TARGETING ZERO PREVENTABLE TRAUMA READMISSIONS

Abstract

Background:

Nearly 1-in-10 trauma patients in the U.S. are readmitted within 30 days of discharge, with a median hospital cost of over $8,000 per readmission. There are national efforts to reduce readmissions in trauma care, but we do not yet understand which are potentially preventable. Our study aims to quantify the Potentially Preventable Readmissions (PPRs) in trauma care to serve as the anchor point for ongoing efforts to curb hospital readmissions and ultimately, bring preventable readmissions to zero.

Methods:

We identified inpatient hospitalizations after trauma as well as readmissions within 90 days in the 2017 National Readmissions Database (NRD). PPRs were defined as the AHRQ-defined Ambulatory Care Sensitive Conditions, in addition to superficial surgical site infection, acute kidney injury/acute renal failure, and aspiration pneumonitis. Mean costs for these admissions were calculated using the NRD. A multivariable logistic regression model was utilized to characterize the relationship between patient characteristics and PPR.

Results:

1,320,083 patients were admitted for trauma care in the 2017 NRD and 137,854 (10.4%) were readmitted within 90 days of discharge. Of these readmissions, 22.7% were potentially preventable. The mean cost was $10,001/PPR, resulting in $313,802,278 in cost to the US healthcare system. Of readmitted trauma patients < 65 years old, Medicaid or Medicare patients had 2.7-fold increased odds of PPRs compared to privately insured patients. Patients of any age with congestive heart failure had 2.9x increased odds of PPR, those with COPD or complicated diabetes had 1.8x increased odds, and those with chronic kidney disease had 1.7x increased odds. Furthermore, as the days from discharge increased, the proportion of readmissions due to PPRs increased.

Conclusions:

One-in-five trauma readmissions are potentially preventable, which account for over $300 million annually in healthcare costs. Improved access to post-discharge ambulatory care may be key to minimizing PPRs, especially for those with certain comorbidities.

Level of Evidence:

Level II, Economic & Value-Based Evaluations

Background:

Readmissions after trauma care are frequent and expensive.¹ Nearly 1-in-10 trauma patients in the U.S. are readmitted within 30 days of discharge, with a median hospital cost of over $8,000 per readmission.¹ While many efforts have been developed to better understand and prevent trauma readmissions,^2–4 the clinical and social heterogeneity of trauma patients further confounds this complex problem. Some readmissions may be preventable with targeted interventions, however other readmissions may simply represent requisite planned clinical care. Understanding these nuances is essential when approaching solutions to this problem. Blunt tools such as policies that aim to reduce all readmissions, may prove ineffective and inappropriate if they do not selectively target those patients at highest risk for preventable readmissions.

Application of the Agency for Healthcare Research and Quality’s (AHRQ) Ambulatory Care Sensitive Conditions (ACSCs) framework can provide insight into which trauma readmissions may be preventable. ACSCs were adopted as AHRQ Prevention Quality Indicators and represent a set of medical conditions that are potentially preventable by timely and effective coordination of outpatient care.^5,6 This framework has previously been used to identify potential preventable readmissions among patients with sepsis,⁷ however the degree to which trauma readmissions could be potentially prevented remains poorly understood. The objective of this study was to identify characteristics of trauma patients whose readmissions are potentially preventable.

In this study, we extended this AHRQ ambulatory care sensitive condition framework to include additional surgically relevant diagnoses, to determine what proportion of readmissions after trauma are potentially preventable and identify those patients at highest risk for potentially preventable readmissions.

Methods:

Study Design

We used a retrospective study design to describe the incidence of Potentially Preventable Readmissions (PPRs) after discharge from a US hospital for trauma care. We utilized a multivariable regression as described in the statistical analysis below to control for inherent differences in our population. Our study period was 2017, the latest data available in the National Readmissions Database (NRD). This most recent year was chosen as the NRD switched to ICD10 codes late in 2016 and the 2017 sample size was adequate for this analysis. The reporting guideline for cohort studies published by the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) was followed.⁸ This study was deemed exempt by the University of Michigan institutional review board given the secondary use of existing data.

