Risk and protective factors for emergency department visits and related hospitalizations among ambulatory adults with traumatic spinal cord injury

Abstract

Objective

To identify the risk and protective factors for emergency department visits (EDV) and inpatient admissions through the ED (EDIP) among a population-based cohort of ambulatory adults with spinal cord injury (SCI).

Design

Prospective self-report cohort study linked to administrative billing data.

Setting

A Medical University in the Southeastern USA.

Participants

656 ambulatory adults (>18 years old) with chronic (>1-year), traumatic SCI identified through a statewide population-based registry.

Outcome Measures

EDV and EDIP in non-federal state hospitals in the year following self-report.

Results

In the final model, lower rates of future EDV were associated with being older (incidence rate ratio, IRR = 0.99), greater time since injury (IRR = 0.94), and higher income (IRR = 0.89). The EDV rate among non-white individuals was 1.49 times as high as white individuals (P = 0.01). Higher rates of future EDV were associated with using prescription medications for pain, spasticity, sleep, and stress (IRR = 1.30), having depression (IRR = 1.16) and a greater number of chronic conditions (IRR = 1.13), and more past year urinary tract infections (IRR = 1.32), upper extremity broken bones (IRR = 1.37), and serious injuries (IRR = 1.20). Factors associated with higher rates of future EDIP included urban location (IRR = 2.07), higher levels of education (IRR = 1.85), and having a current pressure injury (IRR = 2.94). Married or coupled individuals (IRR = 0.48), and those who reported more days per month of inadequate sleep had lower rates of EDIP (IRR = 0.96).

Conclusions

Several preventable health related factors were significantly related to EDV. The study findings may help clinicians predict and prevent EDV, empower stakeholders for better self-management of their health, and inform decisions to promote the reduction of costly EDV and EDIP.

Individuals with spinal cord injury (SCI) experience disproportionately high rates of hospital utilization, including emergency department visits (EDV) and inpatient (IP) admissions within the first year after injury (1–5) and in subsequent years (6–9). Several studies have focused on identifying the causes of and risk factors for hospital utilization after SCI, and recent findings have highlighted important differences in not only personal characteristics that may influence hospital utilization, but also in actual utilization based on ambulatory status (10, 11). However, there has been less focus on ED utilization, and little is known about ED use among those with chronic SCI; most research is based solely on self-report, and few studies have examined utilization in ambulatory adults with SCI. There is a need for better understanding the risk and protective factors for future EDV and ED related hospitalizations among those with chronic SCI who are ambulatory, as they represent a large segment of the SCI population and may experience different predictors of utilization.

Self-report studies of adults with chronic SCI living in the United States (US), have found yearly rates of ED utilization ranging between 37–48%;(8, 9, 12–14) and reported rates are similar to actual utilization as assessed by administrative sources (45%) (8). Comparatively, 22% of US adults report one or more EDV in the past year (15). In a recent study, participants with SCI were found to be 151% more likely to visit the ED than a nationally representative sample (9). Not only do adults with SCI experience higher rates of EDV, but they are also more likely to have an ED related IP admission (EDIP). In two studies, roughly 50% of those who reported an EDV reported an EDV that resulted in a hospital admission (9, 12). Compared to a nationally representative sample, adults with chronic SCI were 249% more likely to have at least one ED related hospitalization (9).

Several factors have been associated with the increased risk of EDV and EDIP, including race-ethnicity, injury severity, ambulatory status, functional independence at initial rehabilitation discharge, education, income, rurality, prescription medication use and misuse, and prior utilization (10, 12, 13, 16–18). While findings from large observational studies of adults with chronic SCI suggest notable differences in hospitalizations and EDV based on ambulatory status, few studies have focused specifically on assessing risk and protective factors for ED utilization among those who are ambulatory (10, 19–21).

Evidence suggests that those who are ambulatory are less likely to be hospitalized and have EDV than non-ambulatory adults (12, 19, 22), however, the rates of utilization remain high. In an earlier study, 44% of ambulatory adults reported a past year EDV, and 34% reported an ED related hospitalization (11). When linked to administrative sources, 36% of ambulatory adults in the population-based sample experienced an EDV in the year after self-report and 11% experienced an EDIP (11). The study also found significant differences in personal characteristics, health status, psychological and behavioral factors, and participation and quality of life, each of which might influence utilization (11). A necessary next step is to better understand how these factors, including self-reported past year utilization may be associated with future ED and EDIP as assessed through administrative sources among ambulatory adults.

