This investigation was a retrospective, observational, longitudinal study of a cohort of adults aged (55+) who underwent a trauma team evaluation in a Level 1 Trauma Center after a fall from May 2020 to July 2021. Cohort members were followed from their initial emergency medical services (EMS) evaluation (when applicable) through discharge from the ED, transfer to another facility, or during their hospitalization from the index ED visit. The investigation involved a combination of data downloaded from a trauma registry and electronic health records (EHR) and toxicological analyses of individual blood samples that had been collected as part of a prior study (20). The study was approved by the Advarra Institutional Review Board (IRB) (#00022129; central IRB) and the study site’s IRB.

Injury Severity Scores (ISS) were obtained from the study site’s trauma registry. ISS ranges from 0 to 75. Zero reflects no apparent injury, and 75 reflects a fatal injury. If an encounter was not entered into this repository, a manual chart review and ISS calculation were conducted by one of the study physicians, who is board-certified in trauma and critical care. Study data abstracted from the EHR included patient demographic characteristics (sex, race, ethnicity), medical history, pertinent timestamps (ED triage, admission, discharge), home medications, initial ED laboratory test results, and final disposition (admission, discharge, transfer, etc.). In addition, trained study staff conducted manual EHR abstraction for the administration of benzodiazepines and opioids in the hospital and pre-hospital settings (via EMS ambulance run reports).

Blood specimens from each cohort member’s initial sampling upon ED arrival were bio-banked and stored refrigerated before being frozen at −80 °C. All samples were analyzed for ethanol, cannabinoids, benzodiazepines, and opioids (illicit and prescription) at the Center for Forensic Science Research and Education (Willow Grove, PA). Benzodiazepines and opioids restricted to medical settings and unlikely to be obtained by patients elsewhere (e.g., midazolam, etc.) were not included in subsequent analyses examining the contribution of benzodiazepines and opioids to fall-related outcomes.

Blood ethanol concentrations were acquired via gas chromatography flame ionization detection (GC-FID). Comprehensive toxicological analysis was conducted for >1200 drug targets, and qualitative drug results for benzodiazepines, cannabinoids, and opioids were acquired via liquid chromatography quadrupole time-of-flight mass spectrometry (LC-QTOF-MS; see Supplementary Material for list). Cannabinoid testing included, but was not limited to, assessing for the presence of delta-9 tetrahydrocannabinol (THC) and metabolites (11-hydroxy-delta-9 tetrahydrocannabinol and 11-nor-9-carboxy-delta-9 tetrahydrocannabinol). Benzodiazepine testing included, but was not limited to, evaluating for the presence of diazepam, alprazolam, lorazepam, and clonazepam, in addition to their metabolites (e.g., nordiazepam, alpha-hydroxy alprazolam, 7-amino clonazepam). Opioid testing included, but was not limited to, assessing for the presence of tramadol, fentanyl, hydrocodone, dihydrocodeine, oxycodone, morphine, codeine, methadone, beta-h-fentanyl, and 4-anilino-N-phenethyl-piperidine (4-ANPP), among appropriate metabolites (e.g., norfentanyl, noroxycodone, etc.).

Numerous factors contribute to falls and fall-related adverse clinical outcomes (21). Many factors are not obtainable (e.g., walking on a slippery floor, transient vertigo, etc.) from the data sources available to this retrospective study or are not known (e.g., frailty, imbalance, etc.) about cohort members. Previous researchers have compiled factors associated with either risk for falls or adverse fall-related clinical outcomes among adults aged 65 + (4,5,8,9). However, these factors have not been categorized by their relative contribution or importance to fall risk or adverse consequences of a fall. Further, the list of covariates that could affect the relationship between the exposures of interest (substance use) and fall outcome is lengthy, and including them in regression analyses for a cohort of 274 participants is not feasible.

To overcome this problem, we conducted propensity score analyses that calculated the propensity of adults aged 55+ to present to the ED for a trauma evaluation after a fall, as opposed to presenting for other mechanisms of injury. We used propensity score adjustment to control for potential confounding in estimating the effect of the exposures of interest on the outcome (22,23). The propensity score was defined as the probability of presenting to the ED for a trauma evaluation for a fall, as opposed to other mechanisms of injury, conditional on the covariates. The propensity scores were used in the models as a moderator of baseline characteristics for examining the relationship between exposures of interest and the outcomes.

The propensity score calculations included the 274 fall cohort members and 1286 adults aged 55+ who underwent a trauma team evaluation for serious injury after another mechanism of injury at this facility from January 2020 to December 2021. In addition to demographic characteristics, the propensity score components included medical history, medications, and abnormal clinical laboratory findings that previous research has associated with either fall risk or adverse clinical outcomes from a fall among adults aged 55+. Except for age and race, all factors were entered into the propensity score model as binary variables (present vs. absent) without any attempt to assign greater weight importance to any variable. Propensity scores were calculated according to age, sex, race, past medical history, medications, and laboratory abnormalities. The resultant propensity scores for the fall cohort were included in the subsequent clinical outcomes analyses.