Data Source

In this study, we used data from the Healthcare Cost and Utilization Project (HCUP) National Readmissions Database (NRD) 2017 participant use file. The NRD is the largest readmissions database in the US accounting for approximately 60% of the total US population as well as total US hospital admissions. The database, assembled annually, includes 15 million inpatient hospital readmissions per year, representing over 35 million readmissions per year nationally. These data include readmissions from 28 geographically dispersed states to represent all 50 states and includes all payers. Patient demographic data, ICD-10 diagnosis codes, and patient linkage numbers to track hospitalizations within state and calendar year, hospital charge and cost-to-charge ratios, are also provided in the NRD. To note, NRD patient linkage numbers cannot be tracked across calendar years or states.

Patient Population

We studied all patients hospitalized in 2017 reported in the NRD with primary diagnosis codes for trauma care (International Classification of Diseases, 10th Revision, Clinical Modification, ICD-10-CM) in accordance with the National Trauma Data Standard.⁹ Primary codes for burns were not included. See Supplemental Materials for ICD-10 codes used. These admissions were deemed the index admission from which all readmissions for that patient linkage number were defined. Only patients who had overlapping admissions from the index trauma admission were excluded.

Outcomes and Variables

The primary outcome we measured was incidence of readmission within 90 days following the index admission for trauma care that we defined as potentially preventable. The definition of a potentially preventable readmission we used included the primary diagnosis code for any of the Ambulatory Care Sensitive Conditions (ACSCs) as defined by the Agency for Healthcare Research and Quality (AHRQ)—short and long term complications of diabetes (DM) including uncontrolled DM and lower extremity amputations of DM patients, chronic obstructive pulmonary disease (COPD) and adult asthma, hypertension (HTN), heart failure (HF), community acquired pneumonia (CAP), and urinary tract infection (UTI)^5,7,10— as well as three specific diagnosis that may be managed with timely ambulatory care for a surgical patient: superficial surgical site infection (SSI), acute kidney injury (AKI), and aspiration pneumonia/pneumonitis. This definition has been established in prior literature.^7,11,12 90 days was chosen to understand the trajectory of PPRs along a longer time course.

Patient demographics included in our analysis were age, sex, insurance status, Injury Severity Score, and comorbidities included in the Charlson Comorbidity Index (CCI).¹³ We calculated the CCI for each patient at their index hospitalization, which we used to assess comorbidities as predictors of PPR. Patients with multiple trauma admissions within a year were uncommon, but only the first index admission was included in the analysis. Overlapping admissions were omitted entirely. In this way, the codes used in the analysis would only be those associated with the index trauma admission. The number of days between the index admission and the first readmission was calculated as the time to event.

The NRD includes hospital location, which is determined by HCUP and based on the Office of Management and Budget assignments. These are categorized into urban/rural settings and teaching/non-teaching status: non-metropolitan, metropolitan non-teaching, and metropolitan teaching. Furthermore, the NRD contains hospital charges—the amount billed to the patient’s insurance company or directly to the patient—as well as cost-to-charge ratios. These ratios are calculated at the hospital level, based on accounting reports collected by CMS. With these two variables, the estimated cost was calculated.

Missing data accounted for less than 0.1%, of which were patients missing primary payer data or hospital location. Given the completeness of the dataset, no imputation was deemed necessary.