While ED utilization has been found to be similar to actual utilization (8), and cross-sectional observational studies have laid the foundation for examining associations between ED utilization and a number of factors after SCI, the linkage of self-reported data with actual future utilization, will more clearly identify sets of factors that contribute to costly ED visits and admissions. Our objective was to identify the risk and protective factors for future EDV and related EDIP among a state population-based cohort of ambulatory adults with chronic SCI. A unique feature of the study is that we linked data from self-report assessments (SRA) with EDV and EDIP in the year after collection of the self-report data. These findings will add to our understanding of the risk and protective factors of ED utilization after SCI.

Methods

All study procedures were approved by an institutional review board.

Participants

Study participants were recruited through the South Carolina SCI Surveillance Registry (SCSCISR), a population-based registry of SCI occurring in the state of South Carolina that identifies all incident cases of SCI (i.e. new injuries) treated within the state at nonmilitary facilities. The SCSCISR methods are detailed in previous studies (11, 19, 23). Initial study inclusion criteria were: (1) ≥ 18 years of age, (2) ≥ 1-year post-injury, and (3) traumatic SCI with residual effects (i.e. not complete recovery). For the current analyses, we excluded those who were non-ambulatory or lived in border counties. Ambulatory status was assessed by participant response to “are you able to walk at all (yes, no)” or response to other ambulation related questions (e.g. distances walked, how often walked, and walking in the home or community) if the direct question was not answered.

As of September 2021, there were 1,160 participants who had completed an SRA and were able to be linked to administrative billing data. Prior to matching cases to the administrative billing data, deceased cases were eliminated, as were those who reported they lived out of state or in border counties (who may use out-of-state hospitals). The final sample for the current study included 656 ambulatory adults with chronic, traumatic SCI.

Data collection procedures

Self-report Assessments. Self-report data were prospectively collected by mail between 2011 and 2018. Participants received a preliminary letter describing the study and notifying them of a forthcoming SRA, which was sent 2–4 weeks later. Non-respondents were sent a second mailing, followed by a phone call. Participants received $50 remuneration for completing the SRA.

Administrative Billing Data. All SRA respondents were matched to the Office of Revenue and Fiscal Affairs, Health, and Demographics (RFAHD) administrative records of hospital utilization, including EDV and EDIP. The administrative billing data source is limited to non-federal hospitals in South Carolina. RFAHD receives medical record and charges data from South Carolina hospitals at least quarterly. Further detail on the data source can be found in previous publications (19, 24). We examined EDV and EDIP in the 12 months after completion of the SRA. General inpatient, specialty hospital inpatient (rehabilitation, psychiatric), and long-term acute care hospitalizations were included in the calculation of hospitalizations.

Measures

The SRA included multiple measures linked to the Theoretical Risk and Prevention Model (TRPM) (25), which was developed in 1996 to help organize risk and protective factors for adverse health outcomes. The TRPM was used to guide the selection of the risk factors from the SRA. Several demographic and injury factors, psychological and socioenvironmental factors, behavioral factors, and health status and health outcomes were assessed. Demographic and injury factors included their age, sex, race (collapsed to white/nonwhite in regression models), and relationship status (dichotomized as married/couple and single), time since injury (years), injury level (categorized into cervical, thoracic, and lumbar/sacral), and ambulatory status (yes/no), based on participant responses to the following question, “Are you able to walk at all?” or response to other ambulation questions if the direct question was not answered.

For the current analysis, the selected psychological and socioenvironmental factors included: urban and rural living status, annual household income from all sources, highest level of education (≤High School, 2 yr. Associates or Vocational, 4 yr. degree, graduate/professional)., insurance status (insured or uninsured), Brief Resilience Scale (BRS) scores, which assess the perceived ability to bounce back or recover from stress (26), several items from the Zuckerman – Kuhlman Personality Questionnaire (ZKPQ), including Activity, Aggression-Hostility, Sociability, Impulsivity-Sensation Seeking, and Neuroticism-Anxiety subscores (27), as well as multiple items of perceived support from the Berlin Social Support Scale (BSSS), including emotional support, instrumental support, need for support, and support seeking (28).

Selected behavioral variables included medication noncompliance (mean of five questions assessing frequency of forgetting medication, skipping doses, etc.), use of prescription medications for pain, spasticity, sleep, or stress, frequency of planned exercise (rarely – 5 + times/week), healthy diet/nutrition (poor – excellent), overall fitness level, hours per day spent out of bed, days per month binge drinking (5 + alcoholic beverages), number of cigarettes smoked per day, and frequency of cannabis use (never-daily).

