteristics on discharge against medical advice using the National Hospital Ambulatory Medical Care Survey (NHAMCS): 2009–2016.

Variable	Unadjusted Odds Ratio
Patient’s Class
Non-Urgent	0.76 (0.31–1.88)
Semi-Urgent	0.57 (0.36–0.91)
Urgent	0.74 (0.52–1.04)
Emergent	0.69 (0.49–0.97)
Immediate	Ref
Weekend Presentation
Yes	0.92 (0.76–1.11)
No	Ref
Health Insurance
No insurance	2.12 (1.73–2.61)
Has insurance	Ref
Previously discharged
Yes	1.35 (0.87–2.09)
No	Ref
Presence of chronic diseases
Yes	1.05 (0.85–1.29)
No	Ref
Therapeutic procedures
Yes	0.37 (0.31–0.46)
No	Ref
Diagnostic procedures
Yes	2.95 (2.46–3.53)
No	Ref
Age categories
18–35 years	2.92 (1.63–5.21)
36–55 years	2.70 (1.50–4.84)
56–75 years	1.96 (1.07–3.62)
>75 years	Ref
Gender
Male	1.05 (0.88–1.25)
Female	Ref
Race/Ethnicity
Non-Hispanic Black	1.74 (1.41–2.15)
Hispanics	1.28 (0.98–1.66)
Other Races	1.23 (0.78–1.96)
Non-Hispanic White	Ref

After adjusting for age, race, and gender, patients classified as emergent were 30% less likely to leave AMA compared to those classified as immediate (Adjusted Odds Ratio (AOR):0.70; 95% CI: 0.50–0.98), and patients without health insurance were two times more likely to leave AMA compared to those with health insurance (AOR: 1.86; 95% CI: 1.51–2.31). The adjusted odds of leaving AMA among patients who had diagnostic procedures remained significantly elevated (AOR: 2.79; 95% CI: 2.32–3.36) while the adjusted odds of leaving AMA remained significantly decreased among trauma patients who had therapeutic procedures (AOR: 0.39; 95% CI: 0.32–0.47), assuming all other factors remained constant (Table 3).

Table 3.

Adjusted odds ratio of the selected patients’ characteristics on discharge against medical advice using the National Hospital Ambulatory Medical Care Survey (NHAMCS): 2009–2016.

Variable	Adjusted Model 1a	Adjusted Model 2b
Patient’s Class
Non-Urgent	0.80 (0.33–1.97)	1.32 (0.53–3.31)
Semi-Urgent	0.63 (0.39–1.00)	1.10 (0.68–1.78)
Urgent	0.80 (0.57–1.13)	1.19 (0.83–1.71)
Emergent	0.70 (0.50–0.98)	0.84 (0.59–1.21)
Immediate	Ref	Ref
Weekend Presentation
Yes	0.92 (0.76–1.11)	0.93 (0.76–1.13)
No	Ref	Ref
Health Insurance
No insurance	1.86 (1.51–2.31)	1.88 (1.51–2.33)
Has insurance	Ref	Ref
Previously discharged
Yes	1.35 (0.87–2.10)	1.25 (0.79–1.98)
No	Ref	Ref
Presence of chronic diseases
Yes	0.96 (0.78–1.18)	0.76 (0.61–0.94)
No	Ref	Ref
Diagnostic procedures
No	2.79 (2.32–3.36)	2.65 (2.19–3.22)
Yes	Ref	Ref
Therapeutic procedures
No	0.39 (0.32–0.47)	0.43 (0.35–0.53)
Yes	Ref	Ref

^aEach variable adjusted for age, gender, and race.

^bAll the variables in the model pooled in a single model in addition to age, gender, and race.

We created a predictive model by inserting all the health and sociodemographic factors to assess the odds of leaving AMA among trauma patients (Table 3). The significant predictors of leave AMA were race/ethnicity, health insurance, the presence of chronic diseases, receipt of diagnostic and therapeutic procedures. Non-Hispanic Blacks had 44% increased odds (AOR: 1.44; 95% CI: 1.15–1.79) of leaving AMA compared to non-Hispanic Whites. Trauma patients without health insurance were 1.9 times (AOR: 1.88; 95% CI: 1.51–2.33) more likely to leave AMA compared to those with health insurance. Trauma patients with co-existing chronic diseases were 24% less likely to leave AMA compared to those without co-existing chronic medical conditions (AOR: 0.76; 96% CI: 0.61–0.94). Trauma patients who received diagnostic procedures had approximately three-fold increased odds of leaving AMA (AOR: 2.65; 95% CI: 2.19–3.22), while those who had therapeutic procedures were 57% less likely to leave AMA (AOR: 0.43; 95% CI: 0.35–0.53). A predictive model that included the interaction effect of the receipt of therapeutic and diagnostic procedures (table not shown) revealed that those that had both diagnostic and therapeutic procedures were 62% less likely to leave AMA, assuming all other factors remained constant ( AOR: 0.38, 95% CI: 0.24–0.62).

4. Discussion

In this study, a small proportion of the trauma patient population left AMA. The odds of leaving AMA among trauma patients decreased with advancing age, and non-Hispanic Blacks were more likely to leave AMA than non-Hispanic Whites. In predicting trauma patients that will leave AMA, non-Hispanic Blacks, patients without health insurance, and those that received diagnostic procedures were more likely to leave AMA, while patients who had co-existing chronic medical conditions and received therapeutic procedures were less likely to leave AMA.