Statistical Analyses

A multivariable logistic regression model was utilized throughout the analysis, unless otherwise indicated. No power analysis was performed given the large, national dataset used. For each patient admitted for trauma care in 2017, the number of total readmissions, PPRs, and non-PPRs were calculated. As we wanted to develop a clinically-relevant model that was applicable at the time of discharge—when it is unknown whether a patient will be readmitted at all, readmitted as a PPR, or readmitted as a non-PPR—we performed our multivariable logistic regression on all readmitted and non-readmitted patients. Covariates used in our model include age, sex, primary insurance payer (Medicare/Medicaid, private, uninsured, other), hospital location/status, Injury Severity Score (ISS), and Charlson Comorbidity Index, to control for these selected confounders in our primary outcome and estimate fixed effects of each. Estimates of fixed effects of patient characteristics are presented in Table 1. Patients’ insurance status over the age of 65 were analyzed separately given the increased likelihood these patients have Medicare as their payer type, which included an interaction term between the primary payer variable and a categorical variable for age ≥65. This is displayed in the Supplementary Materials.

Table 1. National Readmissions Database Trauma Patient Characteristics.

This table displays the characteristics of the 2017 trauma cohort from the National Readmissions Database for all patients without readmission, patients readmitted without a potentially preventable diagnosis, and patients readmitted with a potentially preventable diagnosis.

Patient Characteristic	Without Readmission	With 90-Day Readmission
		Without PPR	PPR
Age, mean years (SD)	60.1 (24.3)	68.8 (19.4)	76.5 (14.6)
Female	50.8	52.9	60.4
Insurance Status
Medicare/Medicaid	64.4	80.3	90.2
Private Insurance	23.7	12.9	6.7
Uninsured	5.2	3.1	1.1
Other	6.6	3.7	2.0
Hospital Metropolitan-Teaching Status
Metropolitan Non-teaching	18.6	22.2	25.5
Metropolitan Teaching	73.8	70.7	64.1
Non-metropolitan	7.5	7.1	10.4
Hospital Bed Size
Small	12.9	14.4	17.8
Medium	26.0	27.9	29.7
Large	61.1	57.8	52.6
Median Household Income Quartile of Patient Zip Code
<25%	29.0	29.4	29.6
25–50%	27.6	27.8	28.3
50–75%	23.7	23.6	23.3
>75%	18.3	18.3	18.0
Injury Severity Score
<8	51.4	51.1	53.2
8–15	26.4	32.0	34.9
16–24	8.9	11.4	9.1
25+	10.9	3.6	1.7
Comorbidities
Acute Myocardial Infarction	4.5	8.1	11.0
Congestive Heart Failure	9.5	19.2	41.4
Peripheral Vascular Disease	4.9	8.4	12.0
Cerebrovascular Disease	4.2	8.4	7.0
Dementia	11.7	18.3	24.2
Chronic Obstructive Pulmonary Disease	15.5	22.9	34.9
Rheumatoid Disease	2.4	3.6	3.7
Peptic Ulcer Disease	0.4	1.6	1.0
Mild Liver Disease	2.0	3.9	2.9
Diabetes	11.3	14.1	16.5
Diabetes with Complications	6.6	12.7	23.1
Hemiplegia or Paraplegia	1.7	2.7	1.5
Chronic Kidney Disease	10.6	20.4	35.7
Cancer	1.9	4.0	3.8
Moderate to Severe Liver Disease	0.4	1.7	0.9
Metastatic Cancer	0.7	2.0	1.5
Acquired Immunodeficiency Syndrome	0.1	0.3	0.2

Data are presented as a proportion (%) unless otherwise indicated

Hospital level variation in PPR rates is presented as a caterpillar plot with 95% confidence intervals. All analyses were weighted using an HCUP-provided hospitalization-level weight, given the complex survey methodology utilized in the NRD. All hypothesis tests were two-sided with a 5% type I error rate. Statistical analyses were performed using STATA statistical software, version 16.1/MP, (StataCorp, LLC) and used the charlson package to calculate the Charlson Comorbidity Index.¹⁴

Sensitivity analysis

To ensure that our findings were not dependent on the specifics of the study design, we performed four sensitivity analyses. First, to identify predictors of PPR from a population of patients who otherwise would not be readmitted, we excluded non-PPR readmissions for all multivariable models.¹¹

Next, as many metrics for hospital readmissions are based on 30-day readmissions, we also performed a sensitivity analysis for PPRs up to 30 days of readmission. The same methods were used as above.

