a b s t r a c t

Background: The purpose of this study was to identify risk factors of mortality for Geriatric patients who fell from ground level at home and had a normal physiological examination at the scene.

Methods: Patients aged 65 and above, who sustained a ground level fall (GLF) with normal scene Glasgow Coma Scale score 15, systolic blood pressure (SBP) N 90 and b160 mmHg, heart rate >= 60 and <=100 beats per min- ute) from the 2012-2014 National Trauma Data Bank data sets were included in the study. Patients’ char- acteristics, existing comorbidities [history of smoking, chronic kidney disease (CKD), cerebrovascular accident (CVA), diabetes mellitus , and hypertension (HTN) requiring medication], injury severity scores (ISS), American College of Surgeons’ (ACS) trauma center designation level, and outcomes were examined for each case. Risks factors of mortality were identified using bivariate analysis and logistic regression modeling.

Results: A total of 40,800 patients satisfied the study inclusion criteria. The findings of the logistic regression model for mortality using the covariates age, sex, race, SBP, ISS, ACS trauma level, smoking status, CKD, CVA, DM, and HTN were associated with a higher risk of mortality (p b .05). The fitted model had an Area under the Curve (AUC) measure of 0.75.

Conclusion: Cases of geriatric patients who look normal after a fall from ground level at home can still be associ- ated with higher risk of in-hospital death, particularly those who are older, male, have certain comorbidities. These higher-risk patients should be triaged to the hospital with proper evaluation and management.

(C) 2019

Introduction

Falls and fall-related injuries and death pose a major challenge to global human health. According to reports from the World Health Orga- nization in January 2018, falls are still the second leading cause of acci- dental or unintentional injury deaths, resulting in around 37 million people worldwide seeking medical care annually, and of that 650,000 die [1]. The largest morbidity and mortality occur in patient aged 65 years and older.

As our geriatric (age >= 65 years) population is on rise so is the inci- dence of Ground level fall (GLF) and death associated with it [2,3]. The patient who survived initial hospital admission and get discharged

? Note. The data was presented (oral) at the Annual Clinical Congress American College of Surgeon, 2018, October 21-25, Boston, MA.

* Corresponding author at: Department of Surgery, Division of Trauma & Surgical Critical Care, Jersey Shore University Medical Center, 1945 State Route 33, Neptune, NJ 07754, USA.

E-mail address: [email protected] (N. Ahmed).

also found to have higher mortality in one year follow up [3,4]. Highest mortality was seen on those patients who were discharged to skilled nursing facility compared to home without any assistance [4].

Previous studies have looked at the risk factors of mortality of all GLF and found old age, high injury severity score (ISS), low Glasgow Coma Scale score, male gender, presence of severe traumatic brain in- jury and solid Organ injuries to be associated with mortality [5,6]. Most of the in-hospital mortality resulted from traumatic injury, not necessarily due to preexisting comorbidities, however, later death showed some correlation with patient chronic health conditions i.e. with the history of cardiac disease, stroke, diabetes etc. [7]. Since in- hospital mortality in geriatric patients following a GLF can be directly re- lated to injury and injury severity, however, comorbidities can be a con- tributing factor. Therefore this study was designed to evaluate cohort of patients who fell from a ground level and who presented with normal GCS score and normal systolic Blood pressure and heart rate at the scene using patient’s characteristics, injury factors and comorbidities to identify the risk factors for mortality. The implication of the study would be once we identified the risk factors in this cohort, perhaps

https://doi.org/10.1016/j.ajem.2019.12.031

0735-6757/(C) 2019

better strategy would be instituted to high risk patients for better triaging the patients at the scene and for better in-patient monitoring and management as well.

Methods

Patient level data from January 1, 2012- December 31, 2014 from the US National Trauma Data Bank (NTDB), which is the largest trauma da- tabase housing millions of data records for injured patients all across the US, was utilized for this study. All patient records in the NTDB are de- identified and contributed on a voluntary basis by the participating health care institutions. Inclusion of these patient records in the NTDB database only requires that the injury diagnosis code satisfy the follow- ing criteria as defined by the International Classification of Diseases, Ninth Revision, Clinical Modifier (ICD-9-CM): 800-959.9, excluding the (ICD-9-CM): 905-909.9 (later complications of injury), 910-924.9 (superficial injury), and 930-939.9 (generic foreign body injuries) [8].