The selected health status and health outcome variables included self-reported past year EDV and IP hospital discharges, the presence of current pressure sores (yes/no), the presence of chronic conditions (0–7 conditions) and UTIs in past year (0–13+), past year broken bones in upper extremity, fevers, sweats and chills, and serious injuries, reported days per month they did not receive enough sleep, and days per month that pain interfered with usual activities. Participants were also asked about the presence of chronic conditions, assessed using standardized questions from the Behavioral Risk Factor Surveillance System (BRFSS) available at the time of the SRA development, including diabetes, heart attack, angina or coronary heart disease, stroke, high blood, high cholesterol, or cancer. The presence of chronic conditions was dichotomized to yes/no. The Patient Health Questionnaire-9 (PHQ-9), a 9-item self-report questionnaire, was used to measure depressive symptomatology; (8) we classified probable major depression as PHQ-9 ≥ 15.

Primary outcomes. We assessed the total number of annual EDV and EDIP in non-federal South Carolina hospitals in the year following the SRA response.

Analyses

All analyses were completed with SAS Version 9.4 (SAS Institute).^a The simple (bivariate) and multiple regression predictive relationships between each variable and EDV and EDIP were calculated. All regression models were generated by Proc Genmod using a negative binomial distribution since the outcome variables are overdispersed (variance > mean) count data. The simple regression results did not influence the development of the multiple regression models, they are provided to demonstrate relationships for descriptive purposes only.

Missing data are common with self-administered surveys due to skip patterns and items that are skipped intentionally or unintentionally; using only those cases with complete data for every predictor (complete case or listwise deletion approach) in the multiple regression model excluded too many observations. Therefore, an alternative approach, multiple imputation, was used to address missing data. Since several missing predictor variables are binary or categorical, the FCS (Fully Conditional Method) option for Proc MI in SAS was used to create 10 imputed datasets. In addition to the 40 variables in the simple regression models, 32 auxiliary variables were used in the imputation phase to increase accuracy. Proc MIANALYZE was used to combine the regression results from the 10 imputed datasets.

Regression analyses were constructed in accordance with the TRPM (25), which is designed for predicting health and longevity after SCI using multiple stages of predictors. These included: (1) demographic and injury factors, (2) psychological and socioenvironmental factors, (3) behaviors, and (4) health status and health outcomes. The final stage of mortality represents the ultimate medical endpoint but is beyond the scope of this study. The model proposes a series of mediational relationships, such that those sets of variables most proximal to the health outcome will mediate the relationships of those more distal to the outcome. In this case, health status variables would be most predictive and, when introduced into the analysis, would reduce the strength of prediction of health behaviors and psychological and socioeconomic factors.

We entered variables in three stages in according to the TRPM, with demographic/injury and psychological/socioenvironmental entered first followed by behavioral factors in stage 2 and health factors in the final stage. There are challenges regarding calculating power with multiple imputation and a negative binomial distribution. For the analyses, we relied on the rule of thumb of 10 or more observations per predictor, therefore our sample size should provide sufficient power despite the large number of predictors. Incidence rate ratios (IRR), 95% confidence intervals (CI) and P-values are provided.

Results

We identified 656 ambulatory participants with self-report data who could be linked to the administrative billing data. On average, participants were 53 years old and 4.6 years post injury; 70% of participants were male, 61% were white, and 35% were Black or African American (Table 1). In the year following completion of the SRA there were a total of 575 EDV and 101 EDIP. The mean EDV was 0.88 (2.02), with a range of 0–20 and the mean EDIP was 0.15 (0.53), with a range of 0–6 (Table 2).

Table 1.

Participant characteristics (N = 656).

Variables	Frequency % or Mean (SD)
Demographic & Injury Variables
Age	53.1 (16.2)
Male	70%
Race
White	61%
Black/AA	35%
Other	4%
Married/partnered	49%
Years post injury	4.6 (5.6)
Spinal injury at cervical level	71%
Psychological & Socioenvironmental Variables
Urban	73%
Income
<$10,000/yr	30%
$10,000-$35,000/yr	39%
>$35,000/yr	32%
High school or less education	56%
Brief Resilience Scale (1–5)	3.3 (0.8)
Behavioral Variables
Never take prescription medications for pain, spasticity, sleep, or stress.	17%
Rarely plan exercise	52%
Fair/poor diet	43%
Poor overall fitness level	20%
Hours/day out of bed	12.4 (4.1)
Never have 5 + alcoholic drinks/day in past month	70%
Don’t smoke any cigarettes/day now	62%
Health Variables
No prior year ED visits (SRA)	55%
No prior year discharges (SRA)	65%
No pressure sore now	93%
No UTIs in past year	77%
No fevers in prior year	63%
No serious injuries in prior year	18%
Major depression (PHQ-9)	23%
Days/month not enough sleep	14.6 (10.4)
Days/month pain interfered with usual activities	16.0 (11.9)
Outcome Variables
No following year IP admits through ED	90%
No following year ED-only visits	64%
No following year IP-only admits	91%

Table 2.