Finally, we used two alternative definitions for PPRs. In our primary analysis we utilized a conservative definition of PPRs that included the 9 potentially preventative conditions—6 from the AHRQ definition and 3 from our surgically-relevant definition—, and therefore performed sensitivity analyses with two alternative definitions. This first alternative definition of PPRs included only the original 6 AHRQ conditions. The second alternative definition of PPRs classified any readmission with a ≤2 day length of stay as potentially preventable.¹¹

Results:

A total of 1,320,083 weighted admissions for trauma care in 2,368 different index hospitals were present in the 2017 NRD. Of these, 137,854 trauma patients were readmitted within 90 days of discharge from their index hospitalization, resulting in a 10.4% readmission rate. Trauma patients with a primary readmission diagnosis code consisting of a PPR accounted for 31,378 90-day readmissions. This resulted in 22.7% of all 90-day readmissions being potentially preventable, and an overall PPR rate of 2.4%.

Table 1 lists the characteristics of trauma patients who were not readmitted, readmitted but not for a PPR, and readmitted for a PPR. The mean age of trauma patients with a PPR was 76.5, while the mean age of our estimation sample was 60.1. Female patients accounted for 60.4% of the PPR population, whereas they consisted of 50.8% of the admitted trauma population. Metropolitan Teaching hospitals have the greatest PPR rate nationally, followed by Metropolitan Non-teaching. There are notable differences in average age, female percentage, Medicaid/Medicare insurance status, and comorbidities between the without readmission and with PPR trauma groups. A summary of the PPR reasons and their frequencies for trauma care is presented in the Supplementary Materials.

Table 2 describes the readmission rates of patients with the characteristics displayed in Table 1, as well as PPR rates and the proportion of readmissions that are PPR. For example, 13.0% of Medicare/Medicaid patients are readmitted with 24.8% of them potentially preventable. 23% of CHF patients are readmitted with 38.9% of them potentially preventable. Higher injury severity score (ISS) was associated with decreased PPRs and increasing number of comorbidities was associated with increasing PPRs (Figure 2). The most common causes for non-PPRs were sepsis (11.9%), infection (3.9%) and pulmonary embolism (1.9%), displayed in the Supplementary Materials.

Table 2. Readmission and Potentially Preventable Readmission Rates by Patient Characteristics.

This table displays the characteristics of the 2017 trauma cohort from the National Readmissions Database by readmission rate, potentially preventable readmission rate, and proportion of readmissions that are potentially preventable.

Patient Characteristic	Readmission Rate (90 days)	Potentially Preventable Readmission Rate (90 days)	Proportion of Readmissions that are PPR
Female	11.2	2.8	25.2%
Insurance Status
Medicare/Medicaid	13.0	3.2	24.8%
Private Insurance	5.3	0.7	13.3%
Uninsured	5.6	0.5	9.5%
Other	5.5	0.8	13.7%
Hospital Metropolitan-Teaching Status
Metropolitan Non-teaching	12.6	3.2	25.3%
Metropolitan Teaching	9.9	2.1	21.1%
Non-metropolitan	10.8	3.3	30.0%
Hospital Bed Size
Small	12.1	3.2	26.6%
Medium	11.3	2.7	23.9%
Large	9.7	2.1	21.1%
Median Household Income Quartile of Patient Zip Code
<25%	10.6	2.4	22.8%
25–50%	10.6	2.4	23.1%
50–75%	10.4	2.3	22.6%
>75%	6.7	2.3	22.5%
Injury Severity Score
<8	10.5	2.5	23.5%
8–15	12.6	3.1	24.3%
16–24	12.5	2.4	19.0%
25+	3.3	0.4	12.4%
Comorbidities
Acute Myocardial Infarction	18.4	5.3	28.7%
Congestive Heart Failure	23.0	8.9	38.9%
Peripheral Vascular Disease	18.3	5.4	29.3%
Cerebrovascular Disease	18.5	3.6	19.6%
Dementia	16.5	4.6	29.0%
Chronic Obstructive Pulmonary Disease	16.2	5.0	31.0%
Rheumatoid Disease	15.1	3.5	22.9%
Peptic Ulcer Disease	27.4	4.0	14.4%
Mild Liver Disease	17.4	3.1	17.9%
Diabetes	13.1	3.4	25.7%
Diabetes with Complications	21.1	7.4	34.9%
Hemiplegia or Paraplegia	14.3	2.0	13.4%
Chronic Kidney Disease	20.8	7.1	34.0%
Cancer	20.0	4.4	21.8%
Moderate to Severe Liver Disease	29.3	3.9	13.5%
Metastatic Cancer	23.2	4.2	17.4%
Acquired Immunodeficiency Syndrome	20.1	3.0	15.1%