For this study, all NTDB patients who experienced a ground level fall

at home and were brought to a hospital, who were 65 years of age or older, who suffered a traumatic injury, and whose initial evaluation at the scene showed a normal systolic blood pressure (SBP) [Reference Range: 90-160 mm Hg], heart rate (HR) [Reference Range: 60-100 beats per minute], and a Glasgow Coma Scale of 15 were included in

95% confidence intervals, as well as estimated odd ratios (OR) with 95% confidence intervals as measures of precision.

All p-values reported were 2-sided and a p-value b .05 was consid- ered statistically significant. All data cleaning and statistical analyses were performed using both “R: A language and environment for statis- tical computing” [9] and STATA15 [10].

Results

Unadjusted analysis

Among a total of 40,800 patients who met study inclusion criteria, 938 (2.3%) patients died in the hospital; the remaining 39,862 (97.7%) patients survived to hospital discharge. There were significant baseline demographic differences found between these two outcome-based pa- tients groups (those who died and those who survived) regarding age (median [IQR]) (82.0 [77.0, 86.0], vs. 80.0 [73.0, 85.0], P b .001) and bi-

ological sex (male: 49.8% vs. 30.6%, P b .001). No significant difference was found regarding Patient race (P = .46). The patients who died pre- sented with a significantly higher ISS (median [IQR]: 9.0 [9.0, 14.0] vs.

Table 1

Bivariate analysis between the groups

the study [8]. Other variables collected for this study included: biologi- cal sex, race and ethnicity, respiratory rate (RR), injury severity score (ISS), existing comorbidities including: current smoking, chronic kidney

Patient characteristics Alive at discharge

(n = 39,862)

Died at discharge (n = 938)

P-value

disease (CKD), cerebrovascular accident/neurologic deficit (CVA), dia-

Age (in years) b0.001

betes mellitus (DM), and hypertension requiring medication (HTN) sta-

tuses. The primary objective of the study was to determine the

Median [IQR] 80.0 [73.0,

85.0]

82.0 [77.0,

86.0]

incidence of in-hospital mortality, and develop an internally validated risk model to be able to better identify high risk cases in this patient

Biological sex, n (%) b0.001

Female 27,665 (69.4) 471 (50.2)

Male 12,197 (30.6) 467 (49.8)

population.	Race, n (%)		0.46
Patient characteristics were first summarized using median and in- American Indian		123 (0.3)	2 (0.2)
terquartile range (IQR), or frequency and percentage as appropriate. Asian		620 (1.5)	11 (1.2)
All study variables were then compared between the patients who sur- Black/AA		1229 (3.1)	20 (2.1)
Latino		763 (1.9)	14 (1.5)
vived to discharge versus those who did not survive using Wilcoxon Native Hawaiian/Pacific Islander		22 (0.1)	0 (0.0)
rank-sum tests for the continuous and ordinal variables, and Pearson’s Other		482 (1.2)	8 (0.9)
chi-squared tests or Fisher’s exact tests for the categorical variables. Fol- White		36,623 (91.9)	883 (94.1)
lowing the initial bivariate comparisons, 80% of the patient data (n = Systolic blood pressure at admission			b0.00
32,640) was then randomly drawn without replacement into a develop-	Median [IQR]	138.0 [124.0,	133.0 [118.0,
ment dataset using statistical software, while the remaining 20% (n =		149.0]	146.0]
8160) was used as an internal validation dataset. More specifically,	b110	2904 (7.3)	116 (12.4)

1

(mm Hg)

this was accomplished using a function in “R” statistical software explic- itly built to take an imported data set and split it into two or more parts, utilizing a random seed value to start off a random number generation process which terminated when the desired number of patients was reached in each subset of the data.

>=110 Heart rate/Pulse at admission	36,958 (92.7)	822 (87.6)	0.04
Median [IQR]	80.0 [72.0,	80.0 [72.0,
	88.0]	90.0]
Respiratory rate at admission Median [IQR]	18.0 [16.0,	18.0 [16.0,	0.01
	18.0]	20.0]
Missing Injury severity score Median [IQR]	919 (2.3) 9.0 [4.0, 9.0]	16 (1.7) 9.0 [9.0, 14.0]	b0.001
Current smoker, n (%) No	37,154 (93.2)	879 (93.7)	0.59
Yes	2708 (6.8)	59 (6.3)
Chronic kidney disease or on dialysis, n (%) No	38,755 (97.2)	868 (92.5)	b0.001
Yes CerebroVascular injury or neurological	1107 (2.8)	70 (7.5)	0.70
deficit, n (%)