TRPM Variables as simple regression predictors for EDV and EDIP among ambulatory adults living in non-border counties (N = 656).

	EDV				EDIP
	IRR	95% CL		P-value	IRR	95% CL		P-value
Demographic & Injury
Age	0.98	0.97	0.99	0.0004	1.01	0.99	1.03	0.33
Sex	1.41	1.02	1.96	0.04	1.02	0.57	1.84	0.94
Race	1.51	1.11	2.06	0.009	1.07	0.61	1.85	0.82
Relationship status	0.64	0.47	0.87	0.004	0.59	0.34	1.01	0.06
Years post injury	0.94	0.90	0.98	0.003	0.99	0.94	1.05	0.81
Spinal injury level	1.10	0.88	1.37	0.42	0.79	0.51	1.20	0.27
Psychological & Socioenvironmental
Urban/rural	0.98	0.70	1.39	0.92	1.72	0.89	3.30	0.11
Income	0.80	0.76	0.85	<.0001	0.89	0.81	0.98	0.03
Education	0.66	0.53	0.82	<.0001	1.24	0.86	1.77	0.44
Insured/uninsured	0.77	0.44	1.33	0.34	0.75	0.27	2.06	0.57
BRS	0.61	0.50	0.74	<.0001	0.95	0.68	1.31	0.74
ZKPQ Activity	0.98	0.92	1.04	0.46	0.97	0.87	1.07	0.54
ZKPQ Aggression-Hostility	1.08	1.02	1.14	0.01	0.99	0.91	1.09	0.9
ZKPQ Sociability	0.99	0.93	1.06	0.75	1.00	0.89	1.13	0.95
ZKPQ Impulsivity-Sensation Seeking	1.07	1.01	1.14	0.03	1.01	0.90	1.12	0.9
ZKPQ Neuroticism-Anxiety	1.13	1.07	1.19	<.0001	1.08	0.99	1.18	0.08
BSSS emotional support	0.75	0.57	0.97	0.03	0.90	0.56	1.43	0.66
BSSS instrumental support	0.70	0.55	0.88	0.002	0.82	0.55	1.24	0.35
BSSS need for support	1.16	0.87	1.55	0.31	1.36	0.81	2.28	0.25
BSSS support seeking	1.12	0.88	1.42	0.37	0.92	0.59	1.44	0.72
Behavioral Variables
Medication noncompliance	1.17	0.93	1.47	0.18	1.05	0.69	1.60	0.81
Use prescription medications for pain, spasticity, sleep, or stress.	1.55	1.30	1.84	<.0001	1.26	0.93	1.72	0.14
Plan exercise	0.92	0.85	0.99	0.04	0.86	0.74	1.00	0.04
Diet/nutrition	0.82	0.70	0.96	0.02	0.88	0.65	1.19	0.4
Overall fitness level	0.83	0.69	0.99	0.04	0.67	0.48	0.94	0.02
Hours/day out of bed	0.94	0.91	0.98	0.002	0.95	0.88	1.02	0.14
5 + alcoholic drinks/day in month	1.02	0.99	1.05	0.3	0.96	0.89	1.04	0.3
Smoke cigarettes/day now	1.23	1.05	1.43	0.008	1.19	0.91	1.54	0.2
Smoked cannabis in last 90 days	1.23	1.07	1.42	0.004	1.02	0.79	1.32	0.86
Health Status & Outcomes
Prior year ED only visits (SRA)	1.40	1.31	1.50	<.0001	1.27	1.11	1.46	0.0006
Prior year IP discharges (SRA)	1.31	1.17	1.47	<.0001	1.43	1.18	1.74	0.0002
Pressure sore now	1.50	0.85	2.66	0.17	2.63	1.10	6.27	0.03
Chronic conditions	1.06	0.95	1.18	0.30	1.23	1.02	1.49	0.03
UTIs in past year (#)	1.45	1.18	1.79	0.0003	1.28	0.91	1.79	0.16
Broken bones in upper extremity in past year	1.78	1.28	2.48	0.0006	1.86	1.14	3.05	0.01
Fevers/sweats/chills in prior year	1.22	1.02	1.46	0.03	1.31	0.96	1.78	0.09
Serious injuries in prior year	1.44	1.28	1.63	<.0001	1.27	1.06	1.53	0.01
Depression (PHQ-9)	1.36	1.22	1.51	<.0001	1.17	0.98	1.41	0.09
Days/month not enough sleep	1.03	1.02	1.05	<.0001	1.00	0.97	1.02	0.78
Days/month pain interfered with usual activities	1.02	1.01	1.04	0.0006	1.02	1.00	1.04	0.12