Figure 2. Potentially Preventable Readmissions by Injury Severity Score & by Charlson Comorbidity Index.

(a) Bar graph of proportion of readmissions that are potentially preventable by increasing Injury Severity Score (ISS) category, and (b) bar graph of proportion of readmissions that are potentially preventable by increasing Charlson Comorbidity Index (CCI).

Mean cost of PPRs to hospitals was $10,001 (95% CI:$9770-$10231), while the median cost of PPR was $7427. Mean cost of non-PPR 90-day readmissions was $12337 with a 95% CI of $12110–12564, while the median cost of non-PPR 90-day readmissions was $8476. Total cost of PPRs in this cohort was $313,802,278 while the total cost of 90-day readmissions in this cohort was $1,799,772,346.

The total number of readmissions per day, from day 0 to 90 after discharge from index admission, as well as the proportion of readmissions for that day that are PPRs, is shown in Figure 1. The number of readmissions peaked at day 12 with 15,763 cumulative patients readmitted, compared to 39.304 total readmissions between days 60–90 cumulative. The proportion of potentially preventable readmissions increased as length of time from index admission discharge increased, with 20% of readmissions being potentially preventable on day 3 from discharge to 30% on day 90.

Figure 1. Readmissions and Potentially Preventable Readmissions by Day After Discharge.

Two-sided graph displaying the number of readmissions by day of discharge (bar graph) as well as the proportion of readmissions that are potentially preventable by day of discharge (line graph).

On multivariable logistic regression (Table 3), older age was associated with increased odds of PPR (1.3 times increased odds for every 10 years in age). For trauma patients <65 years old, those who had public insurance (Medicaid or Medicare) had a 2.7x increased odds of having a PPR compared to those with Private insurance. Uninsured trauma patients <65 had a 1.3x increased odds of having a PPR than their privately insured counterparts. Household income greater than the 25th%tile was associated with decreased PPR. Patients of any age with CHF had 2.9x increased odds of being readmitted for a potentially preventable cause compared to those without, those with COPD or complicated diabetes had 1.8x increased odds, and those with CKD had 1.7x increased odds.

Table 3. Multivariable Logistic Regression Model for Potentially Preventable Readmissions.

Multivariable regression analysis of potentially preventable readmissions adjusted for age, sex, primary insurance payer (Medicare/Medicaid, private, uninsured, other), hospital location/status, Injury Severity Score, and Charlson Comorbidity Index.