multiple logistic regression analysis was then performed using pa- tient records in the development dataset to produce a survival predic- tion model adjusted for age, gender, race, SBP, HR, RR, ISS, smoking status, CKD status, CVA status, DM status, and HTN status. Any patient records with missing or incomplete information for these selected var- iables were excluded during the software modeling process. The pri- mary purpose of building this model was to identify any significant associations between these critical clinical factors and patient survival to discharge. The modeling variable Selection process combined prior literature considerations, clinical input, and both an initial stepwise-

AIC process and a backward-Likelihood Ratio process as confirmation	No	37,537 (94.2)	880 (93.8)
of factor selection. After the regression model was finalized using the	Yes	2325 (5.8)	58 (6.2)

development dataset, it was fit to the validation dataset to calculate and compare the predicted survival probabilities to the true outcomes for the remaining patients. A receiver-operating characteristic (ROC)

Diabetes mellitus, n (%)			0.15
No	30,456 (76.4)	697 (74.3)
Yes	9406 (23.6)	241 (25.7)
Hypertension (req. meds), n (%)			0.003

curve, along with a corresponding area-under-the-curve (AUC) calcula-	No	14,940 (37.5)	307 (32.7)
tion and a Hosmer-Lemeshow goodness-of-Fit test was used to assess	Yes	24,922 (62.5)	631 (67.3)

the overall validity of the final model. Parameter estimates from the

fitted model were summarized using beta coefficient estimates with

Any time P value b 0.05 is considered as statistically significant. All bold values in this table are statistically significant.

9.0 [4.0, 9.0], P b .001), and a significantly higher rate of comorbidities, including chronic kidney disease (CKD) and hypertension requiring medication (HTN), compared to those who survived (7.5% vs. 2.8%, and 67.3% vs. 62.5%, all P b .05), respectively. No significant differences were found between the groups regarding their reported rates of a his- tory of diabetes, smoking, or Cerebrovascular accidents as shown in Table 1.

Looking into the organ injuries of these patient cohort, brain contu- sion and subarachnoid hemorrhage were the most frequent injury found in 21.86% and 21.48% respectively. subdural hematoma was found in 19.93% of patients. Intertrochanteric fracture and femoral head fracture were seen in 13.76% and 2.87% respectively. Rib fractures were seen in 3.63% and c-spine fractures were seen in 1.75% of patients. The group of patients who died sustained a higher frequency of Brain hemorrhage and cervical spine injury. Femoral neck or intertrochanteric fractures were not significantly different in patients who died or who survived.

Adjusted multivariate analysis

A multivariate logistic regression model was then fit using patient level variables that were either shown to be significantly different in the unadjusted analyses or were known to have the potential for clinical impact on the odds of patient mortality. Those variables included: the patient’s age, biological sex, race, ISS, and comorbidities (including CKD and HTN). The regression ? coefficients provide the estimated change in the log- odds of mortality for a one unit or a category change for the predictor variable. The results showed that those with older age, male gender, higher ISS, HR, and RR and lower SBP at presentation, and a history of CKD, DM, and HTN had significantly higher odds of in- hospital mortality as shown in Table 2. We also tested the model to eval- uate whether higher level care (trauma center designation I & II) versus lower level care (trauma center designation III & IV) had any impact on mortality. The risk model did not show any impact based on whether the patients were brought to a higher level trauma center or not, which was consistent with another recent study on ground level falls [7].

To test the performance of the model used to predict the patient in- hospital mortality, a receiver operating characteristic (ROC) curve was fit using the testing/validation data set (Fig. 1), and the area under the curve (AUC) value was 0.748 [95% CI: 0.742, 0.808]. A sensitivity

analysis was also performed to assess the validity of the model by com- paring the observed mortality rates versus the Expected mortality rates based on the fitted regression model (Fig. 2). With a Hosmer-Lemeshow p-value of N0.05 there is no significant evidence for a lack of model fit, which is also supported by Fig. 2, where the predicted mortality rates for each decile of the fitted risk values in this patient sample are clus- tered on or closely around the diagonal line of equivalence.