Simple regression results

The majority of variables were statistically significant in simple regression analyses (Table 1). A greater number of significant relationships were found between the independent variables and EDV compared to EDIP.

Staged multiple regression results

ED only visits in the year following the SRA. Several factors were significantly related to EDV in the year following completion of the SRA (Table 2). In the final model, the rate of EDV was lower for older adults with SCI (IRR = 0.99), those with greater time since injury (IRR = 0.94), and those with higher income (IRR 0.89). The ED rate among those who are non-white was 1.49 times as high as those who are white (P = 0.01). In Table 3, higher rates of EDV were also associated with having depression (IRR = 1.16), using prescription medications for pain, spasticity, sleep, and stress (IRR = 1.30), having a greater number of chronic conditions (IRR = 1.13), having more past year urinary tract infections (IRR = 1.32), having broken bones in the upper extremity in the past year (IRR = 1.37), and having more serious injuries in the prior year (IRR = 1.20).

Table 3.

Multi staged regression model of predictors of EDV among ambulatory adults living in non-border counties (N = 656).

	EDV
	Stage 1		Stage 2		Final Stage 3
	IRR	P value	IRR	P value	IRR	95% CL		P value
Demographic & Injury
Age	0.99	0.01	0.99	0.03	0.99	0.98	1.00	0.04
Sex	1.54	0.01	1.51	0.02	1.30	0.93	1.82	0.13
Race	1.22	0.22	1.30	0.11	1.49	1.08	2.04	0.01
Relationship status	0.90	0.55	0.89	0.52	0.84	0.61	1.17	0.31
Years post injury	0.94	0.003	0.94	0.004	0.94	0.90	0.98	0.002
Psychological & Socioenvironmental Variables
Urban/rural	1.10	0.57	1.12	0.52	1.11	0.79	1.54	0.55
Income	0.87	<.0001	0.87	0.0002	0.89	0.83	0.95	0.0006
Education	0.94	0.60	0.95	0.68	0.90	0.71	1.13	0.35
Insured/uninsured	0.62	0.08	0.64	0.10	0.66	0.39	1.12	0.12
BRS	0.81	0.11	0.84	0.20	0.89	0.69	1.15	0.38
ZKPQ Activity	0.98	0.57	1.01	0.79	1.05	0.98	1.12	0.14
ZKPQ Aggression-Hostility	1.00	0.97	0.98	0.61	0.98	0.92	1.05	0.56
ZKPQ Sociability	1.01	0.83	1.00	0.98	0.99	0.92	1.06	0.73
ZKPQ Impulsivity-Sensation Seeking	1.07	0.07	1.07	0.05	1.07	1.00	1.14	0.07
ZKPQ Neuroticism-Anxiety	1.03	0.48	1.00	0.95	0.94	0.88	1.02	0.12
BSSS emotional support	1.18	0.37	1.19	0.36	1.17	0.81	1.68	0.42
BSSS instrumental support	0.63	0.01	0.64	0.02	0.76	0.52	1.09	0.14
BSSS need for support	0.99	0.96	1.00	0.98	0.90	0.62	1.32	0.59
BSSS support seeking	1.33	0.09	1.32	0.12	1.35	0.96	1.90	0.08
Behavioral Variables
Medication noncompliance			0.93	0.51	0.94	0.75	1.17	0.58
Use prescription medications for pain, spasticity, sleep, or stress.			1.39	0.0005	1.30	1.07	1.57	0.008
Plan exercise			0.96	0.42	0.94	0.86	1.03	0.16
Diet/nutrition			0.96	0.66	0.98	0.82	1.17	0.81
Overall fitness level			1.03	0.82	1.06	0.85	1.31	0.61
Hours/day out of bed			1.00	0.88	1.03	0.98	1.07	0.23
5 + alcoholic drinks/day in month			1.01	0.62	1.00	0.97	1.03	0.92
Smoke cigarettes/day now			1.08	0.32	1.05	0.91	1.23	0.50
Smoked cannabis in last 90 days			1.12	0.12	1.12	0.98	1.28	0.09
Health Status & Outcomes
Pressure sore now					1.53	0.88	2.68	0.13
Chronic conditions					1.13	1.00	1.28	0.05
UTIs in past year (#)					1.32	1.07	1.62	0.009
Broken bones in upper extremity in past year					1.37	1.01	1.86	0.04
Fevers/sweats/chills in prior year					0.94	0.79	1.12	0.52
Serious injuries in prior year					1.20	1.07	1.35	0.001
Depression (PHQ-9)					1.16	1.00	1.35	0.05
Days/month not enough sleep					1.00	0.99	1.02	0.60
Days/month pain interfered with usual activities					0.99	0.98	1.01	0.25