Characteristic	aOR	95% CI	P value
Age, per additional decade	1.29	1.27–1.31	<0.001
Female	0.98	0.95–1.02	0.376
Insurance Status *
For age <65 years
Private Insurance	REF	REF
Medicare/Medicaid	2.71	2.44–3.00	<0.001
Uninsured	1.31	1.09–1.58	0.003
Other	0.83	0.67–1.02	0.077
For age ≥ 65 years
Medicare/Medicaid	REF	REF
Private Insurance	0.78	0.71–0.85	<0.001
Uninsured	0.47	0.29–0.76	0.002
Other	0.86	0.72–1.02	0.019
Hospital Metropolitan-Teaching Status
Metropolitan Non-teaching	REF	REF
Metropolitan Teaching	0.83	0.78–0.87	<0.001
Non-metropolitan	1.00	0.92–1.09	0.992
Hospital Size
Small	REF	REF
Medium	0.92	0.86–0.98	0.012
Large	0.80	0.75–0.85	<0.001
Median Household Income Quartile of Patient Zip Code
<25%	REF	REF
25–50%	0.90	0.86–0.94	<0.001
50–75%	0.87	0.82–0.91	<0.001
>75%	0.84	0.79–0.89	<0.001
Injury Severity Score
<8	REF	REF
8–15	0.96	0.92–0.99	0.024
16–24	0.96	0.90–1.03	0.238
25+	0.14	0.12–0.16	<0.001
Comorbidities
Acute Myocardial Infarction	1.03	0.98–1.09	0.257
Congestive Heart Failure	2.85	2.73–2.97	<0.001
Peripheral Vascular Disease	1.13	1.07–1.19	<0.001
Cerebrovascular Disease	1.01	0.94–1.09	0.742
Dementia	1.20	1.16–1.25	<0.001
Chronic Obstructive Pulmonary Disease	1.82	1.75–1.89	<0.001
Rheumatoid Disease	1.06	0.97–1.16	0.164
Peptic Ulcer Disease	1.15	0.96–1.38	0.138
Mild Liver Disease	1.25	1.13–1.38	<0.001
Diabetes	1.58	1.50–1.65	<0.001
Diabetes with Complications	1.82	1.73–1.92	<0.001
Hemiplegia or Paraplegia	1.06	0.90–1.24	0.495
Chronic Kidney Disease	1.70	1.62–1.78	<0.001
Cancer	1.25	1.14–1.37	<0.001
Moderate to Severe Liver Disease	1.32	1.10–1.59	0.003
Metastatic Cancer	1.39	1.20–1.62	<0.001
Acquired Immunodeficiency Syndrome	1.49	1.05–2.10	0.024

^*Effect of primary payer stratified by age < or ≥ 65 years was estimated with a single model including an interaction term between the covariate for primary payer and a categorial variable for age ≥ 65 years

Abbreviations: aOR, Adjusted Odds Ratio; CI, Confidence Interval

The four multivariable sensitivity analyses—the non-PPR exclusion, 30-day readmissions, and the two alternate definitions of PPRs—showed similar results in PPR associated with age, ISS, income, insurance type, and comorbidities (see Supplemental Materials). Patients in the sensitivity analysis that excluded non-PPRs had slightly increased odds of PPR for CHF, COPD, and complicated diabetes.

Discussion:

The results of this study highlight a novel and quantifiable target for reducing trauma readmissions: reducing potentially preventable readmissions. One out of every five trauma readmissions are potentially preventable. This is a significant economic problem as PPRs in trauma care account for over $300 million in hospital costs yearly. Preventing these readmissions would result in 30,000 fewer readmissions yearly.

Notably, the most important predictors for PPRs were baseline medical comorbidities, not the severity of a patient’s injuries. In fact, patients with higher injury severity have decreased rates of PPRs compared to those patients with lower injury severity. Instead, patients with Medicaid, Medicare, CHF, COPD, and complicated diabetes had 2–3x higher odds of PPRs than those with other comorbidities. These findings suggest that readmission reduction efforts that target severely injured patients may not yield desirable results. Instead, resources should be focused on patients with significant underlying medical comorbidities—especially those that are likely to result in readmission if appropriate ambulatory care is not provided in the vulnerable post-discharge time period.