Discussion

Our study showed that among geriatric patients who fell from a ground level height at home, who had a normal neurological examina- tion and normal vital signs (SBP, HR) and were brought to the hospital there was a 2.3% incidence of in-hospital mortality. While older age, male gender, lower SBP, higher HR, and RR, ISS, and a history of CKD, DM, and HTN requiring medications were associated with a higher risk of in-hospital mortality. However, ACS trauma center designation was not significantly associated with in-hospital mortality.

To most people, falls from a ground level at home may seem to be a

trivial mechanism of injury; however, in geriatric patients, this can re- sult in not only Significant injuries but also in-hospital mortality. Not only is the number of people in the Geriatric population on the rise, but the incidence of fall and fall-related mortality and morbidity is on the rise as well [2,3,11]. Sustaining a traumatic brain injury following a fall has been reported as one of the most common injuries in the geri- atric patients and is a risk factor for mortality, particularly when the pa- tient presents with a depressed GCS level [5,6,12]. Other risk factors for mortality following ground level falls are reported to include old age (particularly >=70 years), high ISS, low GCS, and some comorbidities [5,13,14]. One such study from Padr?n-monedero and colleagues, ana- lyzed the data of around 32,000 patients who sustained Hip fractures after a fall including their comorbidities and the resulting impact on mortality [15]. The authors found a higher odds of mortality among pa- tients who had a history of congestive heart failure (OR: 3.88; 95% CI: 3.42-4.41), metastatic cancer (OR: 3.44; 95% CI: 2.27-5.20), fluid and

Electrolyte disorders (OR: 2.95; 95% CI: 2.47-3.52), coagulopathy (OR:

2.87; 95% CI: 2.08-3.96), and liver disease (OR: 2.40; 95% CI:

1.82-3.17). Additionally, other studies have shown that if the severely injured patients are treated in a higher level trauma center there is a resulting Survival benefit [13,16,17]. However, recent studies did not show a survival benefit for geriatric patients who had sustained a

Table 2

Risk assessment model for in-hospital survival among geriatric patients with an unintentional fall from ground level

Model variablesa	? coefficient	Lower bound of 95% CI for ?	Upper bound of 95% CI for ?	Odds ratio	Lower bound of 95% CI for OR	Upper bound of 95% CI for OR	P-value
(Intercept)	-9.160	-12.230	-7.119	-	-	-	b0.001
Age	0.049	0.037	0.061	1.051	1.038	1.063	b0.001
Gender: Maleb	0.675	0.524	0.825	1.963	1.688	2.283	b0.001
Race: Asianb	0.586	-1.121	3.513	1.798	0.326	33.564	0.583
Race: Blackb	0.513	-1.110	3.418	1.671	0.329	30.514	0.622
Race: Latinob	0.363	-1.342	3.289	1.438	0.261	26.826	0.734
Race: NH or PIb	-9.983	-160.186	-148.668	0.000	0.000	0.000	0.958
Race: Otherb	0.919	-0.811	3.852	2.507	0.444	47.093	0.392
Race: Whiteb	0.918	-0.602	3.793	2.505	0.548	44.387	0.364
SBP	-0.016	-0.021	-0.012	0.984	0.980	0.988	b0.001
Heart rate	0.011	0.004	0.018	1.011	1.004	1.018	0.002
Respiratory rate	0.023	0.005	0.039	1.023	1.005	1.039	0.006
ISS	0.111	0.099	0.122	1.117	1.104	1.130	b0.001
ACS level 3,4, NAb	-0.054	-0.203	0.094	0.947	0.816	1.099	0.474
Smoking statusb	-0.049	-0.384	0.261	0.952	0.681	1.298	0.765
CKD statusb	0.919	0.615	1.204	2.508	1.850	3.334	b0.001
CVA statusb	0.029	-0.287	0.321	1.029	0.751	1.379	0.852
DM statusb	0.177	0.004	0.347	1.194	1.004	1.415	0.042
HTN statusb	0.182	0.022	0.343	1.199	1.023	1.409	0.027

NH = Native Hawaiian; PI = Pacific Islander; SBP = Systolic blood pressure; ISS = injury severity score; CKD = Chronic kidney disease; CVA = Cerebrovascular injury/Neurological Def- icit; DM = Diabetes mellitus; HTN = Hypertension.

^a 935 patients were removed from the modeling process as they did not have a respiratory rate recorded upon admission.