ED Related Inpatient Hospitalizations in the year following the SRA. Married or coupled individuals (IRR = 0.48), and those who reported more days per month of inadequate sleep had lower rates of EDIP (IRR = 0.96) (Table 4). Factors associated with higher rates of EDIP included urban location (IRR = 2.07), higher levels of education (IRR = 1.85), and having a current pressure sure (IRR = 2.94) (Table 4).

Table 4.

Multi-stage regression model of predictors of EDIP among ambulatory adults living in non-border counties (N = 656).

	EDIP
	Stage 1		Stage 2		Final Stage 3
	IRR	p value	IRR	P value	IRR	95% CI		p value
Demographic & Injury
Age	1.01	0.32	1.01	0.28	1.01	0.98	1.03	0.56
Sex	0.69	0.26	0.70	0.30	0.85	0.43	1.69	0.64
Race	1.01	0.96	1.18	0.63	1.36	0.70	2.64	0.36
Relationship status	0.64	0.18	0.59	0.12	0.48	0.25	0.94	0.03
Years post injury	1.00	0.97	1.00	0.90	0.97	0.91	1.03	0.32
Psychological & Socioenvironmental Variables
Urban/rural	1.90	0.06	2.17	0.03	2.07	1.03	4.15	0.04
Income	0.85	0.02	0.89	0.08	0.92	0.80	1.05	0.22
Education	1.77	0.009	1.89	0.006	1.85	1.18	2.89	0.008
Insured/uninsured	0.54	0.24	0.55	0.26	0.58	0.20	1.68	0.31
BRS	1.26	0.32	1.36	0.20	1.22	0.76	1.96	0.41
ZKPQ Activity	0.94	0.30	0.97	0.62	0.99	0.88	1.12	0.91
ZKPQ Aggression-Hostility	0.99	0.86	0.96	0.48	0.99	0.87	1.12	0.81
ZKPQ Sociability	1.07	0.33	1.07	0.36	1.05	0.92	1.20	0.44
ZKPQ Impulsivity-Sensation Seeking	1.01	0.92	1.01	0.87	0.97	0.86	1.11	0.69
ZKPQ Neuroticism-Anxiety	1.14	0.06	1.11	0.14	1.06	0.92	1.21	0.45
BSSS emotional support	1.47	0.34	1.68	0.22	1.71	0.75	3.90	0.21
BSSS instrumental support	0.66	0.27	0.62	0.22	0.65	0.31	1.39	0.27
BSSS need for support	1.73	0.13	1.76	0.13	1.61	0.78	3.31	0.20
BSSS support seeking	0.77	0.42	0.80	0.52	0.89	0.47	1.69	0.72
Behavioral Variables
Medication noncompliance			0.87	0.56	0.99	0.62	1.57	0.95
Use prescription medications for pain, spasticity, sleep, or stress.			1.09	0.65	1.02	0.70	1.49	0.93
Plan exercise			0.88	0.27	0.85	0.68	1.05	0.12
Diet/nutrition			1.08	0.71	1.03	0.69	1.52	0.89
Overall fitness level			0.76	0.23	0.82	0.52	1.30	0.40
Hours/day out of bed			0.99	0.76	1.02	0.94	1.11	0.64
5 + alcoholic drinks/day in month			0.93	0.11	0.91	0.82	1.01	0.08
Smoke cigarettes/day now			1.33	0.07	1.24	0.91	1.70	0.17
Smoked cannabis in last 90 days			1.08	0.60	1.15	0.87	1.53	0.33
Health Status & Outcomes
Pressure sore now					2.94	1.15	7.52	0.02
Chronic conditions					1.17	0.94	1.46	0.17
UTIs in past year					0.94	0.63	1.41	0.77
Broken bones in upper extremity in past year					1.70	0.98	2.95	0.06
Fevers/sweats/chills in prior year					1.12	0.81	1.55	0.48
Serious injuries in prior year					1.10	0.89	1.34	0.38
Depression (PHQ-9)					1.13	0.82	1.55	0.45
Days/month not enough sleep					0.96	0.93	1.00	0.05
Days/month pain interfered with usual activities					1.01	0.98	1.04	0.58