Demographic characteristics, medical history, home medications, and clinical laboratory test results upon ED arrival were summarized for fall cohort members and adults ages 55 and older who underwent a trauma evaluation for other mechanisms of injury; these data were included in the propensity score analyses (Supplement). All analyses were conducted in Stata 17.0 (StataCorp; College Station, TX).

Negative binomial regression models were constructed to examine clinical outcomes through hospitalization (e.g., injury types such as fractures or intracranial hemorrhage; disposition after treatment including admission, transfer to intensive care unit [ICU], and discharge to home, hospice or skilled nursing facility). Logistic regression models for the outcomes of intracranial hemorrhage, hip/pelvic fractures, and spinal fractures; and linear regression models for ED and hospital length of stay were developed. Variables included in these models were the aforementioned propensity score (propensity for presenting for a trauma evaluation after a fall, as opposed to other types of trauma), injury severity score (from the initial trauma team evaluation), and the four classes of substances tested for in the toxicological analyses (ethanol, benzodiazepines, cannabinoids, and opioids).

The models evaluated each of the respective outcomes (e.g., admission) as a function of whether each substance was detected vs. not detected (i.e., binary variable). The relationship assessed in the model between substance detected and each outcome was adjusted for propensity score (i.e., propensity to present to the ED for a fall instead of another type of trauma) and ISS. The ISS were re-categorized as zero to four. The ISS distribution for the fall group was: 9% zero, 49% one, 21% two, 14% three, and 7% four. An ISS of zero was the reference category (i.e., no apparent injury) was the reference. Separate models were formed for each of the four substance classes.

After accounting for drugs administered by emergency medical services or clinical providers at the hospital prior to blood being obtained, prevalence rates of substances detected were 21.2% (95% CI, 16.5–26.5%) opioids, 18.6% (95% CI, 14.2–23.7%) ethanol, 13.9% (95% CI, 10.0–18.5%) benzodiazepines, and 9.1% (95% CI, 6.0–13.2%) cannabinoids. Ethanol concentrations ranged from 30 to 320 mg/dL, with a median concentration of 180 mg/mL and a mean concentration of 170 mg/dL. Multiple substances were detected in some blood samples: 2.3% had benzodiazepines and opioids, 2.0% had cannabinoids and opioids, 1.0% had ethanol and benzodiazepines, and 0.7% had cannabinoids and benzodiazepines (Data not shown).

As shown in Table 5, 42.7% of the fall cohort sustained a fracture, and of these 117 participants with a fracture, 45 (38.5%) had a spinal fracture, and 26 (22.2%) had an extremity fracture. Approximately 17% had an intracranial hemorrhage, and of these 46 participants, 31 (67.4%) had a subarachnoid hemorrhage. Most patients were admitted to the hospital, although few required an intensive care unit stay. Among the 188 (68.6%) admitted to the hospital, 15 (7.8%) were discharged to a rehabilitation center or skilled nursing facility. There were 28 deaths (10.2%) either during the ED visit or hospitalization among the fall cohort. Mean ED length of stay (LOS) exceeded one day (60.2 h), the median was 63 h (IQR 41–70 h), and the range was 27–137 h. Among the 192 admitted to the hospital, the mean hospital LOS was 12 days, the median was 7.6 days (IQR 5.3–10.6), and the range was 1.4–185.1 days.

The prevalence of ethanol, benzodiazepines, cannabinoids, and opioids detected by adverse fall outcomes are portrayed in Table 6. As shown, the occurrence of adverse clinical outcomes (hip/thigh/pelvis fractures, intracranial hemorrhages, ED length of stay, hospital admission, hospital length of stay and death) was similar whether or not ethanol was detected. As shown in the models (Table 7), benzodiazepine use was associated with greater odds of an intracranial hemorrhage and longer ED LOS. Opioids were associated with greater odds of extremity fractures and admission to the hospital. Any substance use was also associated with extremity fractures and ED LOS. Odds of hospital admission and hospital admission LOS were lower when ethanol was present. There were no associations with any of the adverse fall outcomes for cannabinoids.

The findings from this investigation are subject to several limitations. First, the prevalence estimates might not be externally valid to adults aged 55+ who fall but do not require evaluation for a severe injury, are not evaluated in trauma centers, or are not assessed in Level 1 trauma centers with similar catchment area populations. Second, we cannot determine if the substances detected caused the fall. Although identification of these substances in blood samples supports recent use (as compared to urine drug testing, which has a longer window of detection), the timing, frequency, and chronicity of substance ingestion in relation to the fall is not known. In addition, because providers may prescribe certain benzodiazepines and opioids, we could not determine if these substances were used as prescribed or misused. Concerning elapsed time, we could not accurately ascertain the interval between fall and blood specimen acquisition in this study; additionally, patients may have been evaluated by the trauma team after transfer from another hospital, leading to variability in time to presentation.