^b Model reference groups were: Female, American Indian, ACS Level 1 &2, No history of Smoking, No history of CKD, No history of CVA, No history of DM, and No history of HTN.

Fig. 1. Receiver operating characteristic (ROC) curve for the full prognostic model performance in predicting survival in the complete records from the dataset.

ground level fall, sustained mild to moderate injuries, and were brought to a higher level trauma center [8,18,19].

Although the current study echoes some of the findings mentioned above, we specifically looked at all patients who sustained a ground level fall at home who had completely normal physiological measure- ments. In practice, there is occasionally some resistance from geriatric patients, particularly if patient feels “ok”, to go to the hospital for med- ical evaluation. Similarly, over triaging may overwhelm the existing re- sources in the surrounding area hospitals. Therefore, the current study tried to answer the question about who are the patients who need eval- uation after a ground level fall. Our study showed that 20% of patients suffered from brain hemorrhage despite normal GCS score at the scene. This study also showed 2.3% patients died in the hospital after ad- mission following a ground level fall. We developed and validated a risk model to identify factors early that are associated with mortality during the initial hospital stay in this patient cohort. This was achieved by ran- domly selected 80% of the data to develop the model and we used the remaining 20% of the data for validation using multivariate logistic re- gression analysis. The most significant factors associated with increase in in-hospital mortality in this patient population were advanced age, male gender, lower SBP, higher ISS, and a history of CKD, DM, and HTN. In this study, patients only suffered mild to moderate injuries, which may be one of the reasons why patients receiving higher level care did not impact the outcomes as other had suggested [13]. The risk of mortality and readmission both steadily rise after a GLF for

Fig. 2. Observed versus predicted average survival rates for the full prognostic model with the Hosmer-Lemeshow goodness-of-fit test result.

geriatric patients, with a one-year mortality rate that has been reported to be as high as 15% [20,21]. Therefore, triaging these geriatric patients to higher level (ACS level I and level II) trauma centers, where the most up-to-date guidelines and proper resources are available for the care of these patients as mandated by the ACS, should help in minimiz- ing their longer-term mortality and readmission rates as a direct result of the GLF.

Limitations

The current study was performed utilizing an existing national level data set from the NTDB. This is a retrospective study, which carries in- herent challenges; however, it is the largest study performed on this topic to identify high risk patients who may need immediate care after a ground level fall. Regretfully, the detailed information of the pa- tients particular Medication lists and detailed comorbidities are not pro- vided in the NTDB data sets, otherwise we would have included those additional measures. However, we performed a multivariate logistic re- gression analysis to identify the risk factors for mortality among those who sustained a ground level fall using nearly all the clinical factors available to us, so that appropriate patients can be taken for follow-up medical evaluation.

Conclusion

Following a ground level fall, geriatric patients who look and feel normal upon physiological examination at the scene may still be at risk of mortality. Certain patient’s characteristics, and comorbidities are higher risk of mortality and should be triaged to an appropriate hos- pital for follow-up medical evaluation. Normal physiological measures at the scene do not eliminate the risk of in-hospital mortality.

Contributions

First author, Nasim Ahmed 2nd author, Patricia Greenberg conceived and designed the study, First author pulled the data from the NTDB. 2nd author was responsible for all data manipulation and statistical analy- ses. All authors contributed to the manuscript creation.

Compliance with ethical standards

All procedures followed were in accordance with the ethical stan- dards of the Institutional Review Board of Meridian Health and with the Helsinki Declaration of 1975, as revised in 2008. This study analyzed de-identified data from a national database from the American College of Surgeons that is available to researchers and was exempted from IRB review as per policy.

Informed consent

Given that this study utilized a de-identified national database from the American College of Surgeons that is available to all researchers, this study was exempted from IRB review as per policy and no informed consent was required.

Declaration of competing interest

All authors declare no conflict of interest.