Discussion

This study enhances our understanding of factors related to future ED utilization, including EDV and EDIP among ambulatory individuals with SCI by linking self-reported data on multiple predictive factors with administrative data on ED utilization in the following year. The assessment of self-report data allowed us to measure theoretically relevant predictors, rather than solely relying on administrative data, which is often limited in the number and diversity of data available. It also allowed us to identify predictors of future ED utilization, rather than simply correlates of current or previous utilization. The findings are consistent with the TRPM model, with health factors ultimately serving as the optimal predictors. Because there were some differences between EDV and EDIP, we discuss each in turn.

Emergency department visits

The simple regression analyses indicate the importance of multiple types or classes of predictors, as broken down by the TRPM, including demographic and injury characteristics, psychological and socioenvironmental factors, behaviors, and health factors. The findings clearly show important relationships that may be considered in ED prevention, even though the majority of predictors, outside of demographic, injury, and health status factors, drop out of the regression analysis, as predicted by the TRPM (i.e. health factors being most proximal to EDV and other outcomes). Nearly all demographic and injury characteristics were related to EDV, but those that appeared most important and were significant in the final model included race, age, and time since injury; a greater risk of EDV was observed among non-Hispanic Blacks, and both age and years post injury were protective.

Several psychological and socioenvironmental factors were strongly related to EDV in the simple regression analyses, but only income was significant in the final regression model and was protective of EDV. Emotional support was protective in stage two after the addition of behavioral factors and impulsive sensation seeking was maintained as a risk at that stage. It is not surprising that these factors dropped out in the final model as they are the most distal to EDV within the TRPM (i.e. their effects would be explained and less significant after the addition of behavioral and health outcome factors). For prevention, important psychological indicators for risk of EDV would include poor socioeconomics (income, education), lower levels of social support, lower levels of resilience, and personality characteristics that include impulsivity/sensation seeking and aggression hostility. Improving psychological status and socio-economic status may help in reducing risk of health conditions that ultimately lead to and require attention in the ED.

Multiple behavioral factors were also related to EDV in the simple regression analyses, consistent with previous findings (10), and there are clear implications for prevention strategies. Not surprising, cigarette smoking was related to EDV, but so was cannabis use. Because several exercise and activity variables were protective of EDV, it’s clear that avoiding the ED is related to a number of positive and negative health behaviors. In the final model, the only behavioral factor that remained significant was prescription medication use for pain, sleep, depression, or spasticity. Even after controlling for chronic conditions (which were significant in the final model), prescription drug use is a problem. In other studies, we have found high rates of risky prescription opioid use (29), which may place individuals at risk of adverse outcomes including EDV.

As expected, a wide array of health conditions were associated with future EDV, most of which remained significant in the final model, including both chronic conditions and acute secondary health conditions, such as UTIs, serious injuries, and fractures. Prevention strategies for chronic conditions, including depression, would be different than those targeting unintentional injuries and fractures. Considering the health behaviors and psychological factors associated with EDV, it may be that health promotion activities may be used to target prevention of chronic conditions, whereas other interventions would focus at decreasing risk behaviors, such as cigarette smoking, cannabis use, and an elevated use of prescription medications, particularly among those in high risk situations where there are high risk personality characteristics (e.g. sensation seeking) and where there is limited social support.

ED related inpatient hospitalizations

In a previous study of self-reported ED related hospitalizations, greater odds of ED related hospitalizations were found among males, non-Hispanic Blacks, older adults, those who used prescription medications more frequently, and those who misused prescription medications (10). Lower odds of ED hospitalization were found among those with fewer years post injury, those who walk independently, and had an income between $25K and $75K (10). Unfortunately, we did not obtain a great deal of clarity from the current analysis of EDIP. None of the demographic/injury or psychological/socioenvironmental factors and only two behavioral factors (planned exercise and fitness as a proxy for healthy lifestyle) were significant in the simple regression analysis, although several health factors were significant. Contrary to previous findings, in the final analysis, no behavioral factors were significant. We found relationship status, urban rural status, and education became significant, with education being a risk factor for EDIP. Only two health factors were significant in the final model, with current pressure ulcer a risk factor and the number of days of poor sleep a protective factor.

The current analysis is limited in its ability to identify risk factors for EDIP. It seems possible, even likely, that the causes of and need for hospitalization after an ED visit are substantially more complex than that of EDV and idiosyncratic factors may play a greater. Typical psychological, socio-environmental, and behavioral factors seem much more strongly related to EDV. Even with medical factors, the only stable conclusion would be that people who experience pressure ulcers are more likely to have EDIP.

Clinical implications

Hospital utilization places individuals with SCI at risk for poorer health outcomes and comes at a high cost, financially and in terms of lost productivity, decreased participation, reduced quality of life, and increased risk of mortality (30). These findings may be (1) used by clinicians to provide education and raise awareness among ambulatory adults with SCI about the factors related to EDV and EDIP, (2) help rehabilitation professionals and clinicians to better understand how ambulation relates to the health and well-being of those with SCI, and (3) guide the development and tailoring of research and interventions to prevent costly EDV and EDIP and improve care management. Screening questions about pressure sores, UTIs, chronic illnesses, and use of prescription medications may identify patients at higher risk of ED utilization.

Methodologic considerations

There are several important methodologic considerations to note when interpreting the findings. Notably, we used a combination of self-report and administrative billing data sources. Our cohort was identified from population data which captured all civilian cases of SCI within the state. This is a strength compared to studies that utilize single specialty hospital populations, but there still may be response biases in participation among those within the population-based system. The current analysis included a large number of predictors identified from the SRA within the framework of the TRPM but remains within the benchmark of 10 participants per predictor. The large number of predictors may have made it more difficult to identify significant differences because of collinearity between predictors. Utilizing self-report data allowed for inclusion of variables which are not available through the administrative source, but the susceptibility to reporting errors and recall bias is a limitation. We identified the relationships between past year self-reported factors and future ED utilization (EDV and EDIP in the year following the SRA), which is a particular strength when compared with a single cross-sectional measurement. The linkage to administrative UB-04 billing data allowed for examination of all admissions that occurred within non-federal hospitals in South Carolina in the year after the SRA. Additionally hospital data from nearby states are not captured; those in border counties were eliminated to reduce the risk of bias. Our findings are representative of a single geographic region, and may not be generalizable outside of the US, where differences in factors such as individual healthcare systems, insurance, and access, may affect the course of treatment and management, and ultimately need for emergent or planned hospitalization. We also focused on those with ambulatory SCI, so the findings will not necessarily apply to those who have non-ambulatory SCI. We focused on ambulatory because they have a different pattern of health outcomes including different patterns of risk for unintentional injuries. Therefore, there may be qualitative differences in risk of outcomes that lead to ED utilization (i.e. injuries are a primary source of EDV).

Future research

Continued research and focused care are necessary to decrease the risk of EDV and EDIP after SCI. Further study of the associations between personal characteristics and relationships with the leading causes of EDV may enable providers, caregivers, and individuals with SCI to better monitor and manage long-term health of adults with SCI. Detailed cost analyses would also add to the body of literature and provide a better understanding of the true burden of ED utilization.

Conclusions

There were several preventable health factors significantly related to ED utilization. The study findings may help clinicians predict and prevent ED and EDIP utilization, empower stakeholders for better self-management of their health and hospital utilization, and inform policy decisions to target limited resources to reduce costly EDV and related hospitalizations and improve overall quality of life.

List of Abbreviations

ED

Emergency Department

EDIP

ED related Inpatient Admission

EDV

Emergency Department Visit

IP

Inpatient

IRR

Incidence rate ratio

SCI

Spinal cord injury

SCISRS

SCI Surveillance and Registry System

RFAHD

Office of Revenue and Fiscal Affairs, Health, and Demographics

UB-4

Uniform billing discharge data

US

United States

Suppliers

1. SAS for Windows [computer program]. Version 9.4. Cary, NC: SAS Institute; 2015

Disclaimer statements

Contributors None.

Declaration of interest None.

Conflicts of interest Authors have no conflict of interests to declare.

Funding Statement

The contents of this publication were developed under grants from the National Institute on Disability, Independent Living, and Rehabilitation Research (NIDILRR), grant number 90IFRE0028. NIDILRR is a Center within the Administration for Community Living (ACL), Department of Health and Human Services (HHS). The contents of this publication do not necessarily represent the policy of NIDILRR, ADL, or HHS, and you should not assume endorsement by the funding agency or Federal Government.