Previous studies on the pattern of leaving AMA across all specialties have been estimated to range from 1 to 3% U.S. patient population.^2,4,24 However, leaving AMA varies by clinical specialty and diagnosis. Earlier studies have reported leaving AMA rates in excess of 10% among patients managed for substance abuse and HIV,^25,26 while the rates are lower in specialties such as respiratory and neurological medicine. For example, Hoyer et al.²⁷ reported that about three percent of patients with neurological symptoms leave emergency departments AMA. Baptist et al.²⁴ reported that 1.2% and 3.7% of patients admitted for pneumonia and asthma left AMA, respectively. Olufajo et al.,³ using the California State hospital database, reported that 1.7% of U.S. trauma patients between 2007 and 2011 left AMA. Using a nationally representative and weighted sample, we report a similar leaving AMA rate of less than two percent.

This study demonstrated that patients that leave AMA are more likely to be younger than 55 years old and uninsured. Previous studies on trauma patients who leave the ED AMA have found relatively consistent findings in regard to age and insurance status.28, 29, 30 Trauma patients that leave AMA are more likely to be younger than patients that had normal discharge.²⁸ Our finding of the increased rates of leaving AMA among trauma patients less than 55 years is consistent with other research that utilized the Trauma Quality Improvement Program database.³⁰ Patients that were uninsured were also more likely to leave AMA in this study and this is consistent with prior research on patient dispositions post-trauma.^29,30 Haines et al.²⁹ used the National Trauma Data Bank and found that patients that were uninsured were 2.7 times more likely to leave AMA.²⁹ Our study also found that non-Hispanic Black trauma patients had the highest rates of leaving AMA, and these results are consistent with prior research.^28,29

This study reports an interesting association between the odds of leaving AMA and receipt of diagnostic and therapeutic procedures. The latter is associated with decreased odds of leaving AMA while the former is associated with increased odds. A possible explanation of this relationship could lie in the social and behavioral patterns of patients. Patients may have a self-rated opinion of their health status and may be unwilling to engage in a series of tests that will help their caregivers obtain a diagnosis. Upon receiving therapeutic care associated with evidence of improvement, however, they may be unwilling to leave AMA. Earlier studies have reported that patients that leave AMA perceive their clinical state has improved despite clinical evidence of significant pathology³¹ and trauma patients that needed major surgery were less likely to leave AMA.³² Hence, the perception of health status and the severity of their clinical conditions may influence or deter their decision to leave AMA. Faulty perception of health status may underlie the reason for seeking to leave AMA, seek readmission rate, and their increased risk of morbidity and mortality.^3,24

This study identifies the characteristics of trauma patients who leave AMA. Patients’ perception of their clinical diagnosis, the care they receive, and the therapeutic procedures they obtained may influence their decision to leave AMA. The possibility exists that their perception of health status might inform their decision to leave AMA. Effective communication might hold the key to reducing the rates of trauma patients that leave AMA. Assessing the role of effective information dissemination to trauma patients presents areas of future research. Further, negotiating with patients to complete the recommended treatment becomes vital as the patient and the healthcare system are negatively affected by leaving AMA.^2,5,10,33 Leaving AMA is associated with increased readmission rates, fragmented patient care, and healthcare litigation.^2,5,10,33 A qualitative study that assessed physician and patients’ notes on reasons for requesting and denying discharge, respectively, identified the role of strained patient-physician communication in decisions to leave AMA.³⁴ The possibility exists that involving medical social workers to serve as trained advocates for patient completion of care may reduce the leaving AMA rates. Bahadori et al.³⁵ recommended that all cases of leaving against medical advice be attended to by the hospital’s medical social workers. Anticipatory discussion with patients at greater odds of leaving AMA on the challenges associated with and the need to complete care may yield better results than discussing the need to complete care after the patient signifies interest to leave AMA.

This study has its limitations. The retrospective nature of the study makes it impossible to establish causality. Earlier studies have reported waiting time as a factor associated with leaving AMA.²⁷ We were unable to report this variable because of the large proportion of missing data among those who left AMA compared to those who did not. While coverage errors in the NHAMCS are less likely due to its complex three-stage probability sampling method, the possibility of data entry errors and selection bias cannot be eliminated. Although non-response bias cannot be eliminated, the response rate of the NHAMCS is comparable to other national surveys such as the National Health Interview Survey³⁶ and the National Health and Nutrition Examination Survey.³⁷ This study is strengthened by its use of a nationally representative patient population. By pooling several years together and applying survey-weighted analysis, we were able to increase the power of the study, reduce measurement errors, and inferential bias. To our knowledge, there is yet to be any study that used a nationally representative dataset to assess the pattern of leaving AMA among trauma patients in the United States. Further, while previous studies have identified patients’ sociodemographic characteristics that leave AMA, this study reports the role of in-hospital events such as receiving diagnostic and therapeutic procedures on leaving AMA. This information may become useful in designing interventions aimed at reducing leaving AMA rates.

In conclusion, sociodemographic characteristics such as age and race, as well as patient characteristics such as the triage class, presence of health insurance, and the receipt of diagnostic and therapeutic procedures, are significant factors associated with trauma patients leaving AMA. Future research may explore how trauma patients’ self-rated health relates to readmission and mortality rates among trauma patients that leave AMA. Future studies may also explore the role of focused communication-based interventions by trained healthcare personnel on reducing the rates of leaving AMA among trauma and non-trauma patients.

Funding declaration

This research received no specific funding or grant from any funding agency in the public, commercial, or not-for-profit sectors.

Contributions

The authors’ responsibilities were as follows: OJA and SV designed the research plan; OJA analyzed the data; OJA and SV wrote the paper; OJA and SV had primary responsibility for the final content; and all authors read and approved the final manuscript.

Declaration of competing interest

The authors declare no conflict of interest.