References

1. World Health Organization. Falls. URL http://www.who.int/news-room/fact-sheets/ detail/falls; January 2018.
2. Centers for Disease Control and Prevention. Important facts about falls. 10 Feb https://www.cdc.gov/homeandrecreationalsafety/falls/adultfalls.html; 2017.
3. Ayoung-Chee P, McIntyre L, Ebel BE, Mack CD, McCormick W, Maier RV. Long-term outcomes of ground-level falls in the elderly. J Trauma Acute Care Surg 2014;76(2): 498-503 Feb.
4. Ayoung-Chee PR, Rivara FP, Weiser T, Maier RV, Arbabi S. Beyond the hospital doors: improving long-term outcomes for elderly trauma patients. J Trauma Acute Care Surg 2015;78(4):837-43 Apr.
5. Spaniolas K, Cheng JD, Gestring ML, Sangosanya A, Stassen NA, et al. Ground level falls are associated with significant mortality in elderly patients. J Trauma 2010; 69:821-5.
6. Ahmed N, Soroush A, Kuo YH, Davis JM. Risk associated with traumatic intracranial bleed and outcome in patients following a fall from a standing position. Eur J Trauma Emerg Surg 2015;41(3):307-11 Jun.
7. Mangram A, Dzandu J, Harootunian G, Zhou N, Sohn J, et al. Why elderly patients with ground level falls die within 30 DaysAnd beyond? J Gerontol Geriatr Res 2016;5:289. https://doi.org/10.4172/2167-7182.
8. Ahmed N, Greenberg P. Examining outcomes in cases of elderly patients who fell from ground level at home with normal vital signs at the scene: an analysis of the National Trauma Data Bank. J Trauma Acute Care Surg 2019;87:672-7.
9. R Core Team. R: A Language and Environment for Statistical Computing. Vienna, Austria: R Foundation for Statistical Computing; 2018 URL https://www.R-project. org/.
10. StataCorp. Stata Statistical Software: Release 15. StataCorp LLC: College Station, TX; 2017.
11. Padron-Monedero A, Damian J, Martin MP, Fernandez-Cuenca R. Mortality trends for accidental falls in older people in Spain, 2000-2015. BMC Geriatr 2017;17:276 28 Nov.
12. Ahmed N, Bialowas C, Kuo YH, Zawodniak L. Impact of preinjury anticoagulation in patients with traumatic brain injury. South Med J 2009;102:476-80.
13. Garwe T, Cowan LD, Neas B, Cathey T, Danford BC, Greenawalt P. Survival benefit of transfer to Tertiary trauma centers for major trauma patients initially presenting to nontertiary trauma centers. Acad Emerg Med 2010;17(11):1223-32 Nov.
14. Hindmarsh D, Loh M, Finch CF, Hayen A, Close JC. Effect of comorbidity on relative survival following hospitalisation for fall-related hip fracture in older people. Australas J Ageing 2014;33(3):E1-7. https://doi.org/10.1111/j.1741-6612.2012. 00638.x Sep. [Epub 2012 Sep 30].
15. Padron-Monedero A, Lopez-Cuadrado T, Galan I, Martinez-Sanchez EV, Martin P, Fernandez-Cuenca R. Effect of comorbidities on the association between age and hospital mortality after fall-related hip fracture in elderly patients. Osteoporos Int 2017;28(5):1559-68. https://doi.org/10.1007/s00198-017-3926-2 May. [Epub

    2017 Feb 3].

    Metcalfe D, Bouamra O, Parsons NR, Aletrari MO, Lecky FE, Costa ML. Efect of re- gional trauma centralization on volume, injury severity and outcomes of injured pa- tients admitted to trauma centres. Br J Surg 2014;101:959-64 [17].

16. Zacher MT, Kanz K-G, Hanschen M, Haberle S, van Griensven M, Lefering R, et al. As- sociation between volume of severely injured patients and mortality in German trauma hospitals. BJS 2015;102:1213-9.
17. Ahmed N, Kuo YH. Evaluating the outcomes of blunt Thoracic trauma in elderly pa- tients following a fall from a ground level: higher level care institution vs. lower level care institution. Eur J Trauma Emerg Surg 2019. https://doi.org/10.1007/ s00068-019-01230-1.
18. Velmahos GC, Jindal A, Chan LS, Murray JA, Vassiliu P, Berne TV, et al. “Insignificant”

    mechanism of injury: not to be taken lightly. J Am Coll Surg 2001;192(2):147-52.

    Liu SW, Obermeyer Z, Chang Y, Shankar KN. Frequency of ED revisits and death among older adults after a fall. Am J Emerg Med 2015;33(8):1012-8 Aug.

19. Galet C, Zhou Y, Ten Eyck P, Romanowski KS. Fall injuries, associated deaths, and 30- day readmission for subsequent falls are increasing in the elderly US population: a query of the WHO mortality database and National Readmission Database from 2010 to 2014. Clin Epidemiol 2018;10:1627-37.