The Affordable Care Act’s Hospital Readmission Reduction Program (HRRP), first implemented in 2013, has generated significant interest in efforts to reduce readmissions nationwide.^15–18 The HHRP most recently attracted attention when it issued $500 million of penalties to 2,500 hospitals in 2017 for 30-day hospital readmissions.^4,17 Although this policy did not initially include surgical conditions when implemented in 2013, today, two of the six measures are surgical—CABG and total hip arthroplasty—with plans for adding more.^19–21 As more surgical measures are added, 30-day hospital readmissions after emergency surgical procedures are increasingly being scrutinized.^3,22,23 Our study provides additional context for these discussions by naming a pragmatic floor for readmission reduction in trauma care, as well as a potential method for defining pragmatic readmission targets for other specialties. Defining this floor is further relevant because some surgical readmissions are planned^24,25 and if not classified appropriately, could result in inappropriate penalties.

The current body of literature evaluating readmissions among trauma patients have focused on association with particular injury patterns²⁶, patient characteristics²⁷, discharge destination²⁸, time-to-intervention²⁹, and inpatient lab markers³⁰. However, the current study suggests that the readmissions that are most likely to be prevented may not be associated— or may even be inversely associated—with the severity of the patient’s injuries. This may indicate better inpatient trauma care for the injury itself may not decrease readmission rates, but rather better post-discharge ambulatory care. The link between the injury and the readmission, therefore, may be that the injury triggered a physiologically vulnerable state—a temporal vulnerability—leading to more difficult to control comorbidities soon after injury. Furthermore, there has been a recent increase in programs implemented to reduce readmissions, such as post-discharge callback programs^31–33 and interventions to improve readability of discharge paperwork³⁴. One of these callback programs targeted high-risk trauma patients in particular and showed significant readmission reduction rates.³³ This work can help further target the patient populations that are most vulnerable to readmissions that could be potentially prevented.

These results must be interpreted in light of the study’s limitations. The National Readmissions Database (NRD) is ideal for tracking readmissions through different hospital systems but does not allow for a full assessment of the amount or quality of the patient’s clinical interventions. It is also known that trauma outcomes are associated with race/ethnicity^35,36, however this is not captured in the NRD.³⁷ The NRD also does not capture Emergency Department admissions or hospital admissions before the index hospitalization, as well as details on the discharge disposition. Furthermore, comorbidities, access to health care, and post-trauma outcomes are all related to the risk of readmission. As an observational study, it is difficult to determine if comorbidities can be separated out as an intervenable characteristic of the already vulnerable readmitted trauma patient population, or determine how much of an impact that the index trauma had on the readmission. Access to a hospital for readmission is also inherent in this study design—the population not captured in this dataset are trauma patients who were not able to access any care at the time of readmission need, and could change the baseline readmission rate if access to care was improved. A prospective, clinically-based study would allow for more granularity into the effects of care received, comorbidities, and access to care as separate entities.

The implications of this study can be discussed in context of two broad categories for readmission reduction: implications for targets readmission reduction programs and broader systems impact. Programmatic approaches to reducing readmissions, such as those mentioned above, are at the hospital or clinic level and have been implemented in several trauma centers throughout the country. These programs aptly require time and monetary resources that could be even more effective at reducing readmissions if directed towards higher risk patients such as those with Medicaid, Medicare, CHF, COPD, and complicated diabetes. Again, these patients may be at higher risk for readmission after their traumatic injury not due to the quality of the care of the acute injury they received, but the temporal vulnerability of experiencing a in injury while having these co-morbidities. Furthermore, since these potentially preventable readmissions are more likely to be medical rather than surgical or traumatic, programs may also consider focusing on ensuring timely access to PCP follow up care for these particularly high-risk trauma patients. This work could also be used to inform a larger systems-level program intervention, such as a trauma quality initiative to reduce readmissions, to potentially improve outcomes and reduce spending.

Conclusion

One out of every five trauma readmissions is potentially preventable. These potentially preventable readmissions are not driven by severity of a patient’s injuries but rather underlying medical comorbidities which may be exacerbated during post-injury recovery. Potentially preventable trauma readmissions cost the trauma system $300 million annually but may be prevented by improving access to primary post-discharge follow up care. Focusing efforts on high-risk patients may help reduce these high-cost, low-value readmissions and improve patients’ overall outcomes as they recover